34
35/*
36 * Uncomment the following extensions for better performance in a VM,
37 * especially if you have support in the hypervisor.
38 * See http://info.iet.unipi.it/~luigi/netmap/
39 */
40// #define BATCH_DISPATCH
41// #define NIC_SEND_COMBINING
42// #define NIC_PARAVIRT /* enable virtio-like synchronization */
43
44#include "opt_inet.h"
45#include "opt_inet6.h"
46
47#ifdef HAVE_KERNEL_OPTION_HEADERS
48#include "opt_device_polling.h"
49#endif
50
51#include <sys/param.h>
52#include <sys/systm.h>
53#include <sys/buf_ring.h>
54#include <sys/bus.h>
55#include <sys/endian.h>
56#include <sys/kernel.h>
57#include <sys/kthread.h>
58#include <sys/malloc.h>
59#include <sys/mbuf.h>
60#include <sys/module.h>
61#include <sys/rman.h>
62#include <sys/socket.h>
63#include <sys/sockio.h>
64#include <sys/sysctl.h>
65#include <sys/taskqueue.h>
66#include <sys/eventhandler.h>
67#include <machine/bus.h>
68#include <machine/resource.h>
69
70#include <net/bpf.h>
71#include <net/ethernet.h>
72#include <net/if.h>
73#include <net/if_var.h>
74#include <net/if_arp.h>
75#include <net/if_dl.h>
76#include <net/if_media.h>
77
78#include <net/if_types.h>
79#include <net/if_vlan_var.h>
80
81#include <netinet/in_systm.h>
82#include <netinet/in.h>
83#include <netinet/if_ether.h>
84#include <netinet/ip.h>
85#include <netinet/ip6.h>
86#include <netinet/tcp.h>
87#include <netinet/udp.h>
88
89#include <machine/in_cksum.h>
90#include <dev/led/led.h>
91#include <dev/pci/pcivar.h>
92#include <dev/pci/pcireg.h>
93
94#include "e1000_api.h"
95#include "if_lem.h"
96
97/*********************************************************************
98 * Legacy Em Driver version:
99 *********************************************************************/
100char lem_driver_version[] = "1.1.0";
101
102/*********************************************************************
103 * PCI Device ID Table
104 *
105 * Used by probe to select devices to load on
106 * Last field stores an index into e1000_strings
107 * Last entry must be all 0s
108 *
109 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
110 *********************************************************************/
111
112static em_vendor_info_t lem_vendor_info_array[] =
113{
114 /* Intel(R) PRO/1000 Network Connection */
115 { 0x8086, E1000_DEV_ID_82540EM, PCI_ANY_ID, PCI_ANY_ID, 0},
116 { 0x8086, E1000_DEV_ID_82540EM_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
117 { 0x8086, E1000_DEV_ID_82540EP, PCI_ANY_ID, PCI_ANY_ID, 0},
118 { 0x8086, E1000_DEV_ID_82540EP_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
119 { 0x8086, E1000_DEV_ID_82540EP_LP, PCI_ANY_ID, PCI_ANY_ID, 0},
120
121 { 0x8086, E1000_DEV_ID_82541EI, PCI_ANY_ID, PCI_ANY_ID, 0},
122 { 0x8086, E1000_DEV_ID_82541ER, PCI_ANY_ID, PCI_ANY_ID, 0},
123 { 0x8086, E1000_DEV_ID_82541ER_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
124 { 0x8086, E1000_DEV_ID_82541EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
125 { 0x8086, E1000_DEV_ID_82541GI, PCI_ANY_ID, PCI_ANY_ID, 0},
126 { 0x8086, E1000_DEV_ID_82541GI_LF, PCI_ANY_ID, PCI_ANY_ID, 0},
127 { 0x8086, E1000_DEV_ID_82541GI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
128
129 { 0x8086, E1000_DEV_ID_82542, PCI_ANY_ID, PCI_ANY_ID, 0},
130
131 { 0x8086, E1000_DEV_ID_82543GC_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
132 { 0x8086, E1000_DEV_ID_82543GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
133
134 { 0x8086, E1000_DEV_ID_82544EI_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
135 { 0x8086, E1000_DEV_ID_82544EI_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
136 { 0x8086, E1000_DEV_ID_82544GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
137 { 0x8086, E1000_DEV_ID_82544GC_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
138
139 { 0x8086, E1000_DEV_ID_82545EM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
140 { 0x8086, E1000_DEV_ID_82545EM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
141 { 0x8086, E1000_DEV_ID_82545GM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
142 { 0x8086, E1000_DEV_ID_82545GM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
143 { 0x8086, E1000_DEV_ID_82545GM_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
144
145 { 0x8086, E1000_DEV_ID_82546EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
146 { 0x8086, E1000_DEV_ID_82546EB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
147 { 0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
148 { 0x8086, E1000_DEV_ID_82546GB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
149 { 0x8086, E1000_DEV_ID_82546GB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
150 { 0x8086, E1000_DEV_ID_82546GB_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
151 { 0x8086, E1000_DEV_ID_82546GB_PCIE, PCI_ANY_ID, PCI_ANY_ID, 0},
152 { 0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
153 { 0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3,
154 PCI_ANY_ID, PCI_ANY_ID, 0},
155
156 { 0x8086, E1000_DEV_ID_82547EI, PCI_ANY_ID, PCI_ANY_ID, 0},
157 { 0x8086, E1000_DEV_ID_82547EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
158 { 0x8086, E1000_DEV_ID_82547GI, PCI_ANY_ID, PCI_ANY_ID, 0},
159 /* required last entry */
160 { 0, 0, 0, 0, 0}
161};
162
163/*********************************************************************
164 * Table of branding strings for all supported NICs.
165 *********************************************************************/
166
167static char *lem_strings[] = {
168 "Intel(R) PRO/1000 Legacy Network Connection"
169};
170
171/*********************************************************************
172 * Function prototypes
173 *********************************************************************/
174static int lem_probe(device_t);
175static int lem_attach(device_t);
176static int lem_detach(device_t);
177static int lem_shutdown(device_t);
178static int lem_suspend(device_t);
179static int lem_resume(device_t);
180static void lem_start(if_t);
181static void lem_start_locked(if_t ifp);
182static int lem_ioctl(if_t, u_long, caddr_t);
183static uint64_t lem_get_counter(if_t, ift_counter);
184static void lem_init(void *);
185static void lem_init_locked(struct adapter *);
186static void lem_stop(void *);
187static void lem_media_status(if_t, struct ifmediareq *);
188static int lem_media_change(if_t);
189static void lem_identify_hardware(struct adapter *);
190static int lem_allocate_pci_resources(struct adapter *);
191static int lem_allocate_irq(struct adapter *adapter);
192static void lem_free_pci_resources(struct adapter *);
193static void lem_local_timer(void *);
194static int lem_hardware_init(struct adapter *);
195static int lem_setup_interface(device_t, struct adapter *);
196static void lem_setup_transmit_structures(struct adapter *);
197static void lem_initialize_transmit_unit(struct adapter *);
198static int lem_setup_receive_structures(struct adapter *);
199static void lem_initialize_receive_unit(struct adapter *);
200static void lem_enable_intr(struct adapter *);
201static void lem_disable_intr(struct adapter *);
202static void lem_free_transmit_structures(struct adapter *);
203static void lem_free_receive_structures(struct adapter *);
204static void lem_update_stats_counters(struct adapter *);
205static void lem_add_hw_stats(struct adapter *adapter);
206static void lem_txeof(struct adapter *);
207static void lem_tx_purge(struct adapter *);
208static int lem_allocate_receive_structures(struct adapter *);
209static int lem_allocate_transmit_structures(struct adapter *);
210static bool lem_rxeof(struct adapter *, int, int *);
211#ifndef __NO_STRICT_ALIGNMENT
212static int lem_fixup_rx(struct adapter *);
213#endif
214static void lem_receive_checksum(struct adapter *, struct e1000_rx_desc *,
215 struct mbuf *);
216static void lem_transmit_checksum_setup(struct adapter *, struct mbuf *,
217 u32 *, u32 *);
218static void lem_set_promisc(struct adapter *);
219static void lem_disable_promisc(struct adapter *);
220static void lem_set_multi(struct adapter *);
221static void lem_update_link_status(struct adapter *);
222static int lem_get_buf(struct adapter *, int);
223static void lem_register_vlan(void *, if_t, u16);
224static void lem_unregister_vlan(void *, if_t, u16);
225static void lem_setup_vlan_hw_support(struct adapter *);
226static int lem_xmit(struct adapter *, struct mbuf **);
227static void lem_smartspeed(struct adapter *);
228static int lem_82547_fifo_workaround(struct adapter *, int);
229static void lem_82547_update_fifo_head(struct adapter *, int);
230static int lem_82547_tx_fifo_reset(struct adapter *);
231static void lem_82547_move_tail(void *);
232static int lem_dma_malloc(struct adapter *, bus_size_t,
233 struct em_dma_alloc *, int);
234static void lem_dma_free(struct adapter *, struct em_dma_alloc *);
235static int lem_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
236static void lem_print_nvm_info(struct adapter *);
237static int lem_is_valid_ether_addr(u8 *);
238static u32 lem_fill_descriptors (bus_addr_t address, u32 length,
239 PDESC_ARRAY desc_array);
240static int lem_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
241static void lem_add_int_delay_sysctl(struct adapter *, const char *,
242 const char *, struct em_int_delay_info *, int, int);
243static void lem_set_flow_cntrl(struct adapter *, const char *,
244 const char *, int *, int);
245/* Management and WOL Support */
246static void lem_init_manageability(struct adapter *);
247static void lem_release_manageability(struct adapter *);
248static void lem_get_hw_control(struct adapter *);
249static void lem_release_hw_control(struct adapter *);
250static void lem_get_wakeup(device_t);
251static void lem_enable_wakeup(device_t);
252static int lem_enable_phy_wakeup(struct adapter *);
253static void lem_led_func(void *, int);
254
255static void lem_intr(void *);
256static int lem_irq_fast(void *);
257static void lem_handle_rxtx(void *context, int pending);
258static void lem_handle_link(void *context, int pending);
259static void lem_add_rx_process_limit(struct adapter *, const char *,
260 const char *, int *, int);
261
262#ifdef DEVICE_POLLING
263static poll_handler_t lem_poll;
264#endif /* POLLING */
265
266/*********************************************************************
267 * FreeBSD Device Interface Entry Points
268 *********************************************************************/
269
270static device_method_t lem_methods[] = {
271 /* Device interface */
272 DEVMETHOD(device_probe, lem_probe),
273 DEVMETHOD(device_attach, lem_attach),
274 DEVMETHOD(device_detach, lem_detach),
275 DEVMETHOD(device_shutdown, lem_shutdown),
276 DEVMETHOD(device_suspend, lem_suspend),
277 DEVMETHOD(device_resume, lem_resume),
278 DEVMETHOD_END
279};
280
281static driver_t lem_driver = {
282 "em", lem_methods, sizeof(struct adapter),
283};
284
285extern devclass_t em_devclass;
286DRIVER_MODULE(lem, pci, lem_driver, em_devclass, 0, 0);
287MODULE_DEPEND(lem, pci, 1, 1, 1);
288MODULE_DEPEND(lem, ether, 1, 1, 1);
289#ifdef DEV_NETMAP
290MODULE_DEPEND(lem, netmap, 1, 1, 1);
291#endif /* DEV_NETMAP */
292
293/*********************************************************************
294 * Tunable default values.
295 *********************************************************************/
296
297#define EM_TICKS_TO_USECS(ticks) ((1024 * (ticks) + 500) / 1000)
298#define EM_USECS_TO_TICKS(usecs) ((1000 * (usecs) + 512) / 1024)
299
300#define MAX_INTS_PER_SEC 8000
301#define DEFAULT_ITR (1000000000/(MAX_INTS_PER_SEC * 256))
302
303static int lem_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
304static int lem_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
305static int lem_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
306static int lem_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
307/*
308 * increase lem_rxd and lem_txd to at least 2048 in netmap mode
309 * for better performance.
310 */
311static int lem_rxd = EM_DEFAULT_RXD;
312static int lem_txd = EM_DEFAULT_TXD;
313static int lem_smart_pwr_down = FALSE;
314
315/* Controls whether promiscuous also shows bad packets */
316static int lem_debug_sbp = FALSE;
317
318TUNABLE_INT("hw.em.tx_int_delay", &lem_tx_int_delay_dflt);
319TUNABLE_INT("hw.em.rx_int_delay", &lem_rx_int_delay_dflt);
320TUNABLE_INT("hw.em.tx_abs_int_delay", &lem_tx_abs_int_delay_dflt);
321TUNABLE_INT("hw.em.rx_abs_int_delay", &lem_rx_abs_int_delay_dflt);
322TUNABLE_INT("hw.em.rxd", &lem_rxd);
323TUNABLE_INT("hw.em.txd", &lem_txd);
324TUNABLE_INT("hw.em.smart_pwr_down", &lem_smart_pwr_down);
325TUNABLE_INT("hw.em.sbp", &lem_debug_sbp);
326
327/* Interrupt style - default to fast */
328static int lem_use_legacy_irq = 0;
329TUNABLE_INT("hw.em.use_legacy_irq", &lem_use_legacy_irq);
330
331/* How many packets rxeof tries to clean at a time */
332static int lem_rx_process_limit = 100;
333TUNABLE_INT("hw.em.rx_process_limit", &lem_rx_process_limit);
334
335/* Flow control setting - default to FULL */
336static int lem_fc_setting = e1000_fc_full;
337TUNABLE_INT("hw.em.fc_setting", &lem_fc_setting);
338
339/* Global used in WOL setup with multiport cards */
340static int global_quad_port_a = 0;
341
342#ifdef DEV_NETMAP /* see ixgbe.c for details */
343#include <dev/netmap/if_lem_netmap.h>
344#endif /* DEV_NETMAP */
345
346/*********************************************************************
347 * Device identification routine
348 *
349 * em_probe determines if the driver should be loaded on
350 * adapter based on PCI vendor/device id of the adapter.
351 *
352 * return BUS_PROBE_DEFAULT on success, positive on failure
353 *********************************************************************/
354
355static int
356lem_probe(device_t dev)
357{
358 char adapter_name[60];
359 u16 pci_vendor_id = 0;
360 u16 pci_device_id = 0;
361 u16 pci_subvendor_id = 0;
362 u16 pci_subdevice_id = 0;
363 em_vendor_info_t *ent;
364
365 INIT_DEBUGOUT("em_probe: begin");
366
367 pci_vendor_id = pci_get_vendor(dev);
368 if (pci_vendor_id != EM_VENDOR_ID)
369 return (ENXIO);
370
371 pci_device_id = pci_get_device(dev);
372 pci_subvendor_id = pci_get_subvendor(dev);
373 pci_subdevice_id = pci_get_subdevice(dev);
374
375 ent = lem_vendor_info_array;
376 while (ent->vendor_id != 0) {
377 if ((pci_vendor_id == ent->vendor_id) &&
378 (pci_device_id == ent->device_id) &&
379
380 ((pci_subvendor_id == ent->subvendor_id) ||
381 (ent->subvendor_id == PCI_ANY_ID)) &&
382
383 ((pci_subdevice_id == ent->subdevice_id) ||
384 (ent->subdevice_id == PCI_ANY_ID))) {
385 sprintf(adapter_name, "%s %s",
386 lem_strings[ent->index],
387 lem_driver_version);
388 device_set_desc_copy(dev, adapter_name);
389 return (BUS_PROBE_DEFAULT);
390 }
391 ent++;
392 }
393
394 return (ENXIO);
395}
396
397/*********************************************************************
398 * Device initialization routine
399 *
400 * The attach entry point is called when the driver is being loaded.
401 * This routine identifies the type of hardware, allocates all resources
402 * and initializes the hardware.
403 *
404 * return 0 on success, positive on failure
405 *********************************************************************/
406
407static int
408lem_attach(device_t dev)
409{
410 struct adapter *adapter;
411 int tsize, rsize;
412 int error = 0;
413
414 INIT_DEBUGOUT("lem_attach: begin");
415
416 adapter = device_get_softc(dev);
417 adapter->dev = adapter->osdep.dev = dev;
418 EM_CORE_LOCK_INIT(adapter, device_get_nameunit(dev));
419 EM_TX_LOCK_INIT(adapter, device_get_nameunit(dev));
420 EM_RX_LOCK_INIT(adapter, device_get_nameunit(dev));
421
422 /* SYSCTL stuff */
423 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
424 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
425 OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
426 lem_sysctl_nvm_info, "I", "NVM Information");
427
428 callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0);
429 callout_init_mtx(&adapter->tx_fifo_timer, &adapter->tx_mtx, 0);
430
431 /* Determine hardware and mac info */
432 lem_identify_hardware(adapter);
433
434 /* Setup PCI resources */
435 if (lem_allocate_pci_resources(adapter)) {
436 device_printf(dev, "Allocation of PCI resources failed\n");
437 error = ENXIO;
438 goto err_pci;
439 }
440
441 /* Do Shared Code initialization */
442 if (e1000_setup_init_funcs(&adapter->hw, TRUE)) {
443 device_printf(dev, "Setup of Shared code failed\n");
444 error = ENXIO;
445 goto err_pci;
446 }
447
448 e1000_get_bus_info(&adapter->hw);
449
450 /* Set up some sysctls for the tunable interrupt delays */
451 lem_add_int_delay_sysctl(adapter, "rx_int_delay",
452 "receive interrupt delay in usecs", &adapter->rx_int_delay,
453 E1000_REGISTER(&adapter->hw, E1000_RDTR), lem_rx_int_delay_dflt);
454 lem_add_int_delay_sysctl(adapter, "tx_int_delay",
455 "transmit interrupt delay in usecs", &adapter->tx_int_delay,
456 E1000_REGISTER(&adapter->hw, E1000_TIDV), lem_tx_int_delay_dflt);
457 if (adapter->hw.mac.type >= e1000_82540) {
458 lem_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
459 "receive interrupt delay limit in usecs",
460 &adapter->rx_abs_int_delay,
461 E1000_REGISTER(&adapter->hw, E1000_RADV),
462 lem_rx_abs_int_delay_dflt);
463 lem_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
464 "transmit interrupt delay limit in usecs",
465 &adapter->tx_abs_int_delay,
466 E1000_REGISTER(&adapter->hw, E1000_TADV),
467 lem_tx_abs_int_delay_dflt);
468 lem_add_int_delay_sysctl(adapter, "itr",
469 "interrupt delay limit in usecs/4",
470 &adapter->tx_itr,
471 E1000_REGISTER(&adapter->hw, E1000_ITR),
472 DEFAULT_ITR);
473 }
474
475 /* Sysctls for limiting the amount of work done in the taskqueue */
476 lem_add_rx_process_limit(adapter, "rx_processing_limit",
477 "max number of rx packets to process", &adapter->rx_process_limit,
478 lem_rx_process_limit);
479
480#ifdef NIC_SEND_COMBINING
481 /* Sysctls to control mitigation */
482 lem_add_rx_process_limit(adapter, "sc_enable",
483 "driver TDT mitigation", &adapter->sc_enable, 0);
484#endif /* NIC_SEND_COMBINING */
485#ifdef BATCH_DISPATCH
486 lem_add_rx_process_limit(adapter, "batch_enable",
487 "driver rx batch", &adapter->batch_enable, 0);
488#endif /* BATCH_DISPATCH */
489#ifdef NIC_PARAVIRT
490 lem_add_rx_process_limit(adapter, "rx_retries",
491 "driver rx retries", &adapter->rx_retries, 0);
492#endif /* NIC_PARAVIRT */
493
494 /* Sysctl for setting the interface flow control */
495 lem_set_flow_cntrl(adapter, "flow_control",
496 "flow control setting",
497 &adapter->fc_setting, lem_fc_setting);
498
499 /*
500 * Validate number of transmit and receive descriptors. It
501 * must not exceed hardware maximum, and must be multiple
502 * of E1000_DBA_ALIGN.
503 */
504 if (((lem_txd * sizeof(struct e1000_tx_desc)) % EM_DBA_ALIGN) != 0 ||
505 (adapter->hw.mac.type >= e1000_82544 && lem_txd > EM_MAX_TXD) ||
506 (adapter->hw.mac.type < e1000_82544 && lem_txd > EM_MAX_TXD_82543) ||
507 (lem_txd < EM_MIN_TXD)) {
508 device_printf(dev, "Using %d TX descriptors instead of %d!\n",
509 EM_DEFAULT_TXD, lem_txd);
510 adapter->num_tx_desc = EM_DEFAULT_TXD;
511 } else
512 adapter->num_tx_desc = lem_txd;
513 if (((lem_rxd * sizeof(struct e1000_rx_desc)) % EM_DBA_ALIGN) != 0 ||
514 (adapter->hw.mac.type >= e1000_82544 && lem_rxd > EM_MAX_RXD) ||
515 (adapter->hw.mac.type < e1000_82544 && lem_rxd > EM_MAX_RXD_82543) ||
516 (lem_rxd < EM_MIN_RXD)) {
517 device_printf(dev, "Using %d RX descriptors instead of %d!\n",
518 EM_DEFAULT_RXD, lem_rxd);
519 adapter->num_rx_desc = EM_DEFAULT_RXD;
520 } else
521 adapter->num_rx_desc = lem_rxd;
522
523 adapter->hw.mac.autoneg = DO_AUTO_NEG;
524 adapter->hw.phy.autoneg_wait_to_complete = FALSE;
525 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
526 adapter->rx_buffer_len = 2048;
527
528 e1000_init_script_state_82541(&adapter->hw, TRUE);
529 e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
530
531 /* Copper options */
532 if (adapter->hw.phy.media_type == e1000_media_type_copper) {
533 adapter->hw.phy.mdix = AUTO_ALL_MODES;
534 adapter->hw.phy.disable_polarity_correction = FALSE;
535 adapter->hw.phy.ms_type = EM_MASTER_SLAVE;
536 }
537
538 /*
539 * Set the frame limits assuming
540 * standard ethernet sized frames.
541 */
542 adapter->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
543 adapter->min_frame_size = ETH_ZLEN + ETHERNET_FCS_SIZE;
544
545 /*
546 * This controls when hardware reports transmit completion
547 * status.
548 */
549 adapter->hw.mac.report_tx_early = 1;
550
551#ifdef NIC_PARAVIRT
552 device_printf(dev, "driver supports paravirt, subdev 0x%x\n",
553 adapter->hw.subsystem_device_id);
554 if (adapter->hw.subsystem_device_id == E1000_PARA_SUBDEV) {
555 uint64_t bus_addr;
556
557 device_printf(dev, "paravirt support on dev %p\n", adapter);
558 tsize = 4096; // XXX one page for the csb
559 if (lem_dma_malloc(adapter, tsize, &adapter->csb_mem, BUS_DMA_NOWAIT)) {
560 device_printf(dev, "Unable to allocate csb memory\n");
561 error = ENOMEM;
562 goto err_csb;
563 }
564 /* Setup the Base of the CSB */
565 adapter->csb = (struct paravirt_csb *)adapter->csb_mem.dma_vaddr;
566 /* force the first kick */
567 adapter->csb->host_need_txkick = 1; /* txring empty */
568 adapter->csb->guest_need_rxkick = 1; /* no rx packets */
569 bus_addr = adapter->csb_mem.dma_paddr;
570 lem_add_rx_process_limit(adapter, "csb_on",
571 "enable paravirt.", &adapter->csb->guest_csb_on, 0);
572 lem_add_rx_process_limit(adapter, "txc_lim",
573 "txc_lim", &adapter->csb->host_txcycles_lim, 1);
574
575 /* some stats */
576#define PA_SC(name, var, val) \
577 lem_add_rx_process_limit(adapter, name, name, var, val)
578 PA_SC("host_need_txkick",&adapter->csb->host_need_txkick, 1);
579 PA_SC("host_rxkick_at",&adapter->csb->host_rxkick_at, ~0);
580 PA_SC("guest_need_txkick",&adapter->csb->guest_need_txkick, 0);
581 PA_SC("guest_need_rxkick",&adapter->csb->guest_need_rxkick, 1);
582 PA_SC("tdt_reg_count",&adapter->tdt_reg_count, 0);
583 PA_SC("tdt_csb_count",&adapter->tdt_csb_count, 0);
584 PA_SC("tdt_int_count",&adapter->tdt_int_count, 0);
585 PA_SC("guest_need_kick_count",&adapter->guest_need_kick_count, 0);
586 /* tell the host where the block is */
587 E1000_WRITE_REG(&adapter->hw, E1000_CSBAH,
588 (u32)(bus_addr >> 32));
589 E1000_WRITE_REG(&adapter->hw, E1000_CSBAL,
590 (u32)bus_addr);
591 }
592#endif /* NIC_PARAVIRT */
593
594 tsize = roundup2(adapter->num_tx_desc * sizeof(struct e1000_tx_desc),
595 EM_DBA_ALIGN);
596
597 /* Allocate Transmit Descriptor ring */
598 if (lem_dma_malloc(adapter, tsize, &adapter->txdma, BUS_DMA_NOWAIT)) {
599 device_printf(dev, "Unable to allocate tx_desc memory\n");
600 error = ENOMEM;
601 goto err_tx_desc;
602 }
603 adapter->tx_desc_base =
604 (struct e1000_tx_desc *)adapter->txdma.dma_vaddr;
605
606 rsize = roundup2(adapter->num_rx_desc * sizeof(struct e1000_rx_desc),
607 EM_DBA_ALIGN);
608
609 /* Allocate Receive Descriptor ring */
610 if (lem_dma_malloc(adapter, rsize, &adapter->rxdma, BUS_DMA_NOWAIT)) {
611 device_printf(dev, "Unable to allocate rx_desc memory\n");
612 error = ENOMEM;
613 goto err_rx_desc;
614 }
615 adapter->rx_desc_base =
616 (struct e1000_rx_desc *)adapter->rxdma.dma_vaddr;
617
618 /* Allocate multicast array memory. */
619 adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
620 MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
621 if (adapter->mta == NULL) {
622 device_printf(dev, "Can not allocate multicast setup array\n");
623 error = ENOMEM;
624 goto err_hw_init;
625 }
626
627 /*
628 ** Start from a known state, this is
629 ** important in reading the nvm and
630 ** mac from that.
631 */
632 e1000_reset_hw(&adapter->hw);
633
634 /* Make sure we have a good EEPROM before we read from it */
635 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
636 /*
637 ** Some PCI-E parts fail the first check due to
638 ** the link being in sleep state, call it again,
639 ** if it fails a second time its a real issue.
640 */
641 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
642 device_printf(dev,
643 "The EEPROM Checksum Is Not Valid\n");
644 error = EIO;
645 goto err_hw_init;
646 }
647 }
648
649 /* Copy the permanent MAC address out of the EEPROM */
650 if (e1000_read_mac_addr(&adapter->hw) < 0) {
651 device_printf(dev, "EEPROM read error while reading MAC"
652 " address\n");
653 error = EIO;
654 goto err_hw_init;
655 }
656
657 if (!lem_is_valid_ether_addr(adapter->hw.mac.addr)) {
658 device_printf(dev, "Invalid MAC address\n");
659 error = EIO;
660 goto err_hw_init;
661 }
662
663 /* Initialize the hardware */
664 if (lem_hardware_init(adapter)) {
665 device_printf(dev, "Unable to initialize the hardware\n");
666 error = EIO;
667 goto err_hw_init;
668 }
669
670 /* Allocate transmit descriptors and buffers */
671 if (lem_allocate_transmit_structures(adapter)) {
672 device_printf(dev, "Could not setup transmit structures\n");
673 error = ENOMEM;
674 goto err_tx_struct;
675 }
676
677 /* Allocate receive descriptors and buffers */
678 if (lem_allocate_receive_structures(adapter)) {
679 device_printf(dev, "Could not setup receive structures\n");
680 error = ENOMEM;
681 goto err_rx_struct;
682 }
683
684 /*
685 ** Do interrupt configuration
686 */
687 error = lem_allocate_irq(adapter);
688 if (error)
689 goto err_rx_struct;
690
691 /*
692 * Get Wake-on-Lan and Management info for later use
693 */
694 lem_get_wakeup(dev);
695
696 /* Setup OS specific network interface */
697 if (lem_setup_interface(dev, adapter) != 0)
698 goto err_rx_struct;
699
700 /* Initialize statistics */
701 lem_update_stats_counters(adapter);
702
703 adapter->hw.mac.get_link_status = 1;
704 lem_update_link_status(adapter);
705
706 /* Indicate SOL/IDER usage */
707 if (e1000_check_reset_block(&adapter->hw))
708 device_printf(dev,
709 "PHY reset is blocked due to SOL/IDER session.\n");
710
711 /* Do we need workaround for 82544 PCI-X adapter? */
712 if (adapter->hw.bus.type == e1000_bus_type_pcix &&
713 adapter->hw.mac.type == e1000_82544)
714 adapter->pcix_82544 = TRUE;
715 else
716 adapter->pcix_82544 = FALSE;
717
718 /* Register for VLAN events */
719 adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
720 lem_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
721 adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
722 lem_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
723
724 lem_add_hw_stats(adapter);
725
726 /* Non-AMT based hardware can now take control from firmware */
727 if (adapter->has_manage && !adapter->has_amt)
728 lem_get_hw_control(adapter);
729
730 /* Tell the stack that the interface is not active */
731 if_setdrvflagbits(adapter->ifp, 0, IFF_DRV_OACTIVE | IFF_DRV_RUNNING);
732
733 adapter->led_dev = led_create(lem_led_func, adapter,
734 device_get_nameunit(dev));
735
736#ifdef DEV_NETMAP
737 lem_netmap_attach(adapter);
738#endif /* DEV_NETMAP */
739 INIT_DEBUGOUT("lem_attach: end");
740
741 return (0);
742
743err_rx_struct:
744 lem_free_transmit_structures(adapter);
745err_tx_struct:
746err_hw_init:
747 lem_release_hw_control(adapter);
748 lem_dma_free(adapter, &adapter->rxdma);
749err_rx_desc:
750 lem_dma_free(adapter, &adapter->txdma);
751err_tx_desc:
752#ifdef NIC_PARAVIRT
753 lem_dma_free(adapter, &adapter->csb_mem);
754err_csb:
755#endif /* NIC_PARAVIRT */
756
757err_pci:
758 if (adapter->ifp != (void *)NULL)
759 if_free(adapter->ifp);
760 lem_free_pci_resources(adapter);
761 free(adapter->mta, M_DEVBUF);
762 EM_TX_LOCK_DESTROY(adapter);
763 EM_RX_LOCK_DESTROY(adapter);
764 EM_CORE_LOCK_DESTROY(adapter);
765
766 return (error);
767}
768
769/*********************************************************************
770 * Device removal routine
771 *
772 * The detach entry point is called when the driver is being removed.
773 * This routine stops the adapter and deallocates all the resources
774 * that were allocated for driver operation.
775 *
776 * return 0 on success, positive on failure
777 *********************************************************************/
778
779static int
780lem_detach(device_t dev)
781{
782 struct adapter *adapter = device_get_softc(dev);
783 if_t ifp = adapter->ifp;
784
785 INIT_DEBUGOUT("em_detach: begin");
786
787 /* Make sure VLANS are not using driver */
788 if (if_vlantrunkinuse(ifp)) {
789 device_printf(dev,"Vlan in use, detach first\n");
790 return (EBUSY);
791 }
792
793#ifdef DEVICE_POLLING
794 if (if_getcapenable(ifp) & IFCAP_POLLING)
795 ether_poll_deregister(ifp);
796#endif
797
798 if (adapter->led_dev != NULL)
799 led_destroy(adapter->led_dev);
800
801 EM_CORE_LOCK(adapter);
802 EM_TX_LOCK(adapter);
803 adapter->in_detach = 1;
804 lem_stop(adapter);
805 e1000_phy_hw_reset(&adapter->hw);
806
807 lem_release_manageability(adapter);
808
809 EM_TX_UNLOCK(adapter);
810 EM_CORE_UNLOCK(adapter);
811
812 /* Unregister VLAN events */
813 if (adapter->vlan_attach != NULL)
814 EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach);
815 if (adapter->vlan_detach != NULL)
816 EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach);
817
818 ether_ifdetach(adapter->ifp);
819 callout_drain(&adapter->timer);
820 callout_drain(&adapter->tx_fifo_timer);
821
822#ifdef DEV_NETMAP
823 netmap_detach(ifp);
824#endif /* DEV_NETMAP */
825 lem_free_pci_resources(adapter);
826 bus_generic_detach(dev);
827 if_free(ifp);
828
829 lem_free_transmit_structures(adapter);
830 lem_free_receive_structures(adapter);
831
832 /* Free Transmit Descriptor ring */
833 if (adapter->tx_desc_base) {
834 lem_dma_free(adapter, &adapter->txdma);
835 adapter->tx_desc_base = NULL;
836 }
837
838 /* Free Receive Descriptor ring */
839 if (adapter->rx_desc_base) {
840 lem_dma_free(adapter, &adapter->rxdma);
841 adapter->rx_desc_base = NULL;
842 }
843
844#ifdef NIC_PARAVIRT
845 if (adapter->csb) {
846 lem_dma_free(adapter, &adapter->csb_mem);
847 adapter->csb = NULL;
848 }
849#endif /* NIC_PARAVIRT */
850 lem_release_hw_control(adapter);
851 free(adapter->mta, M_DEVBUF);
852 EM_TX_LOCK_DESTROY(adapter);
853 EM_RX_LOCK_DESTROY(adapter);
854 EM_CORE_LOCK_DESTROY(adapter);
855
856 return (0);
857}
858
859/*********************************************************************
860 *
861 * Shutdown entry point
862 *
863 **********************************************************************/
864
865static int
866lem_shutdown(device_t dev)
867{
868 return lem_suspend(dev);
869}
870
871/*
872 * Suspend/resume device methods.
873 */
874static int
875lem_suspend(device_t dev)
876{
877 struct adapter *adapter = device_get_softc(dev);
878
879 EM_CORE_LOCK(adapter);
880
881 lem_release_manageability(adapter);
882 lem_release_hw_control(adapter);
883 lem_enable_wakeup(dev);
884
885 EM_CORE_UNLOCK(adapter);
886
887 return bus_generic_suspend(dev);
888}
889
890static int
891lem_resume(device_t dev)
892{
893 struct adapter *adapter = device_get_softc(dev);
894 if_t ifp = adapter->ifp;
895
896 EM_CORE_LOCK(adapter);
897 lem_init_locked(adapter);
898 lem_init_manageability(adapter);
899 EM_CORE_UNLOCK(adapter);
900 lem_start(ifp);
901
902 return bus_generic_resume(dev);
903}
904
905
906static void
907lem_start_locked(if_t ifp)
908{
909 struct adapter *adapter = if_getsoftc(ifp);
910 struct mbuf *m_head;
911
912 EM_TX_LOCK_ASSERT(adapter);
913
914 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) !=
915 IFF_DRV_RUNNING)
916 return;
917 if (!adapter->link_active)
918 return;
919
920 /*
921 * Force a cleanup if number of TX descriptors
922 * available hits the threshold
923 */
924 if (adapter->num_tx_desc_avail <= EM_TX_CLEANUP_THRESHOLD) {
925 lem_txeof(adapter);
926 /* Now do we at least have a minimal? */
927 if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD) {
928 adapter->no_tx_desc_avail1++;
929 return;
930 }
931 }
932
933 while (!if_sendq_empty(ifp)) {
934 m_head = if_dequeue(ifp);
935
936 if (m_head == NULL)
937 break;
938 /*
939 * Encapsulation can modify our pointer, and or make it
940 * NULL on failure. In that event, we can't requeue.
941 */
942 if (lem_xmit(adapter, &m_head)) {
943 if (m_head == NULL)
944 break;
945 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, 0);
946 if_sendq_prepend(ifp, m_head);
947 break;
948 }
949
950 /* Send a copy of the frame to the BPF listener */
951 if_etherbpfmtap(ifp, m_head);
952
953 /* Set timeout in case hardware has problems transmitting. */
954 adapter->watchdog_check = TRUE;
955 adapter->watchdog_time = ticks;
956 }
957 if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD)
958 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, 0);
959#ifdef NIC_PARAVIRT
960 if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE && adapter->csb &&
961 adapter->csb->guest_csb_on &&
962 !(adapter->csb->guest_need_txkick & 1)) {
963 adapter->csb->guest_need_txkick = 1;
964 adapter->guest_need_kick_count++;
965 // XXX memory barrier
966 lem_txeof(adapter); // XXX possibly clear IFF_DRV_OACTIVE
967 }
968#endif /* NIC_PARAVIRT */
969
970 return;
971}
972
973static void
974lem_start(if_t ifp)
975{
976 struct adapter *adapter = if_getsoftc(ifp);
977
978 EM_TX_LOCK(adapter);
979 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
980 lem_start_locked(ifp);
981 EM_TX_UNLOCK(adapter);
982}
983
984/*********************************************************************
985 * Ioctl entry point
986 *
987 * em_ioctl is called when the user wants to configure the
988 * interface.
989 *
990 * return 0 on success, positive on failure
991 **********************************************************************/
992
993static int
994lem_ioctl(if_t ifp, u_long command, caddr_t data)
995{
996 struct adapter *adapter = if_getsoftc(ifp);
997 struct ifreq *ifr = (struct ifreq *)data;
998#if defined(INET) || defined(INET6)
999 struct ifaddr *ifa = (struct ifaddr *)data;
1000#endif
1001 bool avoid_reset = FALSE;
1002 int error = 0;
1003
1004 if (adapter->in_detach)
1005 return (error);
1006
1007 switch (command) {
1008 case SIOCSIFADDR:
1009#ifdef INET
1010 if (ifa->ifa_addr->sa_family == AF_INET)
1011 avoid_reset = TRUE;
1012#endif
1013#ifdef INET6
1014 if (ifa->ifa_addr->sa_family == AF_INET6)
1015 avoid_reset = TRUE;
1016#endif
1017 /*
1018 ** Calling init results in link renegotiation,
1019 ** so we avoid doing it when possible.
1020 */
1021 if (avoid_reset) {
1022 if_setflagbits(ifp, IFF_UP, 0);
1023 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1024 lem_init(adapter);
1025#ifdef INET
1026 if (!(if_getflags(ifp) & IFF_NOARP))
1027 arp_ifinit(ifp, ifa);
1028#endif
1029 } else
1030 error = ether_ioctl(ifp, command, data);
1031 break;
1032 case SIOCSIFMTU:
1033 {
1034 int max_frame_size;
1035
1036 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1037
1038 EM_CORE_LOCK(adapter);
1039 switch (adapter->hw.mac.type) {
1040 case e1000_82542:
1041 max_frame_size = ETHER_MAX_LEN;
1042 break;
1043 default:
1044 max_frame_size = MAX_JUMBO_FRAME_SIZE;
1045 }
1046 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
1047 ETHER_CRC_LEN) {
1048 EM_CORE_UNLOCK(adapter);
1049 error = EINVAL;
1050 break;
1051 }
1052
1053 if_setmtu(ifp, ifr->ifr_mtu);
1054 adapter->max_frame_size =
1055 if_getmtu(ifp) + ETHER_HDR_LEN + ETHER_CRC_LEN;
1056 lem_init_locked(adapter);
1057 EM_CORE_UNLOCK(adapter);
1058 break;
1059 }
1060 case SIOCSIFFLAGS:
1061 IOCTL_DEBUGOUT("ioctl rcv'd:\
1062 SIOCSIFFLAGS (Set Interface Flags)");
1063 EM_CORE_LOCK(adapter);
1064 if (if_getflags(ifp) & IFF_UP) {
1065 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING)) {
1066 if ((if_getflags(ifp) ^ adapter->if_flags) &
1067 (IFF_PROMISC | IFF_ALLMULTI)) {
1068 lem_disable_promisc(adapter);
1069 lem_set_promisc(adapter);
1070 }
1071 } else
1072 lem_init_locked(adapter);
1073 } else
1074 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1075 EM_TX_LOCK(adapter);
1076 lem_stop(adapter);
1077 EM_TX_UNLOCK(adapter);
1078 }
1079 adapter->if_flags = if_getflags(ifp);
1080 EM_CORE_UNLOCK(adapter);
1081 break;
1082 case SIOCADDMULTI:
1083 case SIOCDELMULTI:
1084 IOCTL_DEBUGOUT("ioctl rcv'd: SIOC(ADD|DEL)MULTI");
1085 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1086 EM_CORE_LOCK(adapter);
1087 lem_disable_intr(adapter);
1088 lem_set_multi(adapter);
1089 if (adapter->hw.mac.type == e1000_82542 &&
1090 adapter->hw.revision_id == E1000_REVISION_2) {
1091 lem_initialize_receive_unit(adapter);
1092 }
1093#ifdef DEVICE_POLLING
1094 if (!(if_getcapenable(ifp) & IFCAP_POLLING))
1095#endif
1096 lem_enable_intr(adapter);
1097 EM_CORE_UNLOCK(adapter);
1098 }
1099 break;
1100 case SIOCSIFMEDIA:
1101 /* Check SOL/IDER usage */
1102 EM_CORE_LOCK(adapter);
1103 if (e1000_check_reset_block(&adapter->hw)) {
1104 EM_CORE_UNLOCK(adapter);
1105 device_printf(adapter->dev, "Media change is"
1106 " blocked due to SOL/IDER session.\n");
1107 break;
1108 }
1109 EM_CORE_UNLOCK(adapter);
1110 case SIOCGIFMEDIA:
1111 IOCTL_DEBUGOUT("ioctl rcv'd: \
1112 SIOCxIFMEDIA (Get/Set Interface Media)");
1113 error = ifmedia_ioctl(ifp, ifr, &adapter->media, command);
1114 break;
1115 case SIOCSIFCAP:
1116 {
1117 int mask, reinit;
1118
1119 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFCAP (Set Capabilities)");
1120 reinit = 0;
1121 mask = ifr->ifr_reqcap ^ if_getcapenable(ifp);
1122#ifdef DEVICE_POLLING
1123 if (mask & IFCAP_POLLING) {
1124 if (ifr->ifr_reqcap & IFCAP_POLLING) {
1125 error = ether_poll_register(lem_poll, ifp);
1126 if (error)
1127 return (error);
1128 EM_CORE_LOCK(adapter);
1129 lem_disable_intr(adapter);
1130 if_setcapenablebit(ifp, IFCAP_POLLING, 0);
1131 EM_CORE_UNLOCK(adapter);
1132 } else {
1133 error = ether_poll_deregister(ifp);
1134 /* Enable interrupt even in error case */
1135 EM_CORE_LOCK(adapter);
1136 lem_enable_intr(adapter);
1137 if_setcapenablebit(ifp, 0, IFCAP_POLLING);
1138 EM_CORE_UNLOCK(adapter);
1139 }
1140 }
1141#endif
1142 if (mask & IFCAP_HWCSUM) {
1143 if_togglecapenable(ifp, IFCAP_HWCSUM);
1144 reinit = 1;
1145 }
1146 if (mask & IFCAP_VLAN_HWTAGGING) {
1147 if_togglecapenable(ifp, IFCAP_VLAN_HWTAGGING);
1148 reinit = 1;
1149 }
1150 if ((mask & IFCAP_WOL) &&
1151 (if_getcapabilities(ifp) & IFCAP_WOL) != 0) {
1152 if (mask & IFCAP_WOL_MCAST)
1153 if_togglecapenable(ifp, IFCAP_WOL_MCAST);
1154 if (mask & IFCAP_WOL_MAGIC)
1155 if_togglecapenable(ifp, IFCAP_WOL_MAGIC);
1156 }
1157 if (reinit && (if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1158 lem_init(adapter);
1159 if_vlancap(ifp);
1160 break;
1161 }
1162
1163 default:
1164 error = ether_ioctl(ifp, command, data);
1165 break;
1166 }
1167
1168 return (error);
1169}
1170
1171
1172/*********************************************************************
1173 * Init entry point
1174 *
1175 * This routine is used in two ways. It is used by the stack as
1176 * init entry point in network interface structure. It is also used
1177 * by the driver as a hw/sw initialization routine to get to a
1178 * consistent state.
1179 *
1180 * return 0 on success, positive on failure
1181 **********************************************************************/
1182
1183static void
1184lem_init_locked(struct adapter *adapter)
1185{
1186 if_t ifp = adapter->ifp;
1187 device_t dev = adapter->dev;
1188 u32 pba;
1189
1190 INIT_DEBUGOUT("lem_init: begin");
1191
1192 EM_CORE_LOCK_ASSERT(adapter);
1193
1194 EM_TX_LOCK(adapter);
1195 lem_stop(adapter);
1196 EM_TX_UNLOCK(adapter);
1197
1198 /*
1199 * Packet Buffer Allocation (PBA)
1200 * Writing PBA sets the receive portion of the buffer
1201 * the remainder is used for the transmit buffer.
1202 *
1203 * Devices before the 82547 had a Packet Buffer of 64K.
1204 * Default allocation: PBA=48K for Rx, leaving 16K for Tx.
1205 * After the 82547 the buffer was reduced to 40K.
1206 * Default allocation: PBA=30K for Rx, leaving 10K for Tx.
1207 * Note: default does not leave enough room for Jumbo Frame >10k.
1208 */
1209 switch (adapter->hw.mac.type) {
1210 case e1000_82547:
1211 case e1000_82547_rev_2: /* 82547: Total Packet Buffer is 40K */
1212 if (adapter->max_frame_size > 8192)
1213 pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
1214 else
1215 pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
1216 adapter->tx_fifo_head = 0;
1217 adapter->tx_head_addr = pba << EM_TX_HEAD_ADDR_SHIFT;
1218 adapter->tx_fifo_size =
1219 (E1000_PBA_40K - pba) << EM_PBA_BYTES_SHIFT;
1220 break;
1221 default:
1222 /* Devices before 82547 had a Packet Buffer of 64K. */
1223 if (adapter->max_frame_size > 8192)
1224 pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
1225 else
1226 pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
1227 }
1228
1229 INIT_DEBUGOUT1("lem_init: pba=%dK",pba);
1230 E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
1231
1232 /* Get the latest mac address, User can use a LAA */
1233 bcopy(if_getlladdr(adapter->ifp), adapter->hw.mac.addr,
1234 ETHER_ADDR_LEN);
1235
1236 /* Put the address into the Receive Address Array */
1237 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1238
1239 /* Initialize the hardware */
1240 if (lem_hardware_init(adapter)) {
1241 device_printf(dev, "Unable to initialize the hardware\n");
1242 return;
1243 }
1244 lem_update_link_status(adapter);
1245
1246 /* Setup VLAN support, basic and offload if available */
1247 E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1248
1249 /* Set hardware offload abilities */
1250 if_clearhwassist(ifp);
1251 if (adapter->hw.mac.type >= e1000_82543) {
1252 if (if_getcapenable(ifp) & IFCAP_TXCSUM)
1253 if_sethwassistbits(ifp, CSUM_TCP | CSUM_UDP, 0);
1254 }
1255
1256 /* Configure for OS presence */
1257 lem_init_manageability(adapter);
1258
1259 /* Prepare transmit descriptors and buffers */
1260 lem_setup_transmit_structures(adapter);
1261 lem_initialize_transmit_unit(adapter);
1262
1263 /* Setup Multicast table */
1264 lem_set_multi(adapter);
1265
1266 /* Prepare receive descriptors and buffers */
1267 if (lem_setup_receive_structures(adapter)) {
1268 device_printf(dev, "Could not setup receive structures\n");
1269 EM_TX_LOCK(adapter);
1270 lem_stop(adapter);
1271 EM_TX_UNLOCK(adapter);
1272 return;
1273 }
1274 lem_initialize_receive_unit(adapter);
1275
1276 /* Use real VLAN Filter support? */
1277 if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1278 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1279 /* Use real VLAN Filter support */
1280 lem_setup_vlan_hw_support(adapter);
1281 else {
1282 u32 ctrl;
1283 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1284 ctrl |= E1000_CTRL_VME;
1285 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1286 }
1287 }
1288
1289 /* Don't lose promiscuous settings */
1290 lem_set_promisc(adapter);
1291
1292 if_setdrvflagbits(ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
1293
1294 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
1295 e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1296
1297#ifdef DEVICE_POLLING
1298 /*
1299 * Only enable interrupts if we are not polling, make sure
1300 * they are off otherwise.
1301 */
1302 if (if_getcapenable(ifp) & IFCAP_POLLING)
1303 lem_disable_intr(adapter);
1304 else
1305#endif /* DEVICE_POLLING */
1306 lem_enable_intr(adapter);
1307
1308 /* AMT based hardware can now take control from firmware */
1309 if (adapter->has_manage && adapter->has_amt)
1310 lem_get_hw_control(adapter);
1311}
1312
1313static void
1314lem_init(void *arg)
1315{
1316 struct adapter *adapter = arg;
1317
1318 EM_CORE_LOCK(adapter);
1319 lem_init_locked(adapter);
1320 EM_CORE_UNLOCK(adapter);
1321}
1322
1323
1324#ifdef DEVICE_POLLING
1325/*********************************************************************
1326 *
1327 * Legacy polling routine
1328 *
1329 *********************************************************************/
1330static int
1331lem_poll(if_t ifp, enum poll_cmd cmd, int count)
1332{
1333 struct adapter *adapter = if_getsoftc(ifp);
1334 u32 reg_icr, rx_done = 0;
1335
1336 EM_CORE_LOCK(adapter);
1337 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0) {
1338 EM_CORE_UNLOCK(adapter);
1339 return (rx_done);
1340 }
1341
1342 if (cmd == POLL_AND_CHECK_STATUS) {
1343 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1344 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1345 callout_stop(&adapter->timer);
1346 adapter->hw.mac.get_link_status = 1;
1347 lem_update_link_status(adapter);
1348 callout_reset(&adapter->timer, hz,
1349 lem_local_timer, adapter);
1350 }
1351 }
1352 EM_CORE_UNLOCK(adapter);
1353
1354 lem_rxeof(adapter, count, &rx_done);
1355
1356 EM_TX_LOCK(adapter);
1357 lem_txeof(adapter);
1358 if(!if_sendq_empty(ifp))
1359 lem_start_locked(ifp);
1360 EM_TX_UNLOCK(adapter);
1361 return (rx_done);
1362}
1363#endif /* DEVICE_POLLING */
1364
1365/*********************************************************************
1366 *
1367 * Legacy Interrupt Service routine
1368 *
1369 *********************************************************************/
1370static void
1371lem_intr(void *arg)
1372{
1373 struct adapter *adapter = arg;
1374 if_t ifp = adapter->ifp;
1375 u32 reg_icr;
1376
1377
1378 if ((if_getcapenable(ifp) & IFCAP_POLLING) ||
1379 ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0))
1380 return;
1381
1382 EM_CORE_LOCK(adapter);
1383 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1384 if (reg_icr & E1000_ICR_RXO)
1385 adapter->rx_overruns++;
1386
1387 if ((reg_icr == 0xffffffff) || (reg_icr == 0)) {
1388 EM_CORE_UNLOCK(adapter);
1389 return;
1390 }
1391
1392 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1393 callout_stop(&adapter->timer);
1394 adapter->hw.mac.get_link_status = 1;
1395 lem_update_link_status(adapter);
1396 /* Deal with TX cruft when link lost */
1397 lem_tx_purge(adapter);
1398 callout_reset(&adapter->timer, hz,
1399 lem_local_timer, adapter);
1400 EM_CORE_UNLOCK(adapter);
1401 return;
1402 }
1403
1404 EM_CORE_UNLOCK(adapter);
1405 lem_rxeof(adapter, -1, NULL);
1406
1407 EM_TX_LOCK(adapter);
1408 lem_txeof(adapter);
1409 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) &&
1410 (!if_sendq_empty(ifp)))
1411 lem_start_locked(ifp);
1412 EM_TX_UNLOCK(adapter);
1413 return;
1414}
1415
1416
1417static void
1418lem_handle_link(void *context, int pending)
1419{
1420 struct adapter *adapter = context;
1421 if_t ifp = adapter->ifp;
1422
1423 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1424 return;
1425
1426 EM_CORE_LOCK(adapter);
1427 callout_stop(&adapter->timer);
1428 lem_update_link_status(adapter);
1429 /* Deal with TX cruft when link lost */
1430 lem_tx_purge(adapter);
1431 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
1432 EM_CORE_UNLOCK(adapter);
1433}
1434
1435
1436/* Combined RX/TX handler, used by Legacy and MSI */
1437static void
1438lem_handle_rxtx(void *context, int pending)
1439{
1440 struct adapter *adapter = context;
1441 if_t ifp = adapter->ifp;
1442
1443
1444 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1445 bool more = lem_rxeof(adapter, adapter->rx_process_limit, NULL);
1446 EM_TX_LOCK(adapter);
1447 lem_txeof(adapter);
1448 if(!if_sendq_empty(ifp))
1449 lem_start_locked(ifp);
1450 EM_TX_UNLOCK(adapter);
1451 if (more) {
1452 taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
1453 return;
1454 }
1455 }
1456
1457 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
1458 lem_enable_intr(adapter);
1459}
1460
1461/*********************************************************************
1462 *
1463 * Fast Legacy/MSI Combined Interrupt Service routine
1464 *
1465 *********************************************************************/
1466static int
1467lem_irq_fast(void *arg)
1468{
1469 struct adapter *adapter = arg;
1470 if_t ifp;
1471 u32 reg_icr;
1472
1473 ifp = adapter->ifp;
1474
1475 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1476
1477 /* Hot eject? */
1478 if (reg_icr == 0xffffffff)
1479 return FILTER_STRAY;
1480
1481 /* Definitely not our interrupt. */
1482 if (reg_icr == 0x0)
1483 return FILTER_STRAY;
1484
1485 /*
1486 * Mask interrupts until the taskqueue is finished running. This is
1487 * cheap, just assume that it is needed. This also works around the
1488 * MSI message reordering errata on certain systems.
1489 */
1490 lem_disable_intr(adapter);
1491 taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
1492
1493 /* Link status change */
1494 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1495 adapter->hw.mac.get_link_status = 1;
1496 taskqueue_enqueue(taskqueue_fast, &adapter->link_task);
1497 }
1498
1499 if (reg_icr & E1000_ICR_RXO)
1500 adapter->rx_overruns++;
1501 return FILTER_HANDLED;
1502}
1503
1504
1505/*********************************************************************
1506 *
1507 * Media Ioctl callback
1508 *
1509 * This routine is called whenever the user queries the status of
1510 * the interface using ifconfig.
1511 *
1512 **********************************************************************/
1513static void
1514lem_media_status(if_t ifp, struct ifmediareq *ifmr)
1515{
1516 struct adapter *adapter = if_getsoftc(ifp);
1517 u_char fiber_type = IFM_1000_SX;
1518
1519 INIT_DEBUGOUT("lem_media_status: begin");
1520
1521 EM_CORE_LOCK(adapter);
1522 lem_update_link_status(adapter);
1523
1524 ifmr->ifm_status = IFM_AVALID;
1525 ifmr->ifm_active = IFM_ETHER;
1526
1527 if (!adapter->link_active) {
1528 EM_CORE_UNLOCK(adapter);
1529 return;
1530 }
1531
1532 ifmr->ifm_status |= IFM_ACTIVE;
1533
1534 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1535 (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1536 if (adapter->hw.mac.type == e1000_82545)
1537 fiber_type = IFM_1000_LX;
1538 ifmr->ifm_active |= fiber_type | IFM_FDX;
1539 } else {
1540 switch (adapter->link_speed) {
1541 case 10:
1542 ifmr->ifm_active |= IFM_10_T;
1543 break;
1544 case 100:
1545 ifmr->ifm_active |= IFM_100_TX;
1546 break;
1547 case 1000:
1548 ifmr->ifm_active |= IFM_1000_T;
1549 break;
1550 }
1551 if (adapter->link_duplex == FULL_DUPLEX)
1552 ifmr->ifm_active |= IFM_FDX;
1553 else
1554 ifmr->ifm_active |= IFM_HDX;
1555 }
1556 EM_CORE_UNLOCK(adapter);
1557}
1558
1559/*********************************************************************
1560 *
1561 * Media Ioctl callback
1562 *
1563 * This routine is called when the user changes speed/duplex using
1564 * media/mediopt option with ifconfig.
1565 *
1566 **********************************************************************/
1567static int
1568lem_media_change(if_t ifp)
1569{
1570 struct adapter *adapter = if_getsoftc(ifp);
1571 struct ifmedia *ifm = &adapter->media;
1572
1573 INIT_DEBUGOUT("lem_media_change: begin");
1574
1575 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1576 return (EINVAL);
1577
1578 EM_CORE_LOCK(adapter);
1579 switch (IFM_SUBTYPE(ifm->ifm_media)) {
1580 case IFM_AUTO:
1581 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1582 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1583 break;
1584 case IFM_1000_LX:
1585 case IFM_1000_SX:
1586 case IFM_1000_T:
1587 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1588 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1589 break;
1590 case IFM_100_TX:
1591 adapter->hw.mac.autoneg = FALSE;
1592 adapter->hw.phy.autoneg_advertised = 0;
1593 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1594 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1595 else
1596 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1597 break;
1598 case IFM_10_T:
1599 adapter->hw.mac.autoneg = FALSE;
1600 adapter->hw.phy.autoneg_advertised = 0;
1601 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1602 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1603 else
1604 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1605 break;
1606 default:
1607 device_printf(adapter->dev, "Unsupported media type\n");
1608 }
1609
1610 lem_init_locked(adapter);
1611 EM_CORE_UNLOCK(adapter);
1612
1613 return (0);
1614}
1615
1616/*********************************************************************
1617 *
1618 * This routine maps the mbufs to tx descriptors.
1619 *
1620 * return 0 on success, positive on failure
1621 **********************************************************************/
1622
1623static int
1624lem_xmit(struct adapter *adapter, struct mbuf **m_headp)
1625{
1626 bus_dma_segment_t segs[EM_MAX_SCATTER];
1627 bus_dmamap_t map;
1628 struct em_buffer *tx_buffer, *tx_buffer_mapped;
1629 struct e1000_tx_desc *ctxd = NULL;
1630 struct mbuf *m_head;
1631 u32 txd_upper, txd_lower, txd_used, txd_saved;
1632 int error, nsegs, i, j, first, last = 0;
1633
1634 m_head = *m_headp;
1635 txd_upper = txd_lower = txd_used = txd_saved = 0;
1636
1637 /*
1638 ** When doing checksum offload, it is critical to
1639 ** make sure the first mbuf has more than header,
1640 ** because that routine expects data to be present.
1641 */
1642 if ((m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD) &&
1643 (m_head->m_len < ETHER_HDR_LEN + sizeof(struct ip))) {
1644 m_head = m_pullup(m_head, ETHER_HDR_LEN + sizeof(struct ip));
1645 *m_headp = m_head;
1646 if (m_head == NULL)
1647 return (ENOBUFS);
1648 }
1649
1650 /*
1651 * Map the packet for DMA
1652 *
1653 * Capture the first descriptor index,
1654 * this descriptor will have the index
1655 * of the EOP which is the only one that
1656 * now gets a DONE bit writeback.
1657 */
1658 first = adapter->next_avail_tx_desc;
1659 tx_buffer = &adapter->tx_buffer_area[first];
1660 tx_buffer_mapped = tx_buffer;
1661 map = tx_buffer->map;
1662
1663 error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
1664 *m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
1665
1666 /*
1667 * There are two types of errors we can (try) to handle:
1668 * - EFBIG means the mbuf chain was too long and bus_dma ran
1669 * out of segments. Defragment the mbuf chain and try again.
1670 * - ENOMEM means bus_dma could not obtain enough bounce buffers
1671 * at this point in time. Defer sending and try again later.
1672 * All other errors, in particular EINVAL, are fatal and prevent the
1673 * mbuf chain from ever going through. Drop it and report error.
1674 */
1675 if (error == EFBIG) {
1676 struct mbuf *m;
1677
1678 m = m_defrag(*m_headp, M_NOWAIT);
1679 if (m == NULL) {
1680 adapter->mbuf_alloc_failed++;
1681 m_freem(*m_headp);
1682 *m_headp = NULL;
1683 return (ENOBUFS);
1684 }
1685 *m_headp = m;
1686
1687 /* Try it again */
1688 error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
1689 *m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
1690
1691 if (error) {
1692 adapter->no_tx_dma_setup++;
1693 m_freem(*m_headp);
1694 *m_headp = NULL;
1695 return (error);
1696 }
1697 } else if (error != 0) {
1698 adapter->no_tx_dma_setup++;
1699 return (error);
1700 }
1701
1702 if (nsegs > (adapter->num_tx_desc_avail - 2)) {
1703 adapter->no_tx_desc_avail2++;
1704 bus_dmamap_unload(adapter->txtag, map);
1705 return (ENOBUFS);
1706 }
1707 m_head = *m_headp;
1708
1709 /* Do hardware assists */
1710 if (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)
1711 lem_transmit_checksum_setup(adapter, m_head,
1712 &txd_upper, &txd_lower);
1713
1714 i = adapter->next_avail_tx_desc;
1715 if (adapter->pcix_82544)
1716 txd_saved = i;
1717
1718 /* Set up our transmit descriptors */
1719 for (j = 0; j < nsegs; j++) {
1720 bus_size_t seg_len;
1721 bus_addr_t seg_addr;
1722 /* If adapter is 82544 and on PCIX bus */
1723 if(adapter->pcix_82544) {
1724 DESC_ARRAY desc_array;
1725 u32 array_elements, counter;
1726 /*
1727 * Check the Address and Length combination and
1728 * split the data accordingly
1729 */
1730 array_elements = lem_fill_descriptors(segs[j].ds_addr,
1731 segs[j].ds_len, &desc_array);
1732 for (counter = 0; counter < array_elements; counter++) {
1733 if (txd_used == adapter->num_tx_desc_avail) {
1734 adapter->next_avail_tx_desc = txd_saved;
1735 adapter->no_tx_desc_avail2++;
1736 bus_dmamap_unload(adapter->txtag, map);
1737 return (ENOBUFS);
1738 }
1739 tx_buffer = &adapter->tx_buffer_area[i];
1740 ctxd = &adapter->tx_desc_base[i];
1741 ctxd->buffer_addr = htole64(
1742 desc_array.descriptor[counter].address);
1743 ctxd->lower.data = htole32(
1744 (adapter->txd_cmd | txd_lower | (u16)
1745 desc_array.descriptor[counter].length));
1746 ctxd->upper.data =
1747 htole32((txd_upper));
1748 last = i;
1749 if (++i == adapter->num_tx_desc)
1750 i = 0;
1751 tx_buffer->m_head = NULL;
1752 tx_buffer->next_eop = -1;
1753 txd_used++;
1754 }
1755 } else {
1756 tx_buffer = &adapter->tx_buffer_area[i];
1757 ctxd = &adapter->tx_desc_base[i];
1758 seg_addr = segs[j].ds_addr;
1759 seg_len = segs[j].ds_len;
1760 ctxd->buffer_addr = htole64(seg_addr);
1761 ctxd->lower.data = htole32(
1762 adapter->txd_cmd | txd_lower | seg_len);
1763 ctxd->upper.data =
1764 htole32(txd_upper);
1765 last = i;
1766 if (++i == adapter->num_tx_desc)
1767 i = 0;
1768 tx_buffer->m_head = NULL;
1769 tx_buffer->next_eop = -1;
1770 }
1771 }
1772
1773 adapter->next_avail_tx_desc = i;
1774
1775 if (adapter->pcix_82544)
1776 adapter->num_tx_desc_avail -= txd_used;
1777 else
1778 adapter->num_tx_desc_avail -= nsegs;
1779
1780 if (m_head->m_flags & M_VLANTAG) {
1781 /* Set the vlan id. */
1782 ctxd->upper.fields.special =
1783 htole16(m_head->m_pkthdr.ether_vtag);
1784 /* Tell hardware to add tag */
1785 ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE);
1786 }
1787
1788 tx_buffer->m_head = m_head;
1789 tx_buffer_mapped->map = tx_buffer->map;
1790 tx_buffer->map = map;
1791 bus_dmamap_sync(adapter->txtag, map, BUS_DMASYNC_PREWRITE);
1792
1793 /*
1794 * Last Descriptor of Packet
1795 * needs End Of Packet (EOP)
1796 * and Report Status (RS)
1797 */
1798 ctxd->lower.data |=
1799 htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
1800 /*
1801 * Keep track in the first buffer which
1802 * descriptor will be written back
1803 */
1804 tx_buffer = &adapter->tx_buffer_area[first];
1805 tx_buffer->next_eop = last;
1806 adapter->watchdog_time = ticks;
1807
1808 /*
1809 * Advance the Transmit Descriptor Tail (TDT), this tells the E1000
1810 * that this frame is available to transmit.
1811 */
1812 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
1813 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1814
1815#ifdef NIC_PARAVIRT
1816 if (adapter->csb) {
1817 adapter->csb->guest_tdt = i;
1818 /* XXX memory barrier ? */
1819 if (adapter->csb->guest_csb_on &&
1820 !(adapter->csb->host_need_txkick & 1)) {
1821 /* XXX maybe useless
1822 * clean the ring. maybe do it before ?
1823 * maybe a little bit of histeresys ?
1824 */
1825 if (adapter->num_tx_desc_avail <= 64) {// XXX
1826 lem_txeof(adapter);
1827 }
1828 return (0);
1829 }
1830 }
1831#endif /* NIC_PARAVIRT */
1832
1833#ifdef NIC_SEND_COMBINING
1834 if (adapter->sc_enable) {
1835 if (adapter->shadow_tdt & MIT_PENDING_INT) {
1836 /* signal intr and data pending */
1837 adapter->shadow_tdt = MIT_PENDING_TDT | (i & 0xffff);
1838 return (0);
1839 } else {
1840 adapter->shadow_tdt = MIT_PENDING_INT;
1841 }
1842 }
1843#endif /* NIC_SEND_COMBINING */
1844
1845 if (adapter->hw.mac.type == e1000_82547 &&
1846 adapter->link_duplex == HALF_DUPLEX)
1847 lem_82547_move_tail(adapter);
1848 else {
1849 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), i);
1850 if (adapter->hw.mac.type == e1000_82547)
1851 lem_82547_update_fifo_head(adapter,
1852 m_head->m_pkthdr.len);
1853 }
1854
1855 return (0);
1856}
1857
1858/*********************************************************************
1859 *
1860 * 82547 workaround to avoid controller hang in half-duplex environment.
1861 * The workaround is to avoid queuing a large packet that would span
1862 * the internal Tx FIFO ring boundary. We need to reset the FIFO pointers
1863 * in this case. We do that only when FIFO is quiescent.
1864 *
1865 **********************************************************************/
1866static void
1867lem_82547_move_tail(void *arg)
1868{
1869 struct adapter *adapter = arg;
1870 struct e1000_tx_desc *tx_desc;
1871 u16 hw_tdt, sw_tdt, length = 0;
1872 bool eop = 0;
1873
1874 EM_TX_LOCK_ASSERT(adapter);
1875
1876 hw_tdt = E1000_READ_REG(&adapter->hw, E1000_TDT(0));
1877 sw_tdt = adapter->next_avail_tx_desc;
1878
1879 while (hw_tdt != sw_tdt) {
1880 tx_desc = &adapter->tx_desc_base[hw_tdt];
1881 length += tx_desc->lower.flags.length;
1882 eop = tx_desc->lower.data & E1000_TXD_CMD_EOP;
1883 if (++hw_tdt == adapter->num_tx_desc)
1884 hw_tdt = 0;
1885
1886 if (eop) {
1887 if (lem_82547_fifo_workaround(adapter, length)) {
1888 adapter->tx_fifo_wrk_cnt++;
1889 callout_reset(&adapter->tx_fifo_timer, 1,
1890 lem_82547_move_tail, adapter);
1891 break;
1892 }
1893 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), hw_tdt);
1894 lem_82547_update_fifo_head(adapter, length);
1895 length = 0;
1896 }
1897 }
1898}
1899
1900static int
1901lem_82547_fifo_workaround(struct adapter *adapter, int len)
1902{
1903 int fifo_space, fifo_pkt_len;
1904
1905 fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1906
1907 if (adapter->link_duplex == HALF_DUPLEX) {
1908 fifo_space = adapter->tx_fifo_size - adapter->tx_fifo_head;
1909
1910 if (fifo_pkt_len >= (EM_82547_PKT_THRESH + fifo_space)) {
1911 if (lem_82547_tx_fifo_reset(adapter))
1912 return (0);
1913 else
1914 return (1);
1915 }
1916 }
1917
1918 return (0);
1919}
1920
1921static void
1922lem_82547_update_fifo_head(struct adapter *adapter, int len)
1923{
1924 int fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1925
1926 /* tx_fifo_head is always 16 byte aligned */
1927 adapter->tx_fifo_head += fifo_pkt_len;
1928 if (adapter->tx_fifo_head >= adapter->tx_fifo_size) {
1929 adapter->tx_fifo_head -= adapter->tx_fifo_size;
1930 }
1931}
1932
1933
1934static int
1935lem_82547_tx_fifo_reset(struct adapter *adapter)
1936{
1937 u32 tctl;
1938
1939 if ((E1000_READ_REG(&adapter->hw, E1000_TDT(0)) ==
1940 E1000_READ_REG(&adapter->hw, E1000_TDH(0))) &&
1941 (E1000_READ_REG(&adapter->hw, E1000_TDFT) ==
1942 E1000_READ_REG(&adapter->hw, E1000_TDFH)) &&
1943 (E1000_READ_REG(&adapter->hw, E1000_TDFTS) ==
1944 E1000_READ_REG(&adapter->hw, E1000_TDFHS)) &&
1945 (E1000_READ_REG(&adapter->hw, E1000_TDFPC) == 0)) {
1946 /* Disable TX unit */
1947 tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
1948 E1000_WRITE_REG(&adapter->hw, E1000_TCTL,
1949 tctl & ~E1000_TCTL_EN);
1950
1951 /* Reset FIFO pointers */
1952 E1000_WRITE_REG(&adapter->hw, E1000_TDFT,
1953 adapter->tx_head_addr);
1954 E1000_WRITE_REG(&adapter->hw, E1000_TDFH,
1955 adapter->tx_head_addr);
1956 E1000_WRITE_REG(&adapter->hw, E1000_TDFTS,
1957 adapter->tx_head_addr);
1958 E1000_WRITE_REG(&adapter->hw, E1000_TDFHS,
1959 adapter->tx_head_addr);
1960
1961 /* Re-enable TX unit */
1962 E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
1963 E1000_WRITE_FLUSH(&adapter->hw);
1964
1965 adapter->tx_fifo_head = 0;
1966 adapter->tx_fifo_reset_cnt++;
1967
1968 return (TRUE);
1969 }
1970 else {
1971 return (FALSE);
1972 }
1973}
1974
1975static void
1976lem_set_promisc(struct adapter *adapter)
1977{
1978 if_t ifp = adapter->ifp;
1979 u32 reg_rctl;
1980
1981 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1982
1983 if (if_getflags(ifp) & IFF_PROMISC) {
1984 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1985 /* Turn this on if you want to see bad packets */
1986 if (lem_debug_sbp)
1987 reg_rctl |= E1000_RCTL_SBP;
1988 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1989 } else if (if_getflags(ifp) & IFF_ALLMULTI) {
1990 reg_rctl |= E1000_RCTL_MPE;
1991 reg_rctl &= ~E1000_RCTL_UPE;
1992 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1993 }
1994}
1995
1996static void
1997lem_disable_promisc(struct adapter *adapter)
1998{
1999 if_t ifp = adapter->ifp;
2000 u32 reg_rctl;
2001 int mcnt = 0;
2002
2003 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2004 reg_rctl &= (~E1000_RCTL_UPE);
2005 if (if_getflags(ifp) & IFF_ALLMULTI)
2006 mcnt = MAX_NUM_MULTICAST_ADDRESSES;
2007 else
2008 mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
2009
2010 /* Don't disable if in MAX groups */
2011 if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
2012 reg_rctl &= (~E1000_RCTL_MPE);
2013 reg_rctl &= (~E1000_RCTL_SBP);
2014 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2015}
2016
2017
2018/*********************************************************************
2019 * Multicast Update
2020 *
2021 * This routine is called whenever multicast address list is updated.
2022 *
2023 **********************************************************************/
2024
2025static void
2026lem_set_multi(struct adapter *adapter)
2027{
2028 if_t ifp = adapter->ifp;
2029 u32 reg_rctl = 0;
2030 u8 *mta; /* Multicast array memory */
2031 int mcnt = 0;
2032
2033 IOCTL_DEBUGOUT("lem_set_multi: begin");
2034
2035 mta = adapter->mta;
2036 bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
2037
2038 if (adapter->hw.mac.type == e1000_82542 &&
2039 adapter->hw.revision_id == E1000_REVISION_2) {
2040 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2041 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
2042 e1000_pci_clear_mwi(&adapter->hw);
2043 reg_rctl |= E1000_RCTL_RST;
2044 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2045 msec_delay(5);
2046 }
2047
2048 if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
2049
2050 if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
2051 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2052 reg_rctl |= E1000_RCTL_MPE;
2053 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2054 } else
2055 e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
2056
2057 if (adapter->hw.mac.type == e1000_82542 &&
2058 adapter->hw.revision_id == E1000_REVISION_2) {
2059 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2060 reg_rctl &= ~E1000_RCTL_RST;
2061 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2062 msec_delay(5);
2063 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
2064 e1000_pci_set_mwi(&adapter->hw);
2065 }
2066}
2067
2068
2069/*********************************************************************
2070 * Timer routine
2071 *
2072 * This routine checks for link status and updates statistics.
2073 *
2074 **********************************************************************/
2075
2076static void
2077lem_local_timer(void *arg)
2078{
2079 struct adapter *adapter = arg;
2080
2081 EM_CORE_LOCK_ASSERT(adapter);
2082
2083 lem_update_link_status(adapter);
2084 lem_update_stats_counters(adapter);
2085
2086 lem_smartspeed(adapter);
2087
2088#ifdef NIC_PARAVIRT
2089 /* recover space if needed */
2090 if (adapter->csb && adapter->csb->guest_csb_on &&
2091 (adapter->watchdog_check == TRUE) &&
2092 (ticks - adapter->watchdog_time > EM_WATCHDOG) &&
2093 (adapter->num_tx_desc_avail != adapter->num_tx_desc) ) {
2094 lem_txeof(adapter);
2095 /*
2096 * lem_txeof() normally (except when space in the queue
2097 * runs low XXX) cleans watchdog_check so that
2098 * we do not hung.
2099 */
2100 }
2101#endif /* NIC_PARAVIRT */
2102 /*
2103 * We check the watchdog: the time since
2104 * the last TX descriptor was cleaned.
2105 * This implies a functional TX engine.
2106 */
2107 if ((adapter->watchdog_check == TRUE) &&
2108 (ticks - adapter->watchdog_time > EM_WATCHDOG))
2109 goto hung;
2110
2111 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
2112 return;
2113hung:
2114 device_printf(adapter->dev, "Watchdog timeout -- resetting\n");
2115 if_setdrvflagbits(adapter->ifp, 0, IFF_DRV_RUNNING);
2116 adapter->watchdog_events++;
2117 lem_init_locked(adapter);
2118}
2119
2120static void
2121lem_update_link_status(struct adapter *adapter)
2122{
2123 struct e1000_hw *hw = &adapter->hw;
2124 if_t ifp = adapter->ifp;
2125 device_t dev = adapter->dev;
2126 u32 link_check = 0;
2127
2128 /* Get the cached link value or read phy for real */
2129 switch (hw->phy.media_type) {
2130 case e1000_media_type_copper:
2131 if (hw->mac.get_link_status) {
2132 /* Do the work to read phy */
2133 e1000_check_for_link(hw);
2134 link_check = !hw->mac.get_link_status;
2135 if (link_check) /* ESB2 fix */
2136 e1000_cfg_on_link_up(hw);
2137 } else
2138 link_check = TRUE;
2139 break;
2140 case e1000_media_type_fiber:
2141 e1000_check_for_link(hw);
2142 link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2143 E1000_STATUS_LU);
2144 break;
2145 case e1000_media_type_internal_serdes:
2146 e1000_check_for_link(hw);
2147 link_check = adapter->hw.mac.serdes_has_link;
2148 break;
2149 default:
2150 case e1000_media_type_unknown:
2151 break;
2152 }
2153
2154 /* Now check for a transition */
2155 if (link_check && (adapter->link_active == 0)) {
2156 e1000_get_speed_and_duplex(hw, &adapter->link_speed,
2157 &adapter->link_duplex);
2158 if (bootverbose)
2159 device_printf(dev, "Link is up %d Mbps %s\n",
2160 adapter->link_speed,
2161 ((adapter->link_duplex == FULL_DUPLEX) ?
2162 "Full Duplex" : "Half Duplex"));
2163 adapter->link_active = 1;
2164 adapter->smartspeed = 0;
2165 if_setbaudrate(ifp, adapter->link_speed * 1000000);
2166 if_link_state_change(ifp, LINK_STATE_UP);
2167 } else if (!link_check && (adapter->link_active == 1)) {
2168 if_setbaudrate(ifp, 0);
2169 adapter->link_speed = 0;
2170 adapter->link_duplex = 0;
2171 if (bootverbose)
2172 device_printf(dev, "Link is Down\n");
2173 adapter->link_active = 0;
2174 /* Link down, disable watchdog */
2175 adapter->watchdog_check = FALSE;
2176 if_link_state_change(ifp, LINK_STATE_DOWN);
2177 }
2178}
2179
2180/*********************************************************************
2181 *
2182 * This routine disables all traffic on the adapter by issuing a
2183 * global reset on the MAC and deallocates TX/RX buffers.
2184 *
2185 * This routine should always be called with BOTH the CORE
2186 * and TX locks.
2187 **********************************************************************/
2188
2189static void
2190lem_stop(void *arg)
2191{
2192 struct adapter *adapter = arg;
2193 if_t ifp = adapter->ifp;
2194
2195 EM_CORE_LOCK_ASSERT(adapter);
2196 EM_TX_LOCK_ASSERT(adapter);
2197
2198 INIT_DEBUGOUT("lem_stop: begin");
2199
2200 lem_disable_intr(adapter);
2201 callout_stop(&adapter->timer);
2202 callout_stop(&adapter->tx_fifo_timer);
2203
2204 /* Tell the stack that the interface is no longer active */
2205 if_setdrvflagbits(ifp, 0, (IFF_DRV_RUNNING | IFF_DRV_OACTIVE));
2206
2207 e1000_reset_hw(&adapter->hw);
2208 if (adapter->hw.mac.type >= e1000_82544)
2209 E1000_WRITE_REG(&adapter->hw, E1000_WUC, 0);
2210
2211 e1000_led_off(&adapter->hw);
2212 e1000_cleanup_led(&adapter->hw);
2213}
2214
2215
2216/*********************************************************************
2217 *
2218 * Determine hardware revision.
2219 *
2220 **********************************************************************/
2221static void
2222lem_identify_hardware(struct adapter *adapter)
2223{
2224 device_t dev = adapter->dev;
2225
2226 /* Make sure our PCI config space has the necessary stuff set */
2227 pci_enable_busmaster(dev);
2228 adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
2229
2230 /* Save off the information about this board */
2231 adapter->hw.vendor_id = pci_get_vendor(dev);
2232 adapter->hw.device_id = pci_get_device(dev);
2233 adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
2234 adapter->hw.subsystem_vendor_id =
2235 pci_read_config(dev, PCIR_SUBVEND_0, 2);
2236 adapter->hw.subsystem_device_id =
2237 pci_read_config(dev, PCIR_SUBDEV_0, 2);
2238
2239 /* Do Shared Code Init and Setup */
2240 if (e1000_set_mac_type(&adapter->hw)) {
2241 device_printf(dev, "Setup init failure\n");
2242 return;
2243 }
2244}
2245
2246static int
2247lem_allocate_pci_resources(struct adapter *adapter)
2248{
2249 device_t dev = adapter->dev;
2250 int val, rid, error = E1000_SUCCESS;
2251
2252 rid = PCIR_BAR(0);
2253 adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2254 &rid, RF_ACTIVE);
2255 if (adapter->memory == NULL) {
2256 device_printf(dev, "Unable to allocate bus resource: memory\n");
2257 return (ENXIO);
2258 }
2259 adapter->osdep.mem_bus_space_tag =
2260 rman_get_bustag(adapter->memory);
2261 adapter->osdep.mem_bus_space_handle =
2262 rman_get_bushandle(adapter->memory);
2263 adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
2264
2265 /* Only older adapters use IO mapping */
2266 if (adapter->hw.mac.type > e1000_82543) {
2267 /* Figure our where our IO BAR is ? */
2268 for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
2269 val = pci_read_config(dev, rid, 4);
2270 if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2271 adapter->io_rid = rid;
2272 break;
2273 }
2274 rid += 4;
2275 /* check for 64bit BAR */
2276 if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2277 rid += 4;
2278 }
2279 if (rid >= PCIR_CIS) {
2280 device_printf(dev, "Unable to locate IO BAR\n");
2281 return (ENXIO);
2282 }
2283 adapter->ioport = bus_alloc_resource_any(dev,
2284 SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
2285 if (adapter->ioport == NULL) {
2286 device_printf(dev, "Unable to allocate bus resource: "
2287 "ioport\n");
2288 return (ENXIO);
2289 }
2290 adapter->hw.io_base = 0;
2291 adapter->osdep.io_bus_space_tag =
2292 rman_get_bustag(adapter->ioport);
2293 adapter->osdep.io_bus_space_handle =
2294 rman_get_bushandle(adapter->ioport);
2295 }
2296
2297 adapter->hw.back = &adapter->osdep;
2298
2299 return (error);
2300}
2301
2302/*********************************************************************
2303 *
2304 * Setup the Legacy or MSI Interrupt handler
2305 *
2306 **********************************************************************/
2307int
2308lem_allocate_irq(struct adapter *adapter)
2309{
2310 device_t dev = adapter->dev;
2311 int error, rid = 0;
2312
2313 /* Manually turn off all interrupts */
2314 E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
2315
2316 /* We allocate a single interrupt resource */
2317 adapter->res[0] = bus_alloc_resource_any(dev,
2318 SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
2319 if (adapter->res[0] == NULL) {
2320 device_printf(dev, "Unable to allocate bus resource: "
2321 "interrupt\n");
2322 return (ENXIO);
2323 }
2324
2325 /* Do Legacy setup? */
2326 if (lem_use_legacy_irq) {
2327 if ((error = bus_setup_intr(dev, adapter->res[0],
2328 INTR_TYPE_NET | INTR_MPSAFE, NULL, lem_intr, adapter,
2329 &adapter->tag[0])) != 0) {
2330 device_printf(dev,
2331 "Failed to register interrupt handler");
2332 return (error);
2333 }
2334 return (0);
2335 }
2336
2337 /*
2338 * Use a Fast interrupt and the associated
2339 * deferred processing contexts.
2340 */
2341 TASK_INIT(&adapter->rxtx_task, 0, lem_handle_rxtx, adapter);
2342 TASK_INIT(&adapter->link_task, 0, lem_handle_link, adapter);
2343 adapter->tq = taskqueue_create_fast("lem_taskq", M_NOWAIT,
2344 taskqueue_thread_enqueue, &adapter->tq);
2345 taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s taskq",
2346 device_get_nameunit(adapter->dev));
2347 if ((error = bus_setup_intr(dev, adapter->res[0],
2348 INTR_TYPE_NET, lem_irq_fast, NULL, adapter,
2349 &adapter->tag[0])) != 0) {
2350 device_printf(dev, "Failed to register fast interrupt "
2351 "handler: %d\n", error);
2352 taskqueue_free(adapter->tq);
2353 adapter->tq = NULL;
2354 return (error);
2355 }
2356
2357 return (0);
2358}
2359
2360
2361static void
2362lem_free_pci_resources(struct adapter *adapter)
2363{
2364 device_t dev = adapter->dev;
2365
2366
2367 if (adapter->tag[0] != NULL) {
2368 bus_teardown_intr(dev, adapter->res[0],
2369 adapter->tag[0]);
2370 adapter->tag[0] = NULL;
2371 }
2372
2373 if (adapter->res[0] != NULL) {
2374 bus_release_resource(dev, SYS_RES_IRQ,
2375 0, adapter->res[0]);
2376 }
2377
2378 if (adapter->memory != NULL)
2379 bus_release_resource(dev, SYS_RES_MEMORY,
2380 PCIR_BAR(0), adapter->memory);
2381
2382 if (adapter->ioport != NULL)
2383 bus_release_resource(dev, SYS_RES_IOPORT,
2384 adapter->io_rid, adapter->ioport);
2385}
2386
2387
2388/*********************************************************************
2389 *
2390 * Initialize the hardware to a configuration
2391 * as specified by the adapter structure.
2392 *
2393 **********************************************************************/
2394static int
2395lem_hardware_init(struct adapter *adapter)
2396{
2397 device_t dev = adapter->dev;
2398 u16 rx_buffer_size;
2399
2400 INIT_DEBUGOUT("lem_hardware_init: begin");
2401
2402 /* Issue a global reset */
2403 e1000_reset_hw(&adapter->hw);
2404
2405 /* When hardware is reset, fifo_head is also reset */
2406 adapter->tx_fifo_head = 0;
2407
2408 /*
2409 * These parameters control the automatic generation (Tx) and
2410 * response (Rx) to Ethernet PAUSE frames.
2411 * - High water mark should allow for at least two frames to be
2412 * received after sending an XOFF.
2413 * - Low water mark works best when it is very near the high water mark.
2414 * This allows the receiver to restart by sending XON when it has
2415 * drained a bit. Here we use an arbitary value of 1500 which will
2416 * restart after one full frame is pulled from the buffer. There
2417 * could be several smaller frames in the buffer and if so they will
2418 * not trigger the XON until their total number reduces the buffer
2419 * by 1500.
2420 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2421 */
2422 rx_buffer_size = ((E1000_READ_REG(&adapter->hw, E1000_PBA) &
2423 0xffff) << 10 );
2424
2425 adapter->hw.fc.high_water = rx_buffer_size -
2426 roundup2(adapter->max_frame_size, 1024);
2427 adapter->hw.fc.low_water = adapter->hw.fc.high_water - 1500;
2428
2429 adapter->hw.fc.pause_time = EM_FC_PAUSE_TIME;
2430 adapter->hw.fc.send_xon = TRUE;
2431
2432 /* Set Flow control, use the tunable location if sane */
2433 if ((lem_fc_setting >= 0) && (lem_fc_setting < 4))
2434 adapter->hw.fc.requested_mode = lem_fc_setting;
2435 else
2436 adapter->hw.fc.requested_mode = e1000_fc_none;
2437
2438 if (e1000_init_hw(&adapter->hw) < 0) {
2439 device_printf(dev, "Hardware Initialization Failed\n");
2440 return (EIO);
2441 }
2442
2443 e1000_check_for_link(&adapter->hw);
2444
2445 return (0);
2446}
2447
2448/*********************************************************************
2449 *
2450 * Setup networking device structure and register an interface.
2451 *
2452 **********************************************************************/
2453static int
2454lem_setup_interface(device_t dev, struct adapter *adapter)
2455{
2456 if_t ifp;
2457
2458 INIT_DEBUGOUT("lem_setup_interface: begin");
2459
2460 ifp = adapter->ifp = if_gethandle(IFT_ETHER);
2461 if (ifp == (void *)NULL) {
2462 device_printf(dev, "can not allocate ifnet structure\n");
2463 return (-1);
2464 }
2465 if_initname(ifp, device_get_name(dev), device_get_unit(dev));
2466 if_setinitfn(ifp, lem_init);
2467 if_setsoftc(ifp, adapter);
2468 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
2469 if_setioctlfn(ifp, lem_ioctl);
2470 if_setstartfn(ifp, lem_start);
2471 if_setgetcounterfn(ifp, lem_get_counter);
2472 if_setsendqlen(ifp, adapter->num_tx_desc - 1);
2473 if_setsendqready(ifp);
2474
2475 ether_ifattach(ifp, adapter->hw.mac.addr);
2476
2477 if_setcapabilities(ifp, 0);
2478
2479 if (adapter->hw.mac.type >= e1000_82543) {
2480 if_setcapabilitiesbit(ifp, IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM, 0);
2481 if_setcapenablebit(ifp, IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM, 0);
2482 }
2483
2484 /*
2485 * Tell the upper layer(s) we support long frames.
2486 */
2487 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2488 if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU, 0);
2489 if_setcapenablebit(ifp, IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU, 0);
2490
2491 /*
2492 ** Dont turn this on by default, if vlans are
2493 ** created on another pseudo device (eg. lagg)
2494 ** then vlan events are not passed thru, breaking
2495 ** operation, but with HW FILTER off it works. If
2496 ** using vlans directly on the em driver you can
2497 ** enable this and get full hardware tag filtering.
2498 */
2499 if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER, 0);
2500
2501#ifdef DEVICE_POLLING
2502 if_setcapabilitiesbit(ifp, IFCAP_POLLING, 0);
2503#endif
2504
2505 /* Enable only WOL MAGIC by default */
2506 if (adapter->wol) {
2507 if_setcapabilitiesbit(ifp, IFCAP_WOL, 0);
2508 if_setcapenablebit(ifp, IFCAP_WOL_MAGIC, 0);
2509 }
2510
2511 /*
2512 * Specify the media types supported by this adapter and register
2513 * callbacks to update media and link information
2514 */
2515 ifmedia_init(&adapter->media, IFM_IMASK,
2516 lem_media_change, lem_media_status);
2517 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2518 (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2519 u_char fiber_type = IFM_1000_SX; /* default type */
2520
2521 if (adapter->hw.mac.type == e1000_82545)
2522 fiber_type = IFM_1000_LX;
2523 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type | IFM_FDX,
2524 0, NULL);
2525 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2526 } else {
2527 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2528 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX,
2529 0, NULL);
2530 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX,
2531 0, NULL);
2532 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
2533 0, NULL);
2534 if (adapter->hw.phy.type != e1000_phy_ife) {
2535 ifmedia_add(&adapter->media,
2536 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2537 ifmedia_add(&adapter->media,
2538 IFM_ETHER | IFM_1000_T, 0, NULL);
2539 }
2540 }
2541 ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2542 ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO);
2543 return (0);
2544}
2545
2546
2547/*********************************************************************
2548 *
2549 * Workaround for SmartSpeed on 82541 and 82547 controllers
2550 *
2551 **********************************************************************/
2552static void
2553lem_smartspeed(struct adapter *adapter)
2554{
2555 u16 phy_tmp;
2556
2557 if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2558 adapter->hw.mac.autoneg == 0 ||
2559 (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2560 return;
2561
2562 if (adapter->smartspeed == 0) {
2563 /* If Master/Slave config fault is asserted twice,
2564 * we assume back-to-back */
2565 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2566 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2567 return;
2568 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2569 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2570 e1000_read_phy_reg(&adapter->hw,
2571 PHY_1000T_CTRL, &phy_tmp);
2572 if(phy_tmp & CR_1000T_MS_ENABLE) {
2573 phy_tmp &= ~CR_1000T_MS_ENABLE;
2574 e1000_write_phy_reg(&adapter->hw,
2575 PHY_1000T_CTRL, phy_tmp);
2576 adapter->smartspeed++;
2577 if(adapter->hw.mac.autoneg &&
2578 !e1000_copper_link_autoneg(&adapter->hw) &&
2579 !e1000_read_phy_reg(&adapter->hw,
2580 PHY_CONTROL, &phy_tmp)) {
2581 phy_tmp |= (MII_CR_AUTO_NEG_EN |
2582 MII_CR_RESTART_AUTO_NEG);
2583 e1000_write_phy_reg(&adapter->hw,
2584 PHY_CONTROL, phy_tmp);
2585 }
2586 }
2587 }
2588 return;
2589 } else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2590 /* If still no link, perhaps using 2/3 pair cable */
2591 e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2592 phy_tmp |= CR_1000T_MS_ENABLE;
2593 e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2594 if(adapter->hw.mac.autoneg &&
2595 !e1000_copper_link_autoneg(&adapter->hw) &&
2596 !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2597 phy_tmp |= (MII_CR_AUTO_NEG_EN |
2598 MII_CR_RESTART_AUTO_NEG);
2599 e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2600 }
2601 }
2602 /* Restart process after EM_SMARTSPEED_MAX iterations */
2603 if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2604 adapter->smartspeed = 0;
2605}
2606
2607
2608/*
2609 * Manage DMA'able memory.
2610 */
2611static void
2612lem_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
2613{
2614 if (error)
2615 return;
2616 *(bus_addr_t *) arg = segs[0].ds_addr;
2617}
2618
2619static int
2620lem_dma_malloc(struct adapter *adapter, bus_size_t size,
2621 struct em_dma_alloc *dma, int mapflags)
2622{
2623 int error;
2624
2625 error = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */
2626 EM_DBA_ALIGN, 0, /* alignment, bounds */
2627 BUS_SPACE_MAXADDR, /* lowaddr */
2628 BUS_SPACE_MAXADDR, /* highaddr */
2629 NULL, NULL, /* filter, filterarg */
2630 size, /* maxsize */
2631 1, /* nsegments */
2632 size, /* maxsegsize */
2633 0, /* flags */
2634 NULL, /* lockfunc */
2635 NULL, /* lockarg */
2636 &dma->dma_tag);
2637 if (error) {
2638 device_printf(adapter->dev,
2639 "%s: bus_dma_tag_create failed: %d\n",
2640 __func__, error);
2641 goto fail_0;
2642 }
2643
2644 error = bus_dmamem_alloc(dma->dma_tag, (void**) &dma->dma_vaddr,
2645 BUS_DMA_NOWAIT | BUS_DMA_COHERENT, &dma->dma_map);
2646 if (error) {
2647 device_printf(adapter->dev,
2648 "%s: bus_dmamem_alloc(%ju) failed: %d\n",
2649 __func__, (uintmax_t)size, error);
2650 goto fail_2;
2651 }
2652
2653 dma->dma_paddr = 0;
2654 error = bus_dmamap_load(dma->dma_tag, dma->dma_map, dma->dma_vaddr,
2655 size, lem_dmamap_cb, &dma->dma_paddr, mapflags | BUS_DMA_NOWAIT);
2656 if (error || dma->dma_paddr == 0) {
2657 device_printf(adapter->dev,
2658 "%s: bus_dmamap_load failed: %d\n",
2659 __func__, error);
2660 goto fail_3;
2661 }
2662
2663 return (0);
2664
2665fail_3:
2666 bus_dmamap_unload(dma->dma_tag, dma->dma_map);
2667fail_2:
2668 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
2669 bus_dma_tag_destroy(dma->dma_tag);
2670fail_0:
2671 dma->dma_tag = NULL;
2672
2673 return (error);
2674}
2675
2676static void
2677lem_dma_free(struct adapter *adapter, struct em_dma_alloc *dma)
2678{
2679 if (dma->dma_tag == NULL)
2680 return;
2681 if (dma->dma_paddr != 0) {
2682 bus_dmamap_sync(dma->dma_tag, dma->dma_map,
2683 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2684 bus_dmamap_unload(dma->dma_tag, dma->dma_map);
2685 dma->dma_paddr = 0;
2686 }
2687 if (dma->dma_vaddr != NULL) {
2688 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
2689 dma->dma_vaddr = NULL;
2690 }
2691 bus_dma_tag_destroy(dma->dma_tag);
2692 dma->dma_tag = NULL;
2693}
2694
2695
2696/*********************************************************************
2697 *
2698 * Allocate memory for tx_buffer structures. The tx_buffer stores all
2699 * the information needed to transmit a packet on the wire.
2700 *
2701 **********************************************************************/
2702static int
2703lem_allocate_transmit_structures(struct adapter *adapter)
2704{
2705 device_t dev = adapter->dev;
2706 struct em_buffer *tx_buffer;
2707 int error;
2708
2709 /*
2710 * Create DMA tags for tx descriptors
2711 */
2712 if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
2713 1, 0, /* alignment, bounds */
2714 BUS_SPACE_MAXADDR, /* lowaddr */
2715 BUS_SPACE_MAXADDR, /* highaddr */
2716 NULL, NULL, /* filter, filterarg */
2717 MCLBYTES * EM_MAX_SCATTER, /* maxsize */
2718 EM_MAX_SCATTER, /* nsegments */
2719 MCLBYTES, /* maxsegsize */
2720 0, /* flags */
2721 NULL, /* lockfunc */
2722 NULL, /* lockarg */
2723 &adapter->txtag)) != 0) {
2724 device_printf(dev, "Unable to allocate TX DMA tag\n");
2725 goto fail;
2726 }
2727
2728 adapter->tx_buffer_area = malloc(sizeof(struct em_buffer) *
2729 adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
2730 if (adapter->tx_buffer_area == NULL) {
2731 device_printf(dev, "Unable to allocate tx_buffer memory\n");
2732 error = ENOMEM;
2733 goto fail;
2734 }
2735
2736 /* Create the descriptor buffer dma maps */
2737 for (int i = 0; i < adapter->num_tx_desc; i++) {
2738 tx_buffer = &adapter->tx_buffer_area[i];
2739 error = bus_dmamap_create(adapter->txtag, 0, &tx_buffer->map);
2740 if (error != 0) {
2741 device_printf(dev, "Unable to create TX DMA map\n");
2742 goto fail;
2743 }
2744 tx_buffer->next_eop = -1;
2745 }
2746
2747 return (0);
2748fail:
2749 lem_free_transmit_structures(adapter);
2750 return (error);
2751}
2752
2753/*********************************************************************
2754 *
2755 * (Re)Initialize transmit structures.
2756 *
2757 **********************************************************************/
2758static void
2759lem_setup_transmit_structures(struct adapter *adapter)
2760{
2761 struct em_buffer *tx_buffer;
2762#ifdef DEV_NETMAP
2763 /* we are already locked */
2764 struct netmap_adapter *na = netmap_getna(adapter->ifp);
2765 struct netmap_slot *slot = netmap_reset(na, NR_TX, 0, 0);
2766#endif /* DEV_NETMAP */
2767
2768 /* Clear the old ring contents */
2769 bzero(adapter->tx_desc_base,
2770 (sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
2771
2772 /* Free any existing TX buffers */
2773 for (int i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) {
2774 tx_buffer = &adapter->tx_buffer_area[i];
2775 bus_dmamap_sync(adapter->txtag, tx_buffer->map,
2776 BUS_DMASYNC_POSTWRITE);
2777 bus_dmamap_unload(adapter->txtag, tx_buffer->map);
2778 m_freem(tx_buffer->m_head);
2779 tx_buffer->m_head = NULL;
2780#ifdef DEV_NETMAP
2781 if (slot) {
2782 /* the i-th NIC entry goes to slot si */
2783 int si = netmap_idx_n2k(&na->tx_rings[0], i);
2784 uint64_t paddr;
2785 void *addr;
2786
2787 addr = PNMB(na, slot + si, &paddr);
2788 adapter->tx_desc_base[i].buffer_addr = htole64(paddr);
2789 /* reload the map for netmap mode */
2790 netmap_load_map(na, adapter->txtag, tx_buffer->map, addr);
2791 }
2792#endif /* DEV_NETMAP */
2793 tx_buffer->next_eop = -1;
2794 }
2795
2796 /* Reset state */
2797 adapter->last_hw_offload = 0;
2798 adapter->next_avail_tx_desc = 0;
2799 adapter->next_tx_to_clean = 0;
2800 adapter->num_tx_desc_avail = adapter->num_tx_desc;
2801
2802 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
2803 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2804
2805 return;
2806}
2807
2808/*********************************************************************
2809 *
2810 * Enable transmit unit.
2811 *
2812 **********************************************************************/
2813static void
2814lem_initialize_transmit_unit(struct adapter *adapter)
2815{
2816 u32 tctl, tipg = 0;
2817 u64 bus_addr;
2818
2819 INIT_DEBUGOUT("lem_initialize_transmit_unit: begin");
2820 /* Setup the Base and Length of the Tx Descriptor Ring */
2821 bus_addr = adapter->txdma.dma_paddr;
2822 E1000_WRITE_REG(&adapter->hw, E1000_TDLEN(0),
2823 adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
2824 E1000_WRITE_REG(&adapter->hw, E1000_TDBAH(0),
2825 (u32)(bus_addr >> 32));
2826 E1000_WRITE_REG(&adapter->hw, E1000_TDBAL(0),
2827 (u32)bus_addr);
2828 /* Setup the HW Tx Head and Tail descriptor pointers */
2829 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), 0);
2830 E1000_WRITE_REG(&adapter->hw, E1000_TDH(0), 0);
2831
2832 HW_DEBUGOUT2("Base = %x, Length = %x\n",
2833 E1000_READ_REG(&adapter->hw, E1000_TDBAL(0)),
2834 E1000_READ_REG(&adapter->hw, E1000_TDLEN(0)));
2835
2836 /* Set the default values for the Tx Inter Packet Gap timer */
2837 switch (adapter->hw.mac.type) {
2838 case e1000_82542:
2839 tipg = DEFAULT_82542_TIPG_IPGT;
2840 tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2841 tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2842 break;
2843 default:
2844 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2845 (adapter->hw.phy.media_type ==
2846 e1000_media_type_internal_serdes))
2847 tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2848 else
2849 tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
2850 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2851 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2852 }
2853
2854 E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
2855 E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
2856 if(adapter->hw.mac.type >= e1000_82540)
2857 E1000_WRITE_REG(&adapter->hw, E1000_TADV,
2858 adapter->tx_abs_int_delay.value);
2859
2860 /* Program the Transmit Control Register */
2861 tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
2862 tctl &= ~E1000_TCTL_CT;
2863 tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
2864 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
2865
2866 /* This write will effectively turn on the transmit unit. */
2867 E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
2868
2869 /* Setup Transmit Descriptor Base Settings */
2870 adapter->txd_cmd = E1000_TXD_CMD_IFCS;
2871
2872 if (adapter->tx_int_delay.value > 0)
2873 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2874}
2875
2876/*********************************************************************
2877 *
2878 * Free all transmit related data structures.
2879 *
2880 **********************************************************************/
2881static void
2882lem_free_transmit_structures(struct adapter *adapter)
2883{
2884 struct em_buffer *tx_buffer;
2885
2886 INIT_DEBUGOUT("free_transmit_structures: begin");
2887
2888 if (adapter->tx_buffer_area != NULL) {
2889 for (int i = 0; i < adapter->num_tx_desc; i++) {
2890 tx_buffer = &adapter->tx_buffer_area[i];
2891 if (tx_buffer->m_head != NULL) {
2892 bus_dmamap_sync(adapter->txtag, tx_buffer->map,
2893 BUS_DMASYNC_POSTWRITE);
2894 bus_dmamap_unload(adapter->txtag,
2895 tx_buffer->map);
2896 m_freem(tx_buffer->m_head);
2897 tx_buffer->m_head = NULL;
2898 } else if (tx_buffer->map != NULL)
2899 bus_dmamap_unload(adapter->txtag,
2900 tx_buffer->map);
2901 if (tx_buffer->map != NULL) {
2902 bus_dmamap_destroy(adapter->txtag,
2903 tx_buffer->map);
2904 tx_buffer->map = NULL;
2905 }
2906 }
2907 }
2908 if (adapter->tx_buffer_area != NULL) {
2909 free(adapter->tx_buffer_area, M_DEVBUF);
2910 adapter->tx_buffer_area = NULL;
2911 }
2912 if (adapter->txtag != NULL) {
2913 bus_dma_tag_destroy(adapter->txtag);
2914 adapter->txtag = NULL;
2915 }
2916}
2917
2918/*********************************************************************
2919 *
2920 * The offload context needs to be set when we transfer the first
2921 * packet of a particular protocol (TCP/UDP). This routine has been
2922 * enhanced to deal with inserted VLAN headers, and IPV6 (not complete)
2923 *
2924 * Added back the old method of keeping the current context type
2925 * and not setting if unnecessary, as this is reported to be a
2926 * big performance win. -jfv
2927 **********************************************************************/
2928static void
2929lem_transmit_checksum_setup(struct adapter *adapter, struct mbuf *mp,
2930 u32 *txd_upper, u32 *txd_lower)
2931{
2932 struct e1000_context_desc *TXD = NULL;
2933 struct em_buffer *tx_buffer;
2934 struct ether_vlan_header *eh;
2935 struct ip *ip = NULL;
2936 struct ip6_hdr *ip6;
2937 int curr_txd, ehdrlen;
2938 u32 cmd, hdr_len, ip_hlen;
2939 u16 etype;
2940 u8 ipproto;
2941
2942
2943 cmd = hdr_len = ipproto = 0;
2944 *txd_upper = *txd_lower = 0;
2945 curr_txd = adapter->next_avail_tx_desc;
2946
2947 /*
2948 * Determine where frame payload starts.
2949 * Jump over vlan headers if already present,
2950 * helpful for QinQ too.
2951 */
2952 eh = mtod(mp, struct ether_vlan_header *);
2953 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
2954 etype = ntohs(eh->evl_proto);
2955 ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
2956 } else {
2957 etype = ntohs(eh->evl_encap_proto);
2958 ehdrlen = ETHER_HDR_LEN;
2959 }
2960
2961 /*
2962 * We only support TCP/UDP for IPv4 and IPv6 for the moment.
2963 * TODO: Support SCTP too when it hits the tree.
2964 */
2965 switch (etype) {
2966 case ETHERTYPE_IP:
2967 ip = (struct ip *)(mp->m_data + ehdrlen);
2968 ip_hlen = ip->ip_hl << 2;
2969
2970 /* Setup of IP header checksum. */
2971 if (mp->m_pkthdr.csum_flags & CSUM_IP) {
2972 /*
2973 * Start offset for header checksum calculation.
2974 * End offset for header checksum calculation.
2975 * Offset of place to put the checksum.
2976 */
2977 TXD = (struct e1000_context_desc *)
2978 &adapter->tx_desc_base[curr_txd];
2979 TXD->lower_setup.ip_fields.ipcss = ehdrlen;
2980 TXD->lower_setup.ip_fields.ipcse =
2981 htole16(ehdrlen + ip_hlen);
2982 TXD->lower_setup.ip_fields.ipcso =
2983 ehdrlen + offsetof(struct ip, ip_sum);
2984 cmd |= E1000_TXD_CMD_IP;
2985 *txd_upper |= E1000_TXD_POPTS_IXSM << 8;
2986 }
2987
2988 hdr_len = ehdrlen + ip_hlen;
2989 ipproto = ip->ip_p;
2990
2991 break;
2992 case ETHERTYPE_IPV6:
2993 ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
2994 ip_hlen = sizeof(struct ip6_hdr); /* XXX: No header stacking. */
2995
2996 /* IPv6 doesn't have a header checksum. */
2997
2998 hdr_len = ehdrlen + ip_hlen;
2999 ipproto = ip6->ip6_nxt;
3000 break;
3001
3002 default:
3003 return;
3004 }
3005
3006 switch (ipproto) {
3007 case IPPROTO_TCP:
3008 if (mp->m_pkthdr.csum_flags & CSUM_TCP) {
3009 *txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
3010 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
3011 /* no need for context if already set */
3012 if (adapter->last_hw_offload == CSUM_TCP)
3013 return;
3014 adapter->last_hw_offload = CSUM_TCP;
3015 /*
3016 * Start offset for payload checksum calculation.
3017 * End offset for payload checksum calculation.
3018 * Offset of place to put the checksum.
3019 */
3020 TXD = (struct e1000_context_desc *)
3021 &adapter->tx_desc_base[curr_txd];
3022 TXD->upper_setup.tcp_fields.tucss = hdr_len;
3023 TXD->upper_setup.tcp_fields.tucse = htole16(0);
3024 TXD->upper_setup.tcp_fields.tucso =
3025 hdr_len + offsetof(struct tcphdr, th_sum);
3026 cmd |= E1000_TXD_CMD_TCP;
3027 }
3028 break;
3029 case IPPROTO_UDP:
3030 {
3031 if (mp->m_pkthdr.csum_flags & CSUM_UDP) {
3032 *txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
3033 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
3034 /* no need for context if already set */
3035 if (adapter->last_hw_offload == CSUM_UDP)
3036 return;
3037 adapter->last_hw_offload = CSUM_UDP;
3038 /*
3039 * Start offset for header checksum calculation.
3040 * End offset for header checksum calculation.
3041 * Offset of place to put the checksum.
3042 */
3043 TXD = (struct e1000_context_desc *)
3044 &adapter->tx_desc_base[curr_txd];
3045 TXD->upper_setup.tcp_fields.tucss = hdr_len;
3046 TXD->upper_setup.tcp_fields.tucse = htole16(0);
3047 TXD->upper_setup.tcp_fields.tucso =
3048 hdr_len + offsetof(struct udphdr, uh_sum);
3049 }
3050 /* Fall Thru */
3051 }
3052 default:
3053 break;
3054 }
3055
3056 if (TXD == NULL)
3057 return;
3058 TXD->tcp_seg_setup.data = htole32(0);
3059 TXD->cmd_and_length =
3060 htole32(adapter->txd_cmd | E1000_TXD_CMD_DEXT | cmd);
3061 tx_buffer = &adapter->tx_buffer_area[curr_txd];
3062 tx_buffer->m_head = NULL;
3063 tx_buffer->next_eop = -1;
3064
3065 if (++curr_txd == adapter->num_tx_desc)
3066 curr_txd = 0;
3067
3068 adapter->num_tx_desc_avail--;
3069 adapter->next_avail_tx_desc = curr_txd;
3070}
3071
3072
3073/**********************************************************************
3074 *
3075 * Examine each tx_buffer in the used queue. If the hardware is done
3076 * processing the packet then free associated resources. The
3077 * tx_buffer is put back on the free queue.
3078 *
3079 **********************************************************************/
3080static void
3081lem_txeof(struct adapter *adapter)
3082{
3083 int first, last, done, num_avail;
3084 struct em_buffer *tx_buffer;
3085 struct e1000_tx_desc *tx_desc, *eop_desc;
3086 if_t ifp = adapter->ifp;
3087
3088 EM_TX_LOCK_ASSERT(adapter);
3089
3090#ifdef DEV_NETMAP
3091 if (netmap_tx_irq(ifp, 0))
3092 return;
3093#endif /* DEV_NETMAP */
3094 if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
3095 return;
3096
3097 num_avail = adapter->num_tx_desc_avail;
3098 first = adapter->next_tx_to_clean;
3099 tx_desc = &adapter->tx_desc_base[first];
3100 tx_buffer = &adapter->tx_buffer_area[first];
3101 last = tx_buffer->next_eop;
3102 eop_desc = &adapter->tx_desc_base[last];
3103
3104 /*
3105 * What this does is get the index of the
3106 * first descriptor AFTER the EOP of the
3107 * first packet, that way we can do the
3108 * simple comparison on the inner while loop.
3109 */
3110 if (++last == adapter->num_tx_desc)
3111 last = 0;
3112 done = last;
3113
3114 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
3115 BUS_DMASYNC_POSTREAD);
3116
3117 while (eop_desc->upper.fields.status & E1000_TXD_STAT_DD) {
3118 /* We clean the range of the packet */
3119 while (first != done) {
3120 tx_desc->upper.data = 0;
3121 tx_desc->lower.data = 0;
3122 tx_desc->buffer_addr = 0;
3123 ++num_avail;
3124
3125 if (tx_buffer->m_head) {
3126 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
3127 bus_dmamap_sync(adapter->txtag,
3128 tx_buffer->map,
3129 BUS_DMASYNC_POSTWRITE);
3130 bus_dmamap_unload(adapter->txtag,
3131 tx_buffer->map);
3132
3133 m_freem(tx_buffer->m_head);
3134 tx_buffer->m_head = NULL;
3135 }
3136 tx_buffer->next_eop = -1;
3137 adapter->watchdog_time = ticks;
3138
3139 if (++first == adapter->num_tx_desc)
3140 first = 0;
3141
3142 tx_buffer = &adapter->tx_buffer_area[first];
3143 tx_desc = &adapter->tx_desc_base[first];
3144 }
3145 /* See if we can continue to the next packet */
3146 last = tx_buffer->next_eop;
3147 if (last != -1) {
3148 eop_desc = &adapter->tx_desc_base[last];
3149 /* Get new done point */
3150 if (++last == adapter->num_tx_desc) last = 0;
3151 done = last;
3152 } else
3153 break;
3154 }
3155 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
3156 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3157
3158 adapter->next_tx_to_clean = first;
3159 adapter->num_tx_desc_avail = num_avail;
3160
3161#ifdef NIC_SEND_COMBINING
3162 if ((adapter->shadow_tdt & MIT_PENDING_TDT) == MIT_PENDING_TDT) {
3163 /* a tdt write is pending, do it */
3164 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0),
3165 0xffff & adapter->shadow_tdt);
3166 adapter->shadow_tdt = MIT_PENDING_INT;
3167 } else {
3168 adapter->shadow_tdt = 0; // disable
3169 }
3170#endif /* NIC_SEND_COMBINING */
3171 /*
3172 * If we have enough room, clear IFF_DRV_OACTIVE to
3173 * tell the stack that it is OK to send packets.
3174 * If there are no pending descriptors, clear the watchdog.
3175 */
3176 if (adapter->num_tx_desc_avail > EM_TX_CLEANUP_THRESHOLD) {
3177 if_setdrvflagbits(ifp, 0, IFF_DRV_OACTIVE);
3178#ifdef NIC_PARAVIRT
3179 if (adapter->csb) { // XXX also csb_on ?
3180 adapter->csb->guest_need_txkick = 2; /* acked */
3181 // XXX memory barrier
3182 }
3183#endif /* NIC_PARAVIRT */
3184 if (adapter->num_tx_desc_avail == adapter->num_tx_desc) {
3185 adapter->watchdog_check = FALSE;
3186 return;
3187 }
3188 }
3189}
3190
3191/*********************************************************************
3192 *
3193 * When Link is lost sometimes there is work still in the TX ring
3194 * which may result in a watchdog, rather than allow that we do an
3195 * attempted cleanup and then reinit here. Note that this has been
3196 * seens mostly with fiber adapters.
3197 *
3198 **********************************************************************/
3199static void
3200lem_tx_purge(struct adapter *adapter)
3201{
3202 if ((!adapter->link_active) && (adapter->watchdog_check)) {
3203 EM_TX_LOCK(adapter);
3204 lem_txeof(adapter);
3205 EM_TX_UNLOCK(adapter);
3206 if (adapter->watchdog_check) /* Still outstanding? */
3207 lem_init_locked(adapter);
3208 }
3209}
3210
3211/*********************************************************************
3212 *
3213 * Get a buffer from system mbuf buffer pool.
3214 *
3215 **********************************************************************/
3216static int
3217lem_get_buf(struct adapter *adapter, int i)
3218{
3219 struct mbuf *m;
3220 bus_dma_segment_t segs[1];
3221 bus_dmamap_t map;
3222 struct em_buffer *rx_buffer;
3223 int error, nsegs;
3224
3225 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
3226 if (m == NULL) {
3227 adapter->mbuf_cluster_failed++;
3228 return (ENOBUFS);
3229 }
3230 m->m_len = m->m_pkthdr.len = MCLBYTES;
3231
3232 if (adapter->max_frame_size <= (MCLBYTES - ETHER_ALIGN))
3233 m_adj(m, ETHER_ALIGN);
3234
3235 /*
3236 * Using memory from the mbuf cluster pool, invoke the
3237 * bus_dma machinery to arrange the memory mapping.
3238 */
3239 error = bus_dmamap_load_mbuf_sg(adapter->rxtag,
3240 adapter->rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
3241 if (error != 0) {
3242 m_free(m);
3243 return (error);
3244 }
3245
3246 /* If nsegs is wrong then the stack is corrupt. */
3247 KASSERT(nsegs == 1, ("Too many segments returned!"));
3248
3249 rx_buffer = &adapter->rx_buffer_area[i];
3250 if (rx_buffer->m_head != NULL)
3251 bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
3252
3253 map = rx_buffer->map;
3254 rx_buffer->map = adapter->rx_sparemap;
3255 adapter->rx_sparemap = map;
3256 bus_dmamap_sync(adapter->rxtag, rx_buffer->map, BUS_DMASYNC_PREREAD);
3257 rx_buffer->m_head = m;
3258
3259 adapter->rx_desc_base[i].buffer_addr = htole64(segs[0].ds_addr);
3260 return (0);
3261}
3262
3263/*********************************************************************
3264 *
3265 * Allocate memory for rx_buffer structures. Since we use one
3266 * rx_buffer per received packet, the maximum number of rx_buffer's
3267 * that we'll need is equal to the number of receive descriptors
3268 * that we've allocated.
3269 *
3270 **********************************************************************/
3271static int
3272lem_allocate_receive_structures(struct adapter *adapter)
3273{
3274 device_t dev = adapter->dev;
3275 struct em_buffer *rx_buffer;
3276 int i, error;
3277
3278 adapter->rx_buffer_area = malloc(sizeof(struct em_buffer) *
3279 adapter->num_rx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
3280 if (adapter->rx_buffer_area == NULL) {
3281 device_printf(dev, "Unable to allocate rx_buffer memory\n");
3282 return (ENOMEM);
3283 }
3284
3285 error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
3286 1, 0, /* alignment, bounds */
3287 BUS_SPACE_MAXADDR, /* lowaddr */
3288 BUS_SPACE_MAXADDR, /* highaddr */
3289 NULL, NULL, /* filter, filterarg */
3290 MCLBYTES, /* maxsize */
3291 1, /* nsegments */
3292 MCLBYTES, /* maxsegsize */
3293 0, /* flags */
3294 NULL, /* lockfunc */
3295 NULL, /* lockarg */
3296 &adapter->rxtag);
3297 if (error) {
3298 device_printf(dev, "%s: bus_dma_tag_create failed %d\n",
3299 __func__, error);
3300 goto fail;
3301 }
3302
3303 /* Create the spare map (used by getbuf) */
3304 error = bus_dmamap_create(adapter->rxtag, 0, &adapter->rx_sparemap);
3305 if (error) {
3306 device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
3307 __func__, error);
3308 goto fail;
3309 }
3310
3311 rx_buffer = adapter->rx_buffer_area;
3312 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3313 error = bus_dmamap_create(adapter->rxtag, 0, &rx_buffer->map);
3314 if (error) {
3315 device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
3316 __func__, error);
3317 goto fail;
3318 }
3319 }
3320
3321 return (0);
3322
3323fail:
3324 lem_free_receive_structures(adapter);
3325 return (error);
3326}
3327
3328/*********************************************************************
3329 *
3330 * (Re)initialize receive structures.
3331 *
3332 **********************************************************************/
3333static int
3334lem_setup_receive_structures(struct adapter *adapter)
3335{
3336 struct em_buffer *rx_buffer;
3337 int i, error;
3338#ifdef DEV_NETMAP
3339 /* we are already under lock */
3340 struct netmap_adapter *na = netmap_getna(adapter->ifp);
3341 struct netmap_slot *slot = netmap_reset(na, NR_RX, 0, 0);
3342#endif
3343
3344 /* Reset descriptor ring */
3345 bzero(adapter->rx_desc_base,
3346 (sizeof(struct e1000_rx_desc)) * adapter->num_rx_desc);
3347
3348 /* Free current RX buffers. */
3349 rx_buffer = adapter->rx_buffer_area;
3350 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3351 if (rx_buffer->m_head != NULL) {
3352 bus_dmamap_sync(adapter->rxtag, rx_buffer->map,
3353 BUS_DMASYNC_POSTREAD);
3354 bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
3355 m_freem(rx_buffer->m_head);
3356 rx_buffer->m_head = NULL;
3357 }
3358 }
3359
3360 /* Allocate new ones. */
3361 for (i = 0; i < adapter->num_rx_desc; i++) {
3362#ifdef DEV_NETMAP
3363 if (slot) {
3364 /* the i-th NIC entry goes to slot si */
3365 int si = netmap_idx_n2k(&na->rx_rings[0], i);
3366 uint64_t paddr;
3367 void *addr;
3368
3369 addr = PNMB(na, slot + si, &paddr);
3370 netmap_load_map(na, adapter->rxtag, rx_buffer->map, addr);
3371 /* Update descriptor */
3372 adapter->rx_desc_base[i].buffer_addr = htole64(paddr);
3373 continue;
3374 }
3375#endif /* DEV_NETMAP */
3376 error = lem_get_buf(adapter, i);
3377 if (error)
3378 return (error);
3379 }
3380
3381 /* Setup our descriptor pointers */
3382 adapter->next_rx_desc_to_check = 0;
3383 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3384 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3385
3386 return (0);
3387}
3388
3389/*********************************************************************
3390 *
3391 * Enable receive unit.
3392 *
3393 **********************************************************************/
3394
3395static void
3396lem_initialize_receive_unit(struct adapter *adapter)
3397{
3398 if_t ifp = adapter->ifp;
3399 u64 bus_addr;
3400 u32 rctl, rxcsum;
3401
3402 INIT_DEBUGOUT("lem_initialize_receive_unit: begin");
3403
3404 /*
3405 * Make sure receives are disabled while setting
3406 * up the descriptor ring
3407 */
3408 rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3409 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3410
3411 if (adapter->hw.mac.type >= e1000_82540) {
3412 E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3413 adapter->rx_abs_int_delay.value);
3414 /*
3415 * Set the interrupt throttling rate. Value is calculated
3416 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3417 */
3418 E1000_WRITE_REG(&adapter->hw, E1000_ITR, DEFAULT_ITR);
3419 }
3420
3421 /* Setup the Base and Length of the Rx Descriptor Ring */
3422 bus_addr = adapter->rxdma.dma_paddr;
3423 E1000_WRITE_REG(&adapter->hw, E1000_RDLEN(0),
3424 adapter->num_rx_desc * sizeof(struct e1000_rx_desc));
3425 E1000_WRITE_REG(&adapter->hw, E1000_RDBAH(0),
3426 (u32)(bus_addr >> 32));
3427 E1000_WRITE_REG(&adapter->hw, E1000_RDBAL(0),
3428 (u32)bus_addr);
3429
3430 /* Setup the Receive Control Register */
3431 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3432 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
3433 E1000_RCTL_RDMTS_HALF |
3434 (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3435
3436 /* Make sure VLAN Filters are off */
3437 rctl &= ~E1000_RCTL_VFE;
3438
3439 if (e1000_tbi_sbp_enabled_82543(&adapter->hw))
3440 rctl |= E1000_RCTL_SBP;
3441 else
3442 rctl &= ~E1000_RCTL_SBP;
3443
3444 switch (adapter->rx_buffer_len) {
3445 default:
3446 case 2048:
3447 rctl |= E1000_RCTL_SZ_2048;
3448 break;
3449 case 4096:
3450 rctl |= E1000_RCTL_SZ_4096 |
3451 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3452 break;
3453 case 8192:
3454 rctl |= E1000_RCTL_SZ_8192 |
3455 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3456 break;
3457 case 16384:
3458 rctl |= E1000_RCTL_SZ_16384 |
3459 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3460 break;
3461 }
3462
3463 if (if_getmtu(ifp) > ETHERMTU)
3464 rctl |= E1000_RCTL_LPE;
3465 else
3466 rctl &= ~E1000_RCTL_LPE;
3467
3468 /* Enable 82543 Receive Checksum Offload for TCP and UDP */
3469 if ((adapter->hw.mac.type >= e1000_82543) &&
3470 (if_getcapenable(ifp) & IFCAP_RXCSUM)) {
3471 rxcsum = E1000_READ_REG(&adapter->hw, E1000_RXCSUM);
3472 rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3473 E1000_WRITE_REG(&adapter->hw, E1000_RXCSUM, rxcsum);
3474 }
3475
3476 /* Enable Receives */
3477 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3478
3479 /*
3480 * Setup the HW Rx Head and
3481 * Tail Descriptor Pointers
3482 */
3483 E1000_WRITE_REG(&adapter->hw, E1000_RDH(0), 0);
3484 rctl = adapter->num_rx_desc - 1; /* default RDT value */
3485#ifdef DEV_NETMAP
3486 /* preserve buffers already made available to clients */
3487 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
3488 struct netmap_adapter *na = netmap_getna(adapter->ifp);
3489 rctl -= nm_kr_rxspace(&na->rx_rings[0]);
3490 }
3491#endif /* DEV_NETMAP */
3492 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), rctl);
3493
3494 return;
3495}
3496
3497/*********************************************************************
3498 *
3499 * Free receive related data structures.
3500 *
3501 **********************************************************************/
3502static void
3503lem_free_receive_structures(struct adapter *adapter)
3504{
3505 struct em_buffer *rx_buffer;
3506 int i;
3507
3508 INIT_DEBUGOUT("free_receive_structures: begin");
3509
3510 if (adapter->rx_sparemap) {
3511 bus_dmamap_destroy(adapter->rxtag, adapter->rx_sparemap);
3512 adapter->rx_sparemap = NULL;
3513 }
3514
3515 /* Cleanup any existing buffers */
3516 if (adapter->rx_buffer_area != NULL) {
3517 rx_buffer = adapter->rx_buffer_area;
3518 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3519 if (rx_buffer->m_head != NULL) {
3520 bus_dmamap_sync(adapter->rxtag, rx_buffer->map,
3521 BUS_DMASYNC_POSTREAD);
3522 bus_dmamap_unload(adapter->rxtag,
3523 rx_buffer->map);
3524 m_freem(rx_buffer->m_head);
3525 rx_buffer->m_head = NULL;
3526 } else if (rx_buffer->map != NULL)
3527 bus_dmamap_unload(adapter->rxtag,
3528 rx_buffer->map);
3529 if (rx_buffer->map != NULL) {
3530 bus_dmamap_destroy(adapter->rxtag,
3531 rx_buffer->map);
3532 rx_buffer->map = NULL;
3533 }
3534 }
3535 }
3536
3537 if (adapter->rx_buffer_area != NULL) {
3538 free(adapter->rx_buffer_area, M_DEVBUF);
3539 adapter->rx_buffer_area = NULL;
3540 }
3541
3542 if (adapter->rxtag != NULL) {
3543 bus_dma_tag_destroy(adapter->rxtag);
3544 adapter->rxtag = NULL;
3545 }
3546}
3547
3548/*********************************************************************
3549 *
3550 * This routine executes in interrupt context. It replenishes
3551 * the mbufs in the descriptor and sends data which has been
3552 * dma'ed into host memory to upper layer.
3553 *
3554 * We loop at most count times if count is > 0, or until done if
3555 * count < 0.
3556 *
3557 * For polling we also now return the number of cleaned packets
3558 *********************************************************************/
3559static bool
3560lem_rxeof(struct adapter *adapter, int count, int *done)
3561{
3562 if_t ifp = adapter->ifp;
3563 struct mbuf *mp;
3564 u8 status = 0, accept_frame = 0, eop = 0;
3565 u16 len, desc_len, prev_len_adj;
3566 int i, rx_sent = 0;
3567 struct e1000_rx_desc *current_desc;
3568
3569#ifdef BATCH_DISPATCH
3570 struct mbuf *mh = NULL, *mt = NULL;
3571#endif /* BATCH_DISPATCH */
3572#ifdef NIC_PARAVIRT
3573 int retries = 0;
3574 struct paravirt_csb* csb = adapter->csb;
3575 int csb_mode = csb && csb->guest_csb_on;
3576
3577 //ND("clear guest_rxkick at %d", adapter->next_rx_desc_to_check);
3578 if (csb_mode && csb->guest_need_rxkick)
3579 csb->guest_need_rxkick = 0;
3580#endif /* NIC_PARAVIRT */
3581 EM_RX_LOCK(adapter);
3582
3583#ifdef BATCH_DISPATCH
3584 batch_again:
3585#endif /* BATCH_DISPATCH */
3586 i = adapter->next_rx_desc_to_check;
3587 current_desc = &adapter->rx_desc_base[i];
3588 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3589 BUS_DMASYNC_POSTREAD);
3590
3591#ifdef DEV_NETMAP
3592 if (netmap_rx_irq(ifp, 0, &rx_sent)) {
3593 EM_RX_UNLOCK(adapter);
3594 return (FALSE);
3595 }
3596#endif /* DEV_NETMAP */
3597
3598#if 1 // XXX optimization ?
3599 if (!((current_desc->status) & E1000_RXD_STAT_DD)) {
3600 if (done != NULL)
3601 *done = rx_sent;
3602 EM_RX_UNLOCK(adapter);
3603 return (FALSE);
3604 }
3605#endif /* 0 */
3606
3607 while (count != 0 && if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
3608 struct mbuf *m = NULL;
3609
3610 status = current_desc->status;
3611 if ((status & E1000_RXD_STAT_DD) == 0) {
3612#ifdef NIC_PARAVIRT
3613 if (csb_mode) {
3614 /* buffer not ready yet. Retry a few times before giving up */
3615 if (++retries <= adapter->rx_retries) {
3616 continue;
3617 }
3618 if (csb->guest_need_rxkick == 0) {
3619 // ND("set guest_rxkick at %d", adapter->next_rx_desc_to_check);
3620 csb->guest_need_rxkick = 1;
3621 // XXX memory barrier, status volatile ?
3622 continue; /* double check */
3623 }
3624 }
3625 /* no buffer ready, give up */
3626#endif /* NIC_PARAVIRT */
3627 break;
3628 }
3629#ifdef NIC_PARAVIRT
3630 if (csb_mode) {
3631 if (csb->guest_need_rxkick)
3632 // ND("clear again guest_rxkick at %d", adapter->next_rx_desc_to_check);
3633 csb->guest_need_rxkick = 0;
3634 retries = 0;
3635 }
3636#endif /* NIC_PARAVIRT */
3637
3638 mp = adapter->rx_buffer_area[i].m_head;
3639 /*
3640 * Can't defer bus_dmamap_sync(9) because TBI_ACCEPT
3641 * needs to access the last received byte in the mbuf.
3642 */
3643 bus_dmamap_sync(adapter->rxtag, adapter->rx_buffer_area[i].map,
3644 BUS_DMASYNC_POSTREAD);
3645
3646 accept_frame = 1;
3647 prev_len_adj = 0;
3648 desc_len = le16toh(current_desc->length);
3649 if (status & E1000_RXD_STAT_EOP) {
3650 count--;
3651 eop = 1;
3652 if (desc_len < ETHER_CRC_LEN) {
3653 len = 0;
3654 prev_len_adj = ETHER_CRC_LEN - desc_len;
3655 } else
3656 len = desc_len - ETHER_CRC_LEN;
3657 } else {
3658 eop = 0;
3659 len = desc_len;
3660 }
3661
3662 if (current_desc->errors & E1000_RXD_ERR_FRAME_ERR_MASK) {
3663 u8 last_byte;
3664 u32 pkt_len = desc_len;
3665
3666 if (adapter->fmp != NULL)
3667 pkt_len += adapter->fmp->m_pkthdr.len;
3668
3669 last_byte = *(mtod(mp, caddr_t) + desc_len - 1);
3670 if (TBI_ACCEPT(&adapter->hw, status,
3671 current_desc->errors, pkt_len, last_byte,
3672 adapter->min_frame_size, adapter->max_frame_size)) {
3673 e1000_tbi_adjust_stats_82543(&adapter->hw,
3674 &adapter->stats, pkt_len,
3675 adapter->hw.mac.addr,
3676 adapter->max_frame_size);
3677 if (len > 0)
3678 len--;
3679 } else
3680 accept_frame = 0;
3681 }
3682
3683 if (accept_frame) {
3684 if (lem_get_buf(adapter, i) != 0) {
3685 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
3686 goto discard;
3687 }
3688
3689 /* Assign correct length to the current fragment */
3690 mp->m_len = len;
3691
3692 if (adapter->fmp == NULL) {
3693 mp->m_pkthdr.len = len;
3694 adapter->fmp = mp; /* Store the first mbuf */
3695 adapter->lmp = mp;
3696 } else {
3697 /* Chain mbuf's together */
3698 mp->m_flags &= ~M_PKTHDR;
3699 /*
3700 * Adjust length of previous mbuf in chain if
3701 * we received less than 4 bytes in the last
3702 * descriptor.
3703 */
3704 if (prev_len_adj > 0) {
3705 adapter->lmp->m_len -= prev_len_adj;
3706 adapter->fmp->m_pkthdr.len -=
3707 prev_len_adj;
3708 }
3709 adapter->lmp->m_next = mp;
3710 adapter->lmp = adapter->lmp->m_next;
3711 adapter->fmp->m_pkthdr.len += len;
3712 }
3713
3714 if (eop) {
3715 if_setrcvif(adapter->fmp, ifp);
3716 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
3717 lem_receive_checksum(adapter, current_desc,
3718 adapter->fmp);
3719#ifndef __NO_STRICT_ALIGNMENT
3720 if (adapter->max_frame_size >
3721 (MCLBYTES - ETHER_ALIGN) &&
3722 lem_fixup_rx(adapter) != 0)
3723 goto skip;
3724#endif
3725 if (status & E1000_RXD_STAT_VP) {
3726 adapter->fmp->m_pkthdr.ether_vtag =
3727 le16toh(current_desc->special);
3728 adapter->fmp->m_flags |= M_VLANTAG;
3729 }
3730#ifndef __NO_STRICT_ALIGNMENT
3731skip:
3732#endif
3733 m = adapter->fmp;
3734 adapter->fmp = NULL;
3735 adapter->lmp = NULL;
3736 }
3737 } else {
3738 adapter->dropped_pkts++;
3739discard:
3740 /* Reuse loaded DMA map and just update mbuf chain */
3741 mp = adapter->rx_buffer_area[i].m_head;
3742 mp->m_len = mp->m_pkthdr.len = MCLBYTES;
3743 mp->m_data = mp->m_ext.ext_buf;
3744 mp->m_next = NULL;
3745 if (adapter->max_frame_size <=
3746 (MCLBYTES - ETHER_ALIGN))
3747 m_adj(mp, ETHER_ALIGN);
3748 if (adapter->fmp != NULL) {
3749 m_freem(adapter->fmp);
3750 adapter->fmp = NULL;
3751 adapter->lmp = NULL;
3752 }
3753 m = NULL;
3754 }
3755
3756 /* Zero out the receive descriptors status. */
3757 current_desc->status = 0;
3758 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3759 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3760
3761#ifdef NIC_PARAVIRT
3762 if (csb_mode) {
3763 /* the buffer at i has been already replaced by lem_get_buf()
3764 * so it is safe to set guest_rdt = i and possibly send a kick.
3765 * XXX see if we can optimize it later.
3766 */
3767 csb->guest_rdt = i;
3768 // XXX memory barrier
3769 if (i == csb->host_rxkick_at)
3770 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), i);
3771 }
3772#endif /* NIC_PARAVIRT */
3773 /* Advance our pointers to the next descriptor. */
3774 if (++i == adapter->num_rx_desc)
3775 i = 0;
3776 /* Call into the stack */
3777 if (m != NULL) {
3778#ifdef BATCH_DISPATCH
3779 if (adapter->batch_enable) {
3780 if (mh == NULL)
3781 mh = mt = m;
3782 else
3783 mt->m_nextpkt = m;
3784 mt = m;
3785 m->m_nextpkt = NULL;
3786 rx_sent++;
3787 current_desc = &adapter->rx_desc_base[i];
3788 continue;
3789 }
3790#endif /* BATCH_DISPATCH */
3791 adapter->next_rx_desc_to_check = i;
3792 EM_RX_UNLOCK(adapter);
3793 if_input(ifp, m);
3794 EM_RX_LOCK(adapter);
3795 rx_sent++;
3796 i = adapter->next_rx_desc_to_check;
3797 }
3798 current_desc = &adapter->rx_desc_base[i];
3799 }
3800 adapter->next_rx_desc_to_check = i;
3801#ifdef BATCH_DISPATCH
3802 if (mh) {
3803 EM_RX_UNLOCK(adapter);
3804 while ( (mt = mh) != NULL) {
3805 mh = mh->m_nextpkt;
3806 mt->m_nextpkt = NULL;
3807 if_input(ifp, mt);
3808 }
3809 EM_RX_LOCK(adapter);
3810 i = adapter->next_rx_desc_to_check; /* in case of interrupts */
3811 if (count > 0)
3812 goto batch_again;
3813 }
3814#endif /* BATCH_DISPATCH */
3815
3816 /* Advance the E1000's Receive Queue #0 "Tail Pointer". */
3817 if (--i < 0)
3818 i = adapter->num_rx_desc - 1;
3819#ifdef NIC_PARAVIRT
3820 if (!csb_mode) /* filter out writes */
3821#endif /* NIC_PARAVIRT */
3822 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), i);
3823 if (done != NULL)
3824 *done = rx_sent;
3825 EM_RX_UNLOCK(adapter);
3826 return ((status & E1000_RXD_STAT_DD) ? TRUE : FALSE);
3827}
3828
3829#ifndef __NO_STRICT_ALIGNMENT
3830/*
3831 * When jumbo frames are enabled we should realign entire payload on
3832 * architecures with strict alignment. This is serious design mistake of 8254x
3833 * as it nullifies DMA operations. 8254x just allows RX buffer size to be
3834 * 2048/4096/8192/16384. What we really want is 2048 - ETHER_ALIGN to align its
3835 * payload. On architecures without strict alignment restrictions 8254x still
3836 * performs unaligned memory access which would reduce the performance too.
3837 * To avoid copying over an entire frame to align, we allocate a new mbuf and
3838 * copy ethernet header to the new mbuf. The new mbuf is prepended into the
3839 * existing mbuf chain.
3840 *
3841 * Be aware, best performance of the 8254x is achived only when jumbo frame is
3842 * not used at all on architectures with strict alignment.
3843 */
3844static int
3845lem_fixup_rx(struct adapter *adapter)
3846{
3847 struct mbuf *m, *n;
3848 int error;
3849
3850 error = 0;
3851 m = adapter->fmp;
3852 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) {
3853 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len);
3854 m->m_data += ETHER_HDR_LEN;
3855 } else {
3856 MGETHDR(n, M_NOWAIT, MT_DATA);
3857 if (n != NULL) {
3858 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN);
3859 m->m_data += ETHER_HDR_LEN;
3860 m->m_len -= ETHER_HDR_LEN;
3861 n->m_len = ETHER_HDR_LEN;
3862 M_MOVE_PKTHDR(n, m);
3863 n->m_next = m;
3864 adapter->fmp = n;
3865 } else {
3866 adapter->dropped_pkts++;
3867 m_freem(adapter->fmp);
3868 adapter->fmp = NULL;
3869 error = ENOMEM;
3870 }
3871 }
3872
3873 return (error);
3874}
3875#endif
3876
3877/*********************************************************************
3878 *
3879 * Verify that the hardware indicated that the checksum is valid.
3880 * Inform the stack about the status of checksum so that stack
3881 * doesn't spend time verifying the checksum.
3882 *
3883 *********************************************************************/
3884static void
3885lem_receive_checksum(struct adapter *adapter,
3886 struct e1000_rx_desc *rx_desc, struct mbuf *mp)
3887{
3888 /* 82543 or newer only */
3889 if ((adapter->hw.mac.type < e1000_82543) ||
3890 /* Ignore Checksum bit is set */
3891 (rx_desc->status & E1000_RXD_STAT_IXSM)) {
3892 mp->m_pkthdr.csum_flags = 0;
3893 return;
3894 }
3895
3896 if (rx_desc->status & E1000_RXD_STAT_IPCS) {
3897 /* Did it pass? */
3898 if (!(rx_desc->errors & E1000_RXD_ERR_IPE)) {
3899 /* IP Checksum Good */
3900 mp->m_pkthdr.csum_flags = CSUM_IP_CHECKED;
3901 mp->m_pkthdr.csum_flags |= CSUM_IP_VALID;
3902
3903 } else {
3904 mp->m_pkthdr.csum_flags = 0;
3905 }
3906 }
3907
3908 if (rx_desc->status & E1000_RXD_STAT_TCPCS) {
3909 /* Did it pass? */
3910 if (!(rx_desc->errors & E1000_RXD_ERR_TCPE)) {
3911 mp->m_pkthdr.csum_flags |=
3912 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
3913 mp->m_pkthdr.csum_data = htons(0xffff);
3914 }
3915 }
3916}
3917
3918/*
3919 * This routine is run via an vlan
3920 * config EVENT
3921 */
3922static void
3923lem_register_vlan(void *arg, if_t ifp, u16 vtag)
3924{
3925 struct adapter *adapter = if_getsoftc(ifp);
3926 u32 index, bit;
3927
3928 if (if_getsoftc(ifp) != arg) /* Not our event */
3929 return;
3930
3931 if ((vtag == 0) || (vtag > 4095)) /* Invalid ID */
3932 return;
3933
3934 EM_CORE_LOCK(adapter);
3935 index = (vtag >> 5) & 0x7F;
3936 bit = vtag & 0x1F;
3937 adapter->shadow_vfta[index] |= (1 << bit);
3938 ++adapter->num_vlans;
3939 /* Re-init to load the changes */
3940 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
3941 lem_init_locked(adapter);
3942 EM_CORE_UNLOCK(adapter);
3943}
3944
3945/*
3946 * This routine is run via an vlan
3947 * unconfig EVENT
3948 */
3949static void
3950lem_unregister_vlan(void *arg, if_t ifp, u16 vtag)
3951{
3952 struct adapter *adapter = if_getsoftc(ifp);
3953 u32 index, bit;
3954
3955 if (if_getsoftc(ifp) != arg)
3956 return;
3957
3958 if ((vtag == 0) || (vtag > 4095)) /* Invalid */
3959 return;
3960
3961 EM_CORE_LOCK(adapter);
3962 index = (vtag >> 5) & 0x7F;
3963 bit = vtag & 0x1F;
3964 adapter->shadow_vfta[index] &= ~(1 << bit);
3965 --adapter->num_vlans;
3966 /* Re-init to load the changes */
3967 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
3968 lem_init_locked(adapter);
3969 EM_CORE_UNLOCK(adapter);
3970}
3971
3972static void
3973lem_setup_vlan_hw_support(struct adapter *adapter)
3974{
3975 struct e1000_hw *hw = &adapter->hw;
3976 u32 reg;
3977
3978 /*
3979 ** We get here thru init_locked, meaning
3980 ** a soft reset, this has already cleared
3981 ** the VFTA and other state, so if there
3982 ** have been no vlan's registered do nothing.
3983 */
3984 if (adapter->num_vlans == 0)
3985 return;
3986
3987 /*
3988 ** A soft reset zero's out the VFTA, so
3989 ** we need to repopulate it now.
3990 */
3991 for (int i = 0; i < EM_VFTA_SIZE; i++)
3992 if (adapter->shadow_vfta[i] != 0)
3993 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3994 i, adapter->shadow_vfta[i]);
3995
3996 reg = E1000_READ_REG(hw, E1000_CTRL);
3997 reg |= E1000_CTRL_VME;
3998 E1000_WRITE_REG(hw, E1000_CTRL, reg);
3999
4000 /* Enable the Filter Table */
4001 reg = E1000_READ_REG(hw, E1000_RCTL);
4002 reg &= ~E1000_RCTL_CFIEN;
4003 reg |= E1000_RCTL_VFE;
4004 E1000_WRITE_REG(hw, E1000_RCTL, reg);
4005}
4006
4007static void
4008lem_enable_intr(struct adapter *adapter)
4009{
4010 struct e1000_hw *hw = &adapter->hw;
4011 u32 ims_mask = IMS_ENABLE_MASK;
4012
4013 E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
4014}
4015
4016static void
4017lem_disable_intr(struct adapter *adapter)
4018{
4019 struct e1000_hw *hw = &adapter->hw;
4020
4021 E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
4022}
4023
4024/*
4025 * Bit of a misnomer, what this really means is
4026 * to enable OS management of the system... aka
4027 * to disable special hardware management features
4028 */
4029static void
4030lem_init_manageability(struct adapter *adapter)
4031{
4032 /* A shared code workaround */
4033 if (adapter->has_manage) {
4034 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
4035 /* disable hardware interception of ARP */
4036 manc &= ~(E1000_MANC_ARP_EN);
4037 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
4038 }
4039}
4040
4041/*
4042 * Give control back to hardware management
4043 * controller if there is one.
4044 */
4045static void
4046lem_release_manageability(struct adapter *adapter)
4047{
4048 if (adapter->has_manage) {
4049 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
4050
4051 /* re-enable hardware interception of ARP */
4052 manc |= E1000_MANC_ARP_EN;
4053 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
4054 }
4055}
4056
4057/*
4058 * lem_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
4059 * For ASF and Pass Through versions of f/w this means
4060 * that the driver is loaded. For AMT version type f/w
4061 * this means that the network i/f is open.
4062 */
4063static void
4064lem_get_hw_control(struct adapter *adapter)
4065{
4066 u32 ctrl_ext;
4067
4068 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4069 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
4070 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
4071 return;
4072}
4073
4074/*
4075 * lem_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
4076 * For ASF and Pass Through versions of f/w this means that
4077 * the driver is no longer loaded. For AMT versions of the
4078 * f/w this means that the network i/f is closed.
4079 */
4080static void
4081lem_release_hw_control(struct adapter *adapter)
4082{
4083 u32 ctrl_ext;
4084
4085 if (!adapter->has_manage)
4086 return;
4087
4088 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4089 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
4090 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
4091 return;
4092}
4093
4094static int
4095lem_is_valid_ether_addr(u8 *addr)
4096{
4097 char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
4098
4099 if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
4100 return (FALSE);
4101 }
4102
4103 return (TRUE);
4104}
4105
4106/*
4107** Parse the interface capabilities with regard
4108** to both system management and wake-on-lan for
4109** later use.
4110*/
4111static void
4112lem_get_wakeup(device_t dev)
4113{
4114 struct adapter *adapter = device_get_softc(dev);
4115 u16 eeprom_data = 0, device_id, apme_mask;
4116
4117 adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
4118 apme_mask = EM_EEPROM_APME;
4119
4120 switch (adapter->hw.mac.type) {
4121 case e1000_82542:
4122 case e1000_82543:
4123 break;
4124 case e1000_82544:
4125 e1000_read_nvm(&adapter->hw,
4126 NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
4127 apme_mask = EM_82544_APME;
4128 break;
4129 case e1000_82546:
4130 case e1000_82546_rev_3:
4131 if (adapter->hw.bus.func == 1) {
4132 e1000_read_nvm(&adapter->hw,
4133 NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
4134 break;
4135 } else
4136 e1000_read_nvm(&adapter->hw,
4137 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4138 break;
4139 default:
4140 e1000_read_nvm(&adapter->hw,
4141 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4142 break;
4143 }
4144 if (eeprom_data & apme_mask)
4145 adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
4146 /*
4147 * We have the eeprom settings, now apply the special cases
4148 * where the eeprom may be wrong or the board won't support
4149 * wake on lan on a particular port
4150 */
4151 device_id = pci_get_device(dev);
4152 switch (device_id) {
4153 case E1000_DEV_ID_82546GB_PCIE:
4154 adapter->wol = 0;
4155 break;
4156 case E1000_DEV_ID_82546EB_FIBER:
4157 case E1000_DEV_ID_82546GB_FIBER:
4158 /* Wake events only supported on port A for dual fiber
4159 * regardless of eeprom setting */
4160 if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
4161 E1000_STATUS_FUNC_1)
4162 adapter->wol = 0;
4163 break;
4164 case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
4165 /* if quad port adapter, disable WoL on all but port A */
4166 if (global_quad_port_a != 0)
4167 adapter->wol = 0;
4168 /* Reset for multiple quad port adapters */
4169 if (++global_quad_port_a == 4)
4170 global_quad_port_a = 0;
4171 break;
4172 }
4173 return;
4174}
4175
4176
4177/*
4178 * Enable PCI Wake On Lan capability
4179 */
4180static void
4181lem_enable_wakeup(device_t dev)
4182{
4183 struct adapter *adapter = device_get_softc(dev);
4184 if_t ifp = adapter->ifp;
4185 u32 pmc, ctrl, ctrl_ext, rctl;
4186 u16 status;
4187
4188 if ((pci_find_cap(dev, PCIY_PMG, &pmc) != 0))
4189 return;
4190
4191 /* Advertise the wakeup capability */
4192 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
4193 ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
4194 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
4195 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
4196
4197 /* Keep the laser running on Fiber adapters */
4198 if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
4199 adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
4200 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4201 ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
4202 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
4203 }
4204
4205 /*
4206 ** Determine type of Wakeup: note that wol
4207 ** is set with all bits on by default.
4208 */
4209 if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
4210 adapter->wol &= ~E1000_WUFC_MAG;
4211
4212 if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
4213 adapter->wol &= ~E1000_WUFC_MC;
4214 else {
4215 rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
4216 rctl |= E1000_RCTL_MPE;
4217 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
4218 }
4219
4220 if (adapter->hw.mac.type == e1000_pchlan) {
4221 if (lem_enable_phy_wakeup(adapter))
4222 return;
4223 } else {
4224 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
4225 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
4226 }
4227
4228
4229 /* Request PME */
4230 status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
4231 status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
4232 if (if_getcapenable(ifp) & IFCAP_WOL)
4233 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
4234 pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
4235
4236 return;
4237}
4238
4239/*
4240** WOL in the newer chipset interfaces (pchlan)
4241** require thing to be copied into the phy
4242*/
4243static int
4244lem_enable_phy_wakeup(struct adapter *adapter)
4245{
4246 struct e1000_hw *hw = &adapter->hw;
4247 u32 mreg, ret = 0;
4248 u16 preg;
4249
4250 /* copy MAC RARs to PHY RARs */
4251 for (int i = 0; i < adapter->hw.mac.rar_entry_count; i++) {
4252 mreg = E1000_READ_REG(hw, E1000_RAL(i));
4253 e1000_write_phy_reg(hw, BM_RAR_L(i), (u16)(mreg & 0xFFFF));
4254 e1000_write_phy_reg(hw, BM_RAR_M(i),
4255 (u16)((mreg >> 16) & 0xFFFF));
4256 mreg = E1000_READ_REG(hw, E1000_RAH(i));
4257 e1000_write_phy_reg(hw, BM_RAR_H(i), (u16)(mreg & 0xFFFF));
4258 e1000_write_phy_reg(hw, BM_RAR_CTRL(i),
4259 (u16)((mreg >> 16) & 0xFFFF));
4260 }
4261
4262 /* copy MAC MTA to PHY MTA */
4263 for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
4264 mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
4265 e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
4266 e1000_write_phy_reg(hw, BM_MTA(i) + 1,
4267 (u16)((mreg >> 16) & 0xFFFF));
4268 }
4269
4270 /* configure PHY Rx Control register */
4271 e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
4272 mreg = E1000_READ_REG(hw, E1000_RCTL);
4273 if (mreg & E1000_RCTL_UPE)
4274 preg |= BM_RCTL_UPE;
4275 if (mreg & E1000_RCTL_MPE)
4276 preg |= BM_RCTL_MPE;
4277 preg &= ~(BM_RCTL_MO_MASK);
4278 if (mreg & E1000_RCTL_MO_3)
4279 preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
4280 << BM_RCTL_MO_SHIFT);
4281 if (mreg & E1000_RCTL_BAM)
4282 preg |= BM_RCTL_BAM;
4283 if (mreg & E1000_RCTL_PMCF)
4284 preg |= BM_RCTL_PMCF;
4285 mreg = E1000_READ_REG(hw, E1000_CTRL);
4286 if (mreg & E1000_CTRL_RFCE)
4287 preg |= BM_RCTL_RFCE;
4288 e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
4289
4290 /* enable PHY wakeup in MAC register */
4291 E1000_WRITE_REG(hw, E1000_WUC,
4292 E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
4293 E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
4294
4295 /* configure and enable PHY wakeup in PHY registers */
4296 e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
4297 e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
4298
4299 /* activate PHY wakeup */
4300 ret = hw->phy.ops.acquire(hw);
4301 if (ret) {
4302 printf("Could not acquire PHY\n");
4303 return ret;
4304 }
4305 e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4306 (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
4307 ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
4308 if (ret) {
4309 printf("Could not read PHY page 769\n");
4310 goto out;
4311 }
4312 preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4313 ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4314 if (ret)
4315 printf("Could not set PHY Host Wakeup bit\n");
4316out:
4317 hw->phy.ops.release(hw);
4318
4319 return ret;
4320}
4321
4322static void
4323lem_led_func(void *arg, int onoff)
4324{
4325 struct adapter *adapter = arg;
4326
4327 EM_CORE_LOCK(adapter);
4328 if (onoff) {
4329 e1000_setup_led(&adapter->hw);
4330 e1000_led_on(&adapter->hw);
4331 } else {
4332 e1000_led_off(&adapter->hw);
4333 e1000_cleanup_led(&adapter->hw);
4334 }
4335 EM_CORE_UNLOCK(adapter);
4336}
4337
4338/*********************************************************************
4339* 82544 Coexistence issue workaround.
4340* There are 2 issues.
4341* 1. Transmit Hang issue.
4342* To detect this issue, following equation can be used...
4343* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
4344* If SUM[3:0] is in between 1 to 4, we will have this issue.
4345*
4346* 2. DAC issue.
4347* To detect this issue, following equation can be used...
4348* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
4349* If SUM[3:0] is in between 9 to c, we will have this issue.
4350*
4351*
4352* WORKAROUND:
4353* Make sure we do not have ending address
4354* as 1,2,3,4(Hang) or 9,a,b,c (DAC)
4355*
4356*************************************************************************/
4357static u32
4358lem_fill_descriptors (bus_addr_t address, u32 length,
4359 PDESC_ARRAY desc_array)
4360{
4361 u32 safe_terminator;
4362
4363 /* Since issue is sensitive to length and address.*/
4364 /* Let us first check the address...*/
4365 if (length <= 4) {
4366 desc_array->descriptor[0].address = address;
4367 desc_array->descriptor[0].length = length;
4368 desc_array->elements = 1;
4369 return (desc_array->elements);
4370 }
4371 safe_terminator = (u32)((((u32)address & 0x7) +
4372 (length & 0xF)) & 0xF);
4373 /* if it does not fall between 0x1 to 0x4 and 0x9 to 0xC then return */
4374 if (safe_terminator == 0 ||
4375 (safe_terminator > 4 &&
4376 safe_terminator < 9) ||
4377 (safe_terminator > 0xC &&
4378 safe_terminator <= 0xF)) {
4379 desc_array->descriptor[0].address = address;
4380 desc_array->descriptor[0].length = length;
4381 desc_array->elements = 1;
4382 return (desc_array->elements);
4383 }
4384
4385 desc_array->descriptor[0].address = address;
4386 desc_array->descriptor[0].length = length - 4;
4387 desc_array->descriptor[1].address = address + (length - 4);
4388 desc_array->descriptor[1].length = 4;
4389 desc_array->elements = 2;
4390 return (desc_array->elements);
4391}
4392
4393/**********************************************************************
4394 *
4395 * Update the board statistics counters.
4396 *
4397 **********************************************************************/
4398static void
4399lem_update_stats_counters(struct adapter *adapter)
4400{
4401
4402 if(adapter->hw.phy.media_type == e1000_media_type_copper ||
4403 (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4404 adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
4405 adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
4406 }
4407 adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
4408 adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
4409 adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
4410 adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
4411
4412 adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
4413 adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
4414 adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
4415 adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
4416 adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
4417 adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
4418 adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
4419 adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
4420 adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
4421 adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
4422 adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
4423 adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
4424 adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
4425 adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
4426 adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
4427 adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
4428 adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
4429 adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
4430 adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
4431 adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
4432
4433 /* For the 64-bit byte counters the low dword must be read first. */
4434 /* Both registers clear on the read of the high dword */
4435
4436 adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
4437 ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
4438 adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
4439 ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
4440
4441 adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
4442 adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
4443 adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
4444 adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
4445 adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
4446
4447 adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
4448 adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
4449
4450 adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
4451 adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
4452 adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
4453 adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
4454 adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
4455 adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
4456 adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
4457 adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
4458 adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
4459 adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
4460
4461 if (adapter->hw.mac.type >= e1000_82543) {
4462 adapter->stats.algnerrc +=
4463 E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4464 adapter->stats.rxerrc +=
4465 E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4466 adapter->stats.tncrs +=
4467 E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4468 adapter->stats.cexterr +=
4469 E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4470 adapter->stats.tsctc +=
4471 E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4472 adapter->stats.tsctfc +=
4473 E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4474 }
4475}
4476
4477static uint64_t
4478lem_get_counter(if_t ifp, ift_counter cnt)
4479{
4480 struct adapter *adapter;
4481
4482 adapter = if_getsoftc(ifp);
4483
4484 switch (cnt) {
4485 case IFCOUNTER_COLLISIONS:
4486 return (adapter->stats.colc);
4487 case IFCOUNTER_IERRORS:
4488 return (adapter->dropped_pkts + adapter->stats.rxerrc +
4489 adapter->stats.crcerrs + adapter->stats.algnerrc +
4490 adapter->stats.ruc + adapter->stats.roc +
4491 adapter->stats.mpc + adapter->stats.cexterr);
4492 case IFCOUNTER_OERRORS:
4493 return (adapter->stats.ecol + adapter->stats.latecol +
4494 adapter->watchdog_events);
4495 default:
4496 return (if_get_counter_default(ifp, cnt));
4497 }
4498}
4499
4500/* Export a single 32-bit register via a read-only sysctl. */
4501static int
4502lem_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4503{
4504 struct adapter *adapter;
4505 u_int val;
4506
4507 adapter = oidp->oid_arg1;
4508 val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4509 return (sysctl_handle_int(oidp, &val, 0, req));
4510}
4511
4512/*
4513 * Add sysctl variables, one per statistic, to the system.
4514 */
4515static void
4516lem_add_hw_stats(struct adapter *adapter)
4517{
4518 device_t dev = adapter->dev;
4519
4520 struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4521 struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4522 struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4523 struct e1000_hw_stats *stats = &adapter->stats;
4524
4525 struct sysctl_oid *stat_node;
4526 struct sysctl_oid_list *stat_list;
4527
4528 /* Driver Statistics */
4529 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_alloc_fail",
4530 CTLFLAG_RD, &adapter->mbuf_alloc_failed,
4531 "Std mbuf failed");
4532 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "cluster_alloc_fail",
4533 CTLFLAG_RD, &adapter->mbuf_cluster_failed,
4534 "Std mbuf cluster failed");
4535 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4536 CTLFLAG_RD, &adapter->dropped_pkts,
4537 "Driver dropped packets");
4538 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
4539 CTLFLAG_RD, &adapter->no_tx_dma_setup,
4540 "Driver tx dma failure in xmit");
4541 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_desc_fail1",
4542 CTLFLAG_RD, &adapter->no_tx_desc_avail1,
4543 "Not enough tx descriptors failure in xmit");
4544 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_desc_fail2",
4545 CTLFLAG_RD, &adapter->no_tx_desc_avail2,
4546 "Not enough tx descriptors failure in xmit");
4547 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4548 CTLFLAG_RD, &adapter->rx_overruns,
4549 "RX overruns");
4550 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4551 CTLFLAG_RD, &adapter->watchdog_events,
4552 "Watchdog timeouts");
4553
4554 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4555 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4556 lem_sysctl_reg_handler, "IU",
4557 "Device Control Register");
4558 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4559 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4560 lem_sysctl_reg_handler, "IU",
4561 "Receiver Control Register");
4562 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4563 CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4564 "Flow Control High Watermark");
4565 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4566 CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4567 "Flow Control Low Watermark");
4568 SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "fifo_workaround",
4569 CTLFLAG_RD, &adapter->tx_fifo_wrk_cnt,
4570 "TX FIFO workaround events");
4571 SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "fifo_reset",
4572 CTLFLAG_RD, &adapter->tx_fifo_reset_cnt,
4573 "TX FIFO resets");
4574
4575 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "txd_head",
4576 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_TDH(0),
4577 lem_sysctl_reg_handler, "IU",
4578 "Transmit Descriptor Head");
4579 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "txd_tail",
4580 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_TDT(0),
4581 lem_sysctl_reg_handler, "IU",
4582 "Transmit Descriptor Tail");
4583 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rxd_head",
4584 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RDH(0),
4585 lem_sysctl_reg_handler, "IU",
4586 "Receive Descriptor Head");
4587 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rxd_tail",
4588 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RDT(0),
4589 lem_sysctl_reg_handler, "IU",
4590 "Receive Descriptor Tail");
4591
4592
4593 /* MAC stats get their own sub node */
4594
4595 stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4596 CTLFLAG_RD, NULL, "Statistics");
4597 stat_list = SYSCTL_CHILDREN(stat_node);
4598
4599 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4600 CTLFLAG_RD, &stats->ecol,
4601 "Excessive collisions");
4602 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4603 CTLFLAG_RD, &stats->scc,
4604 "Single collisions");
4605 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4606 CTLFLAG_RD, &stats->mcc,
4607 "Multiple collisions");
4608 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4609 CTLFLAG_RD, &stats->latecol,
4610 "Late collisions");
4611 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4612 CTLFLAG_RD, &stats->colc,
4613 "Collision Count");
4614 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4615 CTLFLAG_RD, &adapter->stats.symerrs,
4616 "Symbol Errors");
4617 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4618 CTLFLAG_RD, &adapter->stats.sec,
4619 "Sequence Errors");
4620 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4621 CTLFLAG_RD, &adapter->stats.dc,
4622 "Defer Count");
4623 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4624 CTLFLAG_RD, &adapter->stats.mpc,
4625 "Missed Packets");
4626 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4627 CTLFLAG_RD, &adapter->stats.rnbc,
4628 "Receive No Buffers");
4629 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4630 CTLFLAG_RD, &adapter->stats.ruc,
4631 "Receive Undersize");
4632 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4633 CTLFLAG_RD, &adapter->stats.rfc,
4634 "Fragmented Packets Received ");
4635 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4636 CTLFLAG_RD, &adapter->stats.roc,
4637 "Oversized Packets Received");
4638 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4639 CTLFLAG_RD, &adapter->stats.rjc,
4640 "Recevied Jabber");
4641 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4642 CTLFLAG_RD, &adapter->stats.rxerrc,
4643 "Receive Errors");
4644 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4645 CTLFLAG_RD, &adapter->stats.crcerrs,
4646 "CRC errors");
4647 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4648 CTLFLAG_RD, &adapter->stats.algnerrc,
4649 "Alignment Errors");
4650 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4651 CTLFLAG_RD, &adapter->stats.cexterr,
4652 "Collision/Carrier extension errors");
4653 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4654 CTLFLAG_RD, &adapter->stats.xonrxc,
4655 "XON Received");
4656 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4657 CTLFLAG_RD, &adapter->stats.xontxc,
4658 "XON Transmitted");
4659 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4660 CTLFLAG_RD, &adapter->stats.xoffrxc,
4661 "XOFF Received");
4662 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4663 CTLFLAG_RD, &adapter->stats.xofftxc,
4664 "XOFF Transmitted");
4665
4666 /* Packet Reception Stats */
4667 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4668 CTLFLAG_RD, &adapter->stats.tpr,
4669 "Total Packets Received ");
4670 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4671 CTLFLAG_RD, &adapter->stats.gprc,
4672 "Good Packets Received");
4673 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4674 CTLFLAG_RD, &adapter->stats.bprc,
4675 "Broadcast Packets Received");
4676 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4677 CTLFLAG_RD, &adapter->stats.mprc,
4678 "Multicast Packets Received");
4679 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4680 CTLFLAG_RD, &adapter->stats.prc64,
4681 "64 byte frames received ");
4682 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4683 CTLFLAG_RD, &adapter->stats.prc127,
4684 "65-127 byte frames received");
4685 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4686 CTLFLAG_RD, &adapter->stats.prc255,
4687 "128-255 byte frames received");
4688 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4689 CTLFLAG_RD, &adapter->stats.prc511,
4690 "256-511 byte frames received");
4691 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4692 CTLFLAG_RD, &adapter->stats.prc1023,
4693 "512-1023 byte frames received");
4694 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4695 CTLFLAG_RD, &adapter->stats.prc1522,
4696 "1023-1522 byte frames received");
4697 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4698 CTLFLAG_RD, &adapter->stats.gorc,
4699 "Good Octets Received");
4700
4701 /* Packet Transmission Stats */
4702 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4703 CTLFLAG_RD, &adapter->stats.gotc,
4704 "Good Octets Transmitted");
4705 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4706 CTLFLAG_RD, &adapter->stats.tpt,
4707 "Total Packets Transmitted");
4708 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4709 CTLFLAG_RD, &adapter->stats.gptc,
4710 "Good Packets Transmitted");
4711 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4712 CTLFLAG_RD, &adapter->stats.bptc,
4713 "Broadcast Packets Transmitted");
4714 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4715 CTLFLAG_RD, &adapter->stats.mptc,
4716 "Multicast Packets Transmitted");
4717 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4718 CTLFLAG_RD, &adapter->stats.ptc64,
4719 "64 byte frames transmitted ");
4720 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4721 CTLFLAG_RD, &adapter->stats.ptc127,
4722 "65-127 byte frames transmitted");
4723 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4724 CTLFLAG_RD, &adapter->stats.ptc255,
4725 "128-255 byte frames transmitted");
4726 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4727 CTLFLAG_RD, &adapter->stats.ptc511,
4728 "256-511 byte frames transmitted");
4729 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4730 CTLFLAG_RD, &adapter->stats.ptc1023,
4731 "512-1023 byte frames transmitted");
4732 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4733 CTLFLAG_RD, &adapter->stats.ptc1522,
4734 "1024-1522 byte frames transmitted");
4735 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4736 CTLFLAG_RD, &adapter->stats.tsctc,
4737 "TSO Contexts Transmitted");
4738 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4739 CTLFLAG_RD, &adapter->stats.tsctfc,
4740 "TSO Contexts Failed");
4741}
4742
4743/**********************************************************************
4744 *
4745 * This routine provides a way to dump out the adapter eeprom,
4746 * often a useful debug/service tool. This only dumps the first
4747 * 32 words, stuff that matters is in that extent.
4748 *
4749 **********************************************************************/
4750
4751static int
4752lem_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4753{
4754 struct adapter *adapter;
4755 int error;
4756 int result;
4757
4758 result = -1;
4759 error = sysctl_handle_int(oidp, &result, 0, req);
4760
4761 if (error || !req->newptr)
4762 return (error);
4763
4764 /*
4765 * This value will cause a hex dump of the
4766 * first 32 16-bit words of the EEPROM to
4767 * the screen.
4768 */
4769 if (result == 1) {
4770 adapter = (struct adapter *)arg1;
4771 lem_print_nvm_info(adapter);
4772 }
4773
4774 return (error);
4775}
4776
4777static void
4778lem_print_nvm_info(struct adapter *adapter)
4779{
4780 u16 eeprom_data;
4781 int i, j, row = 0;
4782
4783 /* Its a bit crude, but it gets the job done */
4784 printf("\nInterface EEPROM Dump:\n");
4785 printf("Offset\n0x0000 ");
4786 for (i = 0, j = 0; i < 32; i++, j++) {
4787 if (j == 8) { /* Make the offset block */
4788 j = 0; ++row;
4789 printf("\n0x00%x0 ",row);
4790 }
4791 e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4792 printf("%04x ", eeprom_data);
4793 }
4794 printf("\n");
4795}
4796
4797static int
4798lem_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4799{
4800 struct em_int_delay_info *info;
4801 struct adapter *adapter;
4802 u32 regval;
4803 int error;
4804 int usecs;
4805 int ticks;
4806
4807 info = (struct em_int_delay_info *)arg1;
4808 usecs = info->value;
4809 error = sysctl_handle_int(oidp, &usecs, 0, req);
4810 if (error != 0 || req->newptr == NULL)
4811 return (error);
4812 if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4813 return (EINVAL);
4814 info->value = usecs;
4815 ticks = EM_USECS_TO_TICKS(usecs);
4816 if (info->offset == E1000_ITR) /* units are 256ns here */
4817 ticks *= 4;
4818
4819 adapter = info->adapter;
4820
4821 EM_CORE_LOCK(adapter);
4822 regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4823 regval = (regval & ~0xffff) | (ticks & 0xffff);
4824 /* Handle a few special cases. */
4825 switch (info->offset) {
4826 case E1000_RDTR:
4827 break;
4828 case E1000_TIDV:
4829 if (ticks == 0) {
4830 adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4831 /* Don't write 0 into the TIDV register. */
4832 regval++;
4833 } else
4834 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4835 break;
4836 }
4837 E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4838 EM_CORE_UNLOCK(adapter);
4839 return (0);
4840}
4841
4842static void
4843lem_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4844 const char *description, struct em_int_delay_info *info,
4845 int offset, int value)
4846{
4847 info->adapter = adapter;
4848 info->offset = offset;
4849 info->value = value;
4850 SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4851 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4852 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4853 info, 0, lem_sysctl_int_delay, "I", description);
4854}
4855
4856static void
4857lem_set_flow_cntrl(struct adapter *adapter, const char *name,
4858 const char *description, int *limit, int value)
4859{
4860 *limit = value;
4861 SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
4862 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4863 OID_AUTO, name, CTLFLAG_RW, limit, value, description);
4864}
4865
4866static void
4867lem_add_rx_process_limit(struct adapter *adapter, const char *name,
4868 const char *description, int *limit, int value)
4869{
4870 *limit = value;
4871 SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
4872 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4873 OID_AUTO, name, CTLFLAG_RW, limit, value, description);
4874}
| 34
35/*
36 * Uncomment the following extensions for better performance in a VM,
37 * especially if you have support in the hypervisor.
38 * See http://info.iet.unipi.it/~luigi/netmap/
39 */
40// #define BATCH_DISPATCH
41// #define NIC_SEND_COMBINING
42// #define NIC_PARAVIRT /* enable virtio-like synchronization */
43
44#include "opt_inet.h"
45#include "opt_inet6.h"
46
47#ifdef HAVE_KERNEL_OPTION_HEADERS
48#include "opt_device_polling.h"
49#endif
50
51#include <sys/param.h>
52#include <sys/systm.h>
53#include <sys/buf_ring.h>
54#include <sys/bus.h>
55#include <sys/endian.h>
56#include <sys/kernel.h>
57#include <sys/kthread.h>
58#include <sys/malloc.h>
59#include <sys/mbuf.h>
60#include <sys/module.h>
61#include <sys/rman.h>
62#include <sys/socket.h>
63#include <sys/sockio.h>
64#include <sys/sysctl.h>
65#include <sys/taskqueue.h>
66#include <sys/eventhandler.h>
67#include <machine/bus.h>
68#include <machine/resource.h>
69
70#include <net/bpf.h>
71#include <net/ethernet.h>
72#include <net/if.h>
73#include <net/if_var.h>
74#include <net/if_arp.h>
75#include <net/if_dl.h>
76#include <net/if_media.h>
77
78#include <net/if_types.h>
79#include <net/if_vlan_var.h>
80
81#include <netinet/in_systm.h>
82#include <netinet/in.h>
83#include <netinet/if_ether.h>
84#include <netinet/ip.h>
85#include <netinet/ip6.h>
86#include <netinet/tcp.h>
87#include <netinet/udp.h>
88
89#include <machine/in_cksum.h>
90#include <dev/led/led.h>
91#include <dev/pci/pcivar.h>
92#include <dev/pci/pcireg.h>
93
94#include "e1000_api.h"
95#include "if_lem.h"
96
97/*********************************************************************
98 * Legacy Em Driver version:
99 *********************************************************************/
100char lem_driver_version[] = "1.1.0";
101
102/*********************************************************************
103 * PCI Device ID Table
104 *
105 * Used by probe to select devices to load on
106 * Last field stores an index into e1000_strings
107 * Last entry must be all 0s
108 *
109 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
110 *********************************************************************/
111
112static em_vendor_info_t lem_vendor_info_array[] =
113{
114 /* Intel(R) PRO/1000 Network Connection */
115 { 0x8086, E1000_DEV_ID_82540EM, PCI_ANY_ID, PCI_ANY_ID, 0},
116 { 0x8086, E1000_DEV_ID_82540EM_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
117 { 0x8086, E1000_DEV_ID_82540EP, PCI_ANY_ID, PCI_ANY_ID, 0},
118 { 0x8086, E1000_DEV_ID_82540EP_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
119 { 0x8086, E1000_DEV_ID_82540EP_LP, PCI_ANY_ID, PCI_ANY_ID, 0},
120
121 { 0x8086, E1000_DEV_ID_82541EI, PCI_ANY_ID, PCI_ANY_ID, 0},
122 { 0x8086, E1000_DEV_ID_82541ER, PCI_ANY_ID, PCI_ANY_ID, 0},
123 { 0x8086, E1000_DEV_ID_82541ER_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
124 { 0x8086, E1000_DEV_ID_82541EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
125 { 0x8086, E1000_DEV_ID_82541GI, PCI_ANY_ID, PCI_ANY_ID, 0},
126 { 0x8086, E1000_DEV_ID_82541GI_LF, PCI_ANY_ID, PCI_ANY_ID, 0},
127 { 0x8086, E1000_DEV_ID_82541GI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
128
129 { 0x8086, E1000_DEV_ID_82542, PCI_ANY_ID, PCI_ANY_ID, 0},
130
131 { 0x8086, E1000_DEV_ID_82543GC_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
132 { 0x8086, E1000_DEV_ID_82543GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
133
134 { 0x8086, E1000_DEV_ID_82544EI_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
135 { 0x8086, E1000_DEV_ID_82544EI_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
136 { 0x8086, E1000_DEV_ID_82544GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
137 { 0x8086, E1000_DEV_ID_82544GC_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
138
139 { 0x8086, E1000_DEV_ID_82545EM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
140 { 0x8086, E1000_DEV_ID_82545EM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
141 { 0x8086, E1000_DEV_ID_82545GM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
142 { 0x8086, E1000_DEV_ID_82545GM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
143 { 0x8086, E1000_DEV_ID_82545GM_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
144
145 { 0x8086, E1000_DEV_ID_82546EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
146 { 0x8086, E1000_DEV_ID_82546EB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
147 { 0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
148 { 0x8086, E1000_DEV_ID_82546GB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
149 { 0x8086, E1000_DEV_ID_82546GB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
150 { 0x8086, E1000_DEV_ID_82546GB_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
151 { 0x8086, E1000_DEV_ID_82546GB_PCIE, PCI_ANY_ID, PCI_ANY_ID, 0},
152 { 0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
153 { 0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3,
154 PCI_ANY_ID, PCI_ANY_ID, 0},
155
156 { 0x8086, E1000_DEV_ID_82547EI, PCI_ANY_ID, PCI_ANY_ID, 0},
157 { 0x8086, E1000_DEV_ID_82547EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
158 { 0x8086, E1000_DEV_ID_82547GI, PCI_ANY_ID, PCI_ANY_ID, 0},
159 /* required last entry */
160 { 0, 0, 0, 0, 0}
161};
162
163/*********************************************************************
164 * Table of branding strings for all supported NICs.
165 *********************************************************************/
166
167static char *lem_strings[] = {
168 "Intel(R) PRO/1000 Legacy Network Connection"
169};
170
171/*********************************************************************
172 * Function prototypes
173 *********************************************************************/
174static int lem_probe(device_t);
175static int lem_attach(device_t);
176static int lem_detach(device_t);
177static int lem_shutdown(device_t);
178static int lem_suspend(device_t);
179static int lem_resume(device_t);
180static void lem_start(if_t);
181static void lem_start_locked(if_t ifp);
182static int lem_ioctl(if_t, u_long, caddr_t);
183static uint64_t lem_get_counter(if_t, ift_counter);
184static void lem_init(void *);
185static void lem_init_locked(struct adapter *);
186static void lem_stop(void *);
187static void lem_media_status(if_t, struct ifmediareq *);
188static int lem_media_change(if_t);
189static void lem_identify_hardware(struct adapter *);
190static int lem_allocate_pci_resources(struct adapter *);
191static int lem_allocate_irq(struct adapter *adapter);
192static void lem_free_pci_resources(struct adapter *);
193static void lem_local_timer(void *);
194static int lem_hardware_init(struct adapter *);
195static int lem_setup_interface(device_t, struct adapter *);
196static void lem_setup_transmit_structures(struct adapter *);
197static void lem_initialize_transmit_unit(struct adapter *);
198static int lem_setup_receive_structures(struct adapter *);
199static void lem_initialize_receive_unit(struct adapter *);
200static void lem_enable_intr(struct adapter *);
201static void lem_disable_intr(struct adapter *);
202static void lem_free_transmit_structures(struct adapter *);
203static void lem_free_receive_structures(struct adapter *);
204static void lem_update_stats_counters(struct adapter *);
205static void lem_add_hw_stats(struct adapter *adapter);
206static void lem_txeof(struct adapter *);
207static void lem_tx_purge(struct adapter *);
208static int lem_allocate_receive_structures(struct adapter *);
209static int lem_allocate_transmit_structures(struct adapter *);
210static bool lem_rxeof(struct adapter *, int, int *);
211#ifndef __NO_STRICT_ALIGNMENT
212static int lem_fixup_rx(struct adapter *);
213#endif
214static void lem_receive_checksum(struct adapter *, struct e1000_rx_desc *,
215 struct mbuf *);
216static void lem_transmit_checksum_setup(struct adapter *, struct mbuf *,
217 u32 *, u32 *);
218static void lem_set_promisc(struct adapter *);
219static void lem_disable_promisc(struct adapter *);
220static void lem_set_multi(struct adapter *);
221static void lem_update_link_status(struct adapter *);
222static int lem_get_buf(struct adapter *, int);
223static void lem_register_vlan(void *, if_t, u16);
224static void lem_unregister_vlan(void *, if_t, u16);
225static void lem_setup_vlan_hw_support(struct adapter *);
226static int lem_xmit(struct adapter *, struct mbuf **);
227static void lem_smartspeed(struct adapter *);
228static int lem_82547_fifo_workaround(struct adapter *, int);
229static void lem_82547_update_fifo_head(struct adapter *, int);
230static int lem_82547_tx_fifo_reset(struct adapter *);
231static void lem_82547_move_tail(void *);
232static int lem_dma_malloc(struct adapter *, bus_size_t,
233 struct em_dma_alloc *, int);
234static void lem_dma_free(struct adapter *, struct em_dma_alloc *);
235static int lem_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
236static void lem_print_nvm_info(struct adapter *);
237static int lem_is_valid_ether_addr(u8 *);
238static u32 lem_fill_descriptors (bus_addr_t address, u32 length,
239 PDESC_ARRAY desc_array);
240static int lem_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
241static void lem_add_int_delay_sysctl(struct adapter *, const char *,
242 const char *, struct em_int_delay_info *, int, int);
243static void lem_set_flow_cntrl(struct adapter *, const char *,
244 const char *, int *, int);
245/* Management and WOL Support */
246static void lem_init_manageability(struct adapter *);
247static void lem_release_manageability(struct adapter *);
248static void lem_get_hw_control(struct adapter *);
249static void lem_release_hw_control(struct adapter *);
250static void lem_get_wakeup(device_t);
251static void lem_enable_wakeup(device_t);
252static int lem_enable_phy_wakeup(struct adapter *);
253static void lem_led_func(void *, int);
254
255static void lem_intr(void *);
256static int lem_irq_fast(void *);
257static void lem_handle_rxtx(void *context, int pending);
258static void lem_handle_link(void *context, int pending);
259static void lem_add_rx_process_limit(struct adapter *, const char *,
260 const char *, int *, int);
261
262#ifdef DEVICE_POLLING
263static poll_handler_t lem_poll;
264#endif /* POLLING */
265
266/*********************************************************************
267 * FreeBSD Device Interface Entry Points
268 *********************************************************************/
269
270static device_method_t lem_methods[] = {
271 /* Device interface */
272 DEVMETHOD(device_probe, lem_probe),
273 DEVMETHOD(device_attach, lem_attach),
274 DEVMETHOD(device_detach, lem_detach),
275 DEVMETHOD(device_shutdown, lem_shutdown),
276 DEVMETHOD(device_suspend, lem_suspend),
277 DEVMETHOD(device_resume, lem_resume),
278 DEVMETHOD_END
279};
280
281static driver_t lem_driver = {
282 "em", lem_methods, sizeof(struct adapter),
283};
284
285extern devclass_t em_devclass;
286DRIVER_MODULE(lem, pci, lem_driver, em_devclass, 0, 0);
287MODULE_DEPEND(lem, pci, 1, 1, 1);
288MODULE_DEPEND(lem, ether, 1, 1, 1);
289#ifdef DEV_NETMAP
290MODULE_DEPEND(lem, netmap, 1, 1, 1);
291#endif /* DEV_NETMAP */
292
293/*********************************************************************
294 * Tunable default values.
295 *********************************************************************/
296
297#define EM_TICKS_TO_USECS(ticks) ((1024 * (ticks) + 500) / 1000)
298#define EM_USECS_TO_TICKS(usecs) ((1000 * (usecs) + 512) / 1024)
299
300#define MAX_INTS_PER_SEC 8000
301#define DEFAULT_ITR (1000000000/(MAX_INTS_PER_SEC * 256))
302
303static int lem_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
304static int lem_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
305static int lem_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
306static int lem_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
307/*
308 * increase lem_rxd and lem_txd to at least 2048 in netmap mode
309 * for better performance.
310 */
311static int lem_rxd = EM_DEFAULT_RXD;
312static int lem_txd = EM_DEFAULT_TXD;
313static int lem_smart_pwr_down = FALSE;
314
315/* Controls whether promiscuous also shows bad packets */
316static int lem_debug_sbp = FALSE;
317
318TUNABLE_INT("hw.em.tx_int_delay", &lem_tx_int_delay_dflt);
319TUNABLE_INT("hw.em.rx_int_delay", &lem_rx_int_delay_dflt);
320TUNABLE_INT("hw.em.tx_abs_int_delay", &lem_tx_abs_int_delay_dflt);
321TUNABLE_INT("hw.em.rx_abs_int_delay", &lem_rx_abs_int_delay_dflt);
322TUNABLE_INT("hw.em.rxd", &lem_rxd);
323TUNABLE_INT("hw.em.txd", &lem_txd);
324TUNABLE_INT("hw.em.smart_pwr_down", &lem_smart_pwr_down);
325TUNABLE_INT("hw.em.sbp", &lem_debug_sbp);
326
327/* Interrupt style - default to fast */
328static int lem_use_legacy_irq = 0;
329TUNABLE_INT("hw.em.use_legacy_irq", &lem_use_legacy_irq);
330
331/* How many packets rxeof tries to clean at a time */
332static int lem_rx_process_limit = 100;
333TUNABLE_INT("hw.em.rx_process_limit", &lem_rx_process_limit);
334
335/* Flow control setting - default to FULL */
336static int lem_fc_setting = e1000_fc_full;
337TUNABLE_INT("hw.em.fc_setting", &lem_fc_setting);
338
339/* Global used in WOL setup with multiport cards */
340static int global_quad_port_a = 0;
341
342#ifdef DEV_NETMAP /* see ixgbe.c for details */
343#include <dev/netmap/if_lem_netmap.h>
344#endif /* DEV_NETMAP */
345
346/*********************************************************************
347 * Device identification routine
348 *
349 * em_probe determines if the driver should be loaded on
350 * adapter based on PCI vendor/device id of the adapter.
351 *
352 * return BUS_PROBE_DEFAULT on success, positive on failure
353 *********************************************************************/
354
355static int
356lem_probe(device_t dev)
357{
358 char adapter_name[60];
359 u16 pci_vendor_id = 0;
360 u16 pci_device_id = 0;
361 u16 pci_subvendor_id = 0;
362 u16 pci_subdevice_id = 0;
363 em_vendor_info_t *ent;
364
365 INIT_DEBUGOUT("em_probe: begin");
366
367 pci_vendor_id = pci_get_vendor(dev);
368 if (pci_vendor_id != EM_VENDOR_ID)
369 return (ENXIO);
370
371 pci_device_id = pci_get_device(dev);
372 pci_subvendor_id = pci_get_subvendor(dev);
373 pci_subdevice_id = pci_get_subdevice(dev);
374
375 ent = lem_vendor_info_array;
376 while (ent->vendor_id != 0) {
377 if ((pci_vendor_id == ent->vendor_id) &&
378 (pci_device_id == ent->device_id) &&
379
380 ((pci_subvendor_id == ent->subvendor_id) ||
381 (ent->subvendor_id == PCI_ANY_ID)) &&
382
383 ((pci_subdevice_id == ent->subdevice_id) ||
384 (ent->subdevice_id == PCI_ANY_ID))) {
385 sprintf(adapter_name, "%s %s",
386 lem_strings[ent->index],
387 lem_driver_version);
388 device_set_desc_copy(dev, adapter_name);
389 return (BUS_PROBE_DEFAULT);
390 }
391 ent++;
392 }
393
394 return (ENXIO);
395}
396
397/*********************************************************************
398 * Device initialization routine
399 *
400 * The attach entry point is called when the driver is being loaded.
401 * This routine identifies the type of hardware, allocates all resources
402 * and initializes the hardware.
403 *
404 * return 0 on success, positive on failure
405 *********************************************************************/
406
407static int
408lem_attach(device_t dev)
409{
410 struct adapter *adapter;
411 int tsize, rsize;
412 int error = 0;
413
414 INIT_DEBUGOUT("lem_attach: begin");
415
416 adapter = device_get_softc(dev);
417 adapter->dev = adapter->osdep.dev = dev;
418 EM_CORE_LOCK_INIT(adapter, device_get_nameunit(dev));
419 EM_TX_LOCK_INIT(adapter, device_get_nameunit(dev));
420 EM_RX_LOCK_INIT(adapter, device_get_nameunit(dev));
421
422 /* SYSCTL stuff */
423 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
424 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
425 OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
426 lem_sysctl_nvm_info, "I", "NVM Information");
427
428 callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0);
429 callout_init_mtx(&adapter->tx_fifo_timer, &adapter->tx_mtx, 0);
430
431 /* Determine hardware and mac info */
432 lem_identify_hardware(adapter);
433
434 /* Setup PCI resources */
435 if (lem_allocate_pci_resources(adapter)) {
436 device_printf(dev, "Allocation of PCI resources failed\n");
437 error = ENXIO;
438 goto err_pci;
439 }
440
441 /* Do Shared Code initialization */
442 if (e1000_setup_init_funcs(&adapter->hw, TRUE)) {
443 device_printf(dev, "Setup of Shared code failed\n");
444 error = ENXIO;
445 goto err_pci;
446 }
447
448 e1000_get_bus_info(&adapter->hw);
449
450 /* Set up some sysctls for the tunable interrupt delays */
451 lem_add_int_delay_sysctl(adapter, "rx_int_delay",
452 "receive interrupt delay in usecs", &adapter->rx_int_delay,
453 E1000_REGISTER(&adapter->hw, E1000_RDTR), lem_rx_int_delay_dflt);
454 lem_add_int_delay_sysctl(adapter, "tx_int_delay",
455 "transmit interrupt delay in usecs", &adapter->tx_int_delay,
456 E1000_REGISTER(&adapter->hw, E1000_TIDV), lem_tx_int_delay_dflt);
457 if (adapter->hw.mac.type >= e1000_82540) {
458 lem_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
459 "receive interrupt delay limit in usecs",
460 &adapter->rx_abs_int_delay,
461 E1000_REGISTER(&adapter->hw, E1000_RADV),
462 lem_rx_abs_int_delay_dflt);
463 lem_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
464 "transmit interrupt delay limit in usecs",
465 &adapter->tx_abs_int_delay,
466 E1000_REGISTER(&adapter->hw, E1000_TADV),
467 lem_tx_abs_int_delay_dflt);
468 lem_add_int_delay_sysctl(adapter, "itr",
469 "interrupt delay limit in usecs/4",
470 &adapter->tx_itr,
471 E1000_REGISTER(&adapter->hw, E1000_ITR),
472 DEFAULT_ITR);
473 }
474
475 /* Sysctls for limiting the amount of work done in the taskqueue */
476 lem_add_rx_process_limit(adapter, "rx_processing_limit",
477 "max number of rx packets to process", &adapter->rx_process_limit,
478 lem_rx_process_limit);
479
480#ifdef NIC_SEND_COMBINING
481 /* Sysctls to control mitigation */
482 lem_add_rx_process_limit(adapter, "sc_enable",
483 "driver TDT mitigation", &adapter->sc_enable, 0);
484#endif /* NIC_SEND_COMBINING */
485#ifdef BATCH_DISPATCH
486 lem_add_rx_process_limit(adapter, "batch_enable",
487 "driver rx batch", &adapter->batch_enable, 0);
488#endif /* BATCH_DISPATCH */
489#ifdef NIC_PARAVIRT
490 lem_add_rx_process_limit(adapter, "rx_retries",
491 "driver rx retries", &adapter->rx_retries, 0);
492#endif /* NIC_PARAVIRT */
493
494 /* Sysctl for setting the interface flow control */
495 lem_set_flow_cntrl(adapter, "flow_control",
496 "flow control setting",
497 &adapter->fc_setting, lem_fc_setting);
498
499 /*
500 * Validate number of transmit and receive descriptors. It
501 * must not exceed hardware maximum, and must be multiple
502 * of E1000_DBA_ALIGN.
503 */
504 if (((lem_txd * sizeof(struct e1000_tx_desc)) % EM_DBA_ALIGN) != 0 ||
505 (adapter->hw.mac.type >= e1000_82544 && lem_txd > EM_MAX_TXD) ||
506 (adapter->hw.mac.type < e1000_82544 && lem_txd > EM_MAX_TXD_82543) ||
507 (lem_txd < EM_MIN_TXD)) {
508 device_printf(dev, "Using %d TX descriptors instead of %d!\n",
509 EM_DEFAULT_TXD, lem_txd);
510 adapter->num_tx_desc = EM_DEFAULT_TXD;
511 } else
512 adapter->num_tx_desc = lem_txd;
513 if (((lem_rxd * sizeof(struct e1000_rx_desc)) % EM_DBA_ALIGN) != 0 ||
514 (adapter->hw.mac.type >= e1000_82544 && lem_rxd > EM_MAX_RXD) ||
515 (adapter->hw.mac.type < e1000_82544 && lem_rxd > EM_MAX_RXD_82543) ||
516 (lem_rxd < EM_MIN_RXD)) {
517 device_printf(dev, "Using %d RX descriptors instead of %d!\n",
518 EM_DEFAULT_RXD, lem_rxd);
519 adapter->num_rx_desc = EM_DEFAULT_RXD;
520 } else
521 adapter->num_rx_desc = lem_rxd;
522
523 adapter->hw.mac.autoneg = DO_AUTO_NEG;
524 adapter->hw.phy.autoneg_wait_to_complete = FALSE;
525 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
526 adapter->rx_buffer_len = 2048;
527
528 e1000_init_script_state_82541(&adapter->hw, TRUE);
529 e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
530
531 /* Copper options */
532 if (adapter->hw.phy.media_type == e1000_media_type_copper) {
533 adapter->hw.phy.mdix = AUTO_ALL_MODES;
534 adapter->hw.phy.disable_polarity_correction = FALSE;
535 adapter->hw.phy.ms_type = EM_MASTER_SLAVE;
536 }
537
538 /*
539 * Set the frame limits assuming
540 * standard ethernet sized frames.
541 */
542 adapter->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
543 adapter->min_frame_size = ETH_ZLEN + ETHERNET_FCS_SIZE;
544
545 /*
546 * This controls when hardware reports transmit completion
547 * status.
548 */
549 adapter->hw.mac.report_tx_early = 1;
550
551#ifdef NIC_PARAVIRT
552 device_printf(dev, "driver supports paravirt, subdev 0x%x\n",
553 adapter->hw.subsystem_device_id);
554 if (adapter->hw.subsystem_device_id == E1000_PARA_SUBDEV) {
555 uint64_t bus_addr;
556
557 device_printf(dev, "paravirt support on dev %p\n", adapter);
558 tsize = 4096; // XXX one page for the csb
559 if (lem_dma_malloc(adapter, tsize, &adapter->csb_mem, BUS_DMA_NOWAIT)) {
560 device_printf(dev, "Unable to allocate csb memory\n");
561 error = ENOMEM;
562 goto err_csb;
563 }
564 /* Setup the Base of the CSB */
565 adapter->csb = (struct paravirt_csb *)adapter->csb_mem.dma_vaddr;
566 /* force the first kick */
567 adapter->csb->host_need_txkick = 1; /* txring empty */
568 adapter->csb->guest_need_rxkick = 1; /* no rx packets */
569 bus_addr = adapter->csb_mem.dma_paddr;
570 lem_add_rx_process_limit(adapter, "csb_on",
571 "enable paravirt.", &adapter->csb->guest_csb_on, 0);
572 lem_add_rx_process_limit(adapter, "txc_lim",
573 "txc_lim", &adapter->csb->host_txcycles_lim, 1);
574
575 /* some stats */
576#define PA_SC(name, var, val) \
577 lem_add_rx_process_limit(adapter, name, name, var, val)
578 PA_SC("host_need_txkick",&adapter->csb->host_need_txkick, 1);
579 PA_SC("host_rxkick_at",&adapter->csb->host_rxkick_at, ~0);
580 PA_SC("guest_need_txkick",&adapter->csb->guest_need_txkick, 0);
581 PA_SC("guest_need_rxkick",&adapter->csb->guest_need_rxkick, 1);
582 PA_SC("tdt_reg_count",&adapter->tdt_reg_count, 0);
583 PA_SC("tdt_csb_count",&adapter->tdt_csb_count, 0);
584 PA_SC("tdt_int_count",&adapter->tdt_int_count, 0);
585 PA_SC("guest_need_kick_count",&adapter->guest_need_kick_count, 0);
586 /* tell the host where the block is */
587 E1000_WRITE_REG(&adapter->hw, E1000_CSBAH,
588 (u32)(bus_addr >> 32));
589 E1000_WRITE_REG(&adapter->hw, E1000_CSBAL,
590 (u32)bus_addr);
591 }
592#endif /* NIC_PARAVIRT */
593
594 tsize = roundup2(adapter->num_tx_desc * sizeof(struct e1000_tx_desc),
595 EM_DBA_ALIGN);
596
597 /* Allocate Transmit Descriptor ring */
598 if (lem_dma_malloc(adapter, tsize, &adapter->txdma, BUS_DMA_NOWAIT)) {
599 device_printf(dev, "Unable to allocate tx_desc memory\n");
600 error = ENOMEM;
601 goto err_tx_desc;
602 }
603 adapter->tx_desc_base =
604 (struct e1000_tx_desc *)adapter->txdma.dma_vaddr;
605
606 rsize = roundup2(adapter->num_rx_desc * sizeof(struct e1000_rx_desc),
607 EM_DBA_ALIGN);
608
609 /* Allocate Receive Descriptor ring */
610 if (lem_dma_malloc(adapter, rsize, &adapter->rxdma, BUS_DMA_NOWAIT)) {
611 device_printf(dev, "Unable to allocate rx_desc memory\n");
612 error = ENOMEM;
613 goto err_rx_desc;
614 }
615 adapter->rx_desc_base =
616 (struct e1000_rx_desc *)adapter->rxdma.dma_vaddr;
617
618 /* Allocate multicast array memory. */
619 adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
620 MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
621 if (adapter->mta == NULL) {
622 device_printf(dev, "Can not allocate multicast setup array\n");
623 error = ENOMEM;
624 goto err_hw_init;
625 }
626
627 /*
628 ** Start from a known state, this is
629 ** important in reading the nvm and
630 ** mac from that.
631 */
632 e1000_reset_hw(&adapter->hw);
633
634 /* Make sure we have a good EEPROM before we read from it */
635 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
636 /*
637 ** Some PCI-E parts fail the first check due to
638 ** the link being in sleep state, call it again,
639 ** if it fails a second time its a real issue.
640 */
641 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
642 device_printf(dev,
643 "The EEPROM Checksum Is Not Valid\n");
644 error = EIO;
645 goto err_hw_init;
646 }
647 }
648
649 /* Copy the permanent MAC address out of the EEPROM */
650 if (e1000_read_mac_addr(&adapter->hw) < 0) {
651 device_printf(dev, "EEPROM read error while reading MAC"
652 " address\n");
653 error = EIO;
654 goto err_hw_init;
655 }
656
657 if (!lem_is_valid_ether_addr(adapter->hw.mac.addr)) {
658 device_printf(dev, "Invalid MAC address\n");
659 error = EIO;
660 goto err_hw_init;
661 }
662
663 /* Initialize the hardware */
664 if (lem_hardware_init(adapter)) {
665 device_printf(dev, "Unable to initialize the hardware\n");
666 error = EIO;
667 goto err_hw_init;
668 }
669
670 /* Allocate transmit descriptors and buffers */
671 if (lem_allocate_transmit_structures(adapter)) {
672 device_printf(dev, "Could not setup transmit structures\n");
673 error = ENOMEM;
674 goto err_tx_struct;
675 }
676
677 /* Allocate receive descriptors and buffers */
678 if (lem_allocate_receive_structures(adapter)) {
679 device_printf(dev, "Could not setup receive structures\n");
680 error = ENOMEM;
681 goto err_rx_struct;
682 }
683
684 /*
685 ** Do interrupt configuration
686 */
687 error = lem_allocate_irq(adapter);
688 if (error)
689 goto err_rx_struct;
690
691 /*
692 * Get Wake-on-Lan and Management info for later use
693 */
694 lem_get_wakeup(dev);
695
696 /* Setup OS specific network interface */
697 if (lem_setup_interface(dev, adapter) != 0)
698 goto err_rx_struct;
699
700 /* Initialize statistics */
701 lem_update_stats_counters(adapter);
702
703 adapter->hw.mac.get_link_status = 1;
704 lem_update_link_status(adapter);
705
706 /* Indicate SOL/IDER usage */
707 if (e1000_check_reset_block(&adapter->hw))
708 device_printf(dev,
709 "PHY reset is blocked due to SOL/IDER session.\n");
710
711 /* Do we need workaround for 82544 PCI-X adapter? */
712 if (adapter->hw.bus.type == e1000_bus_type_pcix &&
713 adapter->hw.mac.type == e1000_82544)
714 adapter->pcix_82544 = TRUE;
715 else
716 adapter->pcix_82544 = FALSE;
717
718 /* Register for VLAN events */
719 adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
720 lem_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
721 adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
722 lem_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
723
724 lem_add_hw_stats(adapter);
725
726 /* Non-AMT based hardware can now take control from firmware */
727 if (adapter->has_manage && !adapter->has_amt)
728 lem_get_hw_control(adapter);
729
730 /* Tell the stack that the interface is not active */
731 if_setdrvflagbits(adapter->ifp, 0, IFF_DRV_OACTIVE | IFF_DRV_RUNNING);
732
733 adapter->led_dev = led_create(lem_led_func, adapter,
734 device_get_nameunit(dev));
735
736#ifdef DEV_NETMAP
737 lem_netmap_attach(adapter);
738#endif /* DEV_NETMAP */
739 INIT_DEBUGOUT("lem_attach: end");
740
741 return (0);
742
743err_rx_struct:
744 lem_free_transmit_structures(adapter);
745err_tx_struct:
746err_hw_init:
747 lem_release_hw_control(adapter);
748 lem_dma_free(adapter, &adapter->rxdma);
749err_rx_desc:
750 lem_dma_free(adapter, &adapter->txdma);
751err_tx_desc:
752#ifdef NIC_PARAVIRT
753 lem_dma_free(adapter, &adapter->csb_mem);
754err_csb:
755#endif /* NIC_PARAVIRT */
756
757err_pci:
758 if (adapter->ifp != (void *)NULL)
759 if_free(adapter->ifp);
760 lem_free_pci_resources(adapter);
761 free(adapter->mta, M_DEVBUF);
762 EM_TX_LOCK_DESTROY(adapter);
763 EM_RX_LOCK_DESTROY(adapter);
764 EM_CORE_LOCK_DESTROY(adapter);
765
766 return (error);
767}
768
769/*********************************************************************
770 * Device removal routine
771 *
772 * The detach entry point is called when the driver is being removed.
773 * This routine stops the adapter and deallocates all the resources
774 * that were allocated for driver operation.
775 *
776 * return 0 on success, positive on failure
777 *********************************************************************/
778
779static int
780lem_detach(device_t dev)
781{
782 struct adapter *adapter = device_get_softc(dev);
783 if_t ifp = adapter->ifp;
784
785 INIT_DEBUGOUT("em_detach: begin");
786
787 /* Make sure VLANS are not using driver */
788 if (if_vlantrunkinuse(ifp)) {
789 device_printf(dev,"Vlan in use, detach first\n");
790 return (EBUSY);
791 }
792
793#ifdef DEVICE_POLLING
794 if (if_getcapenable(ifp) & IFCAP_POLLING)
795 ether_poll_deregister(ifp);
796#endif
797
798 if (adapter->led_dev != NULL)
799 led_destroy(adapter->led_dev);
800
801 EM_CORE_LOCK(adapter);
802 EM_TX_LOCK(adapter);
803 adapter->in_detach = 1;
804 lem_stop(adapter);
805 e1000_phy_hw_reset(&adapter->hw);
806
807 lem_release_manageability(adapter);
808
809 EM_TX_UNLOCK(adapter);
810 EM_CORE_UNLOCK(adapter);
811
812 /* Unregister VLAN events */
813 if (adapter->vlan_attach != NULL)
814 EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach);
815 if (adapter->vlan_detach != NULL)
816 EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach);
817
818 ether_ifdetach(adapter->ifp);
819 callout_drain(&adapter->timer);
820 callout_drain(&adapter->tx_fifo_timer);
821
822#ifdef DEV_NETMAP
823 netmap_detach(ifp);
824#endif /* DEV_NETMAP */
825 lem_free_pci_resources(adapter);
826 bus_generic_detach(dev);
827 if_free(ifp);
828
829 lem_free_transmit_structures(adapter);
830 lem_free_receive_structures(adapter);
831
832 /* Free Transmit Descriptor ring */
833 if (adapter->tx_desc_base) {
834 lem_dma_free(adapter, &adapter->txdma);
835 adapter->tx_desc_base = NULL;
836 }
837
838 /* Free Receive Descriptor ring */
839 if (adapter->rx_desc_base) {
840 lem_dma_free(adapter, &adapter->rxdma);
841 adapter->rx_desc_base = NULL;
842 }
843
844#ifdef NIC_PARAVIRT
845 if (adapter->csb) {
846 lem_dma_free(adapter, &adapter->csb_mem);
847 adapter->csb = NULL;
848 }
849#endif /* NIC_PARAVIRT */
850 lem_release_hw_control(adapter);
851 free(adapter->mta, M_DEVBUF);
852 EM_TX_LOCK_DESTROY(adapter);
853 EM_RX_LOCK_DESTROY(adapter);
854 EM_CORE_LOCK_DESTROY(adapter);
855
856 return (0);
857}
858
859/*********************************************************************
860 *
861 * Shutdown entry point
862 *
863 **********************************************************************/
864
865static int
866lem_shutdown(device_t dev)
867{
868 return lem_suspend(dev);
869}
870
871/*
872 * Suspend/resume device methods.
873 */
874static int
875lem_suspend(device_t dev)
876{
877 struct adapter *adapter = device_get_softc(dev);
878
879 EM_CORE_LOCK(adapter);
880
881 lem_release_manageability(adapter);
882 lem_release_hw_control(adapter);
883 lem_enable_wakeup(dev);
884
885 EM_CORE_UNLOCK(adapter);
886
887 return bus_generic_suspend(dev);
888}
889
890static int
891lem_resume(device_t dev)
892{
893 struct adapter *adapter = device_get_softc(dev);
894 if_t ifp = adapter->ifp;
895
896 EM_CORE_LOCK(adapter);
897 lem_init_locked(adapter);
898 lem_init_manageability(adapter);
899 EM_CORE_UNLOCK(adapter);
900 lem_start(ifp);
901
902 return bus_generic_resume(dev);
903}
904
905
906static void
907lem_start_locked(if_t ifp)
908{
909 struct adapter *adapter = if_getsoftc(ifp);
910 struct mbuf *m_head;
911
912 EM_TX_LOCK_ASSERT(adapter);
913
914 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) !=
915 IFF_DRV_RUNNING)
916 return;
917 if (!adapter->link_active)
918 return;
919
920 /*
921 * Force a cleanup if number of TX descriptors
922 * available hits the threshold
923 */
924 if (adapter->num_tx_desc_avail <= EM_TX_CLEANUP_THRESHOLD) {
925 lem_txeof(adapter);
926 /* Now do we at least have a minimal? */
927 if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD) {
928 adapter->no_tx_desc_avail1++;
929 return;
930 }
931 }
932
933 while (!if_sendq_empty(ifp)) {
934 m_head = if_dequeue(ifp);
935
936 if (m_head == NULL)
937 break;
938 /*
939 * Encapsulation can modify our pointer, and or make it
940 * NULL on failure. In that event, we can't requeue.
941 */
942 if (lem_xmit(adapter, &m_head)) {
943 if (m_head == NULL)
944 break;
945 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, 0);
946 if_sendq_prepend(ifp, m_head);
947 break;
948 }
949
950 /* Send a copy of the frame to the BPF listener */
951 if_etherbpfmtap(ifp, m_head);
952
953 /* Set timeout in case hardware has problems transmitting. */
954 adapter->watchdog_check = TRUE;
955 adapter->watchdog_time = ticks;
956 }
957 if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD)
958 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, 0);
959#ifdef NIC_PARAVIRT
960 if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE && adapter->csb &&
961 adapter->csb->guest_csb_on &&
962 !(adapter->csb->guest_need_txkick & 1)) {
963 adapter->csb->guest_need_txkick = 1;
964 adapter->guest_need_kick_count++;
965 // XXX memory barrier
966 lem_txeof(adapter); // XXX possibly clear IFF_DRV_OACTIVE
967 }
968#endif /* NIC_PARAVIRT */
969
970 return;
971}
972
973static void
974lem_start(if_t ifp)
975{
976 struct adapter *adapter = if_getsoftc(ifp);
977
978 EM_TX_LOCK(adapter);
979 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
980 lem_start_locked(ifp);
981 EM_TX_UNLOCK(adapter);
982}
983
984/*********************************************************************
985 * Ioctl entry point
986 *
987 * em_ioctl is called when the user wants to configure the
988 * interface.
989 *
990 * return 0 on success, positive on failure
991 **********************************************************************/
992
993static int
994lem_ioctl(if_t ifp, u_long command, caddr_t data)
995{
996 struct adapter *adapter = if_getsoftc(ifp);
997 struct ifreq *ifr = (struct ifreq *)data;
998#if defined(INET) || defined(INET6)
999 struct ifaddr *ifa = (struct ifaddr *)data;
1000#endif
1001 bool avoid_reset = FALSE;
1002 int error = 0;
1003
1004 if (adapter->in_detach)
1005 return (error);
1006
1007 switch (command) {
1008 case SIOCSIFADDR:
1009#ifdef INET
1010 if (ifa->ifa_addr->sa_family == AF_INET)
1011 avoid_reset = TRUE;
1012#endif
1013#ifdef INET6
1014 if (ifa->ifa_addr->sa_family == AF_INET6)
1015 avoid_reset = TRUE;
1016#endif
1017 /*
1018 ** Calling init results in link renegotiation,
1019 ** so we avoid doing it when possible.
1020 */
1021 if (avoid_reset) {
1022 if_setflagbits(ifp, IFF_UP, 0);
1023 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1024 lem_init(adapter);
1025#ifdef INET
1026 if (!(if_getflags(ifp) & IFF_NOARP))
1027 arp_ifinit(ifp, ifa);
1028#endif
1029 } else
1030 error = ether_ioctl(ifp, command, data);
1031 break;
1032 case SIOCSIFMTU:
1033 {
1034 int max_frame_size;
1035
1036 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1037
1038 EM_CORE_LOCK(adapter);
1039 switch (adapter->hw.mac.type) {
1040 case e1000_82542:
1041 max_frame_size = ETHER_MAX_LEN;
1042 break;
1043 default:
1044 max_frame_size = MAX_JUMBO_FRAME_SIZE;
1045 }
1046 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
1047 ETHER_CRC_LEN) {
1048 EM_CORE_UNLOCK(adapter);
1049 error = EINVAL;
1050 break;
1051 }
1052
1053 if_setmtu(ifp, ifr->ifr_mtu);
1054 adapter->max_frame_size =
1055 if_getmtu(ifp) + ETHER_HDR_LEN + ETHER_CRC_LEN;
1056 lem_init_locked(adapter);
1057 EM_CORE_UNLOCK(adapter);
1058 break;
1059 }
1060 case SIOCSIFFLAGS:
1061 IOCTL_DEBUGOUT("ioctl rcv'd:\
1062 SIOCSIFFLAGS (Set Interface Flags)");
1063 EM_CORE_LOCK(adapter);
1064 if (if_getflags(ifp) & IFF_UP) {
1065 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING)) {
1066 if ((if_getflags(ifp) ^ adapter->if_flags) &
1067 (IFF_PROMISC | IFF_ALLMULTI)) {
1068 lem_disable_promisc(adapter);
1069 lem_set_promisc(adapter);
1070 }
1071 } else
1072 lem_init_locked(adapter);
1073 } else
1074 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1075 EM_TX_LOCK(adapter);
1076 lem_stop(adapter);
1077 EM_TX_UNLOCK(adapter);
1078 }
1079 adapter->if_flags = if_getflags(ifp);
1080 EM_CORE_UNLOCK(adapter);
1081 break;
1082 case SIOCADDMULTI:
1083 case SIOCDELMULTI:
1084 IOCTL_DEBUGOUT("ioctl rcv'd: SIOC(ADD|DEL)MULTI");
1085 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1086 EM_CORE_LOCK(adapter);
1087 lem_disable_intr(adapter);
1088 lem_set_multi(adapter);
1089 if (adapter->hw.mac.type == e1000_82542 &&
1090 adapter->hw.revision_id == E1000_REVISION_2) {
1091 lem_initialize_receive_unit(adapter);
1092 }
1093#ifdef DEVICE_POLLING
1094 if (!(if_getcapenable(ifp) & IFCAP_POLLING))
1095#endif
1096 lem_enable_intr(adapter);
1097 EM_CORE_UNLOCK(adapter);
1098 }
1099 break;
1100 case SIOCSIFMEDIA:
1101 /* Check SOL/IDER usage */
1102 EM_CORE_LOCK(adapter);
1103 if (e1000_check_reset_block(&adapter->hw)) {
1104 EM_CORE_UNLOCK(adapter);
1105 device_printf(adapter->dev, "Media change is"
1106 " blocked due to SOL/IDER session.\n");
1107 break;
1108 }
1109 EM_CORE_UNLOCK(adapter);
1110 case SIOCGIFMEDIA:
1111 IOCTL_DEBUGOUT("ioctl rcv'd: \
1112 SIOCxIFMEDIA (Get/Set Interface Media)");
1113 error = ifmedia_ioctl(ifp, ifr, &adapter->media, command);
1114 break;
1115 case SIOCSIFCAP:
1116 {
1117 int mask, reinit;
1118
1119 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFCAP (Set Capabilities)");
1120 reinit = 0;
1121 mask = ifr->ifr_reqcap ^ if_getcapenable(ifp);
1122#ifdef DEVICE_POLLING
1123 if (mask & IFCAP_POLLING) {
1124 if (ifr->ifr_reqcap & IFCAP_POLLING) {
1125 error = ether_poll_register(lem_poll, ifp);
1126 if (error)
1127 return (error);
1128 EM_CORE_LOCK(adapter);
1129 lem_disable_intr(adapter);
1130 if_setcapenablebit(ifp, IFCAP_POLLING, 0);
1131 EM_CORE_UNLOCK(adapter);
1132 } else {
1133 error = ether_poll_deregister(ifp);
1134 /* Enable interrupt even in error case */
1135 EM_CORE_LOCK(adapter);
1136 lem_enable_intr(adapter);
1137 if_setcapenablebit(ifp, 0, IFCAP_POLLING);
1138 EM_CORE_UNLOCK(adapter);
1139 }
1140 }
1141#endif
1142 if (mask & IFCAP_HWCSUM) {
1143 if_togglecapenable(ifp, IFCAP_HWCSUM);
1144 reinit = 1;
1145 }
1146 if (mask & IFCAP_VLAN_HWTAGGING) {
1147 if_togglecapenable(ifp, IFCAP_VLAN_HWTAGGING);
1148 reinit = 1;
1149 }
1150 if ((mask & IFCAP_WOL) &&
1151 (if_getcapabilities(ifp) & IFCAP_WOL) != 0) {
1152 if (mask & IFCAP_WOL_MCAST)
1153 if_togglecapenable(ifp, IFCAP_WOL_MCAST);
1154 if (mask & IFCAP_WOL_MAGIC)
1155 if_togglecapenable(ifp, IFCAP_WOL_MAGIC);
1156 }
1157 if (reinit && (if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1158 lem_init(adapter);
1159 if_vlancap(ifp);
1160 break;
1161 }
1162
1163 default:
1164 error = ether_ioctl(ifp, command, data);
1165 break;
1166 }
1167
1168 return (error);
1169}
1170
1171
1172/*********************************************************************
1173 * Init entry point
1174 *
1175 * This routine is used in two ways. It is used by the stack as
1176 * init entry point in network interface structure. It is also used
1177 * by the driver as a hw/sw initialization routine to get to a
1178 * consistent state.
1179 *
1180 * return 0 on success, positive on failure
1181 **********************************************************************/
1182
1183static void
1184lem_init_locked(struct adapter *adapter)
1185{
1186 if_t ifp = adapter->ifp;
1187 device_t dev = adapter->dev;
1188 u32 pba;
1189
1190 INIT_DEBUGOUT("lem_init: begin");
1191
1192 EM_CORE_LOCK_ASSERT(adapter);
1193
1194 EM_TX_LOCK(adapter);
1195 lem_stop(adapter);
1196 EM_TX_UNLOCK(adapter);
1197
1198 /*
1199 * Packet Buffer Allocation (PBA)
1200 * Writing PBA sets the receive portion of the buffer
1201 * the remainder is used for the transmit buffer.
1202 *
1203 * Devices before the 82547 had a Packet Buffer of 64K.
1204 * Default allocation: PBA=48K for Rx, leaving 16K for Tx.
1205 * After the 82547 the buffer was reduced to 40K.
1206 * Default allocation: PBA=30K for Rx, leaving 10K for Tx.
1207 * Note: default does not leave enough room for Jumbo Frame >10k.
1208 */
1209 switch (adapter->hw.mac.type) {
1210 case e1000_82547:
1211 case e1000_82547_rev_2: /* 82547: Total Packet Buffer is 40K */
1212 if (adapter->max_frame_size > 8192)
1213 pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
1214 else
1215 pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
1216 adapter->tx_fifo_head = 0;
1217 adapter->tx_head_addr = pba << EM_TX_HEAD_ADDR_SHIFT;
1218 adapter->tx_fifo_size =
1219 (E1000_PBA_40K - pba) << EM_PBA_BYTES_SHIFT;
1220 break;
1221 default:
1222 /* Devices before 82547 had a Packet Buffer of 64K. */
1223 if (adapter->max_frame_size > 8192)
1224 pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
1225 else
1226 pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
1227 }
1228
1229 INIT_DEBUGOUT1("lem_init: pba=%dK",pba);
1230 E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
1231
1232 /* Get the latest mac address, User can use a LAA */
1233 bcopy(if_getlladdr(adapter->ifp), adapter->hw.mac.addr,
1234 ETHER_ADDR_LEN);
1235
1236 /* Put the address into the Receive Address Array */
1237 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1238
1239 /* Initialize the hardware */
1240 if (lem_hardware_init(adapter)) {
1241 device_printf(dev, "Unable to initialize the hardware\n");
1242 return;
1243 }
1244 lem_update_link_status(adapter);
1245
1246 /* Setup VLAN support, basic and offload if available */
1247 E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1248
1249 /* Set hardware offload abilities */
1250 if_clearhwassist(ifp);
1251 if (adapter->hw.mac.type >= e1000_82543) {
1252 if (if_getcapenable(ifp) & IFCAP_TXCSUM)
1253 if_sethwassistbits(ifp, CSUM_TCP | CSUM_UDP, 0);
1254 }
1255
1256 /* Configure for OS presence */
1257 lem_init_manageability(adapter);
1258
1259 /* Prepare transmit descriptors and buffers */
1260 lem_setup_transmit_structures(adapter);
1261 lem_initialize_transmit_unit(adapter);
1262
1263 /* Setup Multicast table */
1264 lem_set_multi(adapter);
1265
1266 /* Prepare receive descriptors and buffers */
1267 if (lem_setup_receive_structures(adapter)) {
1268 device_printf(dev, "Could not setup receive structures\n");
1269 EM_TX_LOCK(adapter);
1270 lem_stop(adapter);
1271 EM_TX_UNLOCK(adapter);
1272 return;
1273 }
1274 lem_initialize_receive_unit(adapter);
1275
1276 /* Use real VLAN Filter support? */
1277 if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1278 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1279 /* Use real VLAN Filter support */
1280 lem_setup_vlan_hw_support(adapter);
1281 else {
1282 u32 ctrl;
1283 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1284 ctrl |= E1000_CTRL_VME;
1285 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1286 }
1287 }
1288
1289 /* Don't lose promiscuous settings */
1290 lem_set_promisc(adapter);
1291
1292 if_setdrvflagbits(ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
1293
1294 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
1295 e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1296
1297#ifdef DEVICE_POLLING
1298 /*
1299 * Only enable interrupts if we are not polling, make sure
1300 * they are off otherwise.
1301 */
1302 if (if_getcapenable(ifp) & IFCAP_POLLING)
1303 lem_disable_intr(adapter);
1304 else
1305#endif /* DEVICE_POLLING */
1306 lem_enable_intr(adapter);
1307
1308 /* AMT based hardware can now take control from firmware */
1309 if (adapter->has_manage && adapter->has_amt)
1310 lem_get_hw_control(adapter);
1311}
1312
1313static void
1314lem_init(void *arg)
1315{
1316 struct adapter *adapter = arg;
1317
1318 EM_CORE_LOCK(adapter);
1319 lem_init_locked(adapter);
1320 EM_CORE_UNLOCK(adapter);
1321}
1322
1323
1324#ifdef DEVICE_POLLING
1325/*********************************************************************
1326 *
1327 * Legacy polling routine
1328 *
1329 *********************************************************************/
1330static int
1331lem_poll(if_t ifp, enum poll_cmd cmd, int count)
1332{
1333 struct adapter *adapter = if_getsoftc(ifp);
1334 u32 reg_icr, rx_done = 0;
1335
1336 EM_CORE_LOCK(adapter);
1337 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0) {
1338 EM_CORE_UNLOCK(adapter);
1339 return (rx_done);
1340 }
1341
1342 if (cmd == POLL_AND_CHECK_STATUS) {
1343 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1344 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1345 callout_stop(&adapter->timer);
1346 adapter->hw.mac.get_link_status = 1;
1347 lem_update_link_status(adapter);
1348 callout_reset(&adapter->timer, hz,
1349 lem_local_timer, adapter);
1350 }
1351 }
1352 EM_CORE_UNLOCK(adapter);
1353
1354 lem_rxeof(adapter, count, &rx_done);
1355
1356 EM_TX_LOCK(adapter);
1357 lem_txeof(adapter);
1358 if(!if_sendq_empty(ifp))
1359 lem_start_locked(ifp);
1360 EM_TX_UNLOCK(adapter);
1361 return (rx_done);
1362}
1363#endif /* DEVICE_POLLING */
1364
1365/*********************************************************************
1366 *
1367 * Legacy Interrupt Service routine
1368 *
1369 *********************************************************************/
1370static void
1371lem_intr(void *arg)
1372{
1373 struct adapter *adapter = arg;
1374 if_t ifp = adapter->ifp;
1375 u32 reg_icr;
1376
1377
1378 if ((if_getcapenable(ifp) & IFCAP_POLLING) ||
1379 ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0))
1380 return;
1381
1382 EM_CORE_LOCK(adapter);
1383 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1384 if (reg_icr & E1000_ICR_RXO)
1385 adapter->rx_overruns++;
1386
1387 if ((reg_icr == 0xffffffff) || (reg_icr == 0)) {
1388 EM_CORE_UNLOCK(adapter);
1389 return;
1390 }
1391
1392 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1393 callout_stop(&adapter->timer);
1394 adapter->hw.mac.get_link_status = 1;
1395 lem_update_link_status(adapter);
1396 /* Deal with TX cruft when link lost */
1397 lem_tx_purge(adapter);
1398 callout_reset(&adapter->timer, hz,
1399 lem_local_timer, adapter);
1400 EM_CORE_UNLOCK(adapter);
1401 return;
1402 }
1403
1404 EM_CORE_UNLOCK(adapter);
1405 lem_rxeof(adapter, -1, NULL);
1406
1407 EM_TX_LOCK(adapter);
1408 lem_txeof(adapter);
1409 if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) &&
1410 (!if_sendq_empty(ifp)))
1411 lem_start_locked(ifp);
1412 EM_TX_UNLOCK(adapter);
1413 return;
1414}
1415
1416
1417static void
1418lem_handle_link(void *context, int pending)
1419{
1420 struct adapter *adapter = context;
1421 if_t ifp = adapter->ifp;
1422
1423 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
1424 return;
1425
1426 EM_CORE_LOCK(adapter);
1427 callout_stop(&adapter->timer);
1428 lem_update_link_status(adapter);
1429 /* Deal with TX cruft when link lost */
1430 lem_tx_purge(adapter);
1431 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
1432 EM_CORE_UNLOCK(adapter);
1433}
1434
1435
1436/* Combined RX/TX handler, used by Legacy and MSI */
1437static void
1438lem_handle_rxtx(void *context, int pending)
1439{
1440 struct adapter *adapter = context;
1441 if_t ifp = adapter->ifp;
1442
1443
1444 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
1445 bool more = lem_rxeof(adapter, adapter->rx_process_limit, NULL);
1446 EM_TX_LOCK(adapter);
1447 lem_txeof(adapter);
1448 if(!if_sendq_empty(ifp))
1449 lem_start_locked(ifp);
1450 EM_TX_UNLOCK(adapter);
1451 if (more) {
1452 taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
1453 return;
1454 }
1455 }
1456
1457 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
1458 lem_enable_intr(adapter);
1459}
1460
1461/*********************************************************************
1462 *
1463 * Fast Legacy/MSI Combined Interrupt Service routine
1464 *
1465 *********************************************************************/
1466static int
1467lem_irq_fast(void *arg)
1468{
1469 struct adapter *adapter = arg;
1470 if_t ifp;
1471 u32 reg_icr;
1472
1473 ifp = adapter->ifp;
1474
1475 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1476
1477 /* Hot eject? */
1478 if (reg_icr == 0xffffffff)
1479 return FILTER_STRAY;
1480
1481 /* Definitely not our interrupt. */
1482 if (reg_icr == 0x0)
1483 return FILTER_STRAY;
1484
1485 /*
1486 * Mask interrupts until the taskqueue is finished running. This is
1487 * cheap, just assume that it is needed. This also works around the
1488 * MSI message reordering errata on certain systems.
1489 */
1490 lem_disable_intr(adapter);
1491 taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
1492
1493 /* Link status change */
1494 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1495 adapter->hw.mac.get_link_status = 1;
1496 taskqueue_enqueue(taskqueue_fast, &adapter->link_task);
1497 }
1498
1499 if (reg_icr & E1000_ICR_RXO)
1500 adapter->rx_overruns++;
1501 return FILTER_HANDLED;
1502}
1503
1504
1505/*********************************************************************
1506 *
1507 * Media Ioctl callback
1508 *
1509 * This routine is called whenever the user queries the status of
1510 * the interface using ifconfig.
1511 *
1512 **********************************************************************/
1513static void
1514lem_media_status(if_t ifp, struct ifmediareq *ifmr)
1515{
1516 struct adapter *adapter = if_getsoftc(ifp);
1517 u_char fiber_type = IFM_1000_SX;
1518
1519 INIT_DEBUGOUT("lem_media_status: begin");
1520
1521 EM_CORE_LOCK(adapter);
1522 lem_update_link_status(adapter);
1523
1524 ifmr->ifm_status = IFM_AVALID;
1525 ifmr->ifm_active = IFM_ETHER;
1526
1527 if (!adapter->link_active) {
1528 EM_CORE_UNLOCK(adapter);
1529 return;
1530 }
1531
1532 ifmr->ifm_status |= IFM_ACTIVE;
1533
1534 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1535 (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1536 if (adapter->hw.mac.type == e1000_82545)
1537 fiber_type = IFM_1000_LX;
1538 ifmr->ifm_active |= fiber_type | IFM_FDX;
1539 } else {
1540 switch (adapter->link_speed) {
1541 case 10:
1542 ifmr->ifm_active |= IFM_10_T;
1543 break;
1544 case 100:
1545 ifmr->ifm_active |= IFM_100_TX;
1546 break;
1547 case 1000:
1548 ifmr->ifm_active |= IFM_1000_T;
1549 break;
1550 }
1551 if (adapter->link_duplex == FULL_DUPLEX)
1552 ifmr->ifm_active |= IFM_FDX;
1553 else
1554 ifmr->ifm_active |= IFM_HDX;
1555 }
1556 EM_CORE_UNLOCK(adapter);
1557}
1558
1559/*********************************************************************
1560 *
1561 * Media Ioctl callback
1562 *
1563 * This routine is called when the user changes speed/duplex using
1564 * media/mediopt option with ifconfig.
1565 *
1566 **********************************************************************/
1567static int
1568lem_media_change(if_t ifp)
1569{
1570 struct adapter *adapter = if_getsoftc(ifp);
1571 struct ifmedia *ifm = &adapter->media;
1572
1573 INIT_DEBUGOUT("lem_media_change: begin");
1574
1575 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1576 return (EINVAL);
1577
1578 EM_CORE_LOCK(adapter);
1579 switch (IFM_SUBTYPE(ifm->ifm_media)) {
1580 case IFM_AUTO:
1581 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1582 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1583 break;
1584 case IFM_1000_LX:
1585 case IFM_1000_SX:
1586 case IFM_1000_T:
1587 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1588 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1589 break;
1590 case IFM_100_TX:
1591 adapter->hw.mac.autoneg = FALSE;
1592 adapter->hw.phy.autoneg_advertised = 0;
1593 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1594 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1595 else
1596 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1597 break;
1598 case IFM_10_T:
1599 adapter->hw.mac.autoneg = FALSE;
1600 adapter->hw.phy.autoneg_advertised = 0;
1601 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1602 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1603 else
1604 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1605 break;
1606 default:
1607 device_printf(adapter->dev, "Unsupported media type\n");
1608 }
1609
1610 lem_init_locked(adapter);
1611 EM_CORE_UNLOCK(adapter);
1612
1613 return (0);
1614}
1615
1616/*********************************************************************
1617 *
1618 * This routine maps the mbufs to tx descriptors.
1619 *
1620 * return 0 on success, positive on failure
1621 **********************************************************************/
1622
1623static int
1624lem_xmit(struct adapter *adapter, struct mbuf **m_headp)
1625{
1626 bus_dma_segment_t segs[EM_MAX_SCATTER];
1627 bus_dmamap_t map;
1628 struct em_buffer *tx_buffer, *tx_buffer_mapped;
1629 struct e1000_tx_desc *ctxd = NULL;
1630 struct mbuf *m_head;
1631 u32 txd_upper, txd_lower, txd_used, txd_saved;
1632 int error, nsegs, i, j, first, last = 0;
1633
1634 m_head = *m_headp;
1635 txd_upper = txd_lower = txd_used = txd_saved = 0;
1636
1637 /*
1638 ** When doing checksum offload, it is critical to
1639 ** make sure the first mbuf has more than header,
1640 ** because that routine expects data to be present.
1641 */
1642 if ((m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD) &&
1643 (m_head->m_len < ETHER_HDR_LEN + sizeof(struct ip))) {
1644 m_head = m_pullup(m_head, ETHER_HDR_LEN + sizeof(struct ip));
1645 *m_headp = m_head;
1646 if (m_head == NULL)
1647 return (ENOBUFS);
1648 }
1649
1650 /*
1651 * Map the packet for DMA
1652 *
1653 * Capture the first descriptor index,
1654 * this descriptor will have the index
1655 * of the EOP which is the only one that
1656 * now gets a DONE bit writeback.
1657 */
1658 first = adapter->next_avail_tx_desc;
1659 tx_buffer = &adapter->tx_buffer_area[first];
1660 tx_buffer_mapped = tx_buffer;
1661 map = tx_buffer->map;
1662
1663 error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
1664 *m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
1665
1666 /*
1667 * There are two types of errors we can (try) to handle:
1668 * - EFBIG means the mbuf chain was too long and bus_dma ran
1669 * out of segments. Defragment the mbuf chain and try again.
1670 * - ENOMEM means bus_dma could not obtain enough bounce buffers
1671 * at this point in time. Defer sending and try again later.
1672 * All other errors, in particular EINVAL, are fatal and prevent the
1673 * mbuf chain from ever going through. Drop it and report error.
1674 */
1675 if (error == EFBIG) {
1676 struct mbuf *m;
1677
1678 m = m_defrag(*m_headp, M_NOWAIT);
1679 if (m == NULL) {
1680 adapter->mbuf_alloc_failed++;
1681 m_freem(*m_headp);
1682 *m_headp = NULL;
1683 return (ENOBUFS);
1684 }
1685 *m_headp = m;
1686
1687 /* Try it again */
1688 error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
1689 *m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
1690
1691 if (error) {
1692 adapter->no_tx_dma_setup++;
1693 m_freem(*m_headp);
1694 *m_headp = NULL;
1695 return (error);
1696 }
1697 } else if (error != 0) {
1698 adapter->no_tx_dma_setup++;
1699 return (error);
1700 }
1701
1702 if (nsegs > (adapter->num_tx_desc_avail - 2)) {
1703 adapter->no_tx_desc_avail2++;
1704 bus_dmamap_unload(adapter->txtag, map);
1705 return (ENOBUFS);
1706 }
1707 m_head = *m_headp;
1708
1709 /* Do hardware assists */
1710 if (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)
1711 lem_transmit_checksum_setup(adapter, m_head,
1712 &txd_upper, &txd_lower);
1713
1714 i = adapter->next_avail_tx_desc;
1715 if (adapter->pcix_82544)
1716 txd_saved = i;
1717
1718 /* Set up our transmit descriptors */
1719 for (j = 0; j < nsegs; j++) {
1720 bus_size_t seg_len;
1721 bus_addr_t seg_addr;
1722 /* If adapter is 82544 and on PCIX bus */
1723 if(adapter->pcix_82544) {
1724 DESC_ARRAY desc_array;
1725 u32 array_elements, counter;
1726 /*
1727 * Check the Address and Length combination and
1728 * split the data accordingly
1729 */
1730 array_elements = lem_fill_descriptors(segs[j].ds_addr,
1731 segs[j].ds_len, &desc_array);
1732 for (counter = 0; counter < array_elements; counter++) {
1733 if (txd_used == adapter->num_tx_desc_avail) {
1734 adapter->next_avail_tx_desc = txd_saved;
1735 adapter->no_tx_desc_avail2++;
1736 bus_dmamap_unload(adapter->txtag, map);
1737 return (ENOBUFS);
1738 }
1739 tx_buffer = &adapter->tx_buffer_area[i];
1740 ctxd = &adapter->tx_desc_base[i];
1741 ctxd->buffer_addr = htole64(
1742 desc_array.descriptor[counter].address);
1743 ctxd->lower.data = htole32(
1744 (adapter->txd_cmd | txd_lower | (u16)
1745 desc_array.descriptor[counter].length));
1746 ctxd->upper.data =
1747 htole32((txd_upper));
1748 last = i;
1749 if (++i == adapter->num_tx_desc)
1750 i = 0;
1751 tx_buffer->m_head = NULL;
1752 tx_buffer->next_eop = -1;
1753 txd_used++;
1754 }
1755 } else {
1756 tx_buffer = &adapter->tx_buffer_area[i];
1757 ctxd = &adapter->tx_desc_base[i];
1758 seg_addr = segs[j].ds_addr;
1759 seg_len = segs[j].ds_len;
1760 ctxd->buffer_addr = htole64(seg_addr);
1761 ctxd->lower.data = htole32(
1762 adapter->txd_cmd | txd_lower | seg_len);
1763 ctxd->upper.data =
1764 htole32(txd_upper);
1765 last = i;
1766 if (++i == adapter->num_tx_desc)
1767 i = 0;
1768 tx_buffer->m_head = NULL;
1769 tx_buffer->next_eop = -1;
1770 }
1771 }
1772
1773 adapter->next_avail_tx_desc = i;
1774
1775 if (adapter->pcix_82544)
1776 adapter->num_tx_desc_avail -= txd_used;
1777 else
1778 adapter->num_tx_desc_avail -= nsegs;
1779
1780 if (m_head->m_flags & M_VLANTAG) {
1781 /* Set the vlan id. */
1782 ctxd->upper.fields.special =
1783 htole16(m_head->m_pkthdr.ether_vtag);
1784 /* Tell hardware to add tag */
1785 ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE);
1786 }
1787
1788 tx_buffer->m_head = m_head;
1789 tx_buffer_mapped->map = tx_buffer->map;
1790 tx_buffer->map = map;
1791 bus_dmamap_sync(adapter->txtag, map, BUS_DMASYNC_PREWRITE);
1792
1793 /*
1794 * Last Descriptor of Packet
1795 * needs End Of Packet (EOP)
1796 * and Report Status (RS)
1797 */
1798 ctxd->lower.data |=
1799 htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
1800 /*
1801 * Keep track in the first buffer which
1802 * descriptor will be written back
1803 */
1804 tx_buffer = &adapter->tx_buffer_area[first];
1805 tx_buffer->next_eop = last;
1806 adapter->watchdog_time = ticks;
1807
1808 /*
1809 * Advance the Transmit Descriptor Tail (TDT), this tells the E1000
1810 * that this frame is available to transmit.
1811 */
1812 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
1813 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1814
1815#ifdef NIC_PARAVIRT
1816 if (adapter->csb) {
1817 adapter->csb->guest_tdt = i;
1818 /* XXX memory barrier ? */
1819 if (adapter->csb->guest_csb_on &&
1820 !(adapter->csb->host_need_txkick & 1)) {
1821 /* XXX maybe useless
1822 * clean the ring. maybe do it before ?
1823 * maybe a little bit of histeresys ?
1824 */
1825 if (adapter->num_tx_desc_avail <= 64) {// XXX
1826 lem_txeof(adapter);
1827 }
1828 return (0);
1829 }
1830 }
1831#endif /* NIC_PARAVIRT */
1832
1833#ifdef NIC_SEND_COMBINING
1834 if (adapter->sc_enable) {
1835 if (adapter->shadow_tdt & MIT_PENDING_INT) {
1836 /* signal intr and data pending */
1837 adapter->shadow_tdt = MIT_PENDING_TDT | (i & 0xffff);
1838 return (0);
1839 } else {
1840 adapter->shadow_tdt = MIT_PENDING_INT;
1841 }
1842 }
1843#endif /* NIC_SEND_COMBINING */
1844
1845 if (adapter->hw.mac.type == e1000_82547 &&
1846 adapter->link_duplex == HALF_DUPLEX)
1847 lem_82547_move_tail(adapter);
1848 else {
1849 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), i);
1850 if (adapter->hw.mac.type == e1000_82547)
1851 lem_82547_update_fifo_head(adapter,
1852 m_head->m_pkthdr.len);
1853 }
1854
1855 return (0);
1856}
1857
1858/*********************************************************************
1859 *
1860 * 82547 workaround to avoid controller hang in half-duplex environment.
1861 * The workaround is to avoid queuing a large packet that would span
1862 * the internal Tx FIFO ring boundary. We need to reset the FIFO pointers
1863 * in this case. We do that only when FIFO is quiescent.
1864 *
1865 **********************************************************************/
1866static void
1867lem_82547_move_tail(void *arg)
1868{
1869 struct adapter *adapter = arg;
1870 struct e1000_tx_desc *tx_desc;
1871 u16 hw_tdt, sw_tdt, length = 0;
1872 bool eop = 0;
1873
1874 EM_TX_LOCK_ASSERT(adapter);
1875
1876 hw_tdt = E1000_READ_REG(&adapter->hw, E1000_TDT(0));
1877 sw_tdt = adapter->next_avail_tx_desc;
1878
1879 while (hw_tdt != sw_tdt) {
1880 tx_desc = &adapter->tx_desc_base[hw_tdt];
1881 length += tx_desc->lower.flags.length;
1882 eop = tx_desc->lower.data & E1000_TXD_CMD_EOP;
1883 if (++hw_tdt == adapter->num_tx_desc)
1884 hw_tdt = 0;
1885
1886 if (eop) {
1887 if (lem_82547_fifo_workaround(adapter, length)) {
1888 adapter->tx_fifo_wrk_cnt++;
1889 callout_reset(&adapter->tx_fifo_timer, 1,
1890 lem_82547_move_tail, adapter);
1891 break;
1892 }
1893 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), hw_tdt);
1894 lem_82547_update_fifo_head(adapter, length);
1895 length = 0;
1896 }
1897 }
1898}
1899
1900static int
1901lem_82547_fifo_workaround(struct adapter *adapter, int len)
1902{
1903 int fifo_space, fifo_pkt_len;
1904
1905 fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1906
1907 if (adapter->link_duplex == HALF_DUPLEX) {
1908 fifo_space = adapter->tx_fifo_size - adapter->tx_fifo_head;
1909
1910 if (fifo_pkt_len >= (EM_82547_PKT_THRESH + fifo_space)) {
1911 if (lem_82547_tx_fifo_reset(adapter))
1912 return (0);
1913 else
1914 return (1);
1915 }
1916 }
1917
1918 return (0);
1919}
1920
1921static void
1922lem_82547_update_fifo_head(struct adapter *adapter, int len)
1923{
1924 int fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1925
1926 /* tx_fifo_head is always 16 byte aligned */
1927 adapter->tx_fifo_head += fifo_pkt_len;
1928 if (adapter->tx_fifo_head >= adapter->tx_fifo_size) {
1929 adapter->tx_fifo_head -= adapter->tx_fifo_size;
1930 }
1931}
1932
1933
1934static int
1935lem_82547_tx_fifo_reset(struct adapter *adapter)
1936{
1937 u32 tctl;
1938
1939 if ((E1000_READ_REG(&adapter->hw, E1000_TDT(0)) ==
1940 E1000_READ_REG(&adapter->hw, E1000_TDH(0))) &&
1941 (E1000_READ_REG(&adapter->hw, E1000_TDFT) ==
1942 E1000_READ_REG(&adapter->hw, E1000_TDFH)) &&
1943 (E1000_READ_REG(&adapter->hw, E1000_TDFTS) ==
1944 E1000_READ_REG(&adapter->hw, E1000_TDFHS)) &&
1945 (E1000_READ_REG(&adapter->hw, E1000_TDFPC) == 0)) {
1946 /* Disable TX unit */
1947 tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
1948 E1000_WRITE_REG(&adapter->hw, E1000_TCTL,
1949 tctl & ~E1000_TCTL_EN);
1950
1951 /* Reset FIFO pointers */
1952 E1000_WRITE_REG(&adapter->hw, E1000_TDFT,
1953 adapter->tx_head_addr);
1954 E1000_WRITE_REG(&adapter->hw, E1000_TDFH,
1955 adapter->tx_head_addr);
1956 E1000_WRITE_REG(&adapter->hw, E1000_TDFTS,
1957 adapter->tx_head_addr);
1958 E1000_WRITE_REG(&adapter->hw, E1000_TDFHS,
1959 adapter->tx_head_addr);
1960
1961 /* Re-enable TX unit */
1962 E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
1963 E1000_WRITE_FLUSH(&adapter->hw);
1964
1965 adapter->tx_fifo_head = 0;
1966 adapter->tx_fifo_reset_cnt++;
1967
1968 return (TRUE);
1969 }
1970 else {
1971 return (FALSE);
1972 }
1973}
1974
1975static void
1976lem_set_promisc(struct adapter *adapter)
1977{
1978 if_t ifp = adapter->ifp;
1979 u32 reg_rctl;
1980
1981 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1982
1983 if (if_getflags(ifp) & IFF_PROMISC) {
1984 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1985 /* Turn this on if you want to see bad packets */
1986 if (lem_debug_sbp)
1987 reg_rctl |= E1000_RCTL_SBP;
1988 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1989 } else if (if_getflags(ifp) & IFF_ALLMULTI) {
1990 reg_rctl |= E1000_RCTL_MPE;
1991 reg_rctl &= ~E1000_RCTL_UPE;
1992 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1993 }
1994}
1995
1996static void
1997lem_disable_promisc(struct adapter *adapter)
1998{
1999 if_t ifp = adapter->ifp;
2000 u32 reg_rctl;
2001 int mcnt = 0;
2002
2003 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2004 reg_rctl &= (~E1000_RCTL_UPE);
2005 if (if_getflags(ifp) & IFF_ALLMULTI)
2006 mcnt = MAX_NUM_MULTICAST_ADDRESSES;
2007 else
2008 mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
2009
2010 /* Don't disable if in MAX groups */
2011 if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
2012 reg_rctl &= (~E1000_RCTL_MPE);
2013 reg_rctl &= (~E1000_RCTL_SBP);
2014 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2015}
2016
2017
2018/*********************************************************************
2019 * Multicast Update
2020 *
2021 * This routine is called whenever multicast address list is updated.
2022 *
2023 **********************************************************************/
2024
2025static void
2026lem_set_multi(struct adapter *adapter)
2027{
2028 if_t ifp = adapter->ifp;
2029 u32 reg_rctl = 0;
2030 u8 *mta; /* Multicast array memory */
2031 int mcnt = 0;
2032
2033 IOCTL_DEBUGOUT("lem_set_multi: begin");
2034
2035 mta = adapter->mta;
2036 bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
2037
2038 if (adapter->hw.mac.type == e1000_82542 &&
2039 adapter->hw.revision_id == E1000_REVISION_2) {
2040 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2041 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
2042 e1000_pci_clear_mwi(&adapter->hw);
2043 reg_rctl |= E1000_RCTL_RST;
2044 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2045 msec_delay(5);
2046 }
2047
2048 if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
2049
2050 if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
2051 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2052 reg_rctl |= E1000_RCTL_MPE;
2053 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2054 } else
2055 e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
2056
2057 if (adapter->hw.mac.type == e1000_82542 &&
2058 adapter->hw.revision_id == E1000_REVISION_2) {
2059 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
2060 reg_rctl &= ~E1000_RCTL_RST;
2061 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
2062 msec_delay(5);
2063 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
2064 e1000_pci_set_mwi(&adapter->hw);
2065 }
2066}
2067
2068
2069/*********************************************************************
2070 * Timer routine
2071 *
2072 * This routine checks for link status and updates statistics.
2073 *
2074 **********************************************************************/
2075
2076static void
2077lem_local_timer(void *arg)
2078{
2079 struct adapter *adapter = arg;
2080
2081 EM_CORE_LOCK_ASSERT(adapter);
2082
2083 lem_update_link_status(adapter);
2084 lem_update_stats_counters(adapter);
2085
2086 lem_smartspeed(adapter);
2087
2088#ifdef NIC_PARAVIRT
2089 /* recover space if needed */
2090 if (adapter->csb && adapter->csb->guest_csb_on &&
2091 (adapter->watchdog_check == TRUE) &&
2092 (ticks - adapter->watchdog_time > EM_WATCHDOG) &&
2093 (adapter->num_tx_desc_avail != adapter->num_tx_desc) ) {
2094 lem_txeof(adapter);
2095 /*
2096 * lem_txeof() normally (except when space in the queue
2097 * runs low XXX) cleans watchdog_check so that
2098 * we do not hung.
2099 */
2100 }
2101#endif /* NIC_PARAVIRT */
2102 /*
2103 * We check the watchdog: the time since
2104 * the last TX descriptor was cleaned.
2105 * This implies a functional TX engine.
2106 */
2107 if ((adapter->watchdog_check == TRUE) &&
2108 (ticks - adapter->watchdog_time > EM_WATCHDOG))
2109 goto hung;
2110
2111 callout_reset(&adapter->timer, hz, lem_local_timer, adapter);
2112 return;
2113hung:
2114 device_printf(adapter->dev, "Watchdog timeout -- resetting\n");
2115 if_setdrvflagbits(adapter->ifp, 0, IFF_DRV_RUNNING);
2116 adapter->watchdog_events++;
2117 lem_init_locked(adapter);
2118}
2119
2120static void
2121lem_update_link_status(struct adapter *adapter)
2122{
2123 struct e1000_hw *hw = &adapter->hw;
2124 if_t ifp = adapter->ifp;
2125 device_t dev = adapter->dev;
2126 u32 link_check = 0;
2127
2128 /* Get the cached link value or read phy for real */
2129 switch (hw->phy.media_type) {
2130 case e1000_media_type_copper:
2131 if (hw->mac.get_link_status) {
2132 /* Do the work to read phy */
2133 e1000_check_for_link(hw);
2134 link_check = !hw->mac.get_link_status;
2135 if (link_check) /* ESB2 fix */
2136 e1000_cfg_on_link_up(hw);
2137 } else
2138 link_check = TRUE;
2139 break;
2140 case e1000_media_type_fiber:
2141 e1000_check_for_link(hw);
2142 link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2143 E1000_STATUS_LU);
2144 break;
2145 case e1000_media_type_internal_serdes:
2146 e1000_check_for_link(hw);
2147 link_check = adapter->hw.mac.serdes_has_link;
2148 break;
2149 default:
2150 case e1000_media_type_unknown:
2151 break;
2152 }
2153
2154 /* Now check for a transition */
2155 if (link_check && (adapter->link_active == 0)) {
2156 e1000_get_speed_and_duplex(hw, &adapter->link_speed,
2157 &adapter->link_duplex);
2158 if (bootverbose)
2159 device_printf(dev, "Link is up %d Mbps %s\n",
2160 adapter->link_speed,
2161 ((adapter->link_duplex == FULL_DUPLEX) ?
2162 "Full Duplex" : "Half Duplex"));
2163 adapter->link_active = 1;
2164 adapter->smartspeed = 0;
2165 if_setbaudrate(ifp, adapter->link_speed * 1000000);
2166 if_link_state_change(ifp, LINK_STATE_UP);
2167 } else if (!link_check && (adapter->link_active == 1)) {
2168 if_setbaudrate(ifp, 0);
2169 adapter->link_speed = 0;
2170 adapter->link_duplex = 0;
2171 if (bootverbose)
2172 device_printf(dev, "Link is Down\n");
2173 adapter->link_active = 0;
2174 /* Link down, disable watchdog */
2175 adapter->watchdog_check = FALSE;
2176 if_link_state_change(ifp, LINK_STATE_DOWN);
2177 }
2178}
2179
2180/*********************************************************************
2181 *
2182 * This routine disables all traffic on the adapter by issuing a
2183 * global reset on the MAC and deallocates TX/RX buffers.
2184 *
2185 * This routine should always be called with BOTH the CORE
2186 * and TX locks.
2187 **********************************************************************/
2188
2189static void
2190lem_stop(void *arg)
2191{
2192 struct adapter *adapter = arg;
2193 if_t ifp = adapter->ifp;
2194
2195 EM_CORE_LOCK_ASSERT(adapter);
2196 EM_TX_LOCK_ASSERT(adapter);
2197
2198 INIT_DEBUGOUT("lem_stop: begin");
2199
2200 lem_disable_intr(adapter);
2201 callout_stop(&adapter->timer);
2202 callout_stop(&adapter->tx_fifo_timer);
2203
2204 /* Tell the stack that the interface is no longer active */
2205 if_setdrvflagbits(ifp, 0, (IFF_DRV_RUNNING | IFF_DRV_OACTIVE));
2206
2207 e1000_reset_hw(&adapter->hw);
2208 if (adapter->hw.mac.type >= e1000_82544)
2209 E1000_WRITE_REG(&adapter->hw, E1000_WUC, 0);
2210
2211 e1000_led_off(&adapter->hw);
2212 e1000_cleanup_led(&adapter->hw);
2213}
2214
2215
2216/*********************************************************************
2217 *
2218 * Determine hardware revision.
2219 *
2220 **********************************************************************/
2221static void
2222lem_identify_hardware(struct adapter *adapter)
2223{
2224 device_t dev = adapter->dev;
2225
2226 /* Make sure our PCI config space has the necessary stuff set */
2227 pci_enable_busmaster(dev);
2228 adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
2229
2230 /* Save off the information about this board */
2231 adapter->hw.vendor_id = pci_get_vendor(dev);
2232 adapter->hw.device_id = pci_get_device(dev);
2233 adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
2234 adapter->hw.subsystem_vendor_id =
2235 pci_read_config(dev, PCIR_SUBVEND_0, 2);
2236 adapter->hw.subsystem_device_id =
2237 pci_read_config(dev, PCIR_SUBDEV_0, 2);
2238
2239 /* Do Shared Code Init and Setup */
2240 if (e1000_set_mac_type(&adapter->hw)) {
2241 device_printf(dev, "Setup init failure\n");
2242 return;
2243 }
2244}
2245
2246static int
2247lem_allocate_pci_resources(struct adapter *adapter)
2248{
2249 device_t dev = adapter->dev;
2250 int val, rid, error = E1000_SUCCESS;
2251
2252 rid = PCIR_BAR(0);
2253 adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2254 &rid, RF_ACTIVE);
2255 if (adapter->memory == NULL) {
2256 device_printf(dev, "Unable to allocate bus resource: memory\n");
2257 return (ENXIO);
2258 }
2259 adapter->osdep.mem_bus_space_tag =
2260 rman_get_bustag(adapter->memory);
2261 adapter->osdep.mem_bus_space_handle =
2262 rman_get_bushandle(adapter->memory);
2263 adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
2264
2265 /* Only older adapters use IO mapping */
2266 if (adapter->hw.mac.type > e1000_82543) {
2267 /* Figure our where our IO BAR is ? */
2268 for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
2269 val = pci_read_config(dev, rid, 4);
2270 if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2271 adapter->io_rid = rid;
2272 break;
2273 }
2274 rid += 4;
2275 /* check for 64bit BAR */
2276 if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2277 rid += 4;
2278 }
2279 if (rid >= PCIR_CIS) {
2280 device_printf(dev, "Unable to locate IO BAR\n");
2281 return (ENXIO);
2282 }
2283 adapter->ioport = bus_alloc_resource_any(dev,
2284 SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
2285 if (adapter->ioport == NULL) {
2286 device_printf(dev, "Unable to allocate bus resource: "
2287 "ioport\n");
2288 return (ENXIO);
2289 }
2290 adapter->hw.io_base = 0;
2291 adapter->osdep.io_bus_space_tag =
2292 rman_get_bustag(adapter->ioport);
2293 adapter->osdep.io_bus_space_handle =
2294 rman_get_bushandle(adapter->ioport);
2295 }
2296
2297 adapter->hw.back = &adapter->osdep;
2298
2299 return (error);
2300}
2301
2302/*********************************************************************
2303 *
2304 * Setup the Legacy or MSI Interrupt handler
2305 *
2306 **********************************************************************/
2307int
2308lem_allocate_irq(struct adapter *adapter)
2309{
2310 device_t dev = adapter->dev;
2311 int error, rid = 0;
2312
2313 /* Manually turn off all interrupts */
2314 E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
2315
2316 /* We allocate a single interrupt resource */
2317 adapter->res[0] = bus_alloc_resource_any(dev,
2318 SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
2319 if (adapter->res[0] == NULL) {
2320 device_printf(dev, "Unable to allocate bus resource: "
2321 "interrupt\n");
2322 return (ENXIO);
2323 }
2324
2325 /* Do Legacy setup? */
2326 if (lem_use_legacy_irq) {
2327 if ((error = bus_setup_intr(dev, adapter->res[0],
2328 INTR_TYPE_NET | INTR_MPSAFE, NULL, lem_intr, adapter,
2329 &adapter->tag[0])) != 0) {
2330 device_printf(dev,
2331 "Failed to register interrupt handler");
2332 return (error);
2333 }
2334 return (0);
2335 }
2336
2337 /*
2338 * Use a Fast interrupt and the associated
2339 * deferred processing contexts.
2340 */
2341 TASK_INIT(&adapter->rxtx_task, 0, lem_handle_rxtx, adapter);
2342 TASK_INIT(&adapter->link_task, 0, lem_handle_link, adapter);
2343 adapter->tq = taskqueue_create_fast("lem_taskq", M_NOWAIT,
2344 taskqueue_thread_enqueue, &adapter->tq);
2345 taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s taskq",
2346 device_get_nameunit(adapter->dev));
2347 if ((error = bus_setup_intr(dev, adapter->res[0],
2348 INTR_TYPE_NET, lem_irq_fast, NULL, adapter,
2349 &adapter->tag[0])) != 0) {
2350 device_printf(dev, "Failed to register fast interrupt "
2351 "handler: %d\n", error);
2352 taskqueue_free(adapter->tq);
2353 adapter->tq = NULL;
2354 return (error);
2355 }
2356
2357 return (0);
2358}
2359
2360
2361static void
2362lem_free_pci_resources(struct adapter *adapter)
2363{
2364 device_t dev = adapter->dev;
2365
2366
2367 if (adapter->tag[0] != NULL) {
2368 bus_teardown_intr(dev, adapter->res[0],
2369 adapter->tag[0]);
2370 adapter->tag[0] = NULL;
2371 }
2372
2373 if (adapter->res[0] != NULL) {
2374 bus_release_resource(dev, SYS_RES_IRQ,
2375 0, adapter->res[0]);
2376 }
2377
2378 if (adapter->memory != NULL)
2379 bus_release_resource(dev, SYS_RES_MEMORY,
2380 PCIR_BAR(0), adapter->memory);
2381
2382 if (adapter->ioport != NULL)
2383 bus_release_resource(dev, SYS_RES_IOPORT,
2384 adapter->io_rid, adapter->ioport);
2385}
2386
2387
2388/*********************************************************************
2389 *
2390 * Initialize the hardware to a configuration
2391 * as specified by the adapter structure.
2392 *
2393 **********************************************************************/
2394static int
2395lem_hardware_init(struct adapter *adapter)
2396{
2397 device_t dev = adapter->dev;
2398 u16 rx_buffer_size;
2399
2400 INIT_DEBUGOUT("lem_hardware_init: begin");
2401
2402 /* Issue a global reset */
2403 e1000_reset_hw(&adapter->hw);
2404
2405 /* When hardware is reset, fifo_head is also reset */
2406 adapter->tx_fifo_head = 0;
2407
2408 /*
2409 * These parameters control the automatic generation (Tx) and
2410 * response (Rx) to Ethernet PAUSE frames.
2411 * - High water mark should allow for at least two frames to be
2412 * received after sending an XOFF.
2413 * - Low water mark works best when it is very near the high water mark.
2414 * This allows the receiver to restart by sending XON when it has
2415 * drained a bit. Here we use an arbitary value of 1500 which will
2416 * restart after one full frame is pulled from the buffer. There
2417 * could be several smaller frames in the buffer and if so they will
2418 * not trigger the XON until their total number reduces the buffer
2419 * by 1500.
2420 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2421 */
2422 rx_buffer_size = ((E1000_READ_REG(&adapter->hw, E1000_PBA) &
2423 0xffff) << 10 );
2424
2425 adapter->hw.fc.high_water = rx_buffer_size -
2426 roundup2(adapter->max_frame_size, 1024);
2427 adapter->hw.fc.low_water = adapter->hw.fc.high_water - 1500;
2428
2429 adapter->hw.fc.pause_time = EM_FC_PAUSE_TIME;
2430 adapter->hw.fc.send_xon = TRUE;
2431
2432 /* Set Flow control, use the tunable location if sane */
2433 if ((lem_fc_setting >= 0) && (lem_fc_setting < 4))
2434 adapter->hw.fc.requested_mode = lem_fc_setting;
2435 else
2436 adapter->hw.fc.requested_mode = e1000_fc_none;
2437
2438 if (e1000_init_hw(&adapter->hw) < 0) {
2439 device_printf(dev, "Hardware Initialization Failed\n");
2440 return (EIO);
2441 }
2442
2443 e1000_check_for_link(&adapter->hw);
2444
2445 return (0);
2446}
2447
2448/*********************************************************************
2449 *
2450 * Setup networking device structure and register an interface.
2451 *
2452 **********************************************************************/
2453static int
2454lem_setup_interface(device_t dev, struct adapter *adapter)
2455{
2456 if_t ifp;
2457
2458 INIT_DEBUGOUT("lem_setup_interface: begin");
2459
2460 ifp = adapter->ifp = if_gethandle(IFT_ETHER);
2461 if (ifp == (void *)NULL) {
2462 device_printf(dev, "can not allocate ifnet structure\n");
2463 return (-1);
2464 }
2465 if_initname(ifp, device_get_name(dev), device_get_unit(dev));
2466 if_setinitfn(ifp, lem_init);
2467 if_setsoftc(ifp, adapter);
2468 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
2469 if_setioctlfn(ifp, lem_ioctl);
2470 if_setstartfn(ifp, lem_start);
2471 if_setgetcounterfn(ifp, lem_get_counter);
2472 if_setsendqlen(ifp, adapter->num_tx_desc - 1);
2473 if_setsendqready(ifp);
2474
2475 ether_ifattach(ifp, adapter->hw.mac.addr);
2476
2477 if_setcapabilities(ifp, 0);
2478
2479 if (adapter->hw.mac.type >= e1000_82543) {
2480 if_setcapabilitiesbit(ifp, IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM, 0);
2481 if_setcapenablebit(ifp, IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM, 0);
2482 }
2483
2484 /*
2485 * Tell the upper layer(s) we support long frames.
2486 */
2487 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2488 if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU, 0);
2489 if_setcapenablebit(ifp, IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU, 0);
2490
2491 /*
2492 ** Dont turn this on by default, if vlans are
2493 ** created on another pseudo device (eg. lagg)
2494 ** then vlan events are not passed thru, breaking
2495 ** operation, but with HW FILTER off it works. If
2496 ** using vlans directly on the em driver you can
2497 ** enable this and get full hardware tag filtering.
2498 */
2499 if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER, 0);
2500
2501#ifdef DEVICE_POLLING
2502 if_setcapabilitiesbit(ifp, IFCAP_POLLING, 0);
2503#endif
2504
2505 /* Enable only WOL MAGIC by default */
2506 if (adapter->wol) {
2507 if_setcapabilitiesbit(ifp, IFCAP_WOL, 0);
2508 if_setcapenablebit(ifp, IFCAP_WOL_MAGIC, 0);
2509 }
2510
2511 /*
2512 * Specify the media types supported by this adapter and register
2513 * callbacks to update media and link information
2514 */
2515 ifmedia_init(&adapter->media, IFM_IMASK,
2516 lem_media_change, lem_media_status);
2517 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2518 (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2519 u_char fiber_type = IFM_1000_SX; /* default type */
2520
2521 if (adapter->hw.mac.type == e1000_82545)
2522 fiber_type = IFM_1000_LX;
2523 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type | IFM_FDX,
2524 0, NULL);
2525 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2526 } else {
2527 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2528 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX,
2529 0, NULL);
2530 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX,
2531 0, NULL);
2532 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
2533 0, NULL);
2534 if (adapter->hw.phy.type != e1000_phy_ife) {
2535 ifmedia_add(&adapter->media,
2536 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2537 ifmedia_add(&adapter->media,
2538 IFM_ETHER | IFM_1000_T, 0, NULL);
2539 }
2540 }
2541 ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2542 ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO);
2543 return (0);
2544}
2545
2546
2547/*********************************************************************
2548 *
2549 * Workaround for SmartSpeed on 82541 and 82547 controllers
2550 *
2551 **********************************************************************/
2552static void
2553lem_smartspeed(struct adapter *adapter)
2554{
2555 u16 phy_tmp;
2556
2557 if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2558 adapter->hw.mac.autoneg == 0 ||
2559 (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2560 return;
2561
2562 if (adapter->smartspeed == 0) {
2563 /* If Master/Slave config fault is asserted twice,
2564 * we assume back-to-back */
2565 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2566 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2567 return;
2568 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2569 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2570 e1000_read_phy_reg(&adapter->hw,
2571 PHY_1000T_CTRL, &phy_tmp);
2572 if(phy_tmp & CR_1000T_MS_ENABLE) {
2573 phy_tmp &= ~CR_1000T_MS_ENABLE;
2574 e1000_write_phy_reg(&adapter->hw,
2575 PHY_1000T_CTRL, phy_tmp);
2576 adapter->smartspeed++;
2577 if(adapter->hw.mac.autoneg &&
2578 !e1000_copper_link_autoneg(&adapter->hw) &&
2579 !e1000_read_phy_reg(&adapter->hw,
2580 PHY_CONTROL, &phy_tmp)) {
2581 phy_tmp |= (MII_CR_AUTO_NEG_EN |
2582 MII_CR_RESTART_AUTO_NEG);
2583 e1000_write_phy_reg(&adapter->hw,
2584 PHY_CONTROL, phy_tmp);
2585 }
2586 }
2587 }
2588 return;
2589 } else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2590 /* If still no link, perhaps using 2/3 pair cable */
2591 e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2592 phy_tmp |= CR_1000T_MS_ENABLE;
2593 e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2594 if(adapter->hw.mac.autoneg &&
2595 !e1000_copper_link_autoneg(&adapter->hw) &&
2596 !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2597 phy_tmp |= (MII_CR_AUTO_NEG_EN |
2598 MII_CR_RESTART_AUTO_NEG);
2599 e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2600 }
2601 }
2602 /* Restart process after EM_SMARTSPEED_MAX iterations */
2603 if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2604 adapter->smartspeed = 0;
2605}
2606
2607
2608/*
2609 * Manage DMA'able memory.
2610 */
2611static void
2612lem_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
2613{
2614 if (error)
2615 return;
2616 *(bus_addr_t *) arg = segs[0].ds_addr;
2617}
2618
2619static int
2620lem_dma_malloc(struct adapter *adapter, bus_size_t size,
2621 struct em_dma_alloc *dma, int mapflags)
2622{
2623 int error;
2624
2625 error = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */
2626 EM_DBA_ALIGN, 0, /* alignment, bounds */
2627 BUS_SPACE_MAXADDR, /* lowaddr */
2628 BUS_SPACE_MAXADDR, /* highaddr */
2629 NULL, NULL, /* filter, filterarg */
2630 size, /* maxsize */
2631 1, /* nsegments */
2632 size, /* maxsegsize */
2633 0, /* flags */
2634 NULL, /* lockfunc */
2635 NULL, /* lockarg */
2636 &dma->dma_tag);
2637 if (error) {
2638 device_printf(adapter->dev,
2639 "%s: bus_dma_tag_create failed: %d\n",
2640 __func__, error);
2641 goto fail_0;
2642 }
2643
2644 error = bus_dmamem_alloc(dma->dma_tag, (void**) &dma->dma_vaddr,
2645 BUS_DMA_NOWAIT | BUS_DMA_COHERENT, &dma->dma_map);
2646 if (error) {
2647 device_printf(adapter->dev,
2648 "%s: bus_dmamem_alloc(%ju) failed: %d\n",
2649 __func__, (uintmax_t)size, error);
2650 goto fail_2;
2651 }
2652
2653 dma->dma_paddr = 0;
2654 error = bus_dmamap_load(dma->dma_tag, dma->dma_map, dma->dma_vaddr,
2655 size, lem_dmamap_cb, &dma->dma_paddr, mapflags | BUS_DMA_NOWAIT);
2656 if (error || dma->dma_paddr == 0) {
2657 device_printf(adapter->dev,
2658 "%s: bus_dmamap_load failed: %d\n",
2659 __func__, error);
2660 goto fail_3;
2661 }
2662
2663 return (0);
2664
2665fail_3:
2666 bus_dmamap_unload(dma->dma_tag, dma->dma_map);
2667fail_2:
2668 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
2669 bus_dma_tag_destroy(dma->dma_tag);
2670fail_0:
2671 dma->dma_tag = NULL;
2672
2673 return (error);
2674}
2675
2676static void
2677lem_dma_free(struct adapter *adapter, struct em_dma_alloc *dma)
2678{
2679 if (dma->dma_tag == NULL)
2680 return;
2681 if (dma->dma_paddr != 0) {
2682 bus_dmamap_sync(dma->dma_tag, dma->dma_map,
2683 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2684 bus_dmamap_unload(dma->dma_tag, dma->dma_map);
2685 dma->dma_paddr = 0;
2686 }
2687 if (dma->dma_vaddr != NULL) {
2688 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
2689 dma->dma_vaddr = NULL;
2690 }
2691 bus_dma_tag_destroy(dma->dma_tag);
2692 dma->dma_tag = NULL;
2693}
2694
2695
2696/*********************************************************************
2697 *
2698 * Allocate memory for tx_buffer structures. The tx_buffer stores all
2699 * the information needed to transmit a packet on the wire.
2700 *
2701 **********************************************************************/
2702static int
2703lem_allocate_transmit_structures(struct adapter *adapter)
2704{
2705 device_t dev = adapter->dev;
2706 struct em_buffer *tx_buffer;
2707 int error;
2708
2709 /*
2710 * Create DMA tags for tx descriptors
2711 */
2712 if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
2713 1, 0, /* alignment, bounds */
2714 BUS_SPACE_MAXADDR, /* lowaddr */
2715 BUS_SPACE_MAXADDR, /* highaddr */
2716 NULL, NULL, /* filter, filterarg */
2717 MCLBYTES * EM_MAX_SCATTER, /* maxsize */
2718 EM_MAX_SCATTER, /* nsegments */
2719 MCLBYTES, /* maxsegsize */
2720 0, /* flags */
2721 NULL, /* lockfunc */
2722 NULL, /* lockarg */
2723 &adapter->txtag)) != 0) {
2724 device_printf(dev, "Unable to allocate TX DMA tag\n");
2725 goto fail;
2726 }
2727
2728 adapter->tx_buffer_area = malloc(sizeof(struct em_buffer) *
2729 adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
2730 if (adapter->tx_buffer_area == NULL) {
2731 device_printf(dev, "Unable to allocate tx_buffer memory\n");
2732 error = ENOMEM;
2733 goto fail;
2734 }
2735
2736 /* Create the descriptor buffer dma maps */
2737 for (int i = 0; i < adapter->num_tx_desc; i++) {
2738 tx_buffer = &adapter->tx_buffer_area[i];
2739 error = bus_dmamap_create(adapter->txtag, 0, &tx_buffer->map);
2740 if (error != 0) {
2741 device_printf(dev, "Unable to create TX DMA map\n");
2742 goto fail;
2743 }
2744 tx_buffer->next_eop = -1;
2745 }
2746
2747 return (0);
2748fail:
2749 lem_free_transmit_structures(adapter);
2750 return (error);
2751}
2752
2753/*********************************************************************
2754 *
2755 * (Re)Initialize transmit structures.
2756 *
2757 **********************************************************************/
2758static void
2759lem_setup_transmit_structures(struct adapter *adapter)
2760{
2761 struct em_buffer *tx_buffer;
2762#ifdef DEV_NETMAP
2763 /* we are already locked */
2764 struct netmap_adapter *na = netmap_getna(adapter->ifp);
2765 struct netmap_slot *slot = netmap_reset(na, NR_TX, 0, 0);
2766#endif /* DEV_NETMAP */
2767
2768 /* Clear the old ring contents */
2769 bzero(adapter->tx_desc_base,
2770 (sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
2771
2772 /* Free any existing TX buffers */
2773 for (int i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) {
2774 tx_buffer = &adapter->tx_buffer_area[i];
2775 bus_dmamap_sync(adapter->txtag, tx_buffer->map,
2776 BUS_DMASYNC_POSTWRITE);
2777 bus_dmamap_unload(adapter->txtag, tx_buffer->map);
2778 m_freem(tx_buffer->m_head);
2779 tx_buffer->m_head = NULL;
2780#ifdef DEV_NETMAP
2781 if (slot) {
2782 /* the i-th NIC entry goes to slot si */
2783 int si = netmap_idx_n2k(&na->tx_rings[0], i);
2784 uint64_t paddr;
2785 void *addr;
2786
2787 addr = PNMB(na, slot + si, &paddr);
2788 adapter->tx_desc_base[i].buffer_addr = htole64(paddr);
2789 /* reload the map for netmap mode */
2790 netmap_load_map(na, adapter->txtag, tx_buffer->map, addr);
2791 }
2792#endif /* DEV_NETMAP */
2793 tx_buffer->next_eop = -1;
2794 }
2795
2796 /* Reset state */
2797 adapter->last_hw_offload = 0;
2798 adapter->next_avail_tx_desc = 0;
2799 adapter->next_tx_to_clean = 0;
2800 adapter->num_tx_desc_avail = adapter->num_tx_desc;
2801
2802 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
2803 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2804
2805 return;
2806}
2807
2808/*********************************************************************
2809 *
2810 * Enable transmit unit.
2811 *
2812 **********************************************************************/
2813static void
2814lem_initialize_transmit_unit(struct adapter *adapter)
2815{
2816 u32 tctl, tipg = 0;
2817 u64 bus_addr;
2818
2819 INIT_DEBUGOUT("lem_initialize_transmit_unit: begin");
2820 /* Setup the Base and Length of the Tx Descriptor Ring */
2821 bus_addr = adapter->txdma.dma_paddr;
2822 E1000_WRITE_REG(&adapter->hw, E1000_TDLEN(0),
2823 adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
2824 E1000_WRITE_REG(&adapter->hw, E1000_TDBAH(0),
2825 (u32)(bus_addr >> 32));
2826 E1000_WRITE_REG(&adapter->hw, E1000_TDBAL(0),
2827 (u32)bus_addr);
2828 /* Setup the HW Tx Head and Tail descriptor pointers */
2829 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), 0);
2830 E1000_WRITE_REG(&adapter->hw, E1000_TDH(0), 0);
2831
2832 HW_DEBUGOUT2("Base = %x, Length = %x\n",
2833 E1000_READ_REG(&adapter->hw, E1000_TDBAL(0)),
2834 E1000_READ_REG(&adapter->hw, E1000_TDLEN(0)));
2835
2836 /* Set the default values for the Tx Inter Packet Gap timer */
2837 switch (adapter->hw.mac.type) {
2838 case e1000_82542:
2839 tipg = DEFAULT_82542_TIPG_IPGT;
2840 tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2841 tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2842 break;
2843 default:
2844 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2845 (adapter->hw.phy.media_type ==
2846 e1000_media_type_internal_serdes))
2847 tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2848 else
2849 tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
2850 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2851 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2852 }
2853
2854 E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
2855 E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
2856 if(adapter->hw.mac.type >= e1000_82540)
2857 E1000_WRITE_REG(&adapter->hw, E1000_TADV,
2858 adapter->tx_abs_int_delay.value);
2859
2860 /* Program the Transmit Control Register */
2861 tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
2862 tctl &= ~E1000_TCTL_CT;
2863 tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
2864 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
2865
2866 /* This write will effectively turn on the transmit unit. */
2867 E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
2868
2869 /* Setup Transmit Descriptor Base Settings */
2870 adapter->txd_cmd = E1000_TXD_CMD_IFCS;
2871
2872 if (adapter->tx_int_delay.value > 0)
2873 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2874}
2875
2876/*********************************************************************
2877 *
2878 * Free all transmit related data structures.
2879 *
2880 **********************************************************************/
2881static void
2882lem_free_transmit_structures(struct adapter *adapter)
2883{
2884 struct em_buffer *tx_buffer;
2885
2886 INIT_DEBUGOUT("free_transmit_structures: begin");
2887
2888 if (adapter->tx_buffer_area != NULL) {
2889 for (int i = 0; i < adapter->num_tx_desc; i++) {
2890 tx_buffer = &adapter->tx_buffer_area[i];
2891 if (tx_buffer->m_head != NULL) {
2892 bus_dmamap_sync(adapter->txtag, tx_buffer->map,
2893 BUS_DMASYNC_POSTWRITE);
2894 bus_dmamap_unload(adapter->txtag,
2895 tx_buffer->map);
2896 m_freem(tx_buffer->m_head);
2897 tx_buffer->m_head = NULL;
2898 } else if (tx_buffer->map != NULL)
2899 bus_dmamap_unload(adapter->txtag,
2900 tx_buffer->map);
2901 if (tx_buffer->map != NULL) {
2902 bus_dmamap_destroy(adapter->txtag,
2903 tx_buffer->map);
2904 tx_buffer->map = NULL;
2905 }
2906 }
2907 }
2908 if (adapter->tx_buffer_area != NULL) {
2909 free(adapter->tx_buffer_area, M_DEVBUF);
2910 adapter->tx_buffer_area = NULL;
2911 }
2912 if (adapter->txtag != NULL) {
2913 bus_dma_tag_destroy(adapter->txtag);
2914 adapter->txtag = NULL;
2915 }
2916}
2917
2918/*********************************************************************
2919 *
2920 * The offload context needs to be set when we transfer the first
2921 * packet of a particular protocol (TCP/UDP). This routine has been
2922 * enhanced to deal with inserted VLAN headers, and IPV6 (not complete)
2923 *
2924 * Added back the old method of keeping the current context type
2925 * and not setting if unnecessary, as this is reported to be a
2926 * big performance win. -jfv
2927 **********************************************************************/
2928static void
2929lem_transmit_checksum_setup(struct adapter *adapter, struct mbuf *mp,
2930 u32 *txd_upper, u32 *txd_lower)
2931{
2932 struct e1000_context_desc *TXD = NULL;
2933 struct em_buffer *tx_buffer;
2934 struct ether_vlan_header *eh;
2935 struct ip *ip = NULL;
2936 struct ip6_hdr *ip6;
2937 int curr_txd, ehdrlen;
2938 u32 cmd, hdr_len, ip_hlen;
2939 u16 etype;
2940 u8 ipproto;
2941
2942
2943 cmd = hdr_len = ipproto = 0;
2944 *txd_upper = *txd_lower = 0;
2945 curr_txd = adapter->next_avail_tx_desc;
2946
2947 /*
2948 * Determine where frame payload starts.
2949 * Jump over vlan headers if already present,
2950 * helpful for QinQ too.
2951 */
2952 eh = mtod(mp, struct ether_vlan_header *);
2953 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
2954 etype = ntohs(eh->evl_proto);
2955 ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
2956 } else {
2957 etype = ntohs(eh->evl_encap_proto);
2958 ehdrlen = ETHER_HDR_LEN;
2959 }
2960
2961 /*
2962 * We only support TCP/UDP for IPv4 and IPv6 for the moment.
2963 * TODO: Support SCTP too when it hits the tree.
2964 */
2965 switch (etype) {
2966 case ETHERTYPE_IP:
2967 ip = (struct ip *)(mp->m_data + ehdrlen);
2968 ip_hlen = ip->ip_hl << 2;
2969
2970 /* Setup of IP header checksum. */
2971 if (mp->m_pkthdr.csum_flags & CSUM_IP) {
2972 /*
2973 * Start offset for header checksum calculation.
2974 * End offset for header checksum calculation.
2975 * Offset of place to put the checksum.
2976 */
2977 TXD = (struct e1000_context_desc *)
2978 &adapter->tx_desc_base[curr_txd];
2979 TXD->lower_setup.ip_fields.ipcss = ehdrlen;
2980 TXD->lower_setup.ip_fields.ipcse =
2981 htole16(ehdrlen + ip_hlen);
2982 TXD->lower_setup.ip_fields.ipcso =
2983 ehdrlen + offsetof(struct ip, ip_sum);
2984 cmd |= E1000_TXD_CMD_IP;
2985 *txd_upper |= E1000_TXD_POPTS_IXSM << 8;
2986 }
2987
2988 hdr_len = ehdrlen + ip_hlen;
2989 ipproto = ip->ip_p;
2990
2991 break;
2992 case ETHERTYPE_IPV6:
2993 ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
2994 ip_hlen = sizeof(struct ip6_hdr); /* XXX: No header stacking. */
2995
2996 /* IPv6 doesn't have a header checksum. */
2997
2998 hdr_len = ehdrlen + ip_hlen;
2999 ipproto = ip6->ip6_nxt;
3000 break;
3001
3002 default:
3003 return;
3004 }
3005
3006 switch (ipproto) {
3007 case IPPROTO_TCP:
3008 if (mp->m_pkthdr.csum_flags & CSUM_TCP) {
3009 *txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
3010 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
3011 /* no need for context if already set */
3012 if (adapter->last_hw_offload == CSUM_TCP)
3013 return;
3014 adapter->last_hw_offload = CSUM_TCP;
3015 /*
3016 * Start offset for payload checksum calculation.
3017 * End offset for payload checksum calculation.
3018 * Offset of place to put the checksum.
3019 */
3020 TXD = (struct e1000_context_desc *)
3021 &adapter->tx_desc_base[curr_txd];
3022 TXD->upper_setup.tcp_fields.tucss = hdr_len;
3023 TXD->upper_setup.tcp_fields.tucse = htole16(0);
3024 TXD->upper_setup.tcp_fields.tucso =
3025 hdr_len + offsetof(struct tcphdr, th_sum);
3026 cmd |= E1000_TXD_CMD_TCP;
3027 }
3028 break;
3029 case IPPROTO_UDP:
3030 {
3031 if (mp->m_pkthdr.csum_flags & CSUM_UDP) {
3032 *txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
3033 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
3034 /* no need for context if already set */
3035 if (adapter->last_hw_offload == CSUM_UDP)
3036 return;
3037 adapter->last_hw_offload = CSUM_UDP;
3038 /*
3039 * Start offset for header checksum calculation.
3040 * End offset for header checksum calculation.
3041 * Offset of place to put the checksum.
3042 */
3043 TXD = (struct e1000_context_desc *)
3044 &adapter->tx_desc_base[curr_txd];
3045 TXD->upper_setup.tcp_fields.tucss = hdr_len;
3046 TXD->upper_setup.tcp_fields.tucse = htole16(0);
3047 TXD->upper_setup.tcp_fields.tucso =
3048 hdr_len + offsetof(struct udphdr, uh_sum);
3049 }
3050 /* Fall Thru */
3051 }
3052 default:
3053 break;
3054 }
3055
3056 if (TXD == NULL)
3057 return;
3058 TXD->tcp_seg_setup.data = htole32(0);
3059 TXD->cmd_and_length =
3060 htole32(adapter->txd_cmd | E1000_TXD_CMD_DEXT | cmd);
3061 tx_buffer = &adapter->tx_buffer_area[curr_txd];
3062 tx_buffer->m_head = NULL;
3063 tx_buffer->next_eop = -1;
3064
3065 if (++curr_txd == adapter->num_tx_desc)
3066 curr_txd = 0;
3067
3068 adapter->num_tx_desc_avail--;
3069 adapter->next_avail_tx_desc = curr_txd;
3070}
3071
3072
3073/**********************************************************************
3074 *
3075 * Examine each tx_buffer in the used queue. If the hardware is done
3076 * processing the packet then free associated resources. The
3077 * tx_buffer is put back on the free queue.
3078 *
3079 **********************************************************************/
3080static void
3081lem_txeof(struct adapter *adapter)
3082{
3083 int first, last, done, num_avail;
3084 struct em_buffer *tx_buffer;
3085 struct e1000_tx_desc *tx_desc, *eop_desc;
3086 if_t ifp = adapter->ifp;
3087
3088 EM_TX_LOCK_ASSERT(adapter);
3089
3090#ifdef DEV_NETMAP
3091 if (netmap_tx_irq(ifp, 0))
3092 return;
3093#endif /* DEV_NETMAP */
3094 if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
3095 return;
3096
3097 num_avail = adapter->num_tx_desc_avail;
3098 first = adapter->next_tx_to_clean;
3099 tx_desc = &adapter->tx_desc_base[first];
3100 tx_buffer = &adapter->tx_buffer_area[first];
3101 last = tx_buffer->next_eop;
3102 eop_desc = &adapter->tx_desc_base[last];
3103
3104 /*
3105 * What this does is get the index of the
3106 * first descriptor AFTER the EOP of the
3107 * first packet, that way we can do the
3108 * simple comparison on the inner while loop.
3109 */
3110 if (++last == adapter->num_tx_desc)
3111 last = 0;
3112 done = last;
3113
3114 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
3115 BUS_DMASYNC_POSTREAD);
3116
3117 while (eop_desc->upper.fields.status & E1000_TXD_STAT_DD) {
3118 /* We clean the range of the packet */
3119 while (first != done) {
3120 tx_desc->upper.data = 0;
3121 tx_desc->lower.data = 0;
3122 tx_desc->buffer_addr = 0;
3123 ++num_avail;
3124
3125 if (tx_buffer->m_head) {
3126 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
3127 bus_dmamap_sync(adapter->txtag,
3128 tx_buffer->map,
3129 BUS_DMASYNC_POSTWRITE);
3130 bus_dmamap_unload(adapter->txtag,
3131 tx_buffer->map);
3132
3133 m_freem(tx_buffer->m_head);
3134 tx_buffer->m_head = NULL;
3135 }
3136 tx_buffer->next_eop = -1;
3137 adapter->watchdog_time = ticks;
3138
3139 if (++first == adapter->num_tx_desc)
3140 first = 0;
3141
3142 tx_buffer = &adapter->tx_buffer_area[first];
3143 tx_desc = &adapter->tx_desc_base[first];
3144 }
3145 /* See if we can continue to the next packet */
3146 last = tx_buffer->next_eop;
3147 if (last != -1) {
3148 eop_desc = &adapter->tx_desc_base[last];
3149 /* Get new done point */
3150 if (++last == adapter->num_tx_desc) last = 0;
3151 done = last;
3152 } else
3153 break;
3154 }
3155 bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
3156 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3157
3158 adapter->next_tx_to_clean = first;
3159 adapter->num_tx_desc_avail = num_avail;
3160
3161#ifdef NIC_SEND_COMBINING
3162 if ((adapter->shadow_tdt & MIT_PENDING_TDT) == MIT_PENDING_TDT) {
3163 /* a tdt write is pending, do it */
3164 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0),
3165 0xffff & adapter->shadow_tdt);
3166 adapter->shadow_tdt = MIT_PENDING_INT;
3167 } else {
3168 adapter->shadow_tdt = 0; // disable
3169 }
3170#endif /* NIC_SEND_COMBINING */
3171 /*
3172 * If we have enough room, clear IFF_DRV_OACTIVE to
3173 * tell the stack that it is OK to send packets.
3174 * If there are no pending descriptors, clear the watchdog.
3175 */
3176 if (adapter->num_tx_desc_avail > EM_TX_CLEANUP_THRESHOLD) {
3177 if_setdrvflagbits(ifp, 0, IFF_DRV_OACTIVE);
3178#ifdef NIC_PARAVIRT
3179 if (adapter->csb) { // XXX also csb_on ?
3180 adapter->csb->guest_need_txkick = 2; /* acked */
3181 // XXX memory barrier
3182 }
3183#endif /* NIC_PARAVIRT */
3184 if (adapter->num_tx_desc_avail == adapter->num_tx_desc) {
3185 adapter->watchdog_check = FALSE;
3186 return;
3187 }
3188 }
3189}
3190
3191/*********************************************************************
3192 *
3193 * When Link is lost sometimes there is work still in the TX ring
3194 * which may result in a watchdog, rather than allow that we do an
3195 * attempted cleanup and then reinit here. Note that this has been
3196 * seens mostly with fiber adapters.
3197 *
3198 **********************************************************************/
3199static void
3200lem_tx_purge(struct adapter *adapter)
3201{
3202 if ((!adapter->link_active) && (adapter->watchdog_check)) {
3203 EM_TX_LOCK(adapter);
3204 lem_txeof(adapter);
3205 EM_TX_UNLOCK(adapter);
3206 if (adapter->watchdog_check) /* Still outstanding? */
3207 lem_init_locked(adapter);
3208 }
3209}
3210
3211/*********************************************************************
3212 *
3213 * Get a buffer from system mbuf buffer pool.
3214 *
3215 **********************************************************************/
3216static int
3217lem_get_buf(struct adapter *adapter, int i)
3218{
3219 struct mbuf *m;
3220 bus_dma_segment_t segs[1];
3221 bus_dmamap_t map;
3222 struct em_buffer *rx_buffer;
3223 int error, nsegs;
3224
3225 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
3226 if (m == NULL) {
3227 adapter->mbuf_cluster_failed++;
3228 return (ENOBUFS);
3229 }
3230 m->m_len = m->m_pkthdr.len = MCLBYTES;
3231
3232 if (adapter->max_frame_size <= (MCLBYTES - ETHER_ALIGN))
3233 m_adj(m, ETHER_ALIGN);
3234
3235 /*
3236 * Using memory from the mbuf cluster pool, invoke the
3237 * bus_dma machinery to arrange the memory mapping.
3238 */
3239 error = bus_dmamap_load_mbuf_sg(adapter->rxtag,
3240 adapter->rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
3241 if (error != 0) {
3242 m_free(m);
3243 return (error);
3244 }
3245
3246 /* If nsegs is wrong then the stack is corrupt. */
3247 KASSERT(nsegs == 1, ("Too many segments returned!"));
3248
3249 rx_buffer = &adapter->rx_buffer_area[i];
3250 if (rx_buffer->m_head != NULL)
3251 bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
3252
3253 map = rx_buffer->map;
3254 rx_buffer->map = adapter->rx_sparemap;
3255 adapter->rx_sparemap = map;
3256 bus_dmamap_sync(adapter->rxtag, rx_buffer->map, BUS_DMASYNC_PREREAD);
3257 rx_buffer->m_head = m;
3258
3259 adapter->rx_desc_base[i].buffer_addr = htole64(segs[0].ds_addr);
3260 return (0);
3261}
3262
3263/*********************************************************************
3264 *
3265 * Allocate memory for rx_buffer structures. Since we use one
3266 * rx_buffer per received packet, the maximum number of rx_buffer's
3267 * that we'll need is equal to the number of receive descriptors
3268 * that we've allocated.
3269 *
3270 **********************************************************************/
3271static int
3272lem_allocate_receive_structures(struct adapter *adapter)
3273{
3274 device_t dev = adapter->dev;
3275 struct em_buffer *rx_buffer;
3276 int i, error;
3277
3278 adapter->rx_buffer_area = malloc(sizeof(struct em_buffer) *
3279 adapter->num_rx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
3280 if (adapter->rx_buffer_area == NULL) {
3281 device_printf(dev, "Unable to allocate rx_buffer memory\n");
3282 return (ENOMEM);
3283 }
3284
3285 error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
3286 1, 0, /* alignment, bounds */
3287 BUS_SPACE_MAXADDR, /* lowaddr */
3288 BUS_SPACE_MAXADDR, /* highaddr */
3289 NULL, NULL, /* filter, filterarg */
3290 MCLBYTES, /* maxsize */
3291 1, /* nsegments */
3292 MCLBYTES, /* maxsegsize */
3293 0, /* flags */
3294 NULL, /* lockfunc */
3295 NULL, /* lockarg */
3296 &adapter->rxtag);
3297 if (error) {
3298 device_printf(dev, "%s: bus_dma_tag_create failed %d\n",
3299 __func__, error);
3300 goto fail;
3301 }
3302
3303 /* Create the spare map (used by getbuf) */
3304 error = bus_dmamap_create(adapter->rxtag, 0, &adapter->rx_sparemap);
3305 if (error) {
3306 device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
3307 __func__, error);
3308 goto fail;
3309 }
3310
3311 rx_buffer = adapter->rx_buffer_area;
3312 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3313 error = bus_dmamap_create(adapter->rxtag, 0, &rx_buffer->map);
3314 if (error) {
3315 device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
3316 __func__, error);
3317 goto fail;
3318 }
3319 }
3320
3321 return (0);
3322
3323fail:
3324 lem_free_receive_structures(adapter);
3325 return (error);
3326}
3327
3328/*********************************************************************
3329 *
3330 * (Re)initialize receive structures.
3331 *
3332 **********************************************************************/
3333static int
3334lem_setup_receive_structures(struct adapter *adapter)
3335{
3336 struct em_buffer *rx_buffer;
3337 int i, error;
3338#ifdef DEV_NETMAP
3339 /* we are already under lock */
3340 struct netmap_adapter *na = netmap_getna(adapter->ifp);
3341 struct netmap_slot *slot = netmap_reset(na, NR_RX, 0, 0);
3342#endif
3343
3344 /* Reset descriptor ring */
3345 bzero(adapter->rx_desc_base,
3346 (sizeof(struct e1000_rx_desc)) * adapter->num_rx_desc);
3347
3348 /* Free current RX buffers. */
3349 rx_buffer = adapter->rx_buffer_area;
3350 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3351 if (rx_buffer->m_head != NULL) {
3352 bus_dmamap_sync(adapter->rxtag, rx_buffer->map,
3353 BUS_DMASYNC_POSTREAD);
3354 bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
3355 m_freem(rx_buffer->m_head);
3356 rx_buffer->m_head = NULL;
3357 }
3358 }
3359
3360 /* Allocate new ones. */
3361 for (i = 0; i < adapter->num_rx_desc; i++) {
3362#ifdef DEV_NETMAP
3363 if (slot) {
3364 /* the i-th NIC entry goes to slot si */
3365 int si = netmap_idx_n2k(&na->rx_rings[0], i);
3366 uint64_t paddr;
3367 void *addr;
3368
3369 addr = PNMB(na, slot + si, &paddr);
3370 netmap_load_map(na, adapter->rxtag, rx_buffer->map, addr);
3371 /* Update descriptor */
3372 adapter->rx_desc_base[i].buffer_addr = htole64(paddr);
3373 continue;
3374 }
3375#endif /* DEV_NETMAP */
3376 error = lem_get_buf(adapter, i);
3377 if (error)
3378 return (error);
3379 }
3380
3381 /* Setup our descriptor pointers */
3382 adapter->next_rx_desc_to_check = 0;
3383 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3384 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3385
3386 return (0);
3387}
3388
3389/*********************************************************************
3390 *
3391 * Enable receive unit.
3392 *
3393 **********************************************************************/
3394
3395static void
3396lem_initialize_receive_unit(struct adapter *adapter)
3397{
3398 if_t ifp = adapter->ifp;
3399 u64 bus_addr;
3400 u32 rctl, rxcsum;
3401
3402 INIT_DEBUGOUT("lem_initialize_receive_unit: begin");
3403
3404 /*
3405 * Make sure receives are disabled while setting
3406 * up the descriptor ring
3407 */
3408 rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3409 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3410
3411 if (adapter->hw.mac.type >= e1000_82540) {
3412 E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3413 adapter->rx_abs_int_delay.value);
3414 /*
3415 * Set the interrupt throttling rate. Value is calculated
3416 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3417 */
3418 E1000_WRITE_REG(&adapter->hw, E1000_ITR, DEFAULT_ITR);
3419 }
3420
3421 /* Setup the Base and Length of the Rx Descriptor Ring */
3422 bus_addr = adapter->rxdma.dma_paddr;
3423 E1000_WRITE_REG(&adapter->hw, E1000_RDLEN(0),
3424 adapter->num_rx_desc * sizeof(struct e1000_rx_desc));
3425 E1000_WRITE_REG(&adapter->hw, E1000_RDBAH(0),
3426 (u32)(bus_addr >> 32));
3427 E1000_WRITE_REG(&adapter->hw, E1000_RDBAL(0),
3428 (u32)bus_addr);
3429
3430 /* Setup the Receive Control Register */
3431 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3432 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
3433 E1000_RCTL_RDMTS_HALF |
3434 (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3435
3436 /* Make sure VLAN Filters are off */
3437 rctl &= ~E1000_RCTL_VFE;
3438
3439 if (e1000_tbi_sbp_enabled_82543(&adapter->hw))
3440 rctl |= E1000_RCTL_SBP;
3441 else
3442 rctl &= ~E1000_RCTL_SBP;
3443
3444 switch (adapter->rx_buffer_len) {
3445 default:
3446 case 2048:
3447 rctl |= E1000_RCTL_SZ_2048;
3448 break;
3449 case 4096:
3450 rctl |= E1000_RCTL_SZ_4096 |
3451 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3452 break;
3453 case 8192:
3454 rctl |= E1000_RCTL_SZ_8192 |
3455 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3456 break;
3457 case 16384:
3458 rctl |= E1000_RCTL_SZ_16384 |
3459 E1000_RCTL_BSEX | E1000_RCTL_LPE;
3460 break;
3461 }
3462
3463 if (if_getmtu(ifp) > ETHERMTU)
3464 rctl |= E1000_RCTL_LPE;
3465 else
3466 rctl &= ~E1000_RCTL_LPE;
3467
3468 /* Enable 82543 Receive Checksum Offload for TCP and UDP */
3469 if ((adapter->hw.mac.type >= e1000_82543) &&
3470 (if_getcapenable(ifp) & IFCAP_RXCSUM)) {
3471 rxcsum = E1000_READ_REG(&adapter->hw, E1000_RXCSUM);
3472 rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3473 E1000_WRITE_REG(&adapter->hw, E1000_RXCSUM, rxcsum);
3474 }
3475
3476 /* Enable Receives */
3477 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3478
3479 /*
3480 * Setup the HW Rx Head and
3481 * Tail Descriptor Pointers
3482 */
3483 E1000_WRITE_REG(&adapter->hw, E1000_RDH(0), 0);
3484 rctl = adapter->num_rx_desc - 1; /* default RDT value */
3485#ifdef DEV_NETMAP
3486 /* preserve buffers already made available to clients */
3487 if (if_getcapenable(ifp) & IFCAP_NETMAP) {
3488 struct netmap_adapter *na = netmap_getna(adapter->ifp);
3489 rctl -= nm_kr_rxspace(&na->rx_rings[0]);
3490 }
3491#endif /* DEV_NETMAP */
3492 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), rctl);
3493
3494 return;
3495}
3496
3497/*********************************************************************
3498 *
3499 * Free receive related data structures.
3500 *
3501 **********************************************************************/
3502static void
3503lem_free_receive_structures(struct adapter *adapter)
3504{
3505 struct em_buffer *rx_buffer;
3506 int i;
3507
3508 INIT_DEBUGOUT("free_receive_structures: begin");
3509
3510 if (adapter->rx_sparemap) {
3511 bus_dmamap_destroy(adapter->rxtag, adapter->rx_sparemap);
3512 adapter->rx_sparemap = NULL;
3513 }
3514
3515 /* Cleanup any existing buffers */
3516 if (adapter->rx_buffer_area != NULL) {
3517 rx_buffer = adapter->rx_buffer_area;
3518 for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
3519 if (rx_buffer->m_head != NULL) {
3520 bus_dmamap_sync(adapter->rxtag, rx_buffer->map,
3521 BUS_DMASYNC_POSTREAD);
3522 bus_dmamap_unload(adapter->rxtag,
3523 rx_buffer->map);
3524 m_freem(rx_buffer->m_head);
3525 rx_buffer->m_head = NULL;
3526 } else if (rx_buffer->map != NULL)
3527 bus_dmamap_unload(adapter->rxtag,
3528 rx_buffer->map);
3529 if (rx_buffer->map != NULL) {
3530 bus_dmamap_destroy(adapter->rxtag,
3531 rx_buffer->map);
3532 rx_buffer->map = NULL;
3533 }
3534 }
3535 }
3536
3537 if (adapter->rx_buffer_area != NULL) {
3538 free(adapter->rx_buffer_area, M_DEVBUF);
3539 adapter->rx_buffer_area = NULL;
3540 }
3541
3542 if (adapter->rxtag != NULL) {
3543 bus_dma_tag_destroy(adapter->rxtag);
3544 adapter->rxtag = NULL;
3545 }
3546}
3547
3548/*********************************************************************
3549 *
3550 * This routine executes in interrupt context. It replenishes
3551 * the mbufs in the descriptor and sends data which has been
3552 * dma'ed into host memory to upper layer.
3553 *
3554 * We loop at most count times if count is > 0, or until done if
3555 * count < 0.
3556 *
3557 * For polling we also now return the number of cleaned packets
3558 *********************************************************************/
3559static bool
3560lem_rxeof(struct adapter *adapter, int count, int *done)
3561{
3562 if_t ifp = adapter->ifp;
3563 struct mbuf *mp;
3564 u8 status = 0, accept_frame = 0, eop = 0;
3565 u16 len, desc_len, prev_len_adj;
3566 int i, rx_sent = 0;
3567 struct e1000_rx_desc *current_desc;
3568
3569#ifdef BATCH_DISPATCH
3570 struct mbuf *mh = NULL, *mt = NULL;
3571#endif /* BATCH_DISPATCH */
3572#ifdef NIC_PARAVIRT
3573 int retries = 0;
3574 struct paravirt_csb* csb = adapter->csb;
3575 int csb_mode = csb && csb->guest_csb_on;
3576
3577 //ND("clear guest_rxkick at %d", adapter->next_rx_desc_to_check);
3578 if (csb_mode && csb->guest_need_rxkick)
3579 csb->guest_need_rxkick = 0;
3580#endif /* NIC_PARAVIRT */
3581 EM_RX_LOCK(adapter);
3582
3583#ifdef BATCH_DISPATCH
3584 batch_again:
3585#endif /* BATCH_DISPATCH */
3586 i = adapter->next_rx_desc_to_check;
3587 current_desc = &adapter->rx_desc_base[i];
3588 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3589 BUS_DMASYNC_POSTREAD);
3590
3591#ifdef DEV_NETMAP
3592 if (netmap_rx_irq(ifp, 0, &rx_sent)) {
3593 EM_RX_UNLOCK(adapter);
3594 return (FALSE);
3595 }
3596#endif /* DEV_NETMAP */
3597
3598#if 1 // XXX optimization ?
3599 if (!((current_desc->status) & E1000_RXD_STAT_DD)) {
3600 if (done != NULL)
3601 *done = rx_sent;
3602 EM_RX_UNLOCK(adapter);
3603 return (FALSE);
3604 }
3605#endif /* 0 */
3606
3607 while (count != 0 && if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
3608 struct mbuf *m = NULL;
3609
3610 status = current_desc->status;
3611 if ((status & E1000_RXD_STAT_DD) == 0) {
3612#ifdef NIC_PARAVIRT
3613 if (csb_mode) {
3614 /* buffer not ready yet. Retry a few times before giving up */
3615 if (++retries <= adapter->rx_retries) {
3616 continue;
3617 }
3618 if (csb->guest_need_rxkick == 0) {
3619 // ND("set guest_rxkick at %d", adapter->next_rx_desc_to_check);
3620 csb->guest_need_rxkick = 1;
3621 // XXX memory barrier, status volatile ?
3622 continue; /* double check */
3623 }
3624 }
3625 /* no buffer ready, give up */
3626#endif /* NIC_PARAVIRT */
3627 break;
3628 }
3629#ifdef NIC_PARAVIRT
3630 if (csb_mode) {
3631 if (csb->guest_need_rxkick)
3632 // ND("clear again guest_rxkick at %d", adapter->next_rx_desc_to_check);
3633 csb->guest_need_rxkick = 0;
3634 retries = 0;
3635 }
3636#endif /* NIC_PARAVIRT */
3637
3638 mp = adapter->rx_buffer_area[i].m_head;
3639 /*
3640 * Can't defer bus_dmamap_sync(9) because TBI_ACCEPT
3641 * needs to access the last received byte in the mbuf.
3642 */
3643 bus_dmamap_sync(adapter->rxtag, adapter->rx_buffer_area[i].map,
3644 BUS_DMASYNC_POSTREAD);
3645
3646 accept_frame = 1;
3647 prev_len_adj = 0;
3648 desc_len = le16toh(current_desc->length);
3649 if (status & E1000_RXD_STAT_EOP) {
3650 count--;
3651 eop = 1;
3652 if (desc_len < ETHER_CRC_LEN) {
3653 len = 0;
3654 prev_len_adj = ETHER_CRC_LEN - desc_len;
3655 } else
3656 len = desc_len - ETHER_CRC_LEN;
3657 } else {
3658 eop = 0;
3659 len = desc_len;
3660 }
3661
3662 if (current_desc->errors & E1000_RXD_ERR_FRAME_ERR_MASK) {
3663 u8 last_byte;
3664 u32 pkt_len = desc_len;
3665
3666 if (adapter->fmp != NULL)
3667 pkt_len += adapter->fmp->m_pkthdr.len;
3668
3669 last_byte = *(mtod(mp, caddr_t) + desc_len - 1);
3670 if (TBI_ACCEPT(&adapter->hw, status,
3671 current_desc->errors, pkt_len, last_byte,
3672 adapter->min_frame_size, adapter->max_frame_size)) {
3673 e1000_tbi_adjust_stats_82543(&adapter->hw,
3674 &adapter->stats, pkt_len,
3675 adapter->hw.mac.addr,
3676 adapter->max_frame_size);
3677 if (len > 0)
3678 len--;
3679 } else
3680 accept_frame = 0;
3681 }
3682
3683 if (accept_frame) {
3684 if (lem_get_buf(adapter, i) != 0) {
3685 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
3686 goto discard;
3687 }
3688
3689 /* Assign correct length to the current fragment */
3690 mp->m_len = len;
3691
3692 if (adapter->fmp == NULL) {
3693 mp->m_pkthdr.len = len;
3694 adapter->fmp = mp; /* Store the first mbuf */
3695 adapter->lmp = mp;
3696 } else {
3697 /* Chain mbuf's together */
3698 mp->m_flags &= ~M_PKTHDR;
3699 /*
3700 * Adjust length of previous mbuf in chain if
3701 * we received less than 4 bytes in the last
3702 * descriptor.
3703 */
3704 if (prev_len_adj > 0) {
3705 adapter->lmp->m_len -= prev_len_adj;
3706 adapter->fmp->m_pkthdr.len -=
3707 prev_len_adj;
3708 }
3709 adapter->lmp->m_next = mp;
3710 adapter->lmp = adapter->lmp->m_next;
3711 adapter->fmp->m_pkthdr.len += len;
3712 }
3713
3714 if (eop) {
3715 if_setrcvif(adapter->fmp, ifp);
3716 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
3717 lem_receive_checksum(adapter, current_desc,
3718 adapter->fmp);
3719#ifndef __NO_STRICT_ALIGNMENT
3720 if (adapter->max_frame_size >
3721 (MCLBYTES - ETHER_ALIGN) &&
3722 lem_fixup_rx(adapter) != 0)
3723 goto skip;
3724#endif
3725 if (status & E1000_RXD_STAT_VP) {
3726 adapter->fmp->m_pkthdr.ether_vtag =
3727 le16toh(current_desc->special);
3728 adapter->fmp->m_flags |= M_VLANTAG;
3729 }
3730#ifndef __NO_STRICT_ALIGNMENT
3731skip:
3732#endif
3733 m = adapter->fmp;
3734 adapter->fmp = NULL;
3735 adapter->lmp = NULL;
3736 }
3737 } else {
3738 adapter->dropped_pkts++;
3739discard:
3740 /* Reuse loaded DMA map and just update mbuf chain */
3741 mp = adapter->rx_buffer_area[i].m_head;
3742 mp->m_len = mp->m_pkthdr.len = MCLBYTES;
3743 mp->m_data = mp->m_ext.ext_buf;
3744 mp->m_next = NULL;
3745 if (adapter->max_frame_size <=
3746 (MCLBYTES - ETHER_ALIGN))
3747 m_adj(mp, ETHER_ALIGN);
3748 if (adapter->fmp != NULL) {
3749 m_freem(adapter->fmp);
3750 adapter->fmp = NULL;
3751 adapter->lmp = NULL;
3752 }
3753 m = NULL;
3754 }
3755
3756 /* Zero out the receive descriptors status. */
3757 current_desc->status = 0;
3758 bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
3759 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
3760
3761#ifdef NIC_PARAVIRT
3762 if (csb_mode) {
3763 /* the buffer at i has been already replaced by lem_get_buf()
3764 * so it is safe to set guest_rdt = i and possibly send a kick.
3765 * XXX see if we can optimize it later.
3766 */
3767 csb->guest_rdt = i;
3768 // XXX memory barrier
3769 if (i == csb->host_rxkick_at)
3770 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), i);
3771 }
3772#endif /* NIC_PARAVIRT */
3773 /* Advance our pointers to the next descriptor. */
3774 if (++i == adapter->num_rx_desc)
3775 i = 0;
3776 /* Call into the stack */
3777 if (m != NULL) {
3778#ifdef BATCH_DISPATCH
3779 if (adapter->batch_enable) {
3780 if (mh == NULL)
3781 mh = mt = m;
3782 else
3783 mt->m_nextpkt = m;
3784 mt = m;
3785 m->m_nextpkt = NULL;
3786 rx_sent++;
3787 current_desc = &adapter->rx_desc_base[i];
3788 continue;
3789 }
3790#endif /* BATCH_DISPATCH */
3791 adapter->next_rx_desc_to_check = i;
3792 EM_RX_UNLOCK(adapter);
3793 if_input(ifp, m);
3794 EM_RX_LOCK(adapter);
3795 rx_sent++;
3796 i = adapter->next_rx_desc_to_check;
3797 }
3798 current_desc = &adapter->rx_desc_base[i];
3799 }
3800 adapter->next_rx_desc_to_check = i;
3801#ifdef BATCH_DISPATCH
3802 if (mh) {
3803 EM_RX_UNLOCK(adapter);
3804 while ( (mt = mh) != NULL) {
3805 mh = mh->m_nextpkt;
3806 mt->m_nextpkt = NULL;
3807 if_input(ifp, mt);
3808 }
3809 EM_RX_LOCK(adapter);
3810 i = adapter->next_rx_desc_to_check; /* in case of interrupts */
3811 if (count > 0)
3812 goto batch_again;
3813 }
3814#endif /* BATCH_DISPATCH */
3815
3816 /* Advance the E1000's Receive Queue #0 "Tail Pointer". */
3817 if (--i < 0)
3818 i = adapter->num_rx_desc - 1;
3819#ifdef NIC_PARAVIRT
3820 if (!csb_mode) /* filter out writes */
3821#endif /* NIC_PARAVIRT */
3822 E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), i);
3823 if (done != NULL)
3824 *done = rx_sent;
3825 EM_RX_UNLOCK(adapter);
3826 return ((status & E1000_RXD_STAT_DD) ? TRUE : FALSE);
3827}
3828
3829#ifndef __NO_STRICT_ALIGNMENT
3830/*
3831 * When jumbo frames are enabled we should realign entire payload on
3832 * architecures with strict alignment. This is serious design mistake of 8254x
3833 * as it nullifies DMA operations. 8254x just allows RX buffer size to be
3834 * 2048/4096/8192/16384. What we really want is 2048 - ETHER_ALIGN to align its
3835 * payload. On architecures without strict alignment restrictions 8254x still
3836 * performs unaligned memory access which would reduce the performance too.
3837 * To avoid copying over an entire frame to align, we allocate a new mbuf and
3838 * copy ethernet header to the new mbuf. The new mbuf is prepended into the
3839 * existing mbuf chain.
3840 *
3841 * Be aware, best performance of the 8254x is achived only when jumbo frame is
3842 * not used at all on architectures with strict alignment.
3843 */
3844static int
3845lem_fixup_rx(struct adapter *adapter)
3846{
3847 struct mbuf *m, *n;
3848 int error;
3849
3850 error = 0;
3851 m = adapter->fmp;
3852 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) {
3853 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len);
3854 m->m_data += ETHER_HDR_LEN;
3855 } else {
3856 MGETHDR(n, M_NOWAIT, MT_DATA);
3857 if (n != NULL) {
3858 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN);
3859 m->m_data += ETHER_HDR_LEN;
3860 m->m_len -= ETHER_HDR_LEN;
3861 n->m_len = ETHER_HDR_LEN;
3862 M_MOVE_PKTHDR(n, m);
3863 n->m_next = m;
3864 adapter->fmp = n;
3865 } else {
3866 adapter->dropped_pkts++;
3867 m_freem(adapter->fmp);
3868 adapter->fmp = NULL;
3869 error = ENOMEM;
3870 }
3871 }
3872
3873 return (error);
3874}
3875#endif
3876
3877/*********************************************************************
3878 *
3879 * Verify that the hardware indicated that the checksum is valid.
3880 * Inform the stack about the status of checksum so that stack
3881 * doesn't spend time verifying the checksum.
3882 *
3883 *********************************************************************/
3884static void
3885lem_receive_checksum(struct adapter *adapter,
3886 struct e1000_rx_desc *rx_desc, struct mbuf *mp)
3887{
3888 /* 82543 or newer only */
3889 if ((adapter->hw.mac.type < e1000_82543) ||
3890 /* Ignore Checksum bit is set */
3891 (rx_desc->status & E1000_RXD_STAT_IXSM)) {
3892 mp->m_pkthdr.csum_flags = 0;
3893 return;
3894 }
3895
3896 if (rx_desc->status & E1000_RXD_STAT_IPCS) {
3897 /* Did it pass? */
3898 if (!(rx_desc->errors & E1000_RXD_ERR_IPE)) {
3899 /* IP Checksum Good */
3900 mp->m_pkthdr.csum_flags = CSUM_IP_CHECKED;
3901 mp->m_pkthdr.csum_flags |= CSUM_IP_VALID;
3902
3903 } else {
3904 mp->m_pkthdr.csum_flags = 0;
3905 }
3906 }
3907
3908 if (rx_desc->status & E1000_RXD_STAT_TCPCS) {
3909 /* Did it pass? */
3910 if (!(rx_desc->errors & E1000_RXD_ERR_TCPE)) {
3911 mp->m_pkthdr.csum_flags |=
3912 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
3913 mp->m_pkthdr.csum_data = htons(0xffff);
3914 }
3915 }
3916}
3917
3918/*
3919 * This routine is run via an vlan
3920 * config EVENT
3921 */
3922static void
3923lem_register_vlan(void *arg, if_t ifp, u16 vtag)
3924{
3925 struct adapter *adapter = if_getsoftc(ifp);
3926 u32 index, bit;
3927
3928 if (if_getsoftc(ifp) != arg) /* Not our event */
3929 return;
3930
3931 if ((vtag == 0) || (vtag > 4095)) /* Invalid ID */
3932 return;
3933
3934 EM_CORE_LOCK(adapter);
3935 index = (vtag >> 5) & 0x7F;
3936 bit = vtag & 0x1F;
3937 adapter->shadow_vfta[index] |= (1 << bit);
3938 ++adapter->num_vlans;
3939 /* Re-init to load the changes */
3940 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
3941 lem_init_locked(adapter);
3942 EM_CORE_UNLOCK(adapter);
3943}
3944
3945/*
3946 * This routine is run via an vlan
3947 * unconfig EVENT
3948 */
3949static void
3950lem_unregister_vlan(void *arg, if_t ifp, u16 vtag)
3951{
3952 struct adapter *adapter = if_getsoftc(ifp);
3953 u32 index, bit;
3954
3955 if (if_getsoftc(ifp) != arg)
3956 return;
3957
3958 if ((vtag == 0) || (vtag > 4095)) /* Invalid */
3959 return;
3960
3961 EM_CORE_LOCK(adapter);
3962 index = (vtag >> 5) & 0x7F;
3963 bit = vtag & 0x1F;
3964 adapter->shadow_vfta[index] &= ~(1 << bit);
3965 --adapter->num_vlans;
3966 /* Re-init to load the changes */
3967 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
3968 lem_init_locked(adapter);
3969 EM_CORE_UNLOCK(adapter);
3970}
3971
3972static void
3973lem_setup_vlan_hw_support(struct adapter *adapter)
3974{
3975 struct e1000_hw *hw = &adapter->hw;
3976 u32 reg;
3977
3978 /*
3979 ** We get here thru init_locked, meaning
3980 ** a soft reset, this has already cleared
3981 ** the VFTA and other state, so if there
3982 ** have been no vlan's registered do nothing.
3983 */
3984 if (adapter->num_vlans == 0)
3985 return;
3986
3987 /*
3988 ** A soft reset zero's out the VFTA, so
3989 ** we need to repopulate it now.
3990 */
3991 for (int i = 0; i < EM_VFTA_SIZE; i++)
3992 if (adapter->shadow_vfta[i] != 0)
3993 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3994 i, adapter->shadow_vfta[i]);
3995
3996 reg = E1000_READ_REG(hw, E1000_CTRL);
3997 reg |= E1000_CTRL_VME;
3998 E1000_WRITE_REG(hw, E1000_CTRL, reg);
3999
4000 /* Enable the Filter Table */
4001 reg = E1000_READ_REG(hw, E1000_RCTL);
4002 reg &= ~E1000_RCTL_CFIEN;
4003 reg |= E1000_RCTL_VFE;
4004 E1000_WRITE_REG(hw, E1000_RCTL, reg);
4005}
4006
4007static void
4008lem_enable_intr(struct adapter *adapter)
4009{
4010 struct e1000_hw *hw = &adapter->hw;
4011 u32 ims_mask = IMS_ENABLE_MASK;
4012
4013 E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
4014}
4015
4016static void
4017lem_disable_intr(struct adapter *adapter)
4018{
4019 struct e1000_hw *hw = &adapter->hw;
4020
4021 E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
4022}
4023
4024/*
4025 * Bit of a misnomer, what this really means is
4026 * to enable OS management of the system... aka
4027 * to disable special hardware management features
4028 */
4029static void
4030lem_init_manageability(struct adapter *adapter)
4031{
4032 /* A shared code workaround */
4033 if (adapter->has_manage) {
4034 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
4035 /* disable hardware interception of ARP */
4036 manc &= ~(E1000_MANC_ARP_EN);
4037 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
4038 }
4039}
4040
4041/*
4042 * Give control back to hardware management
4043 * controller if there is one.
4044 */
4045static void
4046lem_release_manageability(struct adapter *adapter)
4047{
4048 if (adapter->has_manage) {
4049 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
4050
4051 /* re-enable hardware interception of ARP */
4052 manc |= E1000_MANC_ARP_EN;
4053 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
4054 }
4055}
4056
4057/*
4058 * lem_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
4059 * For ASF and Pass Through versions of f/w this means
4060 * that the driver is loaded. For AMT version type f/w
4061 * this means that the network i/f is open.
4062 */
4063static void
4064lem_get_hw_control(struct adapter *adapter)
4065{
4066 u32 ctrl_ext;
4067
4068 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4069 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
4070 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
4071 return;
4072}
4073
4074/*
4075 * lem_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
4076 * For ASF and Pass Through versions of f/w this means that
4077 * the driver is no longer loaded. For AMT versions of the
4078 * f/w this means that the network i/f is closed.
4079 */
4080static void
4081lem_release_hw_control(struct adapter *adapter)
4082{
4083 u32 ctrl_ext;
4084
4085 if (!adapter->has_manage)
4086 return;
4087
4088 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4089 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
4090 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
4091 return;
4092}
4093
4094static int
4095lem_is_valid_ether_addr(u8 *addr)
4096{
4097 char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
4098
4099 if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
4100 return (FALSE);
4101 }
4102
4103 return (TRUE);
4104}
4105
4106/*
4107** Parse the interface capabilities with regard
4108** to both system management and wake-on-lan for
4109** later use.
4110*/
4111static void
4112lem_get_wakeup(device_t dev)
4113{
4114 struct adapter *adapter = device_get_softc(dev);
4115 u16 eeprom_data = 0, device_id, apme_mask;
4116
4117 adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
4118 apme_mask = EM_EEPROM_APME;
4119
4120 switch (adapter->hw.mac.type) {
4121 case e1000_82542:
4122 case e1000_82543:
4123 break;
4124 case e1000_82544:
4125 e1000_read_nvm(&adapter->hw,
4126 NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
4127 apme_mask = EM_82544_APME;
4128 break;
4129 case e1000_82546:
4130 case e1000_82546_rev_3:
4131 if (adapter->hw.bus.func == 1) {
4132 e1000_read_nvm(&adapter->hw,
4133 NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
4134 break;
4135 } else
4136 e1000_read_nvm(&adapter->hw,
4137 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4138 break;
4139 default:
4140 e1000_read_nvm(&adapter->hw,
4141 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4142 break;
4143 }
4144 if (eeprom_data & apme_mask)
4145 adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
4146 /*
4147 * We have the eeprom settings, now apply the special cases
4148 * where the eeprom may be wrong or the board won't support
4149 * wake on lan on a particular port
4150 */
4151 device_id = pci_get_device(dev);
4152 switch (device_id) {
4153 case E1000_DEV_ID_82546GB_PCIE:
4154 adapter->wol = 0;
4155 break;
4156 case E1000_DEV_ID_82546EB_FIBER:
4157 case E1000_DEV_ID_82546GB_FIBER:
4158 /* Wake events only supported on port A for dual fiber
4159 * regardless of eeprom setting */
4160 if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
4161 E1000_STATUS_FUNC_1)
4162 adapter->wol = 0;
4163 break;
4164 case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
4165 /* if quad port adapter, disable WoL on all but port A */
4166 if (global_quad_port_a != 0)
4167 adapter->wol = 0;
4168 /* Reset for multiple quad port adapters */
4169 if (++global_quad_port_a == 4)
4170 global_quad_port_a = 0;
4171 break;
4172 }
4173 return;
4174}
4175
4176
4177/*
4178 * Enable PCI Wake On Lan capability
4179 */
4180static void
4181lem_enable_wakeup(device_t dev)
4182{
4183 struct adapter *adapter = device_get_softc(dev);
4184 if_t ifp = adapter->ifp;
4185 u32 pmc, ctrl, ctrl_ext, rctl;
4186 u16 status;
4187
4188 if ((pci_find_cap(dev, PCIY_PMG, &pmc) != 0))
4189 return;
4190
4191 /* Advertise the wakeup capability */
4192 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
4193 ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
4194 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
4195 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
4196
4197 /* Keep the laser running on Fiber adapters */
4198 if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
4199 adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
4200 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
4201 ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
4202 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
4203 }
4204
4205 /*
4206 ** Determine type of Wakeup: note that wol
4207 ** is set with all bits on by default.
4208 */
4209 if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
4210 adapter->wol &= ~E1000_WUFC_MAG;
4211
4212 if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
4213 adapter->wol &= ~E1000_WUFC_MC;
4214 else {
4215 rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
4216 rctl |= E1000_RCTL_MPE;
4217 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
4218 }
4219
4220 if (adapter->hw.mac.type == e1000_pchlan) {
4221 if (lem_enable_phy_wakeup(adapter))
4222 return;
4223 } else {
4224 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
4225 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
4226 }
4227
4228
4229 /* Request PME */
4230 status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
4231 status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
4232 if (if_getcapenable(ifp) & IFCAP_WOL)
4233 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
4234 pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
4235
4236 return;
4237}
4238
4239/*
4240** WOL in the newer chipset interfaces (pchlan)
4241** require thing to be copied into the phy
4242*/
4243static int
4244lem_enable_phy_wakeup(struct adapter *adapter)
4245{
4246 struct e1000_hw *hw = &adapter->hw;
4247 u32 mreg, ret = 0;
4248 u16 preg;
4249
4250 /* copy MAC RARs to PHY RARs */
4251 for (int i = 0; i < adapter->hw.mac.rar_entry_count; i++) {
4252 mreg = E1000_READ_REG(hw, E1000_RAL(i));
4253 e1000_write_phy_reg(hw, BM_RAR_L(i), (u16)(mreg & 0xFFFF));
4254 e1000_write_phy_reg(hw, BM_RAR_M(i),
4255 (u16)((mreg >> 16) & 0xFFFF));
4256 mreg = E1000_READ_REG(hw, E1000_RAH(i));
4257 e1000_write_phy_reg(hw, BM_RAR_H(i), (u16)(mreg & 0xFFFF));
4258 e1000_write_phy_reg(hw, BM_RAR_CTRL(i),
4259 (u16)((mreg >> 16) & 0xFFFF));
4260 }
4261
4262 /* copy MAC MTA to PHY MTA */
4263 for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
4264 mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
4265 e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
4266 e1000_write_phy_reg(hw, BM_MTA(i) + 1,
4267 (u16)((mreg >> 16) & 0xFFFF));
4268 }
4269
4270 /* configure PHY Rx Control register */
4271 e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
4272 mreg = E1000_READ_REG(hw, E1000_RCTL);
4273 if (mreg & E1000_RCTL_UPE)
4274 preg |= BM_RCTL_UPE;
4275 if (mreg & E1000_RCTL_MPE)
4276 preg |= BM_RCTL_MPE;
4277 preg &= ~(BM_RCTL_MO_MASK);
4278 if (mreg & E1000_RCTL_MO_3)
4279 preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
4280 << BM_RCTL_MO_SHIFT);
4281 if (mreg & E1000_RCTL_BAM)
4282 preg |= BM_RCTL_BAM;
4283 if (mreg & E1000_RCTL_PMCF)
4284 preg |= BM_RCTL_PMCF;
4285 mreg = E1000_READ_REG(hw, E1000_CTRL);
4286 if (mreg & E1000_CTRL_RFCE)
4287 preg |= BM_RCTL_RFCE;
4288 e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
4289
4290 /* enable PHY wakeup in MAC register */
4291 E1000_WRITE_REG(hw, E1000_WUC,
4292 E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
4293 E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
4294
4295 /* configure and enable PHY wakeup in PHY registers */
4296 e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
4297 e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
4298
4299 /* activate PHY wakeup */
4300 ret = hw->phy.ops.acquire(hw);
4301 if (ret) {
4302 printf("Could not acquire PHY\n");
4303 return ret;
4304 }
4305 e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4306 (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
4307 ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
4308 if (ret) {
4309 printf("Could not read PHY page 769\n");
4310 goto out;
4311 }
4312 preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4313 ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4314 if (ret)
4315 printf("Could not set PHY Host Wakeup bit\n");
4316out:
4317 hw->phy.ops.release(hw);
4318
4319 return ret;
4320}
4321
4322static void
4323lem_led_func(void *arg, int onoff)
4324{
4325 struct adapter *adapter = arg;
4326
4327 EM_CORE_LOCK(adapter);
4328 if (onoff) {
4329 e1000_setup_led(&adapter->hw);
4330 e1000_led_on(&adapter->hw);
4331 } else {
4332 e1000_led_off(&adapter->hw);
4333 e1000_cleanup_led(&adapter->hw);
4334 }
4335 EM_CORE_UNLOCK(adapter);
4336}
4337
4338/*********************************************************************
4339* 82544 Coexistence issue workaround.
4340* There are 2 issues.
4341* 1. Transmit Hang issue.
4342* To detect this issue, following equation can be used...
4343* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
4344* If SUM[3:0] is in between 1 to 4, we will have this issue.
4345*
4346* 2. DAC issue.
4347* To detect this issue, following equation can be used...
4348* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
4349* If SUM[3:0] is in between 9 to c, we will have this issue.
4350*
4351*
4352* WORKAROUND:
4353* Make sure we do not have ending address
4354* as 1,2,3,4(Hang) or 9,a,b,c (DAC)
4355*
4356*************************************************************************/
4357static u32
4358lem_fill_descriptors (bus_addr_t address, u32 length,
4359 PDESC_ARRAY desc_array)
4360{
4361 u32 safe_terminator;
4362
4363 /* Since issue is sensitive to length and address.*/
4364 /* Let us first check the address...*/
4365 if (length <= 4) {
4366 desc_array->descriptor[0].address = address;
4367 desc_array->descriptor[0].length = length;
4368 desc_array->elements = 1;
4369 return (desc_array->elements);
4370 }
4371 safe_terminator = (u32)((((u32)address & 0x7) +
4372 (length & 0xF)) & 0xF);
4373 /* if it does not fall between 0x1 to 0x4 and 0x9 to 0xC then return */
4374 if (safe_terminator == 0 ||
4375 (safe_terminator > 4 &&
4376 safe_terminator < 9) ||
4377 (safe_terminator > 0xC &&
4378 safe_terminator <= 0xF)) {
4379 desc_array->descriptor[0].address = address;
4380 desc_array->descriptor[0].length = length;
4381 desc_array->elements = 1;
4382 return (desc_array->elements);
4383 }
4384
4385 desc_array->descriptor[0].address = address;
4386 desc_array->descriptor[0].length = length - 4;
4387 desc_array->descriptor[1].address = address + (length - 4);
4388 desc_array->descriptor[1].length = 4;
4389 desc_array->elements = 2;
4390 return (desc_array->elements);
4391}
4392
4393/**********************************************************************
4394 *
4395 * Update the board statistics counters.
4396 *
4397 **********************************************************************/
4398static void
4399lem_update_stats_counters(struct adapter *adapter)
4400{
4401
4402 if(adapter->hw.phy.media_type == e1000_media_type_copper ||
4403 (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4404 adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
4405 adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
4406 }
4407 adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
4408 adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
4409 adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
4410 adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
4411
4412 adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
4413 adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
4414 adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
4415 adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
4416 adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
4417 adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
4418 adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
4419 adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
4420 adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
4421 adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
4422 adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
4423 adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
4424 adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
4425 adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
4426 adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
4427 adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
4428 adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
4429 adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
4430 adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
4431 adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
4432
4433 /* For the 64-bit byte counters the low dword must be read first. */
4434 /* Both registers clear on the read of the high dword */
4435
4436 adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
4437 ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
4438 adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
4439 ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
4440
4441 adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
4442 adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
4443 adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
4444 adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
4445 adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
4446
4447 adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
4448 adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
4449
4450 adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
4451 adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
4452 adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
4453 adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
4454 adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
4455 adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
4456 adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
4457 adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
4458 adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
4459 adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
4460
4461 if (adapter->hw.mac.type >= e1000_82543) {
4462 adapter->stats.algnerrc +=
4463 E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4464 adapter->stats.rxerrc +=
4465 E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4466 adapter->stats.tncrs +=
4467 E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4468 adapter->stats.cexterr +=
4469 E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4470 adapter->stats.tsctc +=
4471 E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4472 adapter->stats.tsctfc +=
4473 E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4474 }
4475}
4476
4477static uint64_t
4478lem_get_counter(if_t ifp, ift_counter cnt)
4479{
4480 struct adapter *adapter;
4481
4482 adapter = if_getsoftc(ifp);
4483
4484 switch (cnt) {
4485 case IFCOUNTER_COLLISIONS:
4486 return (adapter->stats.colc);
4487 case IFCOUNTER_IERRORS:
4488 return (adapter->dropped_pkts + adapter->stats.rxerrc +
4489 adapter->stats.crcerrs + adapter->stats.algnerrc +
4490 adapter->stats.ruc + adapter->stats.roc +
4491 adapter->stats.mpc + adapter->stats.cexterr);
4492 case IFCOUNTER_OERRORS:
4493 return (adapter->stats.ecol + adapter->stats.latecol +
4494 adapter->watchdog_events);
4495 default:
4496 return (if_get_counter_default(ifp, cnt));
4497 }
4498}
4499
4500/* Export a single 32-bit register via a read-only sysctl. */
4501static int
4502lem_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4503{
4504 struct adapter *adapter;
4505 u_int val;
4506
4507 adapter = oidp->oid_arg1;
4508 val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4509 return (sysctl_handle_int(oidp, &val, 0, req));
4510}
4511
4512/*
4513 * Add sysctl variables, one per statistic, to the system.
4514 */
4515static void
4516lem_add_hw_stats(struct adapter *adapter)
4517{
4518 device_t dev = adapter->dev;
4519
4520 struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4521 struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4522 struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4523 struct e1000_hw_stats *stats = &adapter->stats;
4524
4525 struct sysctl_oid *stat_node;
4526 struct sysctl_oid_list *stat_list;
4527
4528 /* Driver Statistics */
4529 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_alloc_fail",
4530 CTLFLAG_RD, &adapter->mbuf_alloc_failed,
4531 "Std mbuf failed");
4532 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "cluster_alloc_fail",
4533 CTLFLAG_RD, &adapter->mbuf_cluster_failed,
4534 "Std mbuf cluster failed");
4535 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4536 CTLFLAG_RD, &adapter->dropped_pkts,
4537 "Driver dropped packets");
4538 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
4539 CTLFLAG_RD, &adapter->no_tx_dma_setup,
4540 "Driver tx dma failure in xmit");
4541 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_desc_fail1",
4542 CTLFLAG_RD, &adapter->no_tx_desc_avail1,
4543 "Not enough tx descriptors failure in xmit");
4544 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_desc_fail2",
4545 CTLFLAG_RD, &adapter->no_tx_desc_avail2,
4546 "Not enough tx descriptors failure in xmit");
4547 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4548 CTLFLAG_RD, &adapter->rx_overruns,
4549 "RX overruns");
4550 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4551 CTLFLAG_RD, &adapter->watchdog_events,
4552 "Watchdog timeouts");
4553
4554 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4555 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4556 lem_sysctl_reg_handler, "IU",
4557 "Device Control Register");
4558 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4559 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4560 lem_sysctl_reg_handler, "IU",
4561 "Receiver Control Register");
4562 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4563 CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4564 "Flow Control High Watermark");
4565 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4566 CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4567 "Flow Control Low Watermark");
4568 SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "fifo_workaround",
4569 CTLFLAG_RD, &adapter->tx_fifo_wrk_cnt,
4570 "TX FIFO workaround events");
4571 SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "fifo_reset",
4572 CTLFLAG_RD, &adapter->tx_fifo_reset_cnt,
4573 "TX FIFO resets");
4574
4575 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "txd_head",
4576 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_TDH(0),
4577 lem_sysctl_reg_handler, "IU",
4578 "Transmit Descriptor Head");
4579 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "txd_tail",
4580 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_TDT(0),
4581 lem_sysctl_reg_handler, "IU",
4582 "Transmit Descriptor Tail");
4583 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rxd_head",
4584 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RDH(0),
4585 lem_sysctl_reg_handler, "IU",
4586 "Receive Descriptor Head");
4587 SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rxd_tail",
4588 CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RDT(0),
4589 lem_sysctl_reg_handler, "IU",
4590 "Receive Descriptor Tail");
4591
4592
4593 /* MAC stats get their own sub node */
4594
4595 stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4596 CTLFLAG_RD, NULL, "Statistics");
4597 stat_list = SYSCTL_CHILDREN(stat_node);
4598
4599 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4600 CTLFLAG_RD, &stats->ecol,
4601 "Excessive collisions");
4602 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4603 CTLFLAG_RD, &stats->scc,
4604 "Single collisions");
4605 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4606 CTLFLAG_RD, &stats->mcc,
4607 "Multiple collisions");
4608 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4609 CTLFLAG_RD, &stats->latecol,
4610 "Late collisions");
4611 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4612 CTLFLAG_RD, &stats->colc,
4613 "Collision Count");
4614 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4615 CTLFLAG_RD, &adapter->stats.symerrs,
4616 "Symbol Errors");
4617 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4618 CTLFLAG_RD, &adapter->stats.sec,
4619 "Sequence Errors");
4620 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4621 CTLFLAG_RD, &adapter->stats.dc,
4622 "Defer Count");
4623 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4624 CTLFLAG_RD, &adapter->stats.mpc,
4625 "Missed Packets");
4626 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4627 CTLFLAG_RD, &adapter->stats.rnbc,
4628 "Receive No Buffers");
4629 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4630 CTLFLAG_RD, &adapter->stats.ruc,
4631 "Receive Undersize");
4632 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4633 CTLFLAG_RD, &adapter->stats.rfc,
4634 "Fragmented Packets Received ");
4635 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4636 CTLFLAG_RD, &adapter->stats.roc,
4637 "Oversized Packets Received");
4638 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4639 CTLFLAG_RD, &adapter->stats.rjc,
4640 "Recevied Jabber");
4641 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4642 CTLFLAG_RD, &adapter->stats.rxerrc,
4643 "Receive Errors");
4644 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4645 CTLFLAG_RD, &adapter->stats.crcerrs,
4646 "CRC errors");
4647 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4648 CTLFLAG_RD, &adapter->stats.algnerrc,
4649 "Alignment Errors");
4650 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4651 CTLFLAG_RD, &adapter->stats.cexterr,
4652 "Collision/Carrier extension errors");
4653 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4654 CTLFLAG_RD, &adapter->stats.xonrxc,
4655 "XON Received");
4656 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4657 CTLFLAG_RD, &adapter->stats.xontxc,
4658 "XON Transmitted");
4659 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4660 CTLFLAG_RD, &adapter->stats.xoffrxc,
4661 "XOFF Received");
4662 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4663 CTLFLAG_RD, &adapter->stats.xofftxc,
4664 "XOFF Transmitted");
4665
4666 /* Packet Reception Stats */
4667 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4668 CTLFLAG_RD, &adapter->stats.tpr,
4669 "Total Packets Received ");
4670 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4671 CTLFLAG_RD, &adapter->stats.gprc,
4672 "Good Packets Received");
4673 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4674 CTLFLAG_RD, &adapter->stats.bprc,
4675 "Broadcast Packets Received");
4676 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4677 CTLFLAG_RD, &adapter->stats.mprc,
4678 "Multicast Packets Received");
4679 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4680 CTLFLAG_RD, &adapter->stats.prc64,
4681 "64 byte frames received ");
4682 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4683 CTLFLAG_RD, &adapter->stats.prc127,
4684 "65-127 byte frames received");
4685 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4686 CTLFLAG_RD, &adapter->stats.prc255,
4687 "128-255 byte frames received");
4688 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4689 CTLFLAG_RD, &adapter->stats.prc511,
4690 "256-511 byte frames received");
4691 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4692 CTLFLAG_RD, &adapter->stats.prc1023,
4693 "512-1023 byte frames received");
4694 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4695 CTLFLAG_RD, &adapter->stats.prc1522,
4696 "1023-1522 byte frames received");
4697 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4698 CTLFLAG_RD, &adapter->stats.gorc,
4699 "Good Octets Received");
4700
4701 /* Packet Transmission Stats */
4702 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4703 CTLFLAG_RD, &adapter->stats.gotc,
4704 "Good Octets Transmitted");
4705 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4706 CTLFLAG_RD, &adapter->stats.tpt,
4707 "Total Packets Transmitted");
4708 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4709 CTLFLAG_RD, &adapter->stats.gptc,
4710 "Good Packets Transmitted");
4711 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4712 CTLFLAG_RD, &adapter->stats.bptc,
4713 "Broadcast Packets Transmitted");
4714 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4715 CTLFLAG_RD, &adapter->stats.mptc,
4716 "Multicast Packets Transmitted");
4717 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4718 CTLFLAG_RD, &adapter->stats.ptc64,
4719 "64 byte frames transmitted ");
4720 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4721 CTLFLAG_RD, &adapter->stats.ptc127,
4722 "65-127 byte frames transmitted");
4723 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4724 CTLFLAG_RD, &adapter->stats.ptc255,
4725 "128-255 byte frames transmitted");
4726 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4727 CTLFLAG_RD, &adapter->stats.ptc511,
4728 "256-511 byte frames transmitted");
4729 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4730 CTLFLAG_RD, &adapter->stats.ptc1023,
4731 "512-1023 byte frames transmitted");
4732 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4733 CTLFLAG_RD, &adapter->stats.ptc1522,
4734 "1024-1522 byte frames transmitted");
4735 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4736 CTLFLAG_RD, &adapter->stats.tsctc,
4737 "TSO Contexts Transmitted");
4738 SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4739 CTLFLAG_RD, &adapter->stats.tsctfc,
4740 "TSO Contexts Failed");
4741}
4742
4743/**********************************************************************
4744 *
4745 * This routine provides a way to dump out the adapter eeprom,
4746 * often a useful debug/service tool. This only dumps the first
4747 * 32 words, stuff that matters is in that extent.
4748 *
4749 **********************************************************************/
4750
4751static int
4752lem_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4753{
4754 struct adapter *adapter;
4755 int error;
4756 int result;
4757
4758 result = -1;
4759 error = sysctl_handle_int(oidp, &result, 0, req);
4760
4761 if (error || !req->newptr)
4762 return (error);
4763
4764 /*
4765 * This value will cause a hex dump of the
4766 * first 32 16-bit words of the EEPROM to
4767 * the screen.
4768 */
4769 if (result == 1) {
4770 adapter = (struct adapter *)arg1;
4771 lem_print_nvm_info(adapter);
4772 }
4773
4774 return (error);
4775}
4776
4777static void
4778lem_print_nvm_info(struct adapter *adapter)
4779{
4780 u16 eeprom_data;
4781 int i, j, row = 0;
4782
4783 /* Its a bit crude, but it gets the job done */
4784 printf("\nInterface EEPROM Dump:\n");
4785 printf("Offset\n0x0000 ");
4786 for (i = 0, j = 0; i < 32; i++, j++) {
4787 if (j == 8) { /* Make the offset block */
4788 j = 0; ++row;
4789 printf("\n0x00%x0 ",row);
4790 }
4791 e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4792 printf("%04x ", eeprom_data);
4793 }
4794 printf("\n");
4795}
4796
4797static int
4798lem_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4799{
4800 struct em_int_delay_info *info;
4801 struct adapter *adapter;
4802 u32 regval;
4803 int error;
4804 int usecs;
4805 int ticks;
4806
4807 info = (struct em_int_delay_info *)arg1;
4808 usecs = info->value;
4809 error = sysctl_handle_int(oidp, &usecs, 0, req);
4810 if (error != 0 || req->newptr == NULL)
4811 return (error);
4812 if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4813 return (EINVAL);
4814 info->value = usecs;
4815 ticks = EM_USECS_TO_TICKS(usecs);
4816 if (info->offset == E1000_ITR) /* units are 256ns here */
4817 ticks *= 4;
4818
4819 adapter = info->adapter;
4820
4821 EM_CORE_LOCK(adapter);
4822 regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4823 regval = (regval & ~0xffff) | (ticks & 0xffff);
4824 /* Handle a few special cases. */
4825 switch (info->offset) {
4826 case E1000_RDTR:
4827 break;
4828 case E1000_TIDV:
4829 if (ticks == 0) {
4830 adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4831 /* Don't write 0 into the TIDV register. */
4832 regval++;
4833 } else
4834 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4835 break;
4836 }
4837 E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4838 EM_CORE_UNLOCK(adapter);
4839 return (0);
4840}
4841
4842static void
4843lem_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4844 const char *description, struct em_int_delay_info *info,
4845 int offset, int value)
4846{
4847 info->adapter = adapter;
4848 info->offset = offset;
4849 info->value = value;
4850 SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4851 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4852 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4853 info, 0, lem_sysctl_int_delay, "I", description);
4854}
4855
4856static void
4857lem_set_flow_cntrl(struct adapter *adapter, const char *name,
4858 const char *description, int *limit, int value)
4859{
4860 *limit = value;
4861 SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
4862 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4863 OID_AUTO, name, CTLFLAG_RW, limit, value, description);
4864}
4865
4866static void
4867lem_add_rx_process_limit(struct adapter *adapter, const char *name,
4868 const char *description, int *limit, int value)
4869{
4870 *limit = value;
4871 SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
4872 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4873 OID_AUTO, name, CTLFLAG_RW, limit, value, description);
4874}
|