if_bge.c revision 258959
1/*-
2 * Copyright (c) 2001 Wind River Systems
3 * Copyright (c) 1997, 1998, 1999, 2001
4 *	Bill Paul <wpaul@windriver.com>.  All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 * 3. All advertising materials mentioning features or use of this software
15 *    must display the following acknowledgement:
16 *	This product includes software developed by Bill Paul.
17 * 4. Neither the name of the author nor the names of any co-contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
31 * THE POSSIBILITY OF SUCH DAMAGE.
32 */
33
34#include <sys/cdefs.h>
35__FBSDID("$FreeBSD: stable/10/sys/dev/bge/if_bge.c 258959 2013-12-05 06:13:59Z yongari $");
36
37/*
38 * Broadcom BCM57xx(x)/BCM590x NetXtreme and NetLink family Ethernet driver
39 *
40 * The Broadcom BCM5700 is based on technology originally developed by
41 * Alteon Networks as part of the Tigon I and Tigon II Gigabit Ethernet
42 * MAC chips. The BCM5700, sometimes referred to as the Tigon III, has
43 * two on-board MIPS R4000 CPUs and can have as much as 16MB of external
44 * SSRAM. The BCM5700 supports TCP, UDP and IP checksum offload, jumbo
45 * frames, highly configurable RX filtering, and 16 RX and TX queues
46 * (which, along with RX filter rules, can be used for QOS applications).
47 * Other features, such as TCP segmentation, may be available as part
48 * of value-added firmware updates. Unlike the Tigon I and Tigon II,
49 * firmware images can be stored in hardware and need not be compiled
50 * into the driver.
51 *
52 * The BCM5700 supports the PCI v2.2 and PCI-X v1.0 standards, and will
53 * function in a 32-bit/64-bit 33/66Mhz bus, or a 64-bit/133Mhz bus.
54 *
55 * The BCM5701 is a single-chip solution incorporating both the BCM5700
56 * MAC and a BCM5401 10/100/1000 PHY. Unlike the BCM5700, the BCM5701
57 * does not support external SSRAM.
58 *
59 * Broadcom also produces a variation of the BCM5700 under the "Altima"
60 * brand name, which is functionally similar but lacks PCI-X support.
61 *
62 * Without external SSRAM, you can only have at most 4 TX rings,
63 * and the use of the mini RX ring is disabled. This seems to imply
64 * that these features are simply not available on the BCM5701. As a
65 * result, this driver does not implement any support for the mini RX
66 * ring.
67 */
68
69#ifdef HAVE_KERNEL_OPTION_HEADERS
70#include "opt_device_polling.h"
71#endif
72
73#include <sys/param.h>
74#include <sys/endian.h>
75#include <sys/systm.h>
76#include <sys/sockio.h>
77#include <sys/mbuf.h>
78#include <sys/malloc.h>
79#include <sys/kernel.h>
80#include <sys/module.h>
81#include <sys/socket.h>
82#include <sys/sysctl.h>
83#include <sys/taskqueue.h>
84
85#include <net/if.h>
86#include <net/if_arp.h>
87#include <net/ethernet.h>
88#include <net/if_dl.h>
89#include <net/if_media.h>
90
91#include <net/bpf.h>
92
93#include <net/if_types.h>
94#include <net/if_vlan_var.h>
95
96#include <netinet/in_systm.h>
97#include <netinet/in.h>
98#include <netinet/ip.h>
99#include <netinet/tcp.h>
100
101#include <machine/bus.h>
102#include <machine/resource.h>
103#include <sys/bus.h>
104#include <sys/rman.h>
105
106#include <dev/mii/mii.h>
107#include <dev/mii/miivar.h>
108#include "miidevs.h"
109#include <dev/mii/brgphyreg.h>
110
111#ifdef __sparc64__
112#include <dev/ofw/ofw_bus.h>
113#include <dev/ofw/openfirm.h>
114#include <machine/ofw_machdep.h>
115#include <machine/ver.h>
116#endif
117
118#include <dev/pci/pcireg.h>
119#include <dev/pci/pcivar.h>
120
121#include <dev/bge/if_bgereg.h>
122
123#define	BGE_CSUM_FEATURES	(CSUM_IP | CSUM_TCP)
124#define	ETHER_MIN_NOPAD		(ETHER_MIN_LEN - ETHER_CRC_LEN) /* i.e., 60 */
125
126MODULE_DEPEND(bge, pci, 1, 1, 1);
127MODULE_DEPEND(bge, ether, 1, 1, 1);
128MODULE_DEPEND(bge, miibus, 1, 1, 1);
129
130/* "device miibus" required.  See GENERIC if you get errors here. */
131#include "miibus_if.h"
132
133/*
134 * Various supported device vendors/types and their names. Note: the
135 * spec seems to indicate that the hardware still has Alteon's vendor
136 * ID burned into it, though it will always be overriden by the vendor
137 * ID in the EEPROM. Just to be safe, we cover all possibilities.
138 */
139static const struct bge_type {
140	uint16_t	bge_vid;
141	uint16_t	bge_did;
142} bge_devs[] = {
143	{ ALTEON_VENDORID,	ALTEON_DEVICEID_BCM5700 },
144	{ ALTEON_VENDORID,	ALTEON_DEVICEID_BCM5701 },
145
146	{ ALTIMA_VENDORID,	ALTIMA_DEVICE_AC1000 },
147	{ ALTIMA_VENDORID,	ALTIMA_DEVICE_AC1002 },
148	{ ALTIMA_VENDORID,	ALTIMA_DEVICE_AC9100 },
149
150	{ APPLE_VENDORID,	APPLE_DEVICE_BCM5701 },
151
152	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5700 },
153	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5701 },
154	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5702 },
155	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5702_ALT },
156	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5702X },
157	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5703 },
158	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5703_ALT },
159	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5703X },
160	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5704C },
161	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5704S },
162	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5704S_ALT },
163	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5705 },
164	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5705F },
165	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5705K },
166	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5705M },
167	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5705M_ALT },
168	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5714C },
169	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5714S },
170	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5715 },
171	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5715S },
172	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5717 },
173	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5718 },
174	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5719 },
175	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5720 },
176	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5721 },
177	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5722 },
178	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5723 },
179	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5725 },
180	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5727 },
181	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5750 },
182	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5750M },
183	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5751 },
184	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5751F },
185	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5751M },
186	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5752 },
187	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5752M },
188	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5753 },
189	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5753F },
190	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5753M },
191	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5754 },
192	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5754M },
193	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5755 },
194	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5755M },
195	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5756 },
196	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5761 },
197	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5761E },
198	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5761S },
199	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5761SE },
200	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5762 },
201	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5764 },
202	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5780 },
203	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5780S },
204	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5781 },
205	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5782 },
206	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5784 },
207	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5785F },
208	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5785G },
209	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5786 },
210	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5787 },
211	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5787F },
212	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5787M },
213	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5788 },
214	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5789 },
215	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5901 },
216	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5901A2 },
217	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5903M },
218	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5906 },
219	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM5906M },
220	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57760 },
221	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57761 },
222	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57762 },
223	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57764 },
224	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57765 },
225	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57766 },
226	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57767 },
227	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57780 },
228	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57781 },
229	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57782 },
230	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57785 },
231	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57786 },
232	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57787 },
233	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57788 },
234	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57790 },
235	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57791 },
236	{ BCOM_VENDORID,	BCOM_DEVICEID_BCM57795 },
237
238	{ SK_VENDORID,		SK_DEVICEID_ALTIMA },
239
240	{ TC_VENDORID,		TC_DEVICEID_3C996 },
241
242	{ FJTSU_VENDORID,	FJTSU_DEVICEID_PW008GE4 },
243	{ FJTSU_VENDORID,	FJTSU_DEVICEID_PW008GE5 },
244	{ FJTSU_VENDORID,	FJTSU_DEVICEID_PP250450 },
245
246	{ 0, 0 }
247};
248
249static const struct bge_vendor {
250	uint16_t	v_id;
251	const char	*v_name;
252} bge_vendors[] = {
253	{ ALTEON_VENDORID,	"Alteon" },
254	{ ALTIMA_VENDORID,	"Altima" },
255	{ APPLE_VENDORID,	"Apple" },
256	{ BCOM_VENDORID,	"Broadcom" },
257	{ SK_VENDORID,		"SysKonnect" },
258	{ TC_VENDORID,		"3Com" },
259	{ FJTSU_VENDORID,	"Fujitsu" },
260
261	{ 0, NULL }
262};
263
264static const struct bge_revision {
265	uint32_t	br_chipid;
266	const char	*br_name;
267} bge_revisions[] = {
268	{ BGE_CHIPID_BCM5700_A0,	"BCM5700 A0" },
269	{ BGE_CHIPID_BCM5700_A1,	"BCM5700 A1" },
270	{ BGE_CHIPID_BCM5700_B0,	"BCM5700 B0" },
271	{ BGE_CHIPID_BCM5700_B1,	"BCM5700 B1" },
272	{ BGE_CHIPID_BCM5700_B2,	"BCM5700 B2" },
273	{ BGE_CHIPID_BCM5700_B3,	"BCM5700 B3" },
274	{ BGE_CHIPID_BCM5700_ALTIMA,	"BCM5700 Altima" },
275	{ BGE_CHIPID_BCM5700_C0,	"BCM5700 C0" },
276	{ BGE_CHIPID_BCM5701_A0,	"BCM5701 A0" },
277	{ BGE_CHIPID_BCM5701_B0,	"BCM5701 B0" },
278	{ BGE_CHIPID_BCM5701_B2,	"BCM5701 B2" },
279	{ BGE_CHIPID_BCM5701_B5,	"BCM5701 B5" },
280	{ BGE_CHIPID_BCM5703_A0,	"BCM5703 A0" },
281	{ BGE_CHIPID_BCM5703_A1,	"BCM5703 A1" },
282	{ BGE_CHIPID_BCM5703_A2,	"BCM5703 A2" },
283	{ BGE_CHIPID_BCM5703_A3,	"BCM5703 A3" },
284	{ BGE_CHIPID_BCM5703_B0,	"BCM5703 B0" },
285	{ BGE_CHIPID_BCM5704_A0,	"BCM5704 A0" },
286	{ BGE_CHIPID_BCM5704_A1,	"BCM5704 A1" },
287	{ BGE_CHIPID_BCM5704_A2,	"BCM5704 A2" },
288	{ BGE_CHIPID_BCM5704_A3,	"BCM5704 A3" },
289	{ BGE_CHIPID_BCM5704_B0,	"BCM5704 B0" },
290	{ BGE_CHIPID_BCM5705_A0,	"BCM5705 A0" },
291	{ BGE_CHIPID_BCM5705_A1,	"BCM5705 A1" },
292	{ BGE_CHIPID_BCM5705_A2,	"BCM5705 A2" },
293	{ BGE_CHIPID_BCM5705_A3,	"BCM5705 A3" },
294	{ BGE_CHIPID_BCM5750_A0,	"BCM5750 A0" },
295	{ BGE_CHIPID_BCM5750_A1,	"BCM5750 A1" },
296	{ BGE_CHIPID_BCM5750_A3,	"BCM5750 A3" },
297	{ BGE_CHIPID_BCM5750_B0,	"BCM5750 B0" },
298	{ BGE_CHIPID_BCM5750_B1,	"BCM5750 B1" },
299	{ BGE_CHIPID_BCM5750_C0,	"BCM5750 C0" },
300	{ BGE_CHIPID_BCM5750_C1,	"BCM5750 C1" },
301	{ BGE_CHIPID_BCM5750_C2,	"BCM5750 C2" },
302	{ BGE_CHIPID_BCM5714_A0,	"BCM5714 A0" },
303	{ BGE_CHIPID_BCM5752_A0,	"BCM5752 A0" },
304	{ BGE_CHIPID_BCM5752_A1,	"BCM5752 A1" },
305	{ BGE_CHIPID_BCM5752_A2,	"BCM5752 A2" },
306	{ BGE_CHIPID_BCM5714_B0,	"BCM5714 B0" },
307	{ BGE_CHIPID_BCM5714_B3,	"BCM5714 B3" },
308	{ BGE_CHIPID_BCM5715_A0,	"BCM5715 A0" },
309	{ BGE_CHIPID_BCM5715_A1,	"BCM5715 A1" },
310	{ BGE_CHIPID_BCM5715_A3,	"BCM5715 A3" },
311	{ BGE_CHIPID_BCM5717_A0,	"BCM5717 A0" },
312	{ BGE_CHIPID_BCM5717_B0,	"BCM5717 B0" },
313	{ BGE_CHIPID_BCM5719_A0,	"BCM5719 A0" },
314	{ BGE_CHIPID_BCM5720_A0,	"BCM5720 A0" },
315	{ BGE_CHIPID_BCM5755_A0,	"BCM5755 A0" },
316	{ BGE_CHIPID_BCM5755_A1,	"BCM5755 A1" },
317	{ BGE_CHIPID_BCM5755_A2,	"BCM5755 A2" },
318	{ BGE_CHIPID_BCM5722_A0,	"BCM5722 A0" },
319	{ BGE_CHIPID_BCM5761_A0,	"BCM5761 A0" },
320	{ BGE_CHIPID_BCM5761_A1,	"BCM5761 A1" },
321	{ BGE_CHIPID_BCM5762_A0,	"BCM5762 A0" },
322	{ BGE_CHIPID_BCM5784_A0,	"BCM5784 A0" },
323	{ BGE_CHIPID_BCM5784_A1,	"BCM5784 A1" },
324	/* 5754 and 5787 share the same ASIC ID */
325	{ BGE_CHIPID_BCM5787_A0,	"BCM5754/5787 A0" },
326	{ BGE_CHIPID_BCM5787_A1,	"BCM5754/5787 A1" },
327	{ BGE_CHIPID_BCM5787_A2,	"BCM5754/5787 A2" },
328	{ BGE_CHIPID_BCM5906_A1,	"BCM5906 A1" },
329	{ BGE_CHIPID_BCM5906_A2,	"BCM5906 A2" },
330	{ BGE_CHIPID_BCM57765_A0,	"BCM57765 A0" },
331	{ BGE_CHIPID_BCM57765_B0,	"BCM57765 B0" },
332	{ BGE_CHIPID_BCM57780_A0,	"BCM57780 A0" },
333	{ BGE_CHIPID_BCM57780_A1,	"BCM57780 A1" },
334
335	{ 0, NULL }
336};
337
338/*
339 * Some defaults for major revisions, so that newer steppings
340 * that we don't know about have a shot at working.
341 */
342static const struct bge_revision bge_majorrevs[] = {
343	{ BGE_ASICREV_BCM5700,		"unknown BCM5700" },
344	{ BGE_ASICREV_BCM5701,		"unknown BCM5701" },
345	{ BGE_ASICREV_BCM5703,		"unknown BCM5703" },
346	{ BGE_ASICREV_BCM5704,		"unknown BCM5704" },
347	{ BGE_ASICREV_BCM5705,		"unknown BCM5705" },
348	{ BGE_ASICREV_BCM5750,		"unknown BCM5750" },
349	{ BGE_ASICREV_BCM5714_A0,	"unknown BCM5714" },
350	{ BGE_ASICREV_BCM5752,		"unknown BCM5752" },
351	{ BGE_ASICREV_BCM5780,		"unknown BCM5780" },
352	{ BGE_ASICREV_BCM5714,		"unknown BCM5714" },
353	{ BGE_ASICREV_BCM5755,		"unknown BCM5755" },
354	{ BGE_ASICREV_BCM5761,		"unknown BCM5761" },
355	{ BGE_ASICREV_BCM5784,		"unknown BCM5784" },
356	{ BGE_ASICREV_BCM5785,		"unknown BCM5785" },
357	/* 5754 and 5787 share the same ASIC ID */
358	{ BGE_ASICREV_BCM5787,		"unknown BCM5754/5787" },
359	{ BGE_ASICREV_BCM5906,		"unknown BCM5906" },
360	{ BGE_ASICREV_BCM57765,		"unknown BCM57765" },
361	{ BGE_ASICREV_BCM57766,		"unknown BCM57766" },
362	{ BGE_ASICREV_BCM57780,		"unknown BCM57780" },
363	{ BGE_ASICREV_BCM5717,		"unknown BCM5717" },
364	{ BGE_ASICREV_BCM5719,		"unknown BCM5719" },
365	{ BGE_ASICREV_BCM5720,		"unknown BCM5720" },
366	{ BGE_ASICREV_BCM5762,		"unknown BCM5762" },
367
368	{ 0, NULL }
369};
370
371#define	BGE_IS_JUMBO_CAPABLE(sc)	((sc)->bge_flags & BGE_FLAG_JUMBO)
372#define	BGE_IS_5700_FAMILY(sc)		((sc)->bge_flags & BGE_FLAG_5700_FAMILY)
373#define	BGE_IS_5705_PLUS(sc)		((sc)->bge_flags & BGE_FLAG_5705_PLUS)
374#define	BGE_IS_5714_FAMILY(sc)		((sc)->bge_flags & BGE_FLAG_5714_FAMILY)
375#define	BGE_IS_575X_PLUS(sc)		((sc)->bge_flags & BGE_FLAG_575X_PLUS)
376#define	BGE_IS_5755_PLUS(sc)		((sc)->bge_flags & BGE_FLAG_5755_PLUS)
377#define	BGE_IS_5717_PLUS(sc)		((sc)->bge_flags & BGE_FLAG_5717_PLUS)
378#define	BGE_IS_57765_PLUS(sc)		((sc)->bge_flags & BGE_FLAG_57765_PLUS)
379
380static uint32_t bge_chipid(device_t);
381static const struct bge_vendor * bge_lookup_vendor(uint16_t);
382static const struct bge_revision * bge_lookup_rev(uint32_t);
383
384typedef int	(*bge_eaddr_fcn_t)(struct bge_softc *, uint8_t[]);
385
386static int bge_probe(device_t);
387static int bge_attach(device_t);
388static int bge_detach(device_t);
389static int bge_suspend(device_t);
390static int bge_resume(device_t);
391static void bge_release_resources(struct bge_softc *);
392static void bge_dma_map_addr(void *, bus_dma_segment_t *, int, int);
393static int bge_dma_alloc(struct bge_softc *);
394static void bge_dma_free(struct bge_softc *);
395static int bge_dma_ring_alloc(struct bge_softc *, bus_size_t, bus_size_t,
396    bus_dma_tag_t *, uint8_t **, bus_dmamap_t *, bus_addr_t *, const char *);
397
398static void bge_devinfo(struct bge_softc *);
399static int bge_mbox_reorder(struct bge_softc *);
400
401static int bge_get_eaddr_fw(struct bge_softc *sc, uint8_t ether_addr[]);
402static int bge_get_eaddr_mem(struct bge_softc *, uint8_t[]);
403static int bge_get_eaddr_nvram(struct bge_softc *, uint8_t[]);
404static int bge_get_eaddr_eeprom(struct bge_softc *, uint8_t[]);
405static int bge_get_eaddr(struct bge_softc *, uint8_t[]);
406
407static void bge_txeof(struct bge_softc *, uint16_t);
408static void bge_rxcsum(struct bge_softc *, struct bge_rx_bd *, struct mbuf *);
409static int bge_rxeof(struct bge_softc *, uint16_t, int);
410
411static void bge_asf_driver_up (struct bge_softc *);
412static void bge_tick(void *);
413static void bge_stats_clear_regs(struct bge_softc *);
414static void bge_stats_update(struct bge_softc *);
415static void bge_stats_update_regs(struct bge_softc *);
416static struct mbuf *bge_check_short_dma(struct mbuf *);
417static struct mbuf *bge_setup_tso(struct bge_softc *, struct mbuf *,
418    uint16_t *, uint16_t *);
419static int bge_encap(struct bge_softc *, struct mbuf **, uint32_t *);
420
421static void bge_intr(void *);
422static int bge_msi_intr(void *);
423static void bge_intr_task(void *, int);
424static void bge_start_locked(struct ifnet *);
425static void bge_start(struct ifnet *);
426static int bge_ioctl(struct ifnet *, u_long, caddr_t);
427static void bge_init_locked(struct bge_softc *);
428static void bge_init(void *);
429static void bge_stop_block(struct bge_softc *, bus_size_t, uint32_t);
430static void bge_stop(struct bge_softc *);
431static void bge_watchdog(struct bge_softc *);
432static int bge_shutdown(device_t);
433static int bge_ifmedia_upd_locked(struct ifnet *);
434static int bge_ifmedia_upd(struct ifnet *);
435static void bge_ifmedia_sts(struct ifnet *, struct ifmediareq *);
436
437static uint8_t bge_nvram_getbyte(struct bge_softc *, int, uint8_t *);
438static int bge_read_nvram(struct bge_softc *, caddr_t, int, int);
439
440static uint8_t bge_eeprom_getbyte(struct bge_softc *, int, uint8_t *);
441static int bge_read_eeprom(struct bge_softc *, caddr_t, int, int);
442
443static void bge_setpromisc(struct bge_softc *);
444static void bge_setmulti(struct bge_softc *);
445static void bge_setvlan(struct bge_softc *);
446
447static __inline void bge_rxreuse_std(struct bge_softc *, int);
448static __inline void bge_rxreuse_jumbo(struct bge_softc *, int);
449static int bge_newbuf_std(struct bge_softc *, int);
450static int bge_newbuf_jumbo(struct bge_softc *, int);
451static int bge_init_rx_ring_std(struct bge_softc *);
452static void bge_free_rx_ring_std(struct bge_softc *);
453static int bge_init_rx_ring_jumbo(struct bge_softc *);
454static void bge_free_rx_ring_jumbo(struct bge_softc *);
455static void bge_free_tx_ring(struct bge_softc *);
456static int bge_init_tx_ring(struct bge_softc *);
457
458static int bge_chipinit(struct bge_softc *);
459static int bge_blockinit(struct bge_softc *);
460static uint32_t bge_dma_swap_options(struct bge_softc *);
461
462static int bge_has_eaddr(struct bge_softc *);
463static uint32_t bge_readmem_ind(struct bge_softc *, int);
464static void bge_writemem_ind(struct bge_softc *, int, int);
465static void bge_writembx(struct bge_softc *, int, int);
466#ifdef notdef
467static uint32_t bge_readreg_ind(struct bge_softc *, int);
468#endif
469static void bge_writemem_direct(struct bge_softc *, int, int);
470static void bge_writereg_ind(struct bge_softc *, int, int);
471
472static int bge_miibus_readreg(device_t, int, int);
473static int bge_miibus_writereg(device_t, int, int, int);
474static void bge_miibus_statchg(device_t);
475#ifdef DEVICE_POLLING
476static int bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count);
477#endif
478
479#define	BGE_RESET_SHUTDOWN	0
480#define	BGE_RESET_START		1
481#define	BGE_RESET_SUSPEND	2
482static void bge_sig_post_reset(struct bge_softc *, int);
483static void bge_sig_legacy(struct bge_softc *, int);
484static void bge_sig_pre_reset(struct bge_softc *, int);
485static void bge_stop_fw(struct bge_softc *);
486static int bge_reset(struct bge_softc *);
487static void bge_link_upd(struct bge_softc *);
488
489static void bge_ape_lock_init(struct bge_softc *);
490static void bge_ape_read_fw_ver(struct bge_softc *);
491static int bge_ape_lock(struct bge_softc *, int);
492static void bge_ape_unlock(struct bge_softc *, int);
493static void bge_ape_send_event(struct bge_softc *, uint32_t);
494static void bge_ape_driver_state_change(struct bge_softc *, int);
495
496/*
497 * The BGE_REGISTER_DEBUG option is only for low-level debugging.  It may
498 * leak information to untrusted users.  It is also known to cause alignment
499 * traps on certain architectures.
500 */
501#ifdef BGE_REGISTER_DEBUG
502static int bge_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
503static int bge_sysctl_reg_read(SYSCTL_HANDLER_ARGS);
504static int bge_sysctl_ape_read(SYSCTL_HANDLER_ARGS);
505static int bge_sysctl_mem_read(SYSCTL_HANDLER_ARGS);
506#endif
507static void bge_add_sysctls(struct bge_softc *);
508static void bge_add_sysctl_stats_regs(struct bge_softc *,
509    struct sysctl_ctx_list *, struct sysctl_oid_list *);
510static void bge_add_sysctl_stats(struct bge_softc *, struct sysctl_ctx_list *,
511    struct sysctl_oid_list *);
512static int bge_sysctl_stats(SYSCTL_HANDLER_ARGS);
513
514static device_method_t bge_methods[] = {
515	/* Device interface */
516	DEVMETHOD(device_probe,		bge_probe),
517	DEVMETHOD(device_attach,	bge_attach),
518	DEVMETHOD(device_detach,	bge_detach),
519	DEVMETHOD(device_shutdown,	bge_shutdown),
520	DEVMETHOD(device_suspend,	bge_suspend),
521	DEVMETHOD(device_resume,	bge_resume),
522
523	/* MII interface */
524	DEVMETHOD(miibus_readreg,	bge_miibus_readreg),
525	DEVMETHOD(miibus_writereg,	bge_miibus_writereg),
526	DEVMETHOD(miibus_statchg,	bge_miibus_statchg),
527
528	DEVMETHOD_END
529};
530
531static driver_t bge_driver = {
532	"bge",
533	bge_methods,
534	sizeof(struct bge_softc)
535};
536
537static devclass_t bge_devclass;
538
539DRIVER_MODULE(bge, pci, bge_driver, bge_devclass, 0, 0);
540DRIVER_MODULE(miibus, bge, miibus_driver, miibus_devclass, 0, 0);
541
542static int bge_allow_asf = 1;
543
544TUNABLE_INT("hw.bge.allow_asf", &bge_allow_asf);
545
546static SYSCTL_NODE(_hw, OID_AUTO, bge, CTLFLAG_RD, 0, "BGE driver parameters");
547SYSCTL_INT(_hw_bge, OID_AUTO, allow_asf, CTLFLAG_RD, &bge_allow_asf, 0,
548	"Allow ASF mode if available");
549
550#define	SPARC64_BLADE_1500_MODEL	"SUNW,Sun-Blade-1500"
551#define	SPARC64_BLADE_1500_PATH_BGE	"/pci@1f,700000/network@2"
552#define	SPARC64_BLADE_2500_MODEL	"SUNW,Sun-Blade-2500"
553#define	SPARC64_BLADE_2500_PATH_BGE	"/pci@1c,600000/network@3"
554#define	SPARC64_OFW_SUBVENDOR		"subsystem-vendor-id"
555
556static int
557bge_has_eaddr(struct bge_softc *sc)
558{
559#ifdef __sparc64__
560	char buf[sizeof(SPARC64_BLADE_1500_PATH_BGE)];
561	device_t dev;
562	uint32_t subvendor;
563
564	dev = sc->bge_dev;
565
566	/*
567	 * The on-board BGEs found in sun4u machines aren't fitted with
568	 * an EEPROM which means that we have to obtain the MAC address
569	 * via OFW and that some tests will always fail.  We distinguish
570	 * such BGEs by the subvendor ID, which also has to be obtained
571	 * from OFW instead of the PCI configuration space as the latter
572	 * indicates Broadcom as the subvendor of the netboot interface.
573	 * For early Blade 1500 and 2500 we even have to check the OFW
574	 * device path as the subvendor ID always defaults to Broadcom
575	 * there.
576	 */
577	if (OF_getprop(ofw_bus_get_node(dev), SPARC64_OFW_SUBVENDOR,
578	    &subvendor, sizeof(subvendor)) == sizeof(subvendor) &&
579	    (subvendor == FJTSU_VENDORID || subvendor == SUN_VENDORID))
580		return (0);
581	memset(buf, 0, sizeof(buf));
582	if (OF_package_to_path(ofw_bus_get_node(dev), buf, sizeof(buf)) > 0) {
583		if (strcmp(sparc64_model, SPARC64_BLADE_1500_MODEL) == 0 &&
584		    strcmp(buf, SPARC64_BLADE_1500_PATH_BGE) == 0)
585			return (0);
586		if (strcmp(sparc64_model, SPARC64_BLADE_2500_MODEL) == 0 &&
587		    strcmp(buf, SPARC64_BLADE_2500_PATH_BGE) == 0)
588			return (0);
589	}
590#endif
591	return (1);
592}
593
594static uint32_t
595bge_readmem_ind(struct bge_softc *sc, int off)
596{
597	device_t dev;
598	uint32_t val;
599
600	if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
601	    off >= BGE_STATS_BLOCK && off < BGE_SEND_RING_1_TO_4)
602		return (0);
603
604	dev = sc->bge_dev;
605
606	pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, off, 4);
607	val = pci_read_config(dev, BGE_PCI_MEMWIN_DATA, 4);
608	pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, 0, 4);
609	return (val);
610}
611
612static void
613bge_writemem_ind(struct bge_softc *sc, int off, int val)
614{
615	device_t dev;
616
617	if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
618	    off >= BGE_STATS_BLOCK && off < BGE_SEND_RING_1_TO_4)
619		return;
620
621	dev = sc->bge_dev;
622
623	pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, off, 4);
624	pci_write_config(dev, BGE_PCI_MEMWIN_DATA, val, 4);
625	pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, 0, 4);
626}
627
628#ifdef notdef
629static uint32_t
630bge_readreg_ind(struct bge_softc *sc, int off)
631{
632	device_t dev;
633
634	dev = sc->bge_dev;
635
636	pci_write_config(dev, BGE_PCI_REG_BASEADDR, off, 4);
637	return (pci_read_config(dev, BGE_PCI_REG_DATA, 4));
638}
639#endif
640
641static void
642bge_writereg_ind(struct bge_softc *sc, int off, int val)
643{
644	device_t dev;
645
646	dev = sc->bge_dev;
647
648	pci_write_config(dev, BGE_PCI_REG_BASEADDR, off, 4);
649	pci_write_config(dev, BGE_PCI_REG_DATA, val, 4);
650}
651
652static void
653bge_writemem_direct(struct bge_softc *sc, int off, int val)
654{
655	CSR_WRITE_4(sc, off, val);
656}
657
658static void
659bge_writembx(struct bge_softc *sc, int off, int val)
660{
661	if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
662		off += BGE_LPMBX_IRQ0_HI - BGE_MBX_IRQ0_HI;
663
664	CSR_WRITE_4(sc, off, val);
665	if ((sc->bge_flags & BGE_FLAG_MBOX_REORDER) != 0)
666		CSR_READ_4(sc, off);
667}
668
669/*
670 * Clear all stale locks and select the lock for this driver instance.
671 */
672static void
673bge_ape_lock_init(struct bge_softc *sc)
674{
675	uint32_t bit, regbase;
676	int i;
677
678	if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
679		regbase = BGE_APE_LOCK_GRANT;
680	else
681		regbase = BGE_APE_PER_LOCK_GRANT;
682
683	/* Clear any stale locks. */
684	for (i = BGE_APE_LOCK_PHY0; i <= BGE_APE_LOCK_GPIO; i++) {
685		switch (i) {
686		case BGE_APE_LOCK_PHY0:
687		case BGE_APE_LOCK_PHY1:
688		case BGE_APE_LOCK_PHY2:
689		case BGE_APE_LOCK_PHY3:
690			bit = BGE_APE_LOCK_GRANT_DRIVER0;
691			break;
692		default:
693			if (sc->bge_func_addr == 0)
694				bit = BGE_APE_LOCK_GRANT_DRIVER0;
695			else
696				bit = (1 << sc->bge_func_addr);
697		}
698		APE_WRITE_4(sc, regbase + 4 * i, bit);
699	}
700
701	/* Select the PHY lock based on the device's function number. */
702	switch (sc->bge_func_addr) {
703	case 0:
704		sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY0;
705		break;
706	case 1:
707		sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY1;
708		break;
709	case 2:
710		sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY2;
711		break;
712	case 3:
713		sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY3;
714		break;
715	default:
716		device_printf(sc->bge_dev,
717		    "PHY lock not supported on this function\n");
718	}
719}
720
721/*
722 * Check for APE firmware, set flags, and print version info.
723 */
724static void
725bge_ape_read_fw_ver(struct bge_softc *sc)
726{
727	const char *fwtype;
728	uint32_t apedata, features;
729
730	/* Check for a valid APE signature in shared memory. */
731	apedata = APE_READ_4(sc, BGE_APE_SEG_SIG);
732	if (apedata != BGE_APE_SEG_SIG_MAGIC) {
733		sc->bge_mfw_flags &= ~ BGE_MFW_ON_APE;
734		return;
735	}
736
737	/* Check if APE firmware is running. */
738	apedata = APE_READ_4(sc, BGE_APE_FW_STATUS);
739	if ((apedata & BGE_APE_FW_STATUS_READY) == 0) {
740		device_printf(sc->bge_dev, "APE signature found "
741		    "but FW status not ready! 0x%08x\n", apedata);
742		return;
743	}
744
745	sc->bge_mfw_flags |= BGE_MFW_ON_APE;
746
747	/* Fetch the APE firwmare type and version. */
748	apedata = APE_READ_4(sc, BGE_APE_FW_VERSION);
749	features = APE_READ_4(sc, BGE_APE_FW_FEATURES);
750	if ((features & BGE_APE_FW_FEATURE_NCSI) != 0) {
751		sc->bge_mfw_flags |= BGE_MFW_TYPE_NCSI;
752		fwtype = "NCSI";
753	} else if ((features & BGE_APE_FW_FEATURE_DASH) != 0) {
754		sc->bge_mfw_flags |= BGE_MFW_TYPE_DASH;
755		fwtype = "DASH";
756	} else
757		fwtype = "UNKN";
758
759	/* Print the APE firmware version. */
760	device_printf(sc->bge_dev, "APE FW version: %s v%d.%d.%d.%d\n",
761	    fwtype,
762	    (apedata & BGE_APE_FW_VERSION_MAJMSK) >> BGE_APE_FW_VERSION_MAJSFT,
763	    (apedata & BGE_APE_FW_VERSION_MINMSK) >> BGE_APE_FW_VERSION_MINSFT,
764	    (apedata & BGE_APE_FW_VERSION_REVMSK) >> BGE_APE_FW_VERSION_REVSFT,
765	    (apedata & BGE_APE_FW_VERSION_BLDMSK));
766}
767
768static int
769bge_ape_lock(struct bge_softc *sc, int locknum)
770{
771	uint32_t bit, gnt, req, status;
772	int i, off;
773
774	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
775		return (0);
776
777	/* Lock request/grant registers have different bases. */
778	if (sc->bge_asicrev == BGE_ASICREV_BCM5761) {
779		req = BGE_APE_LOCK_REQ;
780		gnt = BGE_APE_LOCK_GRANT;
781	} else {
782		req = BGE_APE_PER_LOCK_REQ;
783		gnt = BGE_APE_PER_LOCK_GRANT;
784	}
785
786	off = 4 * locknum;
787
788	switch (locknum) {
789	case BGE_APE_LOCK_GPIO:
790		/* Lock required when using GPIO. */
791		if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
792			return (0);
793		if (sc->bge_func_addr == 0)
794			bit = BGE_APE_LOCK_REQ_DRIVER0;
795		else
796			bit = (1 << sc->bge_func_addr);
797		break;
798	case BGE_APE_LOCK_GRC:
799		/* Lock required to reset the device. */
800		if (sc->bge_func_addr == 0)
801			bit = BGE_APE_LOCK_REQ_DRIVER0;
802		else
803			bit = (1 << sc->bge_func_addr);
804		break;
805	case BGE_APE_LOCK_MEM:
806		/* Lock required when accessing certain APE memory. */
807		if (sc->bge_func_addr == 0)
808			bit = BGE_APE_LOCK_REQ_DRIVER0;
809		else
810			bit = (1 << sc->bge_func_addr);
811		break;
812	case BGE_APE_LOCK_PHY0:
813	case BGE_APE_LOCK_PHY1:
814	case BGE_APE_LOCK_PHY2:
815	case BGE_APE_LOCK_PHY3:
816		/* Lock required when accessing PHYs. */
817		bit = BGE_APE_LOCK_REQ_DRIVER0;
818		break;
819	default:
820		return (EINVAL);
821	}
822
823	/* Request a lock. */
824	APE_WRITE_4(sc, req + off, bit);
825
826	/* Wait up to 1 second to acquire lock. */
827	for (i = 0; i < 20000; i++) {
828		status = APE_READ_4(sc, gnt + off);
829		if (status == bit)
830			break;
831		DELAY(50);
832	}
833
834	/* Handle any errors. */
835	if (status != bit) {
836		device_printf(sc->bge_dev, "APE lock %d request failed! "
837		    "request = 0x%04x[0x%04x], status = 0x%04x[0x%04x]\n",
838		    locknum, req + off, bit & 0xFFFF, gnt + off,
839		    status & 0xFFFF);
840		/* Revoke the lock request. */
841		APE_WRITE_4(sc, gnt + off, bit);
842		return (EBUSY);
843	}
844
845	return (0);
846}
847
848static void
849bge_ape_unlock(struct bge_softc *sc, int locknum)
850{
851	uint32_t bit, gnt;
852	int off;
853
854	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
855		return;
856
857	if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
858		gnt = BGE_APE_LOCK_GRANT;
859	else
860		gnt = BGE_APE_PER_LOCK_GRANT;
861
862	off = 4 * locknum;
863
864	switch (locknum) {
865	case BGE_APE_LOCK_GPIO:
866		if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
867			return;
868		if (sc->bge_func_addr == 0)
869			bit = BGE_APE_LOCK_GRANT_DRIVER0;
870		else
871			bit = (1 << sc->bge_func_addr);
872		break;
873	case BGE_APE_LOCK_GRC:
874		if (sc->bge_func_addr == 0)
875			bit = BGE_APE_LOCK_GRANT_DRIVER0;
876		else
877			bit = (1 << sc->bge_func_addr);
878		break;
879	case BGE_APE_LOCK_MEM:
880		if (sc->bge_func_addr == 0)
881			bit = BGE_APE_LOCK_GRANT_DRIVER0;
882		else
883			bit = (1 << sc->bge_func_addr);
884		break;
885	case BGE_APE_LOCK_PHY0:
886	case BGE_APE_LOCK_PHY1:
887	case BGE_APE_LOCK_PHY2:
888	case BGE_APE_LOCK_PHY3:
889		bit = BGE_APE_LOCK_GRANT_DRIVER0;
890		break;
891	default:
892		return;
893	}
894
895	APE_WRITE_4(sc, gnt + off, bit);
896}
897
898/*
899 * Send an event to the APE firmware.
900 */
901static void
902bge_ape_send_event(struct bge_softc *sc, uint32_t event)
903{
904	uint32_t apedata;
905	int i;
906
907	/* NCSI does not support APE events. */
908	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
909		return;
910
911	/* Wait up to 1ms for APE to service previous event. */
912	for (i = 10; i > 0; i--) {
913		if (bge_ape_lock(sc, BGE_APE_LOCK_MEM) != 0)
914			break;
915		apedata = APE_READ_4(sc, BGE_APE_EVENT_STATUS);
916		if ((apedata & BGE_APE_EVENT_STATUS_EVENT_PENDING) == 0) {
917			APE_WRITE_4(sc, BGE_APE_EVENT_STATUS, event |
918			    BGE_APE_EVENT_STATUS_EVENT_PENDING);
919			bge_ape_unlock(sc, BGE_APE_LOCK_MEM);
920			APE_WRITE_4(sc, BGE_APE_EVENT, BGE_APE_EVENT_1);
921			break;
922		}
923		bge_ape_unlock(sc, BGE_APE_LOCK_MEM);
924		DELAY(100);
925	}
926	if (i == 0)
927		device_printf(sc->bge_dev, "APE event 0x%08x send timed out\n",
928		    event);
929}
930
931static void
932bge_ape_driver_state_change(struct bge_softc *sc, int kind)
933{
934	uint32_t apedata, event;
935
936	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
937		return;
938
939	switch (kind) {
940	case BGE_RESET_START:
941		/* If this is the first load, clear the load counter. */
942		apedata = APE_READ_4(sc, BGE_APE_HOST_SEG_SIG);
943		if (apedata != BGE_APE_HOST_SEG_SIG_MAGIC)
944			APE_WRITE_4(sc, BGE_APE_HOST_INIT_COUNT, 0);
945		else {
946			apedata = APE_READ_4(sc, BGE_APE_HOST_INIT_COUNT);
947			APE_WRITE_4(sc, BGE_APE_HOST_INIT_COUNT, ++apedata);
948		}
949		APE_WRITE_4(sc, BGE_APE_HOST_SEG_SIG,
950		    BGE_APE_HOST_SEG_SIG_MAGIC);
951		APE_WRITE_4(sc, BGE_APE_HOST_SEG_LEN,
952		    BGE_APE_HOST_SEG_LEN_MAGIC);
953
954		/* Add some version info if bge(4) supports it. */
955		APE_WRITE_4(sc, BGE_APE_HOST_DRIVER_ID,
956		    BGE_APE_HOST_DRIVER_ID_MAGIC(1, 0));
957		APE_WRITE_4(sc, BGE_APE_HOST_BEHAVIOR,
958		    BGE_APE_HOST_BEHAV_NO_PHYLOCK);
959		APE_WRITE_4(sc, BGE_APE_HOST_HEARTBEAT_INT_MS,
960		    BGE_APE_HOST_HEARTBEAT_INT_DISABLE);
961		APE_WRITE_4(sc, BGE_APE_HOST_DRVR_STATE,
962		    BGE_APE_HOST_DRVR_STATE_START);
963		event = BGE_APE_EVENT_STATUS_STATE_START;
964		break;
965	case BGE_RESET_SHUTDOWN:
966		APE_WRITE_4(sc, BGE_APE_HOST_DRVR_STATE,
967		    BGE_APE_HOST_DRVR_STATE_UNLOAD);
968		event = BGE_APE_EVENT_STATUS_STATE_UNLOAD;
969		break;
970	case BGE_RESET_SUSPEND:
971		event = BGE_APE_EVENT_STATUS_STATE_SUSPEND;
972		break;
973	default:
974		return;
975	}
976
977	bge_ape_send_event(sc, event | BGE_APE_EVENT_STATUS_DRIVER_EVNT |
978	    BGE_APE_EVENT_STATUS_STATE_CHNGE);
979}
980
981/*
982 * Map a single buffer address.
983 */
984
985static void
986bge_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
987{
988	struct bge_dmamap_arg *ctx;
989
990	if (error)
991		return;
992
993	KASSERT(nseg == 1, ("%s: %d segments returned!", __func__, nseg));
994
995	ctx = arg;
996	ctx->bge_busaddr = segs->ds_addr;
997}
998
999static uint8_t
1000bge_nvram_getbyte(struct bge_softc *sc, int addr, uint8_t *dest)
1001{
1002	uint32_t access, byte = 0;
1003	int i;
1004
1005	/* Lock. */
1006	CSR_WRITE_4(sc, BGE_NVRAM_SWARB, BGE_NVRAMSWARB_SET1);
1007	for (i = 0; i < 8000; i++) {
1008		if (CSR_READ_4(sc, BGE_NVRAM_SWARB) & BGE_NVRAMSWARB_GNT1)
1009			break;
1010		DELAY(20);
1011	}
1012	if (i == 8000)
1013		return (1);
1014
1015	/* Enable access. */
1016	access = CSR_READ_4(sc, BGE_NVRAM_ACCESS);
1017	CSR_WRITE_4(sc, BGE_NVRAM_ACCESS, access | BGE_NVRAMACC_ENABLE);
1018
1019	CSR_WRITE_4(sc, BGE_NVRAM_ADDR, addr & 0xfffffffc);
1020	CSR_WRITE_4(sc, BGE_NVRAM_CMD, BGE_NVRAM_READCMD);
1021	for (i = 0; i < BGE_TIMEOUT * 10; i++) {
1022		DELAY(10);
1023		if (CSR_READ_4(sc, BGE_NVRAM_CMD) & BGE_NVRAMCMD_DONE) {
1024			DELAY(10);
1025			break;
1026		}
1027	}
1028
1029	if (i == BGE_TIMEOUT * 10) {
1030		if_printf(sc->bge_ifp, "nvram read timed out\n");
1031		return (1);
1032	}
1033
1034	/* Get result. */
1035	byte = CSR_READ_4(sc, BGE_NVRAM_RDDATA);
1036
1037	*dest = (bswap32(byte) >> ((addr % 4) * 8)) & 0xFF;
1038
1039	/* Disable access. */
1040	CSR_WRITE_4(sc, BGE_NVRAM_ACCESS, access);
1041
1042	/* Unlock. */
1043	CSR_WRITE_4(sc, BGE_NVRAM_SWARB, BGE_NVRAMSWARB_CLR1);
1044	CSR_READ_4(sc, BGE_NVRAM_SWARB);
1045
1046	return (0);
1047}
1048
1049/*
1050 * Read a sequence of bytes from NVRAM.
1051 */
1052static int
1053bge_read_nvram(struct bge_softc *sc, caddr_t dest, int off, int cnt)
1054{
1055	int err = 0, i;
1056	uint8_t byte = 0;
1057
1058	if (sc->bge_asicrev != BGE_ASICREV_BCM5906)
1059		return (1);
1060
1061	for (i = 0; i < cnt; i++) {
1062		err = bge_nvram_getbyte(sc, off + i, &byte);
1063		if (err)
1064			break;
1065		*(dest + i) = byte;
1066	}
1067
1068	return (err ? 1 : 0);
1069}
1070
1071/*
1072 * Read a byte of data stored in the EEPROM at address 'addr.' The
1073 * BCM570x supports both the traditional bitbang interface and an
1074 * auto access interface for reading the EEPROM. We use the auto
1075 * access method.
1076 */
1077static uint8_t
1078bge_eeprom_getbyte(struct bge_softc *sc, int addr, uint8_t *dest)
1079{
1080	int i;
1081	uint32_t byte = 0;
1082
1083	/*
1084	 * Enable use of auto EEPROM access so we can avoid
1085	 * having to use the bitbang method.
1086	 */
1087	BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_AUTO_EEPROM);
1088
1089	/* Reset the EEPROM, load the clock period. */
1090	CSR_WRITE_4(sc, BGE_EE_ADDR,
1091	    BGE_EEADDR_RESET | BGE_EEHALFCLK(BGE_HALFCLK_384SCL));
1092	DELAY(20);
1093
1094	/* Issue the read EEPROM command. */
1095	CSR_WRITE_4(sc, BGE_EE_ADDR, BGE_EE_READCMD | addr);
1096
1097	/* Wait for completion */
1098	for(i = 0; i < BGE_TIMEOUT * 10; i++) {
1099		DELAY(10);
1100		if (CSR_READ_4(sc, BGE_EE_ADDR) & BGE_EEADDR_DONE)
1101			break;
1102	}
1103
1104	if (i == BGE_TIMEOUT * 10) {
1105		device_printf(sc->bge_dev, "EEPROM read timed out\n");
1106		return (1);
1107	}
1108
1109	/* Get result. */
1110	byte = CSR_READ_4(sc, BGE_EE_DATA);
1111
1112	*dest = (byte >> ((addr % 4) * 8)) & 0xFF;
1113
1114	return (0);
1115}
1116
1117/*
1118 * Read a sequence of bytes from the EEPROM.
1119 */
1120static int
1121bge_read_eeprom(struct bge_softc *sc, caddr_t dest, int off, int cnt)
1122{
1123	int i, error = 0;
1124	uint8_t byte = 0;
1125
1126	for (i = 0; i < cnt; i++) {
1127		error = bge_eeprom_getbyte(sc, off + i, &byte);
1128		if (error)
1129			break;
1130		*(dest + i) = byte;
1131	}
1132
1133	return (error ? 1 : 0);
1134}
1135
1136static int
1137bge_miibus_readreg(device_t dev, int phy, int reg)
1138{
1139	struct bge_softc *sc;
1140	uint32_t val;
1141	int i;
1142
1143	sc = device_get_softc(dev);
1144
1145	if (bge_ape_lock(sc, sc->bge_phy_ape_lock) != 0)
1146		return (0);
1147
1148	/* Clear the autopoll bit if set, otherwise may trigger PCI errors. */
1149	if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
1150		CSR_WRITE_4(sc, BGE_MI_MODE,
1151		    sc->bge_mi_mode & ~BGE_MIMODE_AUTOPOLL);
1152		DELAY(80);
1153	}
1154
1155	CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_READ | BGE_MICOMM_BUSY |
1156	    BGE_MIPHY(phy) | BGE_MIREG(reg));
1157
1158	/* Poll for the PHY register access to complete. */
1159	for (i = 0; i < BGE_TIMEOUT; i++) {
1160		DELAY(10);
1161		val = CSR_READ_4(sc, BGE_MI_COMM);
1162		if ((val & BGE_MICOMM_BUSY) == 0) {
1163			DELAY(5);
1164			val = CSR_READ_4(sc, BGE_MI_COMM);
1165			break;
1166		}
1167	}
1168
1169	if (i == BGE_TIMEOUT) {
1170		device_printf(sc->bge_dev,
1171		    "PHY read timed out (phy %d, reg %d, val 0x%08x)\n",
1172		    phy, reg, val);
1173		val = 0;
1174	}
1175
1176	/* Restore the autopoll bit if necessary. */
1177	if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
1178		CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
1179		DELAY(80);
1180	}
1181
1182	bge_ape_unlock(sc, sc->bge_phy_ape_lock);
1183
1184	if (val & BGE_MICOMM_READFAIL)
1185		return (0);
1186
1187	return (val & 0xFFFF);
1188}
1189
1190static int
1191bge_miibus_writereg(device_t dev, int phy, int reg, int val)
1192{
1193	struct bge_softc *sc;
1194	int i;
1195
1196	sc = device_get_softc(dev);
1197
1198	if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
1199	    (reg == BRGPHY_MII_1000CTL || reg == BRGPHY_MII_AUXCTL))
1200		return (0);
1201
1202	if (bge_ape_lock(sc, sc->bge_phy_ape_lock) != 0)
1203		return (0);
1204
1205	/* Clear the autopoll bit if set, otherwise may trigger PCI errors. */
1206	if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
1207		CSR_WRITE_4(sc, BGE_MI_MODE,
1208		    sc->bge_mi_mode & ~BGE_MIMODE_AUTOPOLL);
1209		DELAY(80);
1210	}
1211
1212	CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_WRITE | BGE_MICOMM_BUSY |
1213	    BGE_MIPHY(phy) | BGE_MIREG(reg) | val);
1214
1215	for (i = 0; i < BGE_TIMEOUT; i++) {
1216		DELAY(10);
1217		if (!(CSR_READ_4(sc, BGE_MI_COMM) & BGE_MICOMM_BUSY)) {
1218			DELAY(5);
1219			CSR_READ_4(sc, BGE_MI_COMM); /* dummy read */
1220			break;
1221		}
1222	}
1223
1224	/* Restore the autopoll bit if necessary. */
1225	if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
1226		CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
1227		DELAY(80);
1228	}
1229
1230	bge_ape_unlock(sc, sc->bge_phy_ape_lock);
1231
1232	if (i == BGE_TIMEOUT)
1233		device_printf(sc->bge_dev,
1234		    "PHY write timed out (phy %d, reg %d, val 0x%04x)\n",
1235		    phy, reg, val);
1236
1237	return (0);
1238}
1239
1240static void
1241bge_miibus_statchg(device_t dev)
1242{
1243	struct bge_softc *sc;
1244	struct mii_data *mii;
1245	uint32_t mac_mode, rx_mode, tx_mode;
1246
1247	sc = device_get_softc(dev);
1248	if ((sc->bge_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1249		return;
1250	mii = device_get_softc(sc->bge_miibus);
1251
1252	if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
1253	    (IFM_ACTIVE | IFM_AVALID)) {
1254		switch (IFM_SUBTYPE(mii->mii_media_active)) {
1255		case IFM_10_T:
1256		case IFM_100_TX:
1257			sc->bge_link = 1;
1258			break;
1259		case IFM_1000_T:
1260		case IFM_1000_SX:
1261		case IFM_2500_SX:
1262			if (sc->bge_asicrev != BGE_ASICREV_BCM5906)
1263				sc->bge_link = 1;
1264			else
1265				sc->bge_link = 0;
1266			break;
1267		default:
1268			sc->bge_link = 0;
1269			break;
1270		}
1271	} else
1272		sc->bge_link = 0;
1273	if (sc->bge_link == 0)
1274		return;
1275
1276	/*
1277	 * APE firmware touches these registers to keep the MAC
1278	 * connected to the outside world.  Try to keep the
1279	 * accesses atomic.
1280	 */
1281
1282	/* Set the port mode (MII/GMII) to match the link speed. */
1283	mac_mode = CSR_READ_4(sc, BGE_MAC_MODE) &
1284	    ~(BGE_MACMODE_PORTMODE | BGE_MACMODE_HALF_DUPLEX);
1285	tx_mode = CSR_READ_4(sc, BGE_TX_MODE);
1286	rx_mode = CSR_READ_4(sc, BGE_RX_MODE);
1287
1288	if (IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_T ||
1289	    IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_SX)
1290		mac_mode |= BGE_PORTMODE_GMII;
1291	else
1292		mac_mode |= BGE_PORTMODE_MII;
1293
1294	/* Set MAC flow control behavior to match link flow control settings. */
1295	tx_mode &= ~BGE_TXMODE_FLOWCTL_ENABLE;
1296	rx_mode &= ~BGE_RXMODE_FLOWCTL_ENABLE;
1297	if ((IFM_OPTIONS(mii->mii_media_active) & IFM_FDX) != 0) {
1298		if ((IFM_OPTIONS(mii->mii_media_active) & IFM_ETH_TXPAUSE) != 0)
1299			tx_mode |= BGE_TXMODE_FLOWCTL_ENABLE;
1300		if ((IFM_OPTIONS(mii->mii_media_active) & IFM_ETH_RXPAUSE) != 0)
1301			rx_mode |= BGE_RXMODE_FLOWCTL_ENABLE;
1302	} else
1303		mac_mode |= BGE_MACMODE_HALF_DUPLEX;
1304
1305	CSR_WRITE_4(sc, BGE_MAC_MODE, mac_mode);
1306	DELAY(40);
1307	CSR_WRITE_4(sc, BGE_TX_MODE, tx_mode);
1308	CSR_WRITE_4(sc, BGE_RX_MODE, rx_mode);
1309}
1310
1311/*
1312 * Intialize a standard receive ring descriptor.
1313 */
1314static int
1315bge_newbuf_std(struct bge_softc *sc, int i)
1316{
1317	struct mbuf *m;
1318	struct bge_rx_bd *r;
1319	bus_dma_segment_t segs[1];
1320	bus_dmamap_t map;
1321	int error, nsegs;
1322
1323	if (sc->bge_flags & BGE_FLAG_JUMBO_STD &&
1324	    (sc->bge_ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN +
1325	    ETHER_VLAN_ENCAP_LEN > (MCLBYTES - ETHER_ALIGN))) {
1326		m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES);
1327		if (m == NULL)
1328			return (ENOBUFS);
1329		m->m_len = m->m_pkthdr.len = MJUM9BYTES;
1330	} else {
1331		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1332		if (m == NULL)
1333			return (ENOBUFS);
1334		m->m_len = m->m_pkthdr.len = MCLBYTES;
1335	}
1336	if ((sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) == 0)
1337		m_adj(m, ETHER_ALIGN);
1338
1339	error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_rx_mtag,
1340	    sc->bge_cdata.bge_rx_std_sparemap, m, segs, &nsegs, 0);
1341	if (error != 0) {
1342		m_freem(m);
1343		return (error);
1344	}
1345	if (sc->bge_cdata.bge_rx_std_chain[i] != NULL) {
1346		bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
1347		    sc->bge_cdata.bge_rx_std_dmamap[i], BUS_DMASYNC_POSTREAD);
1348		bus_dmamap_unload(sc->bge_cdata.bge_rx_mtag,
1349		    sc->bge_cdata.bge_rx_std_dmamap[i]);
1350	}
1351	map = sc->bge_cdata.bge_rx_std_dmamap[i];
1352	sc->bge_cdata.bge_rx_std_dmamap[i] = sc->bge_cdata.bge_rx_std_sparemap;
1353	sc->bge_cdata.bge_rx_std_sparemap = map;
1354	sc->bge_cdata.bge_rx_std_chain[i] = m;
1355	sc->bge_cdata.bge_rx_std_seglen[i] = segs[0].ds_len;
1356	r = &sc->bge_ldata.bge_rx_std_ring[sc->bge_std];
1357	r->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[0].ds_addr);
1358	r->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[0].ds_addr);
1359	r->bge_flags = BGE_RXBDFLAG_END;
1360	r->bge_len = segs[0].ds_len;
1361	r->bge_idx = i;
1362
1363	bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
1364	    sc->bge_cdata.bge_rx_std_dmamap[i], BUS_DMASYNC_PREREAD);
1365
1366	return (0);
1367}
1368
1369/*
1370 * Initialize a jumbo receive ring descriptor. This allocates
1371 * a jumbo buffer from the pool managed internally by the driver.
1372 */
1373static int
1374bge_newbuf_jumbo(struct bge_softc *sc, int i)
1375{
1376	bus_dma_segment_t segs[BGE_NSEG_JUMBO];
1377	bus_dmamap_t map;
1378	struct bge_extrx_bd *r;
1379	struct mbuf *m;
1380	int error, nsegs;
1381
1382	MGETHDR(m, M_NOWAIT, MT_DATA);
1383	if (m == NULL)
1384		return (ENOBUFS);
1385
1386	m_cljget(m, M_NOWAIT, MJUM9BYTES);
1387	if (!(m->m_flags & M_EXT)) {
1388		m_freem(m);
1389		return (ENOBUFS);
1390	}
1391	m->m_len = m->m_pkthdr.len = MJUM9BYTES;
1392	if ((sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) == 0)
1393		m_adj(m, ETHER_ALIGN);
1394
1395	error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag_jumbo,
1396	    sc->bge_cdata.bge_rx_jumbo_sparemap, m, segs, &nsegs, 0);
1397	if (error != 0) {
1398		m_freem(m);
1399		return (error);
1400	}
1401
1402	if (sc->bge_cdata.bge_rx_jumbo_chain[i] != NULL) {
1403		bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
1404		    sc->bge_cdata.bge_rx_jumbo_dmamap[i], BUS_DMASYNC_POSTREAD);
1405		bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
1406		    sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
1407	}
1408	map = sc->bge_cdata.bge_rx_jumbo_dmamap[i];
1409	sc->bge_cdata.bge_rx_jumbo_dmamap[i] =
1410	    sc->bge_cdata.bge_rx_jumbo_sparemap;
1411	sc->bge_cdata.bge_rx_jumbo_sparemap = map;
1412	sc->bge_cdata.bge_rx_jumbo_chain[i] = m;
1413	sc->bge_cdata.bge_rx_jumbo_seglen[i][0] = 0;
1414	sc->bge_cdata.bge_rx_jumbo_seglen[i][1] = 0;
1415	sc->bge_cdata.bge_rx_jumbo_seglen[i][2] = 0;
1416	sc->bge_cdata.bge_rx_jumbo_seglen[i][3] = 0;
1417
1418	/*
1419	 * Fill in the extended RX buffer descriptor.
1420	 */
1421	r = &sc->bge_ldata.bge_rx_jumbo_ring[sc->bge_jumbo];
1422	r->bge_flags = BGE_RXBDFLAG_JUMBO_RING | BGE_RXBDFLAG_END;
1423	r->bge_idx = i;
1424	r->bge_len3 = r->bge_len2 = r->bge_len1 = 0;
1425	switch (nsegs) {
1426	case 4:
1427		r->bge_addr3.bge_addr_lo = BGE_ADDR_LO(segs[3].ds_addr);
1428		r->bge_addr3.bge_addr_hi = BGE_ADDR_HI(segs[3].ds_addr);
1429		r->bge_len3 = segs[3].ds_len;
1430		sc->bge_cdata.bge_rx_jumbo_seglen[i][3] = segs[3].ds_len;
1431	case 3:
1432		r->bge_addr2.bge_addr_lo = BGE_ADDR_LO(segs[2].ds_addr);
1433		r->bge_addr2.bge_addr_hi = BGE_ADDR_HI(segs[2].ds_addr);
1434		r->bge_len2 = segs[2].ds_len;
1435		sc->bge_cdata.bge_rx_jumbo_seglen[i][2] = segs[2].ds_len;
1436	case 2:
1437		r->bge_addr1.bge_addr_lo = BGE_ADDR_LO(segs[1].ds_addr);
1438		r->bge_addr1.bge_addr_hi = BGE_ADDR_HI(segs[1].ds_addr);
1439		r->bge_len1 = segs[1].ds_len;
1440		sc->bge_cdata.bge_rx_jumbo_seglen[i][1] = segs[1].ds_len;
1441	case 1:
1442		r->bge_addr0.bge_addr_lo = BGE_ADDR_LO(segs[0].ds_addr);
1443		r->bge_addr0.bge_addr_hi = BGE_ADDR_HI(segs[0].ds_addr);
1444		r->bge_len0 = segs[0].ds_len;
1445		sc->bge_cdata.bge_rx_jumbo_seglen[i][0] = segs[0].ds_len;
1446		break;
1447	default:
1448		panic("%s: %d segments\n", __func__, nsegs);
1449	}
1450
1451	bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
1452	    sc->bge_cdata.bge_rx_jumbo_dmamap[i], BUS_DMASYNC_PREREAD);
1453
1454	return (0);
1455}
1456
1457static int
1458bge_init_rx_ring_std(struct bge_softc *sc)
1459{
1460	int error, i;
1461
1462	bzero(sc->bge_ldata.bge_rx_std_ring, BGE_STD_RX_RING_SZ);
1463	sc->bge_std = 0;
1464	for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
1465		if ((error = bge_newbuf_std(sc, i)) != 0)
1466			return (error);
1467		BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
1468	}
1469
1470	bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
1471	    sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
1472
1473	sc->bge_std = 0;
1474	bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, BGE_STD_RX_RING_CNT - 1);
1475
1476	return (0);
1477}
1478
1479static void
1480bge_free_rx_ring_std(struct bge_softc *sc)
1481{
1482	int i;
1483
1484	for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
1485		if (sc->bge_cdata.bge_rx_std_chain[i] != NULL) {
1486			bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
1487			    sc->bge_cdata.bge_rx_std_dmamap[i],
1488			    BUS_DMASYNC_POSTREAD);
1489			bus_dmamap_unload(sc->bge_cdata.bge_rx_mtag,
1490			    sc->bge_cdata.bge_rx_std_dmamap[i]);
1491			m_freem(sc->bge_cdata.bge_rx_std_chain[i]);
1492			sc->bge_cdata.bge_rx_std_chain[i] = NULL;
1493		}
1494		bzero((char *)&sc->bge_ldata.bge_rx_std_ring[i],
1495		    sizeof(struct bge_rx_bd));
1496	}
1497}
1498
1499static int
1500bge_init_rx_ring_jumbo(struct bge_softc *sc)
1501{
1502	struct bge_rcb *rcb;
1503	int error, i;
1504
1505	bzero(sc->bge_ldata.bge_rx_jumbo_ring, BGE_JUMBO_RX_RING_SZ);
1506	sc->bge_jumbo = 0;
1507	for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
1508		if ((error = bge_newbuf_jumbo(sc, i)) != 0)
1509			return (error);
1510		BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_RING_CNT);
1511	}
1512
1513	bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
1514	    sc->bge_cdata.bge_rx_jumbo_ring_map, BUS_DMASYNC_PREWRITE);
1515
1516	sc->bge_jumbo = 0;
1517
1518	/* Enable the jumbo receive producer ring. */
1519	rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
1520	rcb->bge_maxlen_flags =
1521	    BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_USE_EXT_RX_BD);
1522	CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_MAXLEN_FLAGS, rcb->bge_maxlen_flags);
1523
1524	bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, BGE_JUMBO_RX_RING_CNT - 1);
1525
1526	return (0);
1527}
1528
1529static void
1530bge_free_rx_ring_jumbo(struct bge_softc *sc)
1531{
1532	int i;
1533
1534	for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
1535		if (sc->bge_cdata.bge_rx_jumbo_chain[i] != NULL) {
1536			bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
1537			    sc->bge_cdata.bge_rx_jumbo_dmamap[i],
1538			    BUS_DMASYNC_POSTREAD);
1539			bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
1540			    sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
1541			m_freem(sc->bge_cdata.bge_rx_jumbo_chain[i]);
1542			sc->bge_cdata.bge_rx_jumbo_chain[i] = NULL;
1543		}
1544		bzero((char *)&sc->bge_ldata.bge_rx_jumbo_ring[i],
1545		    sizeof(struct bge_extrx_bd));
1546	}
1547}
1548
1549static void
1550bge_free_tx_ring(struct bge_softc *sc)
1551{
1552	int i;
1553
1554	if (sc->bge_ldata.bge_tx_ring == NULL)
1555		return;
1556
1557	for (i = 0; i < BGE_TX_RING_CNT; i++) {
1558		if (sc->bge_cdata.bge_tx_chain[i] != NULL) {
1559			bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag,
1560			    sc->bge_cdata.bge_tx_dmamap[i],
1561			    BUS_DMASYNC_POSTWRITE);
1562			bus_dmamap_unload(sc->bge_cdata.bge_tx_mtag,
1563			    sc->bge_cdata.bge_tx_dmamap[i]);
1564			m_freem(sc->bge_cdata.bge_tx_chain[i]);
1565			sc->bge_cdata.bge_tx_chain[i] = NULL;
1566		}
1567		bzero((char *)&sc->bge_ldata.bge_tx_ring[i],
1568		    sizeof(struct bge_tx_bd));
1569	}
1570}
1571
1572static int
1573bge_init_tx_ring(struct bge_softc *sc)
1574{
1575	sc->bge_txcnt = 0;
1576	sc->bge_tx_saved_considx = 0;
1577
1578	bzero(sc->bge_ldata.bge_tx_ring, BGE_TX_RING_SZ);
1579	bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
1580	    sc->bge_cdata.bge_tx_ring_map, BUS_DMASYNC_PREWRITE);
1581
1582	/* Initialize transmit producer index for host-memory send ring. */
1583	sc->bge_tx_prodidx = 0;
1584	bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
1585
1586	/* 5700 b2 errata */
1587	if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
1588		bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
1589
1590	/* NIC-memory send ring not used; initialize to zero. */
1591	bge_writembx(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
1592	/* 5700 b2 errata */
1593	if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
1594		bge_writembx(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
1595
1596	return (0);
1597}
1598
1599static void
1600bge_setpromisc(struct bge_softc *sc)
1601{
1602	struct ifnet *ifp;
1603
1604	BGE_LOCK_ASSERT(sc);
1605
1606	ifp = sc->bge_ifp;
1607
1608	/* Enable or disable promiscuous mode as needed. */
1609	if (ifp->if_flags & IFF_PROMISC)
1610		BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
1611	else
1612		BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
1613}
1614
1615static void
1616bge_setmulti(struct bge_softc *sc)
1617{
1618	struct ifnet *ifp;
1619	struct ifmultiaddr *ifma;
1620	uint32_t hashes[4] = { 0, 0, 0, 0 };
1621	int h, i;
1622
1623	BGE_LOCK_ASSERT(sc);
1624
1625	ifp = sc->bge_ifp;
1626
1627	if (ifp->if_flags & IFF_ALLMULTI || ifp->if_flags & IFF_PROMISC) {
1628		for (i = 0; i < 4; i++)
1629			CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0xFFFFFFFF);
1630		return;
1631	}
1632
1633	/* First, zot all the existing filters. */
1634	for (i = 0; i < 4; i++)
1635		CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0);
1636
1637	/* Now program new ones. */
1638	if_maddr_rlock(ifp);
1639	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1640		if (ifma->ifma_addr->sa_family != AF_LINK)
1641			continue;
1642		h = ether_crc32_le(LLADDR((struct sockaddr_dl *)
1643		    ifma->ifma_addr), ETHER_ADDR_LEN) & 0x7F;
1644		hashes[(h & 0x60) >> 5] |= 1 << (h & 0x1F);
1645	}
1646	if_maddr_runlock(ifp);
1647
1648	for (i = 0; i < 4; i++)
1649		CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), hashes[i]);
1650}
1651
1652static void
1653bge_setvlan(struct bge_softc *sc)
1654{
1655	struct ifnet *ifp;
1656
1657	BGE_LOCK_ASSERT(sc);
1658
1659	ifp = sc->bge_ifp;
1660
1661	/* Enable or disable VLAN tag stripping as needed. */
1662	if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
1663		BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_KEEP_VLAN_DIAG);
1664	else
1665		BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_KEEP_VLAN_DIAG);
1666}
1667
1668static void
1669bge_sig_pre_reset(struct bge_softc *sc, int type)
1670{
1671
1672	/*
1673	 * Some chips don't like this so only do this if ASF is enabled
1674	 */
1675	if (sc->bge_asf_mode)
1676		bge_writemem_ind(sc, BGE_SRAM_FW_MB, BGE_SRAM_FW_MB_MAGIC);
1677
1678	if (sc->bge_asf_mode & ASF_NEW_HANDSHAKE) {
1679		switch (type) {
1680		case BGE_RESET_START:
1681			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1682			    BGE_FW_DRV_STATE_START);
1683			break;
1684		case BGE_RESET_SHUTDOWN:
1685			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1686			    BGE_FW_DRV_STATE_UNLOAD);
1687			break;
1688		case BGE_RESET_SUSPEND:
1689			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1690			    BGE_FW_DRV_STATE_SUSPEND);
1691			break;
1692		}
1693	}
1694
1695	if (type == BGE_RESET_START || type == BGE_RESET_SUSPEND)
1696		bge_ape_driver_state_change(sc, type);
1697}
1698
1699static void
1700bge_sig_post_reset(struct bge_softc *sc, int type)
1701{
1702
1703	if (sc->bge_asf_mode & ASF_NEW_HANDSHAKE) {
1704		switch (type) {
1705		case BGE_RESET_START:
1706			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1707			    BGE_FW_DRV_STATE_START_DONE);
1708			/* START DONE */
1709			break;
1710		case BGE_RESET_SHUTDOWN:
1711			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1712			    BGE_FW_DRV_STATE_UNLOAD_DONE);
1713			break;
1714		}
1715	}
1716	if (type == BGE_RESET_SHUTDOWN)
1717		bge_ape_driver_state_change(sc, type);
1718}
1719
1720static void
1721bge_sig_legacy(struct bge_softc *sc, int type)
1722{
1723
1724	if (sc->bge_asf_mode) {
1725		switch (type) {
1726		case BGE_RESET_START:
1727			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1728			    BGE_FW_DRV_STATE_START);
1729			break;
1730		case BGE_RESET_SHUTDOWN:
1731			bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
1732			    BGE_FW_DRV_STATE_UNLOAD);
1733			break;
1734		}
1735	}
1736}
1737
1738static void
1739bge_stop_fw(struct bge_softc *sc)
1740{
1741	int i;
1742
1743	if (sc->bge_asf_mode) {
1744		bge_writemem_ind(sc, BGE_SRAM_FW_CMD_MB, BGE_FW_CMD_PAUSE);
1745		CSR_WRITE_4(sc, BGE_RX_CPU_EVENT,
1746		    CSR_READ_4(sc, BGE_RX_CPU_EVENT) | BGE_RX_CPU_DRV_EVENT);
1747
1748		for (i = 0; i < 100; i++ ) {
1749			if (!(CSR_READ_4(sc, BGE_RX_CPU_EVENT) &
1750			    BGE_RX_CPU_DRV_EVENT))
1751				break;
1752			DELAY(10);
1753		}
1754	}
1755}
1756
1757static uint32_t
1758bge_dma_swap_options(struct bge_softc *sc)
1759{
1760	uint32_t dma_options;
1761
1762	dma_options = BGE_MODECTL_WORDSWAP_NONFRAME |
1763	    BGE_MODECTL_BYTESWAP_DATA | BGE_MODECTL_WORDSWAP_DATA;
1764#if BYTE_ORDER == BIG_ENDIAN
1765	dma_options |= BGE_MODECTL_BYTESWAP_NONFRAME;
1766#endif
1767	return (dma_options);
1768}
1769
1770/*
1771 * Do endian, PCI and DMA initialization.
1772 */
1773static int
1774bge_chipinit(struct bge_softc *sc)
1775{
1776	uint32_t dma_rw_ctl, misc_ctl, mode_ctl;
1777	uint16_t val;
1778	int i;
1779
1780	/* Set endianness before we access any non-PCI registers. */
1781	misc_ctl = BGE_INIT;
1782	if (sc->bge_flags & BGE_FLAG_TAGGED_STATUS)
1783		misc_ctl |= BGE_PCIMISCCTL_TAGGED_STATUS;
1784	pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL, misc_ctl, 4);
1785
1786	/*
1787	 * Clear the MAC statistics block in the NIC's
1788	 * internal memory.
1789	 */
1790	for (i = BGE_STATS_BLOCK;
1791	    i < BGE_STATS_BLOCK_END + 1; i += sizeof(uint32_t))
1792		BGE_MEMWIN_WRITE(sc, i, 0);
1793
1794	for (i = BGE_STATUS_BLOCK;
1795	    i < BGE_STATUS_BLOCK_END + 1; i += sizeof(uint32_t))
1796		BGE_MEMWIN_WRITE(sc, i, 0);
1797
1798	if (sc->bge_chiprev == BGE_CHIPREV_5704_BX) {
1799		/*
1800		 *  Fix data corruption caused by non-qword write with WB.
1801		 *  Fix master abort in PCI mode.
1802		 *  Fix PCI latency timer.
1803		 */
1804		val = pci_read_config(sc->bge_dev, BGE_PCI_MSI_DATA + 2, 2);
1805		val |= (1 << 10) | (1 << 12) | (1 << 13);
1806		pci_write_config(sc->bge_dev, BGE_PCI_MSI_DATA + 2, val, 2);
1807	}
1808
1809	if (sc->bge_asicrev == BGE_ASICREV_BCM57765 ||
1810	    sc->bge_asicrev == BGE_ASICREV_BCM57766) {
1811		/*
1812		 * For the 57766 and non Ax versions of 57765, bootcode
1813		 * needs to setup the PCIE Fast Training Sequence (FTS)
1814		 * value to prevent transmit hangs.
1815		 */
1816		if (sc->bge_chiprev != BGE_CHIPREV_57765_AX) {
1817			CSR_WRITE_4(sc, BGE_CPMU_PADRNG_CTL,
1818			    CSR_READ_4(sc, BGE_CPMU_PADRNG_CTL) |
1819			    BGE_CPMU_PADRNG_CTL_RDIV2);
1820		}
1821	}
1822
1823	/*
1824	 * Set up the PCI DMA control register.
1825	 */
1826	dma_rw_ctl = BGE_PCIDMARWCTL_RD_CMD_SHIFT(6) |
1827	    BGE_PCIDMARWCTL_WR_CMD_SHIFT(7);
1828	if (sc->bge_flags & BGE_FLAG_PCIE) {
1829		if (sc->bge_mps >= 256)
1830			dma_rw_ctl |= BGE_PCIDMARWCTL_WR_WAT_SHIFT(7);
1831		else
1832			dma_rw_ctl |= BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
1833	} else if (sc->bge_flags & BGE_FLAG_PCIX) {
1834		if (BGE_IS_5714_FAMILY(sc)) {
1835			/* 256 bytes for read and write. */
1836			dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(2) |
1837			    BGE_PCIDMARWCTL_WR_WAT_SHIFT(2);
1838			dma_rw_ctl |= (sc->bge_asicrev == BGE_ASICREV_BCM5780) ?
1839			    BGE_PCIDMARWCTL_ONEDMA_ATONCE_GLOBAL :
1840			    BGE_PCIDMARWCTL_ONEDMA_ATONCE_LOCAL;
1841		} else if (sc->bge_asicrev == BGE_ASICREV_BCM5703) {
1842			/*
1843			 * In the BCM5703, the DMA read watermark should
1844			 * be set to less than or equal to the maximum
1845			 * memory read byte count of the PCI-X command
1846			 * register.
1847			 */
1848			dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(4) |
1849			    BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
1850		} else if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
1851			/* 1536 bytes for read, 384 bytes for write. */
1852			dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(7) |
1853			    BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
1854		} else {
1855			/* 384 bytes for read and write. */
1856			dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(3) |
1857			    BGE_PCIDMARWCTL_WR_WAT_SHIFT(3) |
1858			    0x0F;
1859		}
1860		if (sc->bge_asicrev == BGE_ASICREV_BCM5703 ||
1861		    sc->bge_asicrev == BGE_ASICREV_BCM5704) {
1862			uint32_t tmp;
1863
1864			/* Set ONE_DMA_AT_ONCE for hardware workaround. */
1865			tmp = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1F;
1866			if (tmp == 6 || tmp == 7)
1867				dma_rw_ctl |=
1868				    BGE_PCIDMARWCTL_ONEDMA_ATONCE_GLOBAL;
1869
1870			/* Set PCI-X DMA write workaround. */
1871			dma_rw_ctl |= BGE_PCIDMARWCTL_ASRT_ALL_BE;
1872		}
1873	} else {
1874		/* Conventional PCI bus: 256 bytes for read and write. */
1875		dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(7) |
1876		    BGE_PCIDMARWCTL_WR_WAT_SHIFT(7);
1877
1878		if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
1879		    sc->bge_asicrev != BGE_ASICREV_BCM5750)
1880			dma_rw_ctl |= 0x0F;
1881	}
1882	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
1883	    sc->bge_asicrev == BGE_ASICREV_BCM5701)
1884		dma_rw_ctl |= BGE_PCIDMARWCTL_USE_MRM |
1885		    BGE_PCIDMARWCTL_ASRT_ALL_BE;
1886	if (sc->bge_asicrev == BGE_ASICREV_BCM5703 ||
1887	    sc->bge_asicrev == BGE_ASICREV_BCM5704)
1888		dma_rw_ctl &= ~BGE_PCIDMARWCTL_MINDMA;
1889	if (BGE_IS_5717_PLUS(sc)) {
1890		dma_rw_ctl &= ~BGE_PCIDMARWCTL_DIS_CACHE_ALIGNMENT;
1891		if (sc->bge_chipid == BGE_CHIPID_BCM57765_A0)
1892			dma_rw_ctl &= ~BGE_PCIDMARWCTL_CRDRDR_RDMA_MRRS_MSK;
1893		/*
1894		 * Enable HW workaround for controllers that misinterpret
1895		 * a status tag update and leave interrupts permanently
1896		 * disabled.
1897		 */
1898		if (!BGE_IS_57765_PLUS(sc) &&
1899		    sc->bge_asicrev != BGE_ASICREV_BCM5717 &&
1900		    sc->bge_asicrev != BGE_ASICREV_BCM5762)
1901			dma_rw_ctl |= BGE_PCIDMARWCTL_TAGGED_STATUS_WA;
1902	}
1903	pci_write_config(sc->bge_dev, BGE_PCI_DMA_RW_CTL, dma_rw_ctl, 4);
1904
1905	/*
1906	 * Set up general mode register.
1907	 */
1908	mode_ctl = bge_dma_swap_options(sc);
1909	if (sc->bge_asicrev == BGE_ASICREV_BCM5720 ||
1910	    sc->bge_asicrev == BGE_ASICREV_BCM5762) {
1911		/* Retain Host-2-BMC settings written by APE firmware. */
1912		mode_ctl |= CSR_READ_4(sc, BGE_MODE_CTL) &
1913		    (BGE_MODECTL_BYTESWAP_B2HRX_DATA |
1914		    BGE_MODECTL_WORDSWAP_B2HRX_DATA |
1915		    BGE_MODECTL_B2HRX_ENABLE | BGE_MODECTL_HTX2B_ENABLE);
1916	}
1917	mode_ctl |= BGE_MODECTL_MAC_ATTN_INTR | BGE_MODECTL_HOST_SEND_BDS |
1918	    BGE_MODECTL_TX_NO_PHDR_CSUM;
1919
1920	/*
1921	 * BCM5701 B5 have a bug causing data corruption when using
1922	 * 64-bit DMA reads, which can be terminated early and then
1923	 * completed later as 32-bit accesses, in combination with
1924	 * certain bridges.
1925	 */
1926	if (sc->bge_asicrev == BGE_ASICREV_BCM5701 &&
1927	    sc->bge_chipid == BGE_CHIPID_BCM5701_B5)
1928		mode_ctl |= BGE_MODECTL_FORCE_PCI32;
1929
1930	/*
1931	 * Tell the firmware the driver is running
1932	 */
1933	if (sc->bge_asf_mode & ASF_STACKUP)
1934		mode_ctl |= BGE_MODECTL_STACKUP;
1935
1936	CSR_WRITE_4(sc, BGE_MODE_CTL, mode_ctl);
1937
1938	/*
1939	 * Disable memory write invalidate.  Apparently it is not supported
1940	 * properly by these devices.  Also ensure that INTx isn't disabled,
1941	 * as these chips need it even when using MSI.
1942	 */
1943	PCI_CLRBIT(sc->bge_dev, BGE_PCI_CMD,
1944	    PCIM_CMD_INTxDIS | PCIM_CMD_MWIEN, 4);
1945
1946	/* Set the timer prescaler (always 66 MHz). */
1947	CSR_WRITE_4(sc, BGE_MISC_CFG, BGE_32BITTIME_66MHZ);
1948
1949	/* XXX: The Linux tg3 driver does this at the start of brgphy_reset. */
1950	if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
1951		DELAY(40);	/* XXX */
1952
1953		/* Put PHY into ready state */
1954		BGE_CLRBIT(sc, BGE_MISC_CFG, BGE_MISCCFG_EPHY_IDDQ);
1955		CSR_READ_4(sc, BGE_MISC_CFG); /* Flush */
1956		DELAY(40);
1957	}
1958
1959	return (0);
1960}
1961
1962static int
1963bge_blockinit(struct bge_softc *sc)
1964{
1965	struct bge_rcb *rcb;
1966	bus_size_t vrcb;
1967	bge_hostaddr taddr;
1968	uint32_t dmactl, rdmareg, val;
1969	int i, limit;
1970
1971	/*
1972	 * Initialize the memory window pointer register so that
1973	 * we can access the first 32K of internal NIC RAM. This will
1974	 * allow us to set up the TX send ring RCBs and the RX return
1975	 * ring RCBs, plus other things which live in NIC memory.
1976	 */
1977	CSR_WRITE_4(sc, BGE_PCI_MEMWIN_BASEADDR, 0);
1978
1979	/* Note: the BCM5704 has a smaller mbuf space than other chips. */
1980
1981	if (!(BGE_IS_5705_PLUS(sc))) {
1982		/* Configure mbuf memory pool */
1983		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_BASEADDR, BGE_BUFFPOOL_1);
1984		if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
1985			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x10000);
1986		else
1987			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x18000);
1988
1989		/* Configure DMA resource pool */
1990		CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_BASEADDR,
1991		    BGE_DMA_DESCRIPTORS);
1992		CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_LEN, 0x2000);
1993	}
1994
1995	/* Configure mbuf pool watermarks */
1996	if (BGE_IS_5717_PLUS(sc)) {
1997		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
1998		if (sc->bge_ifp->if_mtu > ETHERMTU) {
1999			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x7e);
2000			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0xea);
2001		} else {
2002			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x2a);
2003			CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0xa0);
2004		}
2005	} else if (!BGE_IS_5705_PLUS(sc)) {
2006		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x50);
2007		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x20);
2008		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x60);
2009	} else if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
2010		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
2011		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x04);
2012		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x10);
2013	} else {
2014		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
2015		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x10);
2016		CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x60);
2017	}
2018
2019	/* Configure DMA resource watermarks */
2020	CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_LOWAT, 5);
2021	CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_HIWAT, 10);
2022
2023	/* Enable buffer manager */
2024	val = BGE_BMANMODE_ENABLE | BGE_BMANMODE_LOMBUF_ATTN;
2025	/*
2026	 * Change the arbitration algorithm of TXMBUF read request to
2027	 * round-robin instead of priority based for BCM5719.  When
2028	 * TXFIFO is almost empty, RDMA will hold its request until
2029	 * TXFIFO is not almost empty.
2030	 */
2031	if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
2032		val |= BGE_BMANMODE_NO_TX_UNDERRUN;
2033	CSR_WRITE_4(sc, BGE_BMAN_MODE, val);
2034
2035	/* Poll for buffer manager start indication */
2036	for (i = 0; i < BGE_TIMEOUT; i++) {
2037		DELAY(10);
2038		if (CSR_READ_4(sc, BGE_BMAN_MODE) & BGE_BMANMODE_ENABLE)
2039			break;
2040	}
2041
2042	if (i == BGE_TIMEOUT) {
2043		device_printf(sc->bge_dev, "buffer manager failed to start\n");
2044		return (ENXIO);
2045	}
2046
2047	/* Enable flow-through queues */
2048	CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
2049	CSR_WRITE_4(sc, BGE_FTQ_RESET, 0);
2050
2051	/* Wait until queue initialization is complete */
2052	for (i = 0; i < BGE_TIMEOUT; i++) {
2053		DELAY(10);
2054		if (CSR_READ_4(sc, BGE_FTQ_RESET) == 0)
2055			break;
2056	}
2057
2058	if (i == BGE_TIMEOUT) {
2059		device_printf(sc->bge_dev, "flow-through queue init failed\n");
2060		return (ENXIO);
2061	}
2062
2063	/*
2064	 * Summary of rings supported by the controller:
2065	 *
2066	 * Standard Receive Producer Ring
2067	 * - This ring is used to feed receive buffers for "standard"
2068	 *   sized frames (typically 1536 bytes) to the controller.
2069	 *
2070	 * Jumbo Receive Producer Ring
2071	 * - This ring is used to feed receive buffers for jumbo sized
2072	 *   frames (i.e. anything bigger than the "standard" frames)
2073	 *   to the controller.
2074	 *
2075	 * Mini Receive Producer Ring
2076	 * - This ring is used to feed receive buffers for "mini"
2077	 *   sized frames to the controller.
2078	 * - This feature required external memory for the controller
2079	 *   but was never used in a production system.  Should always
2080	 *   be disabled.
2081	 *
2082	 * Receive Return Ring
2083	 * - After the controller has placed an incoming frame into a
2084	 *   receive buffer that buffer is moved into a receive return
2085	 *   ring.  The driver is then responsible to passing the
2086	 *   buffer up to the stack.  Many versions of the controller
2087	 *   support multiple RR rings.
2088	 *
2089	 * Send Ring
2090	 * - This ring is used for outgoing frames.  Many versions of
2091	 *   the controller support multiple send rings.
2092	 */
2093
2094	/* Initialize the standard receive producer ring control block. */
2095	rcb = &sc->bge_ldata.bge_info.bge_std_rx_rcb;
2096	rcb->bge_hostaddr.bge_addr_lo =
2097	    BGE_ADDR_LO(sc->bge_ldata.bge_rx_std_ring_paddr);
2098	rcb->bge_hostaddr.bge_addr_hi =
2099	    BGE_ADDR_HI(sc->bge_ldata.bge_rx_std_ring_paddr);
2100	bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
2101	    sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREREAD);
2102	if (BGE_IS_5717_PLUS(sc)) {
2103		/*
2104		 * Bits 31-16: Programmable ring size (2048, 1024, 512, .., 32)
2105		 * Bits 15-2 : Maximum RX frame size
2106		 * Bit 1     : 1 = Ring Disabled, 0 = Ring ENabled
2107		 * Bit 0     : Reserved
2108		 */
2109		rcb->bge_maxlen_flags =
2110		    BGE_RCB_MAXLEN_FLAGS(512, BGE_MAX_FRAMELEN << 2);
2111	} else if (BGE_IS_5705_PLUS(sc)) {
2112		/*
2113		 * Bits 31-16: Programmable ring size (512, 256, 128, 64, 32)
2114		 * Bits 15-2 : Reserved (should be 0)
2115		 * Bit 1     : 1 = Ring Disabled, 0 = Ring Enabled
2116		 * Bit 0     : Reserved
2117		 */
2118		rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(512, 0);
2119	} else {
2120		/*
2121		 * Ring size is always XXX entries
2122		 * Bits 31-16: Maximum RX frame size
2123		 * Bits 15-2 : Reserved (should be 0)
2124		 * Bit 1     : 1 = Ring Disabled, 0 = Ring Enabled
2125		 * Bit 0     : Reserved
2126		 */
2127		rcb->bge_maxlen_flags =
2128		    BGE_RCB_MAXLEN_FLAGS(BGE_MAX_FRAMELEN, 0);
2129	}
2130	if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
2131	    sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
2132	    sc->bge_asicrev == BGE_ASICREV_BCM5720)
2133		rcb->bge_nicaddr = BGE_STD_RX_RINGS_5717;
2134	else
2135		rcb->bge_nicaddr = BGE_STD_RX_RINGS;
2136	/* Write the standard receive producer ring control block. */
2137	CSR_WRITE_4(sc, BGE_RX_STD_RCB_HADDR_HI, rcb->bge_hostaddr.bge_addr_hi);
2138	CSR_WRITE_4(sc, BGE_RX_STD_RCB_HADDR_LO, rcb->bge_hostaddr.bge_addr_lo);
2139	CSR_WRITE_4(sc, BGE_RX_STD_RCB_MAXLEN_FLAGS, rcb->bge_maxlen_flags);
2140	CSR_WRITE_4(sc, BGE_RX_STD_RCB_NICADDR, rcb->bge_nicaddr);
2141
2142	/* Reset the standard receive producer ring producer index. */
2143	bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, 0);
2144
2145	/*
2146	 * Initialize the jumbo RX producer ring control
2147	 * block.  We set the 'ring disabled' bit in the
2148	 * flags field until we're actually ready to start
2149	 * using this ring (i.e. once we set the MTU
2150	 * high enough to require it).
2151	 */
2152	if (BGE_IS_JUMBO_CAPABLE(sc)) {
2153		rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
2154		/* Get the jumbo receive producer ring RCB parameters. */
2155		rcb->bge_hostaddr.bge_addr_lo =
2156		    BGE_ADDR_LO(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
2157		rcb->bge_hostaddr.bge_addr_hi =
2158		    BGE_ADDR_HI(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
2159		bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
2160		    sc->bge_cdata.bge_rx_jumbo_ring_map,
2161		    BUS_DMASYNC_PREREAD);
2162		rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0,
2163		    BGE_RCB_FLAG_USE_EXT_RX_BD | BGE_RCB_FLAG_RING_DISABLED);
2164		if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
2165		    sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
2166		    sc->bge_asicrev == BGE_ASICREV_BCM5720)
2167			rcb->bge_nicaddr = BGE_JUMBO_RX_RINGS_5717;
2168		else
2169			rcb->bge_nicaddr = BGE_JUMBO_RX_RINGS;
2170		CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_HADDR_HI,
2171		    rcb->bge_hostaddr.bge_addr_hi);
2172		CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_HADDR_LO,
2173		    rcb->bge_hostaddr.bge_addr_lo);
2174		/* Program the jumbo receive producer ring RCB parameters. */
2175		CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_MAXLEN_FLAGS,
2176		    rcb->bge_maxlen_flags);
2177		CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_NICADDR, rcb->bge_nicaddr);
2178		/* Reset the jumbo receive producer ring producer index. */
2179		bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, 0);
2180	}
2181
2182	/* Disable the mini receive producer ring RCB. */
2183	if (BGE_IS_5700_FAMILY(sc)) {
2184		rcb = &sc->bge_ldata.bge_info.bge_mini_rx_rcb;
2185		rcb->bge_maxlen_flags =
2186		    BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED);
2187		CSR_WRITE_4(sc, BGE_RX_MINI_RCB_MAXLEN_FLAGS,
2188		    rcb->bge_maxlen_flags);
2189		/* Reset the mini receive producer ring producer index. */
2190		bge_writembx(sc, BGE_MBX_RX_MINI_PROD_LO, 0);
2191	}
2192
2193	/* Choose de-pipeline mode for BCM5906 A0, A1 and A2. */
2194	if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
2195		if (sc->bge_chipid == BGE_CHIPID_BCM5906_A0 ||
2196		    sc->bge_chipid == BGE_CHIPID_BCM5906_A1 ||
2197		    sc->bge_chipid == BGE_CHIPID_BCM5906_A2)
2198			CSR_WRITE_4(sc, BGE_ISO_PKT_TX,
2199			    (CSR_READ_4(sc, BGE_ISO_PKT_TX) & ~3) | 2);
2200	}
2201	/*
2202	 * The BD ring replenish thresholds control how often the
2203	 * hardware fetches new BD's from the producer rings in host
2204	 * memory.  Setting the value too low on a busy system can
2205	 * starve the hardware and recue the throughpout.
2206	 *
2207	 * Set the BD ring replentish thresholds. The recommended
2208	 * values are 1/8th the number of descriptors allocated to
2209	 * each ring.
2210	 * XXX The 5754 requires a lower threshold, so it might be a
2211	 * requirement of all 575x family chips.  The Linux driver sets
2212	 * the lower threshold for all 5705 family chips as well, but there
2213	 * are reports that it might not need to be so strict.
2214	 *
2215	 * XXX Linux does some extra fiddling here for the 5906 parts as
2216	 * well.
2217	 */
2218	if (BGE_IS_5705_PLUS(sc))
2219		val = 8;
2220	else
2221		val = BGE_STD_RX_RING_CNT / 8;
2222	CSR_WRITE_4(sc, BGE_RBDI_STD_REPL_THRESH, val);
2223	if (BGE_IS_JUMBO_CAPABLE(sc))
2224		CSR_WRITE_4(sc, BGE_RBDI_JUMBO_REPL_THRESH,
2225		    BGE_JUMBO_RX_RING_CNT/8);
2226	if (BGE_IS_5717_PLUS(sc)) {
2227		CSR_WRITE_4(sc, BGE_STD_REPLENISH_LWM, 32);
2228		CSR_WRITE_4(sc, BGE_JMB_REPLENISH_LWM, 16);
2229	}
2230
2231	/*
2232	 * Disable all send rings by setting the 'ring disabled' bit
2233	 * in the flags field of all the TX send ring control blocks,
2234	 * located in NIC memory.
2235	 */
2236	if (!BGE_IS_5705_PLUS(sc))
2237		/* 5700 to 5704 had 16 send rings. */
2238		limit = BGE_TX_RINGS_EXTSSRAM_MAX;
2239	else if (BGE_IS_57765_PLUS(sc) ||
2240	    sc->bge_asicrev == BGE_ASICREV_BCM5762)
2241		limit = 2;
2242	else if (BGE_IS_5717_PLUS(sc))
2243		limit = 4;
2244	else
2245		limit = 1;
2246	vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
2247	for (i = 0; i < limit; i++) {
2248		RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
2249		    BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED));
2250		RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
2251		vrcb += sizeof(struct bge_rcb);
2252	}
2253
2254	/* Configure send ring RCB 0 (we use only the first ring) */
2255	vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
2256	BGE_HOSTADDR(taddr, sc->bge_ldata.bge_tx_ring_paddr);
2257	RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
2258	RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
2259	if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
2260	    sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
2261	    sc->bge_asicrev == BGE_ASICREV_BCM5720)
2262		RCB_WRITE_4(sc, vrcb, bge_nicaddr, BGE_SEND_RING_5717);
2263	else
2264		RCB_WRITE_4(sc, vrcb, bge_nicaddr,
2265		    BGE_NIC_TXRING_ADDR(0, BGE_TX_RING_CNT));
2266	RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
2267	    BGE_RCB_MAXLEN_FLAGS(BGE_TX_RING_CNT, 0));
2268
2269	/*
2270	 * Disable all receive return rings by setting the
2271	 * 'ring diabled' bit in the flags field of all the receive
2272	 * return ring control blocks, located in NIC memory.
2273	 */
2274	if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
2275	    sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
2276	    sc->bge_asicrev == BGE_ASICREV_BCM5720) {
2277		/* Should be 17, use 16 until we get an SRAM map. */
2278		limit = 16;
2279	} else if (!BGE_IS_5705_PLUS(sc))
2280		limit = BGE_RX_RINGS_MAX;
2281	else if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
2282	    sc->bge_asicrev == BGE_ASICREV_BCM5762 ||
2283	    BGE_IS_57765_PLUS(sc))
2284		limit = 4;
2285	else
2286		limit = 1;
2287	/* Disable all receive return rings. */
2288	vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
2289	for (i = 0; i < limit; i++) {
2290		RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, 0);
2291		RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, 0);
2292		RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
2293		    BGE_RCB_FLAG_RING_DISABLED);
2294		RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
2295		bge_writembx(sc, BGE_MBX_RX_CONS0_LO +
2296		    (i * (sizeof(uint64_t))), 0);
2297		vrcb += sizeof(struct bge_rcb);
2298	}
2299
2300	/*
2301	 * Set up receive return ring 0.  Note that the NIC address
2302	 * for RX return rings is 0x0.  The return rings live entirely
2303	 * within the host, so the nicaddr field in the RCB isn't used.
2304	 */
2305	vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
2306	BGE_HOSTADDR(taddr, sc->bge_ldata.bge_rx_return_ring_paddr);
2307	RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
2308	RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
2309	RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
2310	RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
2311	    BGE_RCB_MAXLEN_FLAGS(sc->bge_return_ring_cnt, 0));
2312
2313	/* Set random backoff seed for TX */
2314	CSR_WRITE_4(sc, BGE_TX_RANDOM_BACKOFF,
2315	    (IF_LLADDR(sc->bge_ifp)[0] + IF_LLADDR(sc->bge_ifp)[1] +
2316	    IF_LLADDR(sc->bge_ifp)[2] + IF_LLADDR(sc->bge_ifp)[3] +
2317	    IF_LLADDR(sc->bge_ifp)[4] + IF_LLADDR(sc->bge_ifp)[5]) &
2318	    BGE_TX_BACKOFF_SEED_MASK);
2319
2320	/* Set inter-packet gap */
2321	val = 0x2620;
2322	if (sc->bge_asicrev == BGE_ASICREV_BCM5720 ||
2323	    sc->bge_asicrev == BGE_ASICREV_BCM5762)
2324		val |= CSR_READ_4(sc, BGE_TX_LENGTHS) &
2325		    (BGE_TXLEN_JMB_FRM_LEN_MSK | BGE_TXLEN_CNT_DN_VAL_MSK);
2326	CSR_WRITE_4(sc, BGE_TX_LENGTHS, val);
2327
2328	/*
2329	 * Specify which ring to use for packets that don't match
2330	 * any RX rules.
2331	 */
2332	CSR_WRITE_4(sc, BGE_RX_RULES_CFG, 0x08);
2333
2334	/*
2335	 * Configure number of RX lists. One interrupt distribution
2336	 * list, sixteen active lists, one bad frames class.
2337	 */
2338	CSR_WRITE_4(sc, BGE_RXLP_CFG, 0x181);
2339
2340	/* Inialize RX list placement stats mask. */
2341	CSR_WRITE_4(sc, BGE_RXLP_STATS_ENABLE_MASK, 0x007FFFFF);
2342	CSR_WRITE_4(sc, BGE_RXLP_STATS_CTL, 0x1);
2343
2344	/* Disable host coalescing until we get it set up */
2345	CSR_WRITE_4(sc, BGE_HCC_MODE, 0x00000000);
2346
2347	/* Poll to make sure it's shut down. */
2348	for (i = 0; i < BGE_TIMEOUT; i++) {
2349		DELAY(10);
2350		if (!(CSR_READ_4(sc, BGE_HCC_MODE) & BGE_HCCMODE_ENABLE))
2351			break;
2352	}
2353
2354	if (i == BGE_TIMEOUT) {
2355		device_printf(sc->bge_dev,
2356		    "host coalescing engine failed to idle\n");
2357		return (ENXIO);
2358	}
2359
2360	/* Set up host coalescing defaults */
2361	CSR_WRITE_4(sc, BGE_HCC_RX_COAL_TICKS, sc->bge_rx_coal_ticks);
2362	CSR_WRITE_4(sc, BGE_HCC_TX_COAL_TICKS, sc->bge_tx_coal_ticks);
2363	CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS, sc->bge_rx_max_coal_bds);
2364	CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS, sc->bge_tx_max_coal_bds);
2365	if (!(BGE_IS_5705_PLUS(sc))) {
2366		CSR_WRITE_4(sc, BGE_HCC_RX_COAL_TICKS_INT, 0);
2367		CSR_WRITE_4(sc, BGE_HCC_TX_COAL_TICKS_INT, 0);
2368	}
2369	CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
2370	CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 1);
2371
2372	/* Set up address of statistics block */
2373	if (!(BGE_IS_5705_PLUS(sc))) {
2374		CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_HI,
2375		    BGE_ADDR_HI(sc->bge_ldata.bge_stats_paddr));
2376		CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_LO,
2377		    BGE_ADDR_LO(sc->bge_ldata.bge_stats_paddr));
2378		CSR_WRITE_4(sc, BGE_HCC_STATS_BASEADDR, BGE_STATS_BLOCK);
2379		CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_BASEADDR, BGE_STATUS_BLOCK);
2380		CSR_WRITE_4(sc, BGE_HCC_STATS_TICKS, sc->bge_stat_ticks);
2381	}
2382
2383	/* Set up address of status block */
2384	CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_HI,
2385	    BGE_ADDR_HI(sc->bge_ldata.bge_status_block_paddr));
2386	CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_LO,
2387	    BGE_ADDR_LO(sc->bge_ldata.bge_status_block_paddr));
2388
2389	/* Set up status block size. */
2390	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
2391	    sc->bge_chipid != BGE_CHIPID_BCM5700_C0) {
2392		val = BGE_STATBLKSZ_FULL;
2393		bzero(sc->bge_ldata.bge_status_block, BGE_STATUS_BLK_SZ);
2394	} else {
2395		val = BGE_STATBLKSZ_32BYTE;
2396		bzero(sc->bge_ldata.bge_status_block, 32);
2397	}
2398	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
2399	    sc->bge_cdata.bge_status_map,
2400	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2401
2402	/* Turn on host coalescing state machine */
2403	CSR_WRITE_4(sc, BGE_HCC_MODE, val | BGE_HCCMODE_ENABLE);
2404
2405	/* Turn on RX BD completion state machine and enable attentions */
2406	CSR_WRITE_4(sc, BGE_RBDC_MODE,
2407	    BGE_RBDCMODE_ENABLE | BGE_RBDCMODE_ATTN);
2408
2409	/* Turn on RX list placement state machine */
2410	CSR_WRITE_4(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
2411
2412	/* Turn on RX list selector state machine. */
2413	if (!(BGE_IS_5705_PLUS(sc)))
2414		CSR_WRITE_4(sc, BGE_RXLS_MODE, BGE_RXLSMODE_ENABLE);
2415
2416	/* Turn on DMA, clear stats. */
2417	val = BGE_MACMODE_TXDMA_ENB | BGE_MACMODE_RXDMA_ENB |
2418	    BGE_MACMODE_RX_STATS_CLEAR | BGE_MACMODE_TX_STATS_CLEAR |
2419	    BGE_MACMODE_RX_STATS_ENB | BGE_MACMODE_TX_STATS_ENB |
2420	    BGE_MACMODE_FRMHDR_DMA_ENB;
2421
2422	if (sc->bge_flags & BGE_FLAG_TBI)
2423		val |= BGE_PORTMODE_TBI;
2424	else if (sc->bge_flags & BGE_FLAG_MII_SERDES)
2425		val |= BGE_PORTMODE_GMII;
2426	else
2427		val |= BGE_PORTMODE_MII;
2428
2429	/* Allow APE to send/receive frames. */
2430	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
2431		val |= BGE_MACMODE_APE_RX_EN | BGE_MACMODE_APE_TX_EN;
2432
2433	CSR_WRITE_4(sc, BGE_MAC_MODE, val);
2434	DELAY(40);
2435
2436	/* Set misc. local control, enable interrupts on attentions */
2437	BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_ONATTN);
2438
2439#ifdef notdef
2440	/* Assert GPIO pins for PHY reset */
2441	BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_MISCIO_OUT0 |
2442	    BGE_MLC_MISCIO_OUT1 | BGE_MLC_MISCIO_OUT2);
2443	BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_MISCIO_OUTEN0 |
2444	    BGE_MLC_MISCIO_OUTEN1 | BGE_MLC_MISCIO_OUTEN2);
2445#endif
2446
2447	/* Turn on DMA completion state machine */
2448	if (!(BGE_IS_5705_PLUS(sc)))
2449		CSR_WRITE_4(sc, BGE_DMAC_MODE, BGE_DMACMODE_ENABLE);
2450
2451	val = BGE_WDMAMODE_ENABLE | BGE_WDMAMODE_ALL_ATTNS;
2452
2453	/* Enable host coalescing bug fix. */
2454	if (BGE_IS_5755_PLUS(sc))
2455		val |= BGE_WDMAMODE_STATUS_TAG_FIX;
2456
2457	/* Request larger DMA burst size to get better performance. */
2458	if (sc->bge_asicrev == BGE_ASICREV_BCM5785)
2459		val |= BGE_WDMAMODE_BURST_ALL_DATA;
2460
2461	/* Turn on write DMA state machine */
2462	CSR_WRITE_4(sc, BGE_WDMA_MODE, val);
2463	DELAY(40);
2464
2465	/* Turn on read DMA state machine */
2466	val = BGE_RDMAMODE_ENABLE | BGE_RDMAMODE_ALL_ATTNS;
2467
2468	if (sc->bge_asicrev == BGE_ASICREV_BCM5717)
2469		val |= BGE_RDMAMODE_MULT_DMA_RD_DIS;
2470
2471	if (sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
2472	    sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
2473	    sc->bge_asicrev == BGE_ASICREV_BCM57780)
2474		val |= BGE_RDMAMODE_BD_SBD_CRPT_ATTN |
2475		    BGE_RDMAMODE_MBUF_RBD_CRPT_ATTN |
2476		    BGE_RDMAMODE_MBUF_SBD_CRPT_ATTN;
2477	if (sc->bge_flags & BGE_FLAG_PCIE)
2478		val |= BGE_RDMAMODE_FIFO_LONG_BURST;
2479	if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) {
2480		val |= BGE_RDMAMODE_TSO4_ENABLE;
2481		if (sc->bge_flags & BGE_FLAG_TSO3 ||
2482		    sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
2483		    sc->bge_asicrev == BGE_ASICREV_BCM57780)
2484			val |= BGE_RDMAMODE_TSO6_ENABLE;
2485	}
2486
2487	if (sc->bge_asicrev == BGE_ASICREV_BCM5720 ||
2488	    sc->bge_asicrev == BGE_ASICREV_BCM5762) {
2489		val |= CSR_READ_4(sc, BGE_RDMA_MODE) &
2490			BGE_RDMAMODE_H2BNC_VLAN_DET;
2491		/*
2492		 * Allow multiple outstanding read requests from
2493		 * non-LSO read DMA engine.
2494		 */
2495		val &= ~BGE_RDMAMODE_MULT_DMA_RD_DIS;
2496	}
2497
2498	if (sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
2499	    sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
2500	    sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
2501	    sc->bge_asicrev == BGE_ASICREV_BCM57780 ||
2502	    BGE_IS_5717_PLUS(sc) || BGE_IS_57765_PLUS(sc)) {
2503		if (sc->bge_asicrev == BGE_ASICREV_BCM5762)
2504			rdmareg = BGE_RDMA_RSRVCTRL_REG2;
2505		else
2506			rdmareg = BGE_RDMA_RSRVCTRL;
2507		dmactl = CSR_READ_4(sc, rdmareg);
2508		/*
2509		 * Adjust tx margin to prevent TX data corruption and
2510		 * fix internal FIFO overflow.
2511		 */
2512		if (sc->bge_chipid == BGE_CHIPID_BCM5719_A0 ||
2513		    sc->bge_asicrev == BGE_ASICREV_BCM5762) {
2514			dmactl &= ~(BGE_RDMA_RSRVCTRL_FIFO_LWM_MASK |
2515			    BGE_RDMA_RSRVCTRL_FIFO_HWM_MASK |
2516			    BGE_RDMA_RSRVCTRL_TXMRGN_MASK);
2517			dmactl |= BGE_RDMA_RSRVCTRL_FIFO_LWM_1_5K |
2518			    BGE_RDMA_RSRVCTRL_FIFO_HWM_1_5K |
2519			    BGE_RDMA_RSRVCTRL_TXMRGN_320B;
2520		}
2521		/*
2522		 * Enable fix for read DMA FIFO overruns.
2523		 * The fix is to limit the number of RX BDs
2524		 * the hardware would fetch at a fime.
2525		 */
2526		CSR_WRITE_4(sc, rdmareg, dmactl |
2527		    BGE_RDMA_RSRVCTRL_FIFO_OFLW_FIX);
2528	}
2529
2530	if (sc->bge_asicrev == BGE_ASICREV_BCM5719) {
2531		CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL,
2532		    CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL) |
2533		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_BD_4K |
2534		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_LSO_4K);
2535	} else if (sc->bge_asicrev == BGE_ASICREV_BCM5720) {
2536		/*
2537		 * Allow 4KB burst length reads for non-LSO frames.
2538		 * Enable 512B burst length reads for buffer descriptors.
2539		 */
2540		CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL,
2541		    CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL) |
2542		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_BD_512 |
2543		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_LSO_4K);
2544	} else if (sc->bge_asicrev == BGE_ASICREV_BCM5762) {
2545		CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL_REG2,
2546		    CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL_REG2) |
2547		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_BD_4K |
2548		    BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_LSO_4K);
2549	}
2550
2551	CSR_WRITE_4(sc, BGE_RDMA_MODE, val);
2552	DELAY(40);
2553
2554	if (sc->bge_flags & BGE_FLAG_RDMA_BUG) {
2555		for (i = 0; i < BGE_NUM_RDMA_CHANNELS / 2; i++) {
2556			val = CSR_READ_4(sc, BGE_RDMA_LENGTH + i * 4);
2557			if ((val & 0xFFFF) > BGE_FRAMELEN)
2558				break;
2559			if (((val >> 16) & 0xFFFF) > BGE_FRAMELEN)
2560				break;
2561		}
2562		if (i != BGE_NUM_RDMA_CHANNELS / 2) {
2563			val = CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL);
2564			if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
2565				val |= BGE_RDMA_TX_LENGTH_WA_5719;
2566			else
2567				val |= BGE_RDMA_TX_LENGTH_WA_5720;
2568			CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL, val);
2569		}
2570	}
2571
2572	/* Turn on RX data completion state machine */
2573	CSR_WRITE_4(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
2574
2575	/* Turn on RX BD initiator state machine */
2576	CSR_WRITE_4(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
2577
2578	/* Turn on RX data and RX BD initiator state machine */
2579	CSR_WRITE_4(sc, BGE_RDBDI_MODE, BGE_RDBDIMODE_ENABLE);
2580
2581	/* Turn on Mbuf cluster free state machine */
2582	if (!(BGE_IS_5705_PLUS(sc)))
2583		CSR_WRITE_4(sc, BGE_MBCF_MODE, BGE_MBCFMODE_ENABLE);
2584
2585	/* Turn on send BD completion state machine */
2586	CSR_WRITE_4(sc, BGE_SBDC_MODE, BGE_SBDCMODE_ENABLE);
2587
2588	/* Turn on send data completion state machine */
2589	val = BGE_SDCMODE_ENABLE;
2590	if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
2591		val |= BGE_SDCMODE_CDELAY;
2592	CSR_WRITE_4(sc, BGE_SDC_MODE, val);
2593
2594	/* Turn on send data initiator state machine */
2595	if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3))
2596		CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE |
2597		    BGE_SDIMODE_HW_LSO_PRE_DMA);
2598	else
2599		CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE);
2600
2601	/* Turn on send BD initiator state machine */
2602	CSR_WRITE_4(sc, BGE_SBDI_MODE, BGE_SBDIMODE_ENABLE);
2603
2604	/* Turn on send BD selector state machine */
2605	CSR_WRITE_4(sc, BGE_SRS_MODE, BGE_SRSMODE_ENABLE);
2606
2607	CSR_WRITE_4(sc, BGE_SDI_STATS_ENABLE_MASK, 0x007FFFFF);
2608	CSR_WRITE_4(sc, BGE_SDI_STATS_CTL,
2609	    BGE_SDISTATSCTL_ENABLE | BGE_SDISTATSCTL_FASTER);
2610
2611	/* ack/clear link change events */
2612	CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
2613	    BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
2614	    BGE_MACSTAT_LINK_CHANGED);
2615	CSR_WRITE_4(sc, BGE_MI_STS, 0);
2616
2617	/*
2618	 * Enable attention when the link has changed state for
2619	 * devices that use auto polling.
2620	 */
2621	if (sc->bge_flags & BGE_FLAG_TBI) {
2622		CSR_WRITE_4(sc, BGE_MI_STS, BGE_MISTS_LINK);
2623	} else {
2624		if (sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) {
2625			CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
2626			DELAY(80);
2627		}
2628		if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
2629		    sc->bge_chipid != BGE_CHIPID_BCM5700_B2)
2630			CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
2631			    BGE_EVTENB_MI_INTERRUPT);
2632	}
2633
2634	/*
2635	 * Clear any pending link state attention.
2636	 * Otherwise some link state change events may be lost until attention
2637	 * is cleared by bge_intr() -> bge_link_upd() sequence.
2638	 * It's not necessary on newer BCM chips - perhaps enabling link
2639	 * state change attentions implies clearing pending attention.
2640	 */
2641	CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
2642	    BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
2643	    BGE_MACSTAT_LINK_CHANGED);
2644
2645	/* Enable link state change attentions. */
2646	BGE_SETBIT(sc, BGE_MAC_EVT_ENB, BGE_EVTENB_LINK_CHANGED);
2647
2648	return (0);
2649}
2650
2651static const struct bge_revision *
2652bge_lookup_rev(uint32_t chipid)
2653{
2654	const struct bge_revision *br;
2655
2656	for (br = bge_revisions; br->br_name != NULL; br++) {
2657		if (br->br_chipid == chipid)
2658			return (br);
2659	}
2660
2661	for (br = bge_majorrevs; br->br_name != NULL; br++) {
2662		if (br->br_chipid == BGE_ASICREV(chipid))
2663			return (br);
2664	}
2665
2666	return (NULL);
2667}
2668
2669static const struct bge_vendor *
2670bge_lookup_vendor(uint16_t vid)
2671{
2672	const struct bge_vendor *v;
2673
2674	for (v = bge_vendors; v->v_name != NULL; v++)
2675		if (v->v_id == vid)
2676			return (v);
2677
2678	return (NULL);
2679}
2680
2681static uint32_t
2682bge_chipid(device_t dev)
2683{
2684	uint32_t id;
2685
2686	id = pci_read_config(dev, BGE_PCI_MISC_CTL, 4) >>
2687	    BGE_PCIMISCCTL_ASICREV_SHIFT;
2688	if (BGE_ASICREV(id) == BGE_ASICREV_USE_PRODID_REG) {
2689		/*
2690		 * Find the ASCI revision.  Different chips use different
2691		 * registers.
2692		 */
2693		switch (pci_get_device(dev)) {
2694		case BCOM_DEVICEID_BCM5717:
2695		case BCOM_DEVICEID_BCM5718:
2696		case BCOM_DEVICEID_BCM5719:
2697		case BCOM_DEVICEID_BCM5720:
2698		case BCOM_DEVICEID_BCM5725:
2699		case BCOM_DEVICEID_BCM5727:
2700		case BCOM_DEVICEID_BCM5762:
2701		case BCOM_DEVICEID_BCM57764:
2702		case BCOM_DEVICEID_BCM57767:
2703		case BCOM_DEVICEID_BCM57787:
2704			id = pci_read_config(dev,
2705			    BGE_PCI_GEN2_PRODID_ASICREV, 4);
2706			break;
2707		case BCOM_DEVICEID_BCM57761:
2708		case BCOM_DEVICEID_BCM57762:
2709		case BCOM_DEVICEID_BCM57765:
2710		case BCOM_DEVICEID_BCM57766:
2711		case BCOM_DEVICEID_BCM57781:
2712		case BCOM_DEVICEID_BCM57782:
2713		case BCOM_DEVICEID_BCM57785:
2714		case BCOM_DEVICEID_BCM57786:
2715		case BCOM_DEVICEID_BCM57791:
2716		case BCOM_DEVICEID_BCM57795:
2717			id = pci_read_config(dev,
2718			    BGE_PCI_GEN15_PRODID_ASICREV, 4);
2719			break;
2720		default:
2721			id = pci_read_config(dev, BGE_PCI_PRODID_ASICREV, 4);
2722		}
2723	}
2724	return (id);
2725}
2726
2727/*
2728 * Probe for a Broadcom chip. Check the PCI vendor and device IDs
2729 * against our list and return its name if we find a match.
2730 *
2731 * Note that since the Broadcom controller contains VPD support, we
2732 * try to get the device name string from the controller itself instead
2733 * of the compiled-in string. It guarantees we'll always announce the
2734 * right product name. We fall back to the compiled-in string when
2735 * VPD is unavailable or corrupt.
2736 */
2737static int
2738bge_probe(device_t dev)
2739{
2740	char buf[96];
2741	char model[64];
2742	const struct bge_revision *br;
2743	const char *pname;
2744	struct bge_softc *sc;
2745	const struct bge_type *t = bge_devs;
2746	const struct bge_vendor *v;
2747	uint32_t id;
2748	uint16_t did, vid;
2749
2750	sc = device_get_softc(dev);
2751	sc->bge_dev = dev;
2752	vid = pci_get_vendor(dev);
2753	did = pci_get_device(dev);
2754	while(t->bge_vid != 0) {
2755		if ((vid == t->bge_vid) && (did == t->bge_did)) {
2756			id = bge_chipid(dev);
2757			br = bge_lookup_rev(id);
2758			if (bge_has_eaddr(sc) &&
2759			    pci_get_vpd_ident(dev, &pname) == 0)
2760				snprintf(model, sizeof(model), "%s", pname);
2761			else {
2762				v = bge_lookup_vendor(vid);
2763				snprintf(model, sizeof(model), "%s %s",
2764				    v != NULL ? v->v_name : "Unknown",
2765				    br != NULL ? br->br_name :
2766				    "NetXtreme/NetLink Ethernet Controller");
2767			}
2768			snprintf(buf, sizeof(buf), "%s, %sASIC rev. %#08x",
2769			    model, br != NULL ? "" : "unknown ", id);
2770			device_set_desc_copy(dev, buf);
2771			return (BUS_PROBE_DEFAULT);
2772		}
2773		t++;
2774	}
2775
2776	return (ENXIO);
2777}
2778
2779static void
2780bge_dma_free(struct bge_softc *sc)
2781{
2782	int i;
2783
2784	/* Destroy DMA maps for RX buffers. */
2785	for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
2786		if (sc->bge_cdata.bge_rx_std_dmamap[i])
2787			bus_dmamap_destroy(sc->bge_cdata.bge_rx_mtag,
2788			    sc->bge_cdata.bge_rx_std_dmamap[i]);
2789	}
2790	if (sc->bge_cdata.bge_rx_std_sparemap)
2791		bus_dmamap_destroy(sc->bge_cdata.bge_rx_mtag,
2792		    sc->bge_cdata.bge_rx_std_sparemap);
2793
2794	/* Destroy DMA maps for jumbo RX buffers. */
2795	for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
2796		if (sc->bge_cdata.bge_rx_jumbo_dmamap[i])
2797			bus_dmamap_destroy(sc->bge_cdata.bge_mtag_jumbo,
2798			    sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
2799	}
2800	if (sc->bge_cdata.bge_rx_jumbo_sparemap)
2801		bus_dmamap_destroy(sc->bge_cdata.bge_mtag_jumbo,
2802		    sc->bge_cdata.bge_rx_jumbo_sparemap);
2803
2804	/* Destroy DMA maps for TX buffers. */
2805	for (i = 0; i < BGE_TX_RING_CNT; i++) {
2806		if (sc->bge_cdata.bge_tx_dmamap[i])
2807			bus_dmamap_destroy(sc->bge_cdata.bge_tx_mtag,
2808			    sc->bge_cdata.bge_tx_dmamap[i]);
2809	}
2810
2811	if (sc->bge_cdata.bge_rx_mtag)
2812		bus_dma_tag_destroy(sc->bge_cdata.bge_rx_mtag);
2813	if (sc->bge_cdata.bge_mtag_jumbo)
2814		bus_dma_tag_destroy(sc->bge_cdata.bge_mtag_jumbo);
2815	if (sc->bge_cdata.bge_tx_mtag)
2816		bus_dma_tag_destroy(sc->bge_cdata.bge_tx_mtag);
2817
2818	/* Destroy standard RX ring. */
2819	if (sc->bge_cdata.bge_rx_std_ring_map)
2820		bus_dmamap_unload(sc->bge_cdata.bge_rx_std_ring_tag,
2821		    sc->bge_cdata.bge_rx_std_ring_map);
2822	if (sc->bge_cdata.bge_rx_std_ring_map && sc->bge_ldata.bge_rx_std_ring)
2823		bus_dmamem_free(sc->bge_cdata.bge_rx_std_ring_tag,
2824		    sc->bge_ldata.bge_rx_std_ring,
2825		    sc->bge_cdata.bge_rx_std_ring_map);
2826
2827	if (sc->bge_cdata.bge_rx_std_ring_tag)
2828		bus_dma_tag_destroy(sc->bge_cdata.bge_rx_std_ring_tag);
2829
2830	/* Destroy jumbo RX ring. */
2831	if (sc->bge_cdata.bge_rx_jumbo_ring_map)
2832		bus_dmamap_unload(sc->bge_cdata.bge_rx_jumbo_ring_tag,
2833		    sc->bge_cdata.bge_rx_jumbo_ring_map);
2834
2835	if (sc->bge_cdata.bge_rx_jumbo_ring_map &&
2836	    sc->bge_ldata.bge_rx_jumbo_ring)
2837		bus_dmamem_free(sc->bge_cdata.bge_rx_jumbo_ring_tag,
2838		    sc->bge_ldata.bge_rx_jumbo_ring,
2839		    sc->bge_cdata.bge_rx_jumbo_ring_map);
2840
2841	if (sc->bge_cdata.bge_rx_jumbo_ring_tag)
2842		bus_dma_tag_destroy(sc->bge_cdata.bge_rx_jumbo_ring_tag);
2843
2844	/* Destroy RX return ring. */
2845	if (sc->bge_cdata.bge_rx_return_ring_map)
2846		bus_dmamap_unload(sc->bge_cdata.bge_rx_return_ring_tag,
2847		    sc->bge_cdata.bge_rx_return_ring_map);
2848
2849	if (sc->bge_cdata.bge_rx_return_ring_map &&
2850	    sc->bge_ldata.bge_rx_return_ring)
2851		bus_dmamem_free(sc->bge_cdata.bge_rx_return_ring_tag,
2852		    sc->bge_ldata.bge_rx_return_ring,
2853		    sc->bge_cdata.bge_rx_return_ring_map);
2854
2855	if (sc->bge_cdata.bge_rx_return_ring_tag)
2856		bus_dma_tag_destroy(sc->bge_cdata.bge_rx_return_ring_tag);
2857
2858	/* Destroy TX ring. */
2859	if (sc->bge_cdata.bge_tx_ring_map)
2860		bus_dmamap_unload(sc->bge_cdata.bge_tx_ring_tag,
2861		    sc->bge_cdata.bge_tx_ring_map);
2862
2863	if (sc->bge_cdata.bge_tx_ring_map && sc->bge_ldata.bge_tx_ring)
2864		bus_dmamem_free(sc->bge_cdata.bge_tx_ring_tag,
2865		    sc->bge_ldata.bge_tx_ring,
2866		    sc->bge_cdata.bge_tx_ring_map);
2867
2868	if (sc->bge_cdata.bge_tx_ring_tag)
2869		bus_dma_tag_destroy(sc->bge_cdata.bge_tx_ring_tag);
2870
2871	/* Destroy status block. */
2872	if (sc->bge_cdata.bge_status_map)
2873		bus_dmamap_unload(sc->bge_cdata.bge_status_tag,
2874		    sc->bge_cdata.bge_status_map);
2875
2876	if (sc->bge_cdata.bge_status_map && sc->bge_ldata.bge_status_block)
2877		bus_dmamem_free(sc->bge_cdata.bge_status_tag,
2878		    sc->bge_ldata.bge_status_block,
2879		    sc->bge_cdata.bge_status_map);
2880
2881	if (sc->bge_cdata.bge_status_tag)
2882		bus_dma_tag_destroy(sc->bge_cdata.bge_status_tag);
2883
2884	/* Destroy statistics block. */
2885	if (sc->bge_cdata.bge_stats_map)
2886		bus_dmamap_unload(sc->bge_cdata.bge_stats_tag,
2887		    sc->bge_cdata.bge_stats_map);
2888
2889	if (sc->bge_cdata.bge_stats_map && sc->bge_ldata.bge_stats)
2890		bus_dmamem_free(sc->bge_cdata.bge_stats_tag,
2891		    sc->bge_ldata.bge_stats,
2892		    sc->bge_cdata.bge_stats_map);
2893
2894	if (sc->bge_cdata.bge_stats_tag)
2895		bus_dma_tag_destroy(sc->bge_cdata.bge_stats_tag);
2896
2897	if (sc->bge_cdata.bge_buffer_tag)
2898		bus_dma_tag_destroy(sc->bge_cdata.bge_buffer_tag);
2899
2900	/* Destroy the parent tag. */
2901	if (sc->bge_cdata.bge_parent_tag)
2902		bus_dma_tag_destroy(sc->bge_cdata.bge_parent_tag);
2903}
2904
2905static int
2906bge_dma_ring_alloc(struct bge_softc *sc, bus_size_t alignment,
2907    bus_size_t maxsize, bus_dma_tag_t *tag, uint8_t **ring, bus_dmamap_t *map,
2908    bus_addr_t *paddr, const char *msg)
2909{
2910	struct bge_dmamap_arg ctx;
2911	int error;
2912
2913	error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
2914	    alignment, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
2915	    NULL, maxsize, 1, maxsize, 0, NULL, NULL, tag);
2916	if (error != 0) {
2917		device_printf(sc->bge_dev,
2918		    "could not create %s dma tag\n", msg);
2919		return (ENOMEM);
2920	}
2921	/* Allocate DMA'able memory for ring. */
2922	error = bus_dmamem_alloc(*tag, (void **)ring,
2923	    BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT, map);
2924	if (error != 0) {
2925		device_printf(sc->bge_dev,
2926		    "could not allocate DMA'able memory for %s\n", msg);
2927		return (ENOMEM);
2928	}
2929	/* Load the address of the ring. */
2930	ctx.bge_busaddr = 0;
2931	error = bus_dmamap_load(*tag, *map, *ring, maxsize, bge_dma_map_addr,
2932	    &ctx, BUS_DMA_NOWAIT);
2933	if (error != 0) {
2934		device_printf(sc->bge_dev,
2935		    "could not load DMA'able memory for %s\n", msg);
2936		return (ENOMEM);
2937	}
2938	*paddr = ctx.bge_busaddr;
2939	return (0);
2940}
2941
2942static int
2943bge_dma_alloc(struct bge_softc *sc)
2944{
2945	bus_addr_t lowaddr;
2946	bus_size_t rxmaxsegsz, sbsz, txsegsz, txmaxsegsz;
2947	int i, error;
2948
2949	lowaddr = BUS_SPACE_MAXADDR;
2950	if ((sc->bge_flags & BGE_FLAG_40BIT_BUG) != 0)
2951		lowaddr = BGE_DMA_MAXADDR;
2952	/*
2953	 * Allocate the parent bus DMA tag appropriate for PCI.
2954	 */
2955	error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev),
2956	    1, 0, lowaddr, BUS_SPACE_MAXADDR, NULL,
2957	    NULL, BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT,
2958	    0, NULL, NULL, &sc->bge_cdata.bge_parent_tag);
2959	if (error != 0) {
2960		device_printf(sc->bge_dev,
2961		    "could not allocate parent dma tag\n");
2962		return (ENOMEM);
2963	}
2964
2965	/* Create tag for standard RX ring. */
2966	error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_STD_RX_RING_SZ,
2967	    &sc->bge_cdata.bge_rx_std_ring_tag,
2968	    (uint8_t **)&sc->bge_ldata.bge_rx_std_ring,
2969	    &sc->bge_cdata.bge_rx_std_ring_map,
2970	    &sc->bge_ldata.bge_rx_std_ring_paddr, "RX ring");
2971	if (error)
2972		return (error);
2973
2974	/* Create tag for RX return ring. */
2975	error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_RX_RTN_RING_SZ(sc),
2976	    &sc->bge_cdata.bge_rx_return_ring_tag,
2977	    (uint8_t **)&sc->bge_ldata.bge_rx_return_ring,
2978	    &sc->bge_cdata.bge_rx_return_ring_map,
2979	    &sc->bge_ldata.bge_rx_return_ring_paddr, "RX return ring");
2980	if (error)
2981		return (error);
2982
2983	/* Create tag for TX ring. */
2984	error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_TX_RING_SZ,
2985	    &sc->bge_cdata.bge_tx_ring_tag,
2986	    (uint8_t **)&sc->bge_ldata.bge_tx_ring,
2987	    &sc->bge_cdata.bge_tx_ring_map,
2988	    &sc->bge_ldata.bge_tx_ring_paddr, "TX ring");
2989	if (error)
2990		return (error);
2991
2992	/*
2993	 * Create tag for status block.
2994	 * Because we only use single Tx/Rx/Rx return ring, use
2995	 * minimum status block size except BCM5700 AX/BX which
2996	 * seems to want to see full status block size regardless
2997	 * of configured number of ring.
2998	 */
2999	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
3000	    sc->bge_chipid != BGE_CHIPID_BCM5700_C0)
3001		sbsz = BGE_STATUS_BLK_SZ;
3002	else
3003		sbsz = 32;
3004	error = bge_dma_ring_alloc(sc, PAGE_SIZE, sbsz,
3005	    &sc->bge_cdata.bge_status_tag,
3006	    (uint8_t **)&sc->bge_ldata.bge_status_block,
3007	    &sc->bge_cdata.bge_status_map,
3008	    &sc->bge_ldata.bge_status_block_paddr, "status block");
3009	if (error)
3010		return (error);
3011
3012	/* Create tag for statistics block. */
3013	error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_STATS_SZ,
3014	    &sc->bge_cdata.bge_stats_tag,
3015	    (uint8_t **)&sc->bge_ldata.bge_stats,
3016	    &sc->bge_cdata.bge_stats_map,
3017	    &sc->bge_ldata.bge_stats_paddr, "statistics block");
3018	if (error)
3019		return (error);
3020
3021	/* Create tag for jumbo RX ring. */
3022	if (BGE_IS_JUMBO_CAPABLE(sc)) {
3023		error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_JUMBO_RX_RING_SZ,
3024		    &sc->bge_cdata.bge_rx_jumbo_ring_tag,
3025		    (uint8_t **)&sc->bge_ldata.bge_rx_jumbo_ring,
3026		    &sc->bge_cdata.bge_rx_jumbo_ring_map,
3027		    &sc->bge_ldata.bge_rx_jumbo_ring_paddr, "jumbo RX ring");
3028		if (error)
3029			return (error);
3030	}
3031
3032	/* Create parent tag for buffers. */
3033	if ((sc->bge_flags & BGE_FLAG_4G_BNDRY_BUG) != 0) {
3034		/*
3035		 * XXX
3036		 * watchdog timeout issue was observed on BCM5704 which
3037		 * lives behind PCI-X bridge(e.g AMD 8131 PCI-X bridge).
3038		 * Both limiting DMA address space to 32bits and flushing
3039		 * mailbox write seem to address the issue.
3040		 */
3041		if (sc->bge_pcixcap != 0)
3042			lowaddr = BUS_SPACE_MAXADDR_32BIT;
3043	}
3044	error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev), 1, 0, lowaddr,
3045	    BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE_32BIT, 0,
3046	    BUS_SPACE_MAXSIZE_32BIT, 0, NULL, NULL,
3047	    &sc->bge_cdata.bge_buffer_tag);
3048	if (error != 0) {
3049		device_printf(sc->bge_dev,
3050		    "could not allocate buffer dma tag\n");
3051		return (ENOMEM);
3052	}
3053	/* Create tag for Tx mbufs. */
3054	if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) {
3055		txsegsz = BGE_TSOSEG_SZ;
3056		txmaxsegsz = 65535 + sizeof(struct ether_vlan_header);
3057	} else {
3058		txsegsz = MCLBYTES;
3059		txmaxsegsz = MCLBYTES * BGE_NSEG_NEW;
3060	}
3061	error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_tag, 1,
3062	    0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
3063	    txmaxsegsz, BGE_NSEG_NEW, txsegsz, 0, NULL, NULL,
3064	    &sc->bge_cdata.bge_tx_mtag);
3065
3066	if (error) {
3067		device_printf(sc->bge_dev, "could not allocate TX dma tag\n");
3068		return (ENOMEM);
3069	}
3070
3071	/* Create tag for Rx mbufs. */
3072	if (sc->bge_flags & BGE_FLAG_JUMBO_STD)
3073		rxmaxsegsz = MJUM9BYTES;
3074	else
3075		rxmaxsegsz = MCLBYTES;
3076	error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_tag, 1, 0,
3077	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, rxmaxsegsz, 1,
3078	    rxmaxsegsz, 0, NULL, NULL, &sc->bge_cdata.bge_rx_mtag);
3079
3080	if (error) {
3081		device_printf(sc->bge_dev, "could not allocate RX dma tag\n");
3082		return (ENOMEM);
3083	}
3084
3085	/* Create DMA maps for RX buffers. */
3086	error = bus_dmamap_create(sc->bge_cdata.bge_rx_mtag, 0,
3087	    &sc->bge_cdata.bge_rx_std_sparemap);
3088	if (error) {
3089		device_printf(sc->bge_dev,
3090		    "can't create spare DMA map for RX\n");
3091		return (ENOMEM);
3092	}
3093	for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
3094		error = bus_dmamap_create(sc->bge_cdata.bge_rx_mtag, 0,
3095			    &sc->bge_cdata.bge_rx_std_dmamap[i]);
3096		if (error) {
3097			device_printf(sc->bge_dev,
3098			    "can't create DMA map for RX\n");
3099			return (ENOMEM);
3100		}
3101	}
3102
3103	/* Create DMA maps for TX buffers. */
3104	for (i = 0; i < BGE_TX_RING_CNT; i++) {
3105		error = bus_dmamap_create(sc->bge_cdata.bge_tx_mtag, 0,
3106			    &sc->bge_cdata.bge_tx_dmamap[i]);
3107		if (error) {
3108			device_printf(sc->bge_dev,
3109			    "can't create DMA map for TX\n");
3110			return (ENOMEM);
3111		}
3112	}
3113
3114	/* Create tags for jumbo RX buffers. */
3115	if (BGE_IS_JUMBO_CAPABLE(sc)) {
3116		error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_tag,
3117		    1, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
3118		    NULL, MJUM9BYTES, BGE_NSEG_JUMBO, PAGE_SIZE,
3119		    0, NULL, NULL, &sc->bge_cdata.bge_mtag_jumbo);
3120		if (error) {
3121			device_printf(sc->bge_dev,
3122			    "could not allocate jumbo dma tag\n");
3123			return (ENOMEM);
3124		}
3125		/* Create DMA maps for jumbo RX buffers. */
3126		error = bus_dmamap_create(sc->bge_cdata.bge_mtag_jumbo,
3127		    0, &sc->bge_cdata.bge_rx_jumbo_sparemap);
3128		if (error) {
3129			device_printf(sc->bge_dev,
3130			    "can't create spare DMA map for jumbo RX\n");
3131			return (ENOMEM);
3132		}
3133		for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
3134			error = bus_dmamap_create(sc->bge_cdata.bge_mtag_jumbo,
3135				    0, &sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
3136			if (error) {
3137				device_printf(sc->bge_dev,
3138				    "can't create DMA map for jumbo RX\n");
3139				return (ENOMEM);
3140			}
3141		}
3142	}
3143
3144	return (0);
3145}
3146
3147/*
3148 * Return true if this device has more than one port.
3149 */
3150static int
3151bge_has_multiple_ports(struct bge_softc *sc)
3152{
3153	device_t dev = sc->bge_dev;
3154	u_int b, d, f, fscan, s;
3155
3156	d = pci_get_domain(dev);
3157	b = pci_get_bus(dev);
3158	s = pci_get_slot(dev);
3159	f = pci_get_function(dev);
3160	for (fscan = 0; fscan <= PCI_FUNCMAX; fscan++)
3161		if (fscan != f && pci_find_dbsf(d, b, s, fscan) != NULL)
3162			return (1);
3163	return (0);
3164}
3165
3166/*
3167 * Return true if MSI can be used with this device.
3168 */
3169static int
3170bge_can_use_msi(struct bge_softc *sc)
3171{
3172	int can_use_msi = 0;
3173
3174	if (sc->bge_msi == 0)
3175		return (0);
3176
3177	/* Disable MSI for polling(4). */
3178#ifdef DEVICE_POLLING
3179	return (0);
3180#endif
3181	switch (sc->bge_asicrev) {
3182	case BGE_ASICREV_BCM5714_A0:
3183	case BGE_ASICREV_BCM5714:
3184		/*
3185		 * Apparently, MSI doesn't work when these chips are
3186		 * configured in single-port mode.
3187		 */
3188		if (bge_has_multiple_ports(sc))
3189			can_use_msi = 1;
3190		break;
3191	case BGE_ASICREV_BCM5750:
3192		if (sc->bge_chiprev != BGE_CHIPREV_5750_AX &&
3193		    sc->bge_chiprev != BGE_CHIPREV_5750_BX)
3194			can_use_msi = 1;
3195		break;
3196	default:
3197		if (BGE_IS_575X_PLUS(sc))
3198			can_use_msi = 1;
3199	}
3200	return (can_use_msi);
3201}
3202
3203static int
3204bge_mbox_reorder(struct bge_softc *sc)
3205{
3206	/* Lists of PCI bridges that are known to reorder mailbox writes. */
3207	static const struct mbox_reorder {
3208		const uint16_t vendor;
3209		const uint16_t device;
3210		const char *desc;
3211	} mbox_reorder_lists[] = {
3212		{ 0x1022, 0x7450, "AMD-8131 PCI-X Bridge" },
3213	};
3214	devclass_t pci, pcib;
3215	device_t bus, dev;
3216	int i;
3217
3218	pci = devclass_find("pci");
3219	pcib = devclass_find("pcib");
3220	dev = sc->bge_dev;
3221	bus = device_get_parent(dev);
3222	for (;;) {
3223		dev = device_get_parent(bus);
3224		bus = device_get_parent(dev);
3225		if (device_get_devclass(dev) != pcib)
3226			break;
3227		for (i = 0; i < nitems(mbox_reorder_lists); i++) {
3228			if (pci_get_vendor(dev) ==
3229			    mbox_reorder_lists[i].vendor &&
3230			    pci_get_device(dev) ==
3231			    mbox_reorder_lists[i].device) {
3232				device_printf(sc->bge_dev,
3233				    "enabling MBOX workaround for %s\n",
3234				    mbox_reorder_lists[i].desc);
3235				return (1);
3236			}
3237		}
3238		if (device_get_devclass(bus) != pci)
3239			break;
3240	}
3241	return (0);
3242}
3243
3244static void
3245bge_devinfo(struct bge_softc *sc)
3246{
3247	uint32_t cfg, clk;
3248
3249	device_printf(sc->bge_dev,
3250	    "CHIP ID 0x%08x; ASIC REV 0x%02x; CHIP REV 0x%02x; ",
3251	    sc->bge_chipid, sc->bge_asicrev, sc->bge_chiprev);
3252	if (sc->bge_flags & BGE_FLAG_PCIE)
3253		printf("PCI-E\n");
3254	else if (sc->bge_flags & BGE_FLAG_PCIX) {
3255		printf("PCI-X ");
3256		cfg = CSR_READ_4(sc, BGE_MISC_CFG) & BGE_MISCCFG_BOARD_ID_MASK;
3257		if (cfg == BGE_MISCCFG_BOARD_ID_5704CIOBE)
3258			clk = 133;
3259		else {
3260			clk = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1F;
3261			switch (clk) {
3262			case 0:
3263				clk = 33;
3264				break;
3265			case 2:
3266				clk = 50;
3267				break;
3268			case 4:
3269				clk = 66;
3270				break;
3271			case 6:
3272				clk = 100;
3273				break;
3274			case 7:
3275				clk = 133;
3276				break;
3277			}
3278		}
3279		printf("%u MHz\n", clk);
3280	} else {
3281		if (sc->bge_pcixcap != 0)
3282			printf("PCI on PCI-X ");
3283		else
3284			printf("PCI ");
3285		cfg = pci_read_config(sc->bge_dev, BGE_PCI_PCISTATE, 4);
3286		if (cfg & BGE_PCISTATE_PCI_BUSSPEED)
3287			clk = 66;
3288		else
3289			clk = 33;
3290		if (cfg & BGE_PCISTATE_32BIT_BUS)
3291			printf("%u MHz; 32bit\n", clk);
3292		else
3293			printf("%u MHz; 64bit\n", clk);
3294	}
3295}
3296
3297static int
3298bge_attach(device_t dev)
3299{
3300	struct ifnet *ifp;
3301	struct bge_softc *sc;
3302	uint32_t hwcfg = 0, misccfg, pcistate;
3303	u_char eaddr[ETHER_ADDR_LEN];
3304	int capmask, error, msicount, reg, rid, trys;
3305
3306	sc = device_get_softc(dev);
3307	sc->bge_dev = dev;
3308
3309	BGE_LOCK_INIT(sc, device_get_nameunit(dev));
3310	TASK_INIT(&sc->bge_intr_task, 0, bge_intr_task, sc);
3311	callout_init_mtx(&sc->bge_stat_ch, &sc->bge_mtx, 0);
3312
3313	/*
3314	 * Map control/status registers.
3315	 */
3316	pci_enable_busmaster(dev);
3317
3318	rid = PCIR_BAR(0);
3319	sc->bge_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
3320	    RF_ACTIVE);
3321
3322	if (sc->bge_res == NULL) {
3323		device_printf (sc->bge_dev, "couldn't map BAR0 memory\n");
3324		error = ENXIO;
3325		goto fail;
3326	}
3327
3328	/* Save various chip information. */
3329	sc->bge_func_addr = pci_get_function(dev);
3330	sc->bge_chipid = bge_chipid(dev);
3331	sc->bge_asicrev = BGE_ASICREV(sc->bge_chipid);
3332	sc->bge_chiprev = BGE_CHIPREV(sc->bge_chipid);
3333
3334	/* Set default PHY address. */
3335	sc->bge_phy_addr = 1;
3336	 /*
3337	  * PHY address mapping for various devices.
3338	  *
3339	  *          | F0 Cu | F0 Sr | F1 Cu | F1 Sr |
3340	  * ---------+-------+-------+-------+-------+
3341	  * BCM57XX  |   1   |   X   |   X   |   X   |
3342	  * BCM5704  |   1   |   X   |   1   |   X   |
3343	  * BCM5717  |   1   |   8   |   2   |   9   |
3344	  * BCM5719  |   1   |   8   |   2   |   9   |
3345	  * BCM5720  |   1   |   8   |   2   |   9   |
3346	  *
3347	  *          | F2 Cu | F2 Sr | F3 Cu | F3 Sr |
3348	  * ---------+-------+-------+-------+-------+
3349	  * BCM57XX  |   X   |   X   |   X   |   X   |
3350	  * BCM5704  |   X   |   X   |   X   |   X   |
3351	  * BCM5717  |   X   |   X   |   X   |   X   |
3352	  * BCM5719  |   3   |   10  |   4   |   11  |
3353	  * BCM5720  |   X   |   X   |   X   |   X   |
3354	  *
3355	  * Other addresses may respond but they are not
3356	  * IEEE compliant PHYs and should be ignored.
3357	  */
3358	if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
3359	    sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
3360	    sc->bge_asicrev == BGE_ASICREV_BCM5720) {
3361		if (sc->bge_chipid != BGE_CHIPID_BCM5717_A0) {
3362			if (CSR_READ_4(sc, BGE_SGDIG_STS) &
3363			    BGE_SGDIGSTS_IS_SERDES)
3364				sc->bge_phy_addr = sc->bge_func_addr + 8;
3365			else
3366				sc->bge_phy_addr = sc->bge_func_addr + 1;
3367		} else {
3368			if (CSR_READ_4(sc, BGE_CPMU_PHY_STRAP) &
3369			    BGE_CPMU_PHY_STRAP_IS_SERDES)
3370				sc->bge_phy_addr = sc->bge_func_addr + 8;
3371			else
3372				sc->bge_phy_addr = sc->bge_func_addr + 1;
3373		}
3374	}
3375
3376	if (bge_has_eaddr(sc))
3377		sc->bge_flags |= BGE_FLAG_EADDR;
3378
3379	/* Save chipset family. */
3380	switch (sc->bge_asicrev) {
3381	case BGE_ASICREV_BCM5762:
3382	case BGE_ASICREV_BCM57765:
3383	case BGE_ASICREV_BCM57766:
3384		sc->bge_flags |= BGE_FLAG_57765_PLUS;
3385		/* FALLTHROUGH */
3386	case BGE_ASICREV_BCM5717:
3387	case BGE_ASICREV_BCM5719:
3388	case BGE_ASICREV_BCM5720:
3389		sc->bge_flags |= BGE_FLAG_5717_PLUS | BGE_FLAG_5755_PLUS |
3390		    BGE_FLAG_575X_PLUS | BGE_FLAG_5705_PLUS | BGE_FLAG_JUMBO |
3391		    BGE_FLAG_JUMBO_FRAME;
3392		if (sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
3393		    sc->bge_asicrev == BGE_ASICREV_BCM5720) {
3394			/*
3395			 * Enable work around for DMA engine miscalculation
3396			 * of TXMBUF available space.
3397			 */
3398			sc->bge_flags |= BGE_FLAG_RDMA_BUG;
3399			if (sc->bge_asicrev == BGE_ASICREV_BCM5719 &&
3400			    sc->bge_chipid == BGE_CHIPID_BCM5719_A0) {
3401				/* Jumbo frame on BCM5719 A0 does not work. */
3402				sc->bge_flags &= ~BGE_FLAG_JUMBO;
3403			}
3404		}
3405		break;
3406	case BGE_ASICREV_BCM5755:
3407	case BGE_ASICREV_BCM5761:
3408	case BGE_ASICREV_BCM5784:
3409	case BGE_ASICREV_BCM5785:
3410	case BGE_ASICREV_BCM5787:
3411	case BGE_ASICREV_BCM57780:
3412		sc->bge_flags |= BGE_FLAG_5755_PLUS | BGE_FLAG_575X_PLUS |
3413		    BGE_FLAG_5705_PLUS;
3414		break;
3415	case BGE_ASICREV_BCM5700:
3416	case BGE_ASICREV_BCM5701:
3417	case BGE_ASICREV_BCM5703:
3418	case BGE_ASICREV_BCM5704:
3419		sc->bge_flags |= BGE_FLAG_5700_FAMILY | BGE_FLAG_JUMBO;
3420		break;
3421	case BGE_ASICREV_BCM5714_A0:
3422	case BGE_ASICREV_BCM5780:
3423	case BGE_ASICREV_BCM5714:
3424		sc->bge_flags |= BGE_FLAG_5714_FAMILY | BGE_FLAG_JUMBO_STD;
3425		/* FALLTHROUGH */
3426	case BGE_ASICREV_BCM5750:
3427	case BGE_ASICREV_BCM5752:
3428	case BGE_ASICREV_BCM5906:
3429		sc->bge_flags |= BGE_FLAG_575X_PLUS;
3430		/* FALLTHROUGH */
3431	case BGE_ASICREV_BCM5705:
3432		sc->bge_flags |= BGE_FLAG_5705_PLUS;
3433		break;
3434	}
3435
3436	/* Identify chips with APE processor. */
3437	switch (sc->bge_asicrev) {
3438	case BGE_ASICREV_BCM5717:
3439	case BGE_ASICREV_BCM5719:
3440	case BGE_ASICREV_BCM5720:
3441	case BGE_ASICREV_BCM5761:
3442	case BGE_ASICREV_BCM5762:
3443		sc->bge_flags |= BGE_FLAG_APE;
3444		break;
3445	}
3446
3447	/* Chips with APE need BAR2 access for APE registers/memory. */
3448	if ((sc->bge_flags & BGE_FLAG_APE) != 0) {
3449		rid = PCIR_BAR(2);
3450		sc->bge_res2 = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
3451		    RF_ACTIVE);
3452		if (sc->bge_res2 == NULL) {
3453			device_printf (sc->bge_dev,
3454			    "couldn't map BAR2 memory\n");
3455			error = ENXIO;
3456			goto fail;
3457		}
3458
3459		/* Enable APE register/memory access by host driver. */
3460		pcistate = pci_read_config(dev, BGE_PCI_PCISTATE, 4);
3461		pcistate |= BGE_PCISTATE_ALLOW_APE_CTLSPC_WR |
3462		    BGE_PCISTATE_ALLOW_APE_SHMEM_WR |
3463		    BGE_PCISTATE_ALLOW_APE_PSPACE_WR;
3464		pci_write_config(dev, BGE_PCI_PCISTATE, pcistate, 4);
3465
3466		bge_ape_lock_init(sc);
3467		bge_ape_read_fw_ver(sc);
3468	}
3469
3470	/* Add SYSCTLs, requires the chipset family to be set. */
3471	bge_add_sysctls(sc);
3472
3473	/* Identify the chips that use an CPMU. */
3474	if (BGE_IS_5717_PLUS(sc) ||
3475	    sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
3476	    sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
3477	    sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
3478	    sc->bge_asicrev == BGE_ASICREV_BCM57780)
3479		sc->bge_flags |= BGE_FLAG_CPMU_PRESENT;
3480	if ((sc->bge_flags & BGE_FLAG_CPMU_PRESENT) != 0)
3481		sc->bge_mi_mode = BGE_MIMODE_500KHZ_CONST;
3482	else
3483		sc->bge_mi_mode = BGE_MIMODE_BASE;
3484	/* Enable auto polling for BCM570[0-5]. */
3485	if (BGE_IS_5700_FAMILY(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5705)
3486		sc->bge_mi_mode |= BGE_MIMODE_AUTOPOLL;
3487
3488	/*
3489	 * All Broadcom controllers have 4GB boundary DMA bug.
3490	 * Whenever an address crosses a multiple of the 4GB boundary
3491	 * (including 4GB, 8Gb, 12Gb, etc.) and makes the transition
3492	 * from 0xX_FFFF_FFFF to 0x(X+1)_0000_0000 an internal DMA
3493	 * state machine will lockup and cause the device to hang.
3494	 */
3495	sc->bge_flags |= BGE_FLAG_4G_BNDRY_BUG;
3496
3497	/* BCM5755 or higher and BCM5906 have short DMA bug. */
3498	if (BGE_IS_5755_PLUS(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5906)
3499		sc->bge_flags |= BGE_FLAG_SHORT_DMA_BUG;
3500
3501	/*
3502	 * BCM5719 cannot handle DMA requests for DMA segments that
3503	 * have larger than 4KB in size.  However the maximum DMA
3504	 * segment size created in DMA tag is 4KB for TSO, so we
3505	 * wouldn't encounter the issue here.
3506	 */
3507	if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
3508		sc->bge_flags |= BGE_FLAG_4K_RDMA_BUG;
3509
3510	misccfg = CSR_READ_4(sc, BGE_MISC_CFG) & BGE_MISCCFG_BOARD_ID_MASK;
3511	if (sc->bge_asicrev == BGE_ASICREV_BCM5705) {
3512		if (misccfg == BGE_MISCCFG_BOARD_ID_5788 ||
3513		    misccfg == BGE_MISCCFG_BOARD_ID_5788M)
3514			sc->bge_flags |= BGE_FLAG_5788;
3515	}
3516
3517	capmask = BMSR_DEFCAPMASK;
3518	if ((sc->bge_asicrev == BGE_ASICREV_BCM5703 &&
3519	    (misccfg == 0x4000 || misccfg == 0x8000)) ||
3520	    (sc->bge_asicrev == BGE_ASICREV_BCM5705 &&
3521	    pci_get_vendor(dev) == BCOM_VENDORID &&
3522	    (pci_get_device(dev) == BCOM_DEVICEID_BCM5901 ||
3523	    pci_get_device(dev) == BCOM_DEVICEID_BCM5901A2 ||
3524	    pci_get_device(dev) == BCOM_DEVICEID_BCM5705F)) ||
3525	    (pci_get_vendor(dev) == BCOM_VENDORID &&
3526	    (pci_get_device(dev) == BCOM_DEVICEID_BCM5751F ||
3527	    pci_get_device(dev) == BCOM_DEVICEID_BCM5753F ||
3528	    pci_get_device(dev) == BCOM_DEVICEID_BCM5787F)) ||
3529	    pci_get_device(dev) == BCOM_DEVICEID_BCM57790 ||
3530	    pci_get_device(dev) == BCOM_DEVICEID_BCM57791 ||
3531	    pci_get_device(dev) == BCOM_DEVICEID_BCM57795 ||
3532	    sc->bge_asicrev == BGE_ASICREV_BCM5906) {
3533		/* These chips are 10/100 only. */
3534		capmask &= ~BMSR_EXTSTAT;
3535		sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
3536	}
3537
3538	/*
3539	 * Some controllers seem to require a special firmware to use
3540	 * TSO. But the firmware is not available to FreeBSD and Linux
3541	 * claims that the TSO performed by the firmware is slower than
3542	 * hardware based TSO. Moreover the firmware based TSO has one
3543	 * known bug which can't handle TSO if Ethernet header + IP/TCP
3544	 * header is greater than 80 bytes. A workaround for the TSO
3545	 * bug exist but it seems it's too expensive than not using
3546	 * TSO at all. Some hardwares also have the TSO bug so limit
3547	 * the TSO to the controllers that are not affected TSO issues
3548	 * (e.g. 5755 or higher).
3549	 */
3550	if (BGE_IS_5717_PLUS(sc)) {
3551		/* BCM5717 requires different TSO configuration. */
3552		sc->bge_flags |= BGE_FLAG_TSO3;
3553		if (sc->bge_asicrev == BGE_ASICREV_BCM5719 &&
3554		    sc->bge_chipid == BGE_CHIPID_BCM5719_A0) {
3555			/* TSO on BCM5719 A0 does not work. */
3556			sc->bge_flags &= ~BGE_FLAG_TSO3;
3557		}
3558	} else if (BGE_IS_5755_PLUS(sc)) {
3559		/*
3560		 * BCM5754 and BCM5787 shares the same ASIC id so
3561		 * explicit device id check is required.
3562		 * Due to unknown reason TSO does not work on BCM5755M.
3563		 */
3564		if (pci_get_device(dev) != BCOM_DEVICEID_BCM5754 &&
3565		    pci_get_device(dev) != BCOM_DEVICEID_BCM5754M &&
3566		    pci_get_device(dev) != BCOM_DEVICEID_BCM5755M)
3567			sc->bge_flags |= BGE_FLAG_TSO;
3568	}
3569
3570	/*
3571	 * Check if this is a PCI-X or PCI Express device.
3572	 */
3573	if (pci_find_cap(dev, PCIY_EXPRESS, &reg) == 0) {
3574		/*
3575		 * Found a PCI Express capabilities register, this
3576		 * must be a PCI Express device.
3577		 */
3578		sc->bge_flags |= BGE_FLAG_PCIE;
3579		sc->bge_expcap = reg;
3580		/* Extract supported maximum payload size. */
3581		sc->bge_mps = pci_read_config(dev, sc->bge_expcap +
3582		    PCIER_DEVICE_CAP, 2);
3583		sc->bge_mps = 128 << (sc->bge_mps & PCIEM_CAP_MAX_PAYLOAD);
3584		if (sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
3585		    sc->bge_asicrev == BGE_ASICREV_BCM5720)
3586			sc->bge_expmrq = 2048;
3587		else
3588			sc->bge_expmrq = 4096;
3589		pci_set_max_read_req(dev, sc->bge_expmrq);
3590	} else {
3591		/*
3592		 * Check if the device is in PCI-X Mode.
3593		 * (This bit is not valid on PCI Express controllers.)
3594		 */
3595		if (pci_find_cap(dev, PCIY_PCIX, &reg) == 0)
3596			sc->bge_pcixcap = reg;
3597		if ((pci_read_config(dev, BGE_PCI_PCISTATE, 4) &
3598		    BGE_PCISTATE_PCI_BUSMODE) == 0)
3599			sc->bge_flags |= BGE_FLAG_PCIX;
3600	}
3601
3602	/*
3603	 * The 40bit DMA bug applies to the 5714/5715 controllers and is
3604	 * not actually a MAC controller bug but an issue with the embedded
3605	 * PCIe to PCI-X bridge in the device. Use 40bit DMA workaround.
3606	 */
3607	if (BGE_IS_5714_FAMILY(sc) && (sc->bge_flags & BGE_FLAG_PCIX))
3608		sc->bge_flags |= BGE_FLAG_40BIT_BUG;
3609	/*
3610	 * Some PCI-X bridges are known to trigger write reordering to
3611	 * the mailbox registers. Typical phenomena is watchdog timeouts
3612	 * caused by out-of-order TX completions.  Enable workaround for
3613	 * PCI-X devices that live behind these bridges.
3614	 * Note, PCI-X controllers can run in PCI mode so we can't use
3615	 * BGE_FLAG_PCIX flag to detect PCI-X controllers.
3616	 */
3617	if (sc->bge_pcixcap != 0 && bge_mbox_reorder(sc) != 0)
3618		sc->bge_flags |= BGE_FLAG_MBOX_REORDER;
3619	/*
3620	 * Allocate the interrupt, using MSI if possible.  These devices
3621	 * support 8 MSI messages, but only the first one is used in
3622	 * normal operation.
3623	 */
3624	rid = 0;
3625	if (pci_find_cap(sc->bge_dev, PCIY_MSI, &reg) == 0) {
3626		sc->bge_msicap = reg;
3627		if (bge_can_use_msi(sc)) {
3628			msicount = pci_msi_count(dev);
3629			if (msicount > 1)
3630				msicount = 1;
3631		} else
3632			msicount = 0;
3633		if (msicount == 1 && pci_alloc_msi(dev, &msicount) == 0) {
3634			rid = 1;
3635			sc->bge_flags |= BGE_FLAG_MSI;
3636		}
3637	}
3638
3639	/*
3640	 * All controllers except BCM5700 supports tagged status but
3641	 * we use tagged status only for MSI case on BCM5717. Otherwise
3642	 * MSI on BCM5717 does not work.
3643	 */
3644#ifndef DEVICE_POLLING
3645	if (sc->bge_flags & BGE_FLAG_MSI && BGE_IS_5717_PLUS(sc))
3646		sc->bge_flags |= BGE_FLAG_TAGGED_STATUS;
3647#endif
3648
3649	sc->bge_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
3650	    RF_SHAREABLE | RF_ACTIVE);
3651
3652	if (sc->bge_irq == NULL) {
3653		device_printf(sc->bge_dev, "couldn't map interrupt\n");
3654		error = ENXIO;
3655		goto fail;
3656	}
3657
3658	bge_devinfo(sc);
3659
3660	sc->bge_asf_mode = 0;
3661	/* No ASF if APE present. */
3662	if ((sc->bge_flags & BGE_FLAG_APE) == 0) {
3663		if (bge_allow_asf && (bge_readmem_ind(sc, BGE_SRAM_DATA_SIG) ==
3664		    BGE_SRAM_DATA_SIG_MAGIC)) {
3665			if (bge_readmem_ind(sc, BGE_SRAM_DATA_CFG) &
3666			    BGE_HWCFG_ASF) {
3667				sc->bge_asf_mode |= ASF_ENABLE;
3668				sc->bge_asf_mode |= ASF_STACKUP;
3669				if (BGE_IS_575X_PLUS(sc))
3670					sc->bge_asf_mode |= ASF_NEW_HANDSHAKE;
3671			}
3672		}
3673	}
3674
3675	bge_stop_fw(sc);
3676	bge_sig_pre_reset(sc, BGE_RESET_SHUTDOWN);
3677	if (bge_reset(sc)) {
3678		device_printf(sc->bge_dev, "chip reset failed\n");
3679		error = ENXIO;
3680		goto fail;
3681	}
3682
3683	bge_sig_legacy(sc, BGE_RESET_SHUTDOWN);
3684	bge_sig_post_reset(sc, BGE_RESET_SHUTDOWN);
3685
3686	if (bge_chipinit(sc)) {
3687		device_printf(sc->bge_dev, "chip initialization failed\n");
3688		error = ENXIO;
3689		goto fail;
3690	}
3691
3692	error = bge_get_eaddr(sc, eaddr);
3693	if (error) {
3694		device_printf(sc->bge_dev,
3695		    "failed to read station address\n");
3696		error = ENXIO;
3697		goto fail;
3698	}
3699
3700	/* 5705 limits RX return ring to 512 entries. */
3701	if (BGE_IS_5717_PLUS(sc))
3702		sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT;
3703	else if (BGE_IS_5705_PLUS(sc))
3704		sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT_5705;
3705	else
3706		sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT;
3707
3708	if (bge_dma_alloc(sc)) {
3709		device_printf(sc->bge_dev,
3710		    "failed to allocate DMA resources\n");
3711		error = ENXIO;
3712		goto fail;
3713	}
3714
3715	/* Set default tuneable values. */
3716	sc->bge_stat_ticks = BGE_TICKS_PER_SEC;
3717	sc->bge_rx_coal_ticks = 150;
3718	sc->bge_tx_coal_ticks = 150;
3719	sc->bge_rx_max_coal_bds = 10;
3720	sc->bge_tx_max_coal_bds = 10;
3721
3722	/* Initialize checksum features to use. */
3723	sc->bge_csum_features = BGE_CSUM_FEATURES;
3724	if (sc->bge_forced_udpcsum != 0)
3725		sc->bge_csum_features |= CSUM_UDP;
3726
3727	/* Set up ifnet structure */
3728	ifp = sc->bge_ifp = if_alloc(IFT_ETHER);
3729	if (ifp == NULL) {
3730		device_printf(sc->bge_dev, "failed to if_alloc()\n");
3731		error = ENXIO;
3732		goto fail;
3733	}
3734	ifp->if_softc = sc;
3735	if_initname(ifp, device_get_name(dev), device_get_unit(dev));
3736	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
3737	ifp->if_ioctl = bge_ioctl;
3738	ifp->if_start = bge_start;
3739	ifp->if_init = bge_init;
3740	ifp->if_snd.ifq_drv_maxlen = BGE_TX_RING_CNT - 1;
3741	IFQ_SET_MAXLEN(&ifp->if_snd, ifp->if_snd.ifq_drv_maxlen);
3742	IFQ_SET_READY(&ifp->if_snd);
3743	ifp->if_hwassist = sc->bge_csum_features;
3744	ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING |
3745	    IFCAP_VLAN_MTU;
3746	if ((sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) != 0) {
3747		ifp->if_hwassist |= CSUM_TSO;
3748		ifp->if_capabilities |= IFCAP_TSO4 | IFCAP_VLAN_HWTSO;
3749	}
3750#ifdef IFCAP_VLAN_HWCSUM
3751	ifp->if_capabilities |= IFCAP_VLAN_HWCSUM;
3752#endif
3753	ifp->if_capenable = ifp->if_capabilities;
3754#ifdef DEVICE_POLLING
3755	ifp->if_capabilities |= IFCAP_POLLING;
3756#endif
3757
3758	/*
3759	 * 5700 B0 chips do not support checksumming correctly due
3760	 * to hardware bugs.
3761	 */
3762	if (sc->bge_chipid == BGE_CHIPID_BCM5700_B0) {
3763		ifp->if_capabilities &= ~IFCAP_HWCSUM;
3764		ifp->if_capenable &= ~IFCAP_HWCSUM;
3765		ifp->if_hwassist = 0;
3766	}
3767
3768	/*
3769	 * Figure out what sort of media we have by checking the
3770	 * hardware config word in the first 32k of NIC internal memory,
3771	 * or fall back to examining the EEPROM if necessary.
3772	 * Note: on some BCM5700 cards, this value appears to be unset.
3773	 * If that's the case, we have to rely on identifying the NIC
3774	 * by its PCI subsystem ID, as we do below for the SysKonnect
3775	 * SK-9D41.
3776	 */
3777	if (bge_readmem_ind(sc, BGE_SRAM_DATA_SIG) == BGE_SRAM_DATA_SIG_MAGIC)
3778		hwcfg = bge_readmem_ind(sc, BGE_SRAM_DATA_CFG);
3779	else if ((sc->bge_flags & BGE_FLAG_EADDR) &&
3780	    (sc->bge_asicrev != BGE_ASICREV_BCM5906)) {
3781		if (bge_read_eeprom(sc, (caddr_t)&hwcfg, BGE_EE_HWCFG_OFFSET,
3782		    sizeof(hwcfg))) {
3783			device_printf(sc->bge_dev, "failed to read EEPROM\n");
3784			error = ENXIO;
3785			goto fail;
3786		}
3787		hwcfg = ntohl(hwcfg);
3788	}
3789
3790	/* The SysKonnect SK-9D41 is a 1000baseSX card. */
3791	if ((pci_read_config(dev, BGE_PCI_SUBSYS, 4) >> 16) ==
3792	    SK_SUBSYSID_9D41 || (hwcfg & BGE_HWCFG_MEDIA) == BGE_MEDIA_FIBER) {
3793		if (BGE_IS_5705_PLUS(sc)) {
3794			sc->bge_flags |= BGE_FLAG_MII_SERDES;
3795			sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
3796		} else
3797			sc->bge_flags |= BGE_FLAG_TBI;
3798	}
3799
3800	/* Set various PHY bug flags. */
3801	if (sc->bge_chipid == BGE_CHIPID_BCM5701_A0 ||
3802	    sc->bge_chipid == BGE_CHIPID_BCM5701_B0)
3803		sc->bge_phy_flags |= BGE_PHY_CRC_BUG;
3804	if (sc->bge_chiprev == BGE_CHIPREV_5703_AX ||
3805	    sc->bge_chiprev == BGE_CHIPREV_5704_AX)
3806		sc->bge_phy_flags |= BGE_PHY_ADC_BUG;
3807	if (sc->bge_chipid == BGE_CHIPID_BCM5704_A0)
3808		sc->bge_phy_flags |= BGE_PHY_5704_A0_BUG;
3809	if (pci_get_subvendor(dev) == DELL_VENDORID)
3810		sc->bge_phy_flags |= BGE_PHY_NO_3LED;
3811	if ((BGE_IS_5705_PLUS(sc)) &&
3812	    sc->bge_asicrev != BGE_ASICREV_BCM5906 &&
3813	    sc->bge_asicrev != BGE_ASICREV_BCM5785 &&
3814	    sc->bge_asicrev != BGE_ASICREV_BCM57780 &&
3815	    !BGE_IS_5717_PLUS(sc)) {
3816		if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
3817		    sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
3818		    sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
3819		    sc->bge_asicrev == BGE_ASICREV_BCM5787) {
3820			if (pci_get_device(dev) != BCOM_DEVICEID_BCM5722 &&
3821			    pci_get_device(dev) != BCOM_DEVICEID_BCM5756)
3822				sc->bge_phy_flags |= BGE_PHY_JITTER_BUG;
3823			if (pci_get_device(dev) == BCOM_DEVICEID_BCM5755M)
3824				sc->bge_phy_flags |= BGE_PHY_ADJUST_TRIM;
3825		} else
3826			sc->bge_phy_flags |= BGE_PHY_BER_BUG;
3827	}
3828
3829	/*
3830	 * Don't enable Ethernet@WireSpeed for the 5700 or the
3831	 * 5705 A0 and A1 chips.
3832	 */
3833	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
3834	    (sc->bge_asicrev == BGE_ASICREV_BCM5705 &&
3835	    (sc->bge_chipid != BGE_CHIPID_BCM5705_A0 &&
3836	    sc->bge_chipid != BGE_CHIPID_BCM5705_A1)))
3837		sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
3838
3839	if (sc->bge_flags & BGE_FLAG_TBI) {
3840		ifmedia_init(&sc->bge_ifmedia, IFM_IMASK, bge_ifmedia_upd,
3841		    bge_ifmedia_sts);
3842		ifmedia_add(&sc->bge_ifmedia, IFM_ETHER | IFM_1000_SX, 0, NULL);
3843		ifmedia_add(&sc->bge_ifmedia, IFM_ETHER | IFM_1000_SX | IFM_FDX,
3844		    0, NULL);
3845		ifmedia_add(&sc->bge_ifmedia, IFM_ETHER | IFM_AUTO, 0, NULL);
3846		ifmedia_set(&sc->bge_ifmedia, IFM_ETHER | IFM_AUTO);
3847		sc->bge_ifmedia.ifm_media = sc->bge_ifmedia.ifm_cur->ifm_media;
3848	} else {
3849		/*
3850		 * Do transceiver setup and tell the firmware the
3851		 * driver is down so we can try to get access the
3852		 * probe if ASF is running.  Retry a couple of times
3853		 * if we get a conflict with the ASF firmware accessing
3854		 * the PHY.
3855		 */
3856		trys = 0;
3857		BGE_CLRBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
3858again:
3859		bge_asf_driver_up(sc);
3860
3861		error = mii_attach(dev, &sc->bge_miibus, ifp, bge_ifmedia_upd,
3862		    bge_ifmedia_sts, capmask, sc->bge_phy_addr, MII_OFFSET_ANY,
3863		    MIIF_DOPAUSE);
3864		if (error != 0) {
3865			if (trys++ < 4) {
3866				device_printf(sc->bge_dev, "Try again\n");
3867				bge_miibus_writereg(sc->bge_dev,
3868				    sc->bge_phy_addr, MII_BMCR, BMCR_RESET);
3869				goto again;
3870			}
3871			device_printf(sc->bge_dev, "attaching PHYs failed\n");
3872			goto fail;
3873		}
3874
3875		/*
3876		 * Now tell the firmware we are going up after probing the PHY
3877		 */
3878		if (sc->bge_asf_mode & ASF_STACKUP)
3879			BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
3880	}
3881
3882	/*
3883	 * When using the BCM5701 in PCI-X mode, data corruption has
3884	 * been observed in the first few bytes of some received packets.
3885	 * Aligning the packet buffer in memory eliminates the corruption.
3886	 * Unfortunately, this misaligns the packet payloads.  On platforms
3887	 * which do not support unaligned accesses, we will realign the
3888	 * payloads by copying the received packets.
3889	 */
3890	if (sc->bge_asicrev == BGE_ASICREV_BCM5701 &&
3891	    sc->bge_flags & BGE_FLAG_PCIX)
3892                sc->bge_flags |= BGE_FLAG_RX_ALIGNBUG;
3893
3894	/*
3895	 * Call MI attach routine.
3896	 */
3897	ether_ifattach(ifp, eaddr);
3898
3899	/* Tell upper layer we support long frames. */
3900	ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
3901
3902	/*
3903	 * Hookup IRQ last.
3904	 */
3905	if (BGE_IS_5755_PLUS(sc) && sc->bge_flags & BGE_FLAG_MSI) {
3906		/* Take advantage of single-shot MSI. */
3907		CSR_WRITE_4(sc, BGE_MSI_MODE, CSR_READ_4(sc, BGE_MSI_MODE) &
3908		    ~BGE_MSIMODE_ONE_SHOT_DISABLE);
3909		sc->bge_tq = taskqueue_create_fast("bge_taskq", M_WAITOK,
3910		    taskqueue_thread_enqueue, &sc->bge_tq);
3911		if (sc->bge_tq == NULL) {
3912			device_printf(dev, "could not create taskqueue.\n");
3913			ether_ifdetach(ifp);
3914			error = ENOMEM;
3915			goto fail;
3916		}
3917		error = taskqueue_start_threads(&sc->bge_tq, 1, PI_NET,
3918		    "%s taskq", device_get_nameunit(sc->bge_dev));
3919		if (error != 0) {
3920			device_printf(dev, "could not start threads.\n");
3921			ether_ifdetach(ifp);
3922			goto fail;
3923		}
3924		error = bus_setup_intr(dev, sc->bge_irq,
3925		    INTR_TYPE_NET | INTR_MPSAFE, bge_msi_intr, NULL, sc,
3926		    &sc->bge_intrhand);
3927	} else
3928		error = bus_setup_intr(dev, sc->bge_irq,
3929		    INTR_TYPE_NET | INTR_MPSAFE, NULL, bge_intr, sc,
3930		    &sc->bge_intrhand);
3931
3932	if (error) {
3933		ether_ifdetach(ifp);
3934		device_printf(sc->bge_dev, "couldn't set up irq\n");
3935	}
3936
3937fail:
3938	if (error)
3939		bge_detach(dev);
3940	return (error);
3941}
3942
3943static int
3944bge_detach(device_t dev)
3945{
3946	struct bge_softc *sc;
3947	struct ifnet *ifp;
3948
3949	sc = device_get_softc(dev);
3950	ifp = sc->bge_ifp;
3951
3952#ifdef DEVICE_POLLING
3953	if (ifp->if_capenable & IFCAP_POLLING)
3954		ether_poll_deregister(ifp);
3955#endif
3956
3957	if (device_is_attached(dev)) {
3958		ether_ifdetach(ifp);
3959		BGE_LOCK(sc);
3960		bge_stop(sc);
3961		BGE_UNLOCK(sc);
3962		callout_drain(&sc->bge_stat_ch);
3963	}
3964
3965	if (sc->bge_tq)
3966		taskqueue_drain(sc->bge_tq, &sc->bge_intr_task);
3967
3968	if (sc->bge_flags & BGE_FLAG_TBI)
3969		ifmedia_removeall(&sc->bge_ifmedia);
3970	else if (sc->bge_miibus != NULL) {
3971		bus_generic_detach(dev);
3972		device_delete_child(dev, sc->bge_miibus);
3973	}
3974
3975	bge_release_resources(sc);
3976
3977	return (0);
3978}
3979
3980static void
3981bge_release_resources(struct bge_softc *sc)
3982{
3983	device_t dev;
3984
3985	dev = sc->bge_dev;
3986
3987	if (sc->bge_tq != NULL)
3988		taskqueue_free(sc->bge_tq);
3989
3990	if (sc->bge_intrhand != NULL)
3991		bus_teardown_intr(dev, sc->bge_irq, sc->bge_intrhand);
3992
3993	if (sc->bge_irq != NULL)
3994		bus_release_resource(dev, SYS_RES_IRQ,
3995		    sc->bge_flags & BGE_FLAG_MSI ? 1 : 0, sc->bge_irq);
3996
3997	if (sc->bge_flags & BGE_FLAG_MSI)
3998		pci_release_msi(dev);
3999
4000	if (sc->bge_res != NULL)
4001		bus_release_resource(dev, SYS_RES_MEMORY,
4002		    PCIR_BAR(0), sc->bge_res);
4003
4004	if (sc->bge_res2 != NULL)
4005		bus_release_resource(dev, SYS_RES_MEMORY,
4006		    PCIR_BAR(2), sc->bge_res2);
4007
4008	if (sc->bge_ifp != NULL)
4009		if_free(sc->bge_ifp);
4010
4011	bge_dma_free(sc);
4012
4013	if (mtx_initialized(&sc->bge_mtx))	/* XXX */
4014		BGE_LOCK_DESTROY(sc);
4015}
4016
4017static int
4018bge_reset(struct bge_softc *sc)
4019{
4020	device_t dev;
4021	uint32_t cachesize, command, mac_mode, mac_mode_mask, reset, val;
4022	void (*write_op)(struct bge_softc *, int, int);
4023	uint16_t devctl;
4024	int i;
4025
4026	dev = sc->bge_dev;
4027
4028	mac_mode_mask = BGE_MACMODE_HALF_DUPLEX | BGE_MACMODE_PORTMODE;
4029	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
4030		mac_mode_mask |= BGE_MACMODE_APE_RX_EN | BGE_MACMODE_APE_TX_EN;
4031	mac_mode = CSR_READ_4(sc, BGE_MAC_MODE) & mac_mode_mask;
4032
4033	if (BGE_IS_575X_PLUS(sc) && !BGE_IS_5714_FAMILY(sc) &&
4034	    (sc->bge_asicrev != BGE_ASICREV_BCM5906)) {
4035		if (sc->bge_flags & BGE_FLAG_PCIE)
4036			write_op = bge_writemem_direct;
4037		else
4038			write_op = bge_writemem_ind;
4039	} else
4040		write_op = bge_writereg_ind;
4041
4042	if (sc->bge_asicrev != BGE_ASICREV_BCM5700 &&
4043	    sc->bge_asicrev != BGE_ASICREV_BCM5701) {
4044		CSR_WRITE_4(sc, BGE_NVRAM_SWARB, BGE_NVRAMSWARB_SET1);
4045		for (i = 0; i < 8000; i++) {
4046			if (CSR_READ_4(sc, BGE_NVRAM_SWARB) &
4047			    BGE_NVRAMSWARB_GNT1)
4048				break;
4049			DELAY(20);
4050		}
4051		if (i == 8000) {
4052			if (bootverbose)
4053				device_printf(dev, "NVRAM lock timedout!\n");
4054		}
4055	}
4056	/* Take APE lock when performing reset. */
4057	bge_ape_lock(sc, BGE_APE_LOCK_GRC);
4058
4059	/* Save some important PCI state. */
4060	cachesize = pci_read_config(dev, BGE_PCI_CACHESZ, 4);
4061	command = pci_read_config(dev, BGE_PCI_CMD, 4);
4062
4063	pci_write_config(dev, BGE_PCI_MISC_CTL,
4064	    BGE_PCIMISCCTL_INDIRECT_ACCESS | BGE_PCIMISCCTL_MASK_PCI_INTR |
4065	    BGE_HIF_SWAP_OPTIONS | BGE_PCIMISCCTL_PCISTATE_RW, 4);
4066
4067	/* Disable fastboot on controllers that support it. */
4068	if (sc->bge_asicrev == BGE_ASICREV_BCM5752 ||
4069	    BGE_IS_5755_PLUS(sc)) {
4070		if (bootverbose)
4071			device_printf(dev, "Disabling fastboot\n");
4072		CSR_WRITE_4(sc, BGE_FASTBOOT_PC, 0x0);
4073	}
4074
4075	/*
4076	 * Write the magic number to SRAM at offset 0xB50.
4077	 * When firmware finishes its initialization it will
4078	 * write ~BGE_SRAM_FW_MB_MAGIC to the same location.
4079	 */
4080	bge_writemem_ind(sc, BGE_SRAM_FW_MB, BGE_SRAM_FW_MB_MAGIC);
4081
4082	reset = BGE_MISCCFG_RESET_CORE_CLOCKS | BGE_32BITTIME_66MHZ;
4083
4084	/* XXX: Broadcom Linux driver. */
4085	if (sc->bge_flags & BGE_FLAG_PCIE) {
4086		if (sc->bge_asicrev != BGE_ASICREV_BCM5785 &&
4087		    (sc->bge_flags & BGE_FLAG_5717_PLUS) == 0) {
4088			if (CSR_READ_4(sc, 0x7E2C) == 0x60)	/* PCIE 1.0 */
4089				CSR_WRITE_4(sc, 0x7E2C, 0x20);
4090		}
4091		if (sc->bge_chipid != BGE_CHIPID_BCM5750_A0) {
4092			/* Prevent PCIE link training during global reset */
4093			CSR_WRITE_4(sc, BGE_MISC_CFG, 1 << 29);
4094			reset |= 1 << 29;
4095		}
4096	}
4097
4098	if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
4099		val = CSR_READ_4(sc, BGE_VCPU_STATUS);
4100		CSR_WRITE_4(sc, BGE_VCPU_STATUS,
4101		    val | BGE_VCPU_STATUS_DRV_RESET);
4102		val = CSR_READ_4(sc, BGE_VCPU_EXT_CTRL);
4103		CSR_WRITE_4(sc, BGE_VCPU_EXT_CTRL,
4104		    val & ~BGE_VCPU_EXT_CTRL_HALT_CPU);
4105	}
4106
4107	/*
4108	 * Set GPHY Power Down Override to leave GPHY
4109	 * powered up in D0 uninitialized.
4110	 */
4111	if (BGE_IS_5705_PLUS(sc) &&
4112	    (sc->bge_flags & BGE_FLAG_CPMU_PRESENT) == 0)
4113		reset |= BGE_MISCCFG_GPHY_PD_OVERRIDE;
4114
4115	/* Issue global reset */
4116	write_op(sc, BGE_MISC_CFG, reset);
4117
4118	if (sc->bge_flags & BGE_FLAG_PCIE)
4119		DELAY(100 * 1000);
4120	else
4121		DELAY(1000);
4122
4123	/* XXX: Broadcom Linux driver. */
4124	if (sc->bge_flags & BGE_FLAG_PCIE) {
4125		if (sc->bge_chipid == BGE_CHIPID_BCM5750_A0) {
4126			DELAY(500000); /* wait for link training to complete */
4127			val = pci_read_config(dev, 0xC4, 4);
4128			pci_write_config(dev, 0xC4, val | (1 << 15), 4);
4129		}
4130		devctl = pci_read_config(dev,
4131		    sc->bge_expcap + PCIER_DEVICE_CTL, 2);
4132		/* Clear enable no snoop and disable relaxed ordering. */
4133		devctl &= ~(PCIEM_CTL_RELAXED_ORD_ENABLE |
4134		    PCIEM_CTL_NOSNOOP_ENABLE);
4135		pci_write_config(dev, sc->bge_expcap + PCIER_DEVICE_CTL,
4136		    devctl, 2);
4137		pci_set_max_read_req(dev, sc->bge_expmrq);
4138		/* Clear error status. */
4139		pci_write_config(dev, sc->bge_expcap + PCIER_DEVICE_STA,
4140		    PCIEM_STA_CORRECTABLE_ERROR |
4141		    PCIEM_STA_NON_FATAL_ERROR | PCIEM_STA_FATAL_ERROR |
4142		    PCIEM_STA_UNSUPPORTED_REQ, 2);
4143	}
4144
4145	/* Reset some of the PCI state that got zapped by reset. */
4146	pci_write_config(dev, BGE_PCI_MISC_CTL,
4147	    BGE_PCIMISCCTL_INDIRECT_ACCESS | BGE_PCIMISCCTL_MASK_PCI_INTR |
4148	    BGE_HIF_SWAP_OPTIONS | BGE_PCIMISCCTL_PCISTATE_RW, 4);
4149	val = BGE_PCISTATE_ROM_ENABLE | BGE_PCISTATE_ROM_RETRY_ENABLE;
4150	if (sc->bge_chipid == BGE_CHIPID_BCM5704_A0 &&
4151	    (sc->bge_flags & BGE_FLAG_PCIX) != 0)
4152		val |= BGE_PCISTATE_RETRY_SAME_DMA;
4153	if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
4154		val |= BGE_PCISTATE_ALLOW_APE_CTLSPC_WR |
4155		    BGE_PCISTATE_ALLOW_APE_SHMEM_WR |
4156		    BGE_PCISTATE_ALLOW_APE_PSPACE_WR;
4157	pci_write_config(dev, BGE_PCI_PCISTATE, val, 4);
4158	pci_write_config(dev, BGE_PCI_CACHESZ, cachesize, 4);
4159	pci_write_config(dev, BGE_PCI_CMD, command, 4);
4160	/*
4161	 * Disable PCI-X relaxed ordering to ensure status block update
4162	 * comes first then packet buffer DMA. Otherwise driver may
4163	 * read stale status block.
4164	 */
4165	if (sc->bge_flags & BGE_FLAG_PCIX) {
4166		devctl = pci_read_config(dev,
4167		    sc->bge_pcixcap + PCIXR_COMMAND, 2);
4168		devctl &= ~PCIXM_COMMAND_ERO;
4169		if (sc->bge_asicrev == BGE_ASICREV_BCM5703) {
4170			devctl &= ~PCIXM_COMMAND_MAX_READ;
4171			devctl |= PCIXM_COMMAND_MAX_READ_2048;
4172		} else if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
4173			devctl &= ~(PCIXM_COMMAND_MAX_SPLITS |
4174			    PCIXM_COMMAND_MAX_READ);
4175			devctl |= PCIXM_COMMAND_MAX_READ_2048;
4176		}
4177		pci_write_config(dev, sc->bge_pcixcap + PCIXR_COMMAND,
4178		    devctl, 2);
4179	}
4180	/* Re-enable MSI, if necessary, and enable the memory arbiter. */
4181	if (BGE_IS_5714_FAMILY(sc)) {
4182		/* This chip disables MSI on reset. */
4183		if (sc->bge_flags & BGE_FLAG_MSI) {
4184			val = pci_read_config(dev,
4185			    sc->bge_msicap + PCIR_MSI_CTRL, 2);
4186			pci_write_config(dev,
4187			    sc->bge_msicap + PCIR_MSI_CTRL,
4188			    val | PCIM_MSICTRL_MSI_ENABLE, 2);
4189			val = CSR_READ_4(sc, BGE_MSI_MODE);
4190			CSR_WRITE_4(sc, BGE_MSI_MODE,
4191			    val | BGE_MSIMODE_ENABLE);
4192		}
4193		val = CSR_READ_4(sc, BGE_MARB_MODE);
4194		CSR_WRITE_4(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE | val);
4195	} else
4196		CSR_WRITE_4(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE);
4197
4198	/* Fix up byte swapping. */
4199	CSR_WRITE_4(sc, BGE_MODE_CTL, bge_dma_swap_options(sc));
4200
4201	val = CSR_READ_4(sc, BGE_MAC_MODE);
4202	val = (val & ~mac_mode_mask) | mac_mode;
4203	CSR_WRITE_4(sc, BGE_MAC_MODE, val);
4204	DELAY(40);
4205
4206	bge_ape_unlock(sc, BGE_APE_LOCK_GRC);
4207
4208	if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
4209		for (i = 0; i < BGE_TIMEOUT; i++) {
4210			val = CSR_READ_4(sc, BGE_VCPU_STATUS);
4211			if (val & BGE_VCPU_STATUS_INIT_DONE)
4212				break;
4213			DELAY(100);
4214		}
4215		if (i == BGE_TIMEOUT) {
4216			device_printf(dev, "reset timed out\n");
4217			return (1);
4218		}
4219	} else {
4220		/*
4221		 * Poll until we see the 1's complement of the magic number.
4222		 * This indicates that the firmware initialization is complete.
4223		 * We expect this to fail if no chip containing the Ethernet
4224		 * address is fitted though.
4225		 */
4226		for (i = 0; i < BGE_TIMEOUT; i++) {
4227			DELAY(10);
4228			val = bge_readmem_ind(sc, BGE_SRAM_FW_MB);
4229			if (val == ~BGE_SRAM_FW_MB_MAGIC)
4230				break;
4231		}
4232
4233		if ((sc->bge_flags & BGE_FLAG_EADDR) && i == BGE_TIMEOUT)
4234			device_printf(dev,
4235			    "firmware handshake timed out, found 0x%08x\n",
4236			    val);
4237		/* BCM57765 A0 needs additional time before accessing. */
4238		if (sc->bge_chipid == BGE_CHIPID_BCM57765_A0)
4239			DELAY(10 * 1000);	/* XXX */
4240	}
4241
4242	/*
4243	 * The 5704 in TBI mode apparently needs some special
4244	 * adjustment to insure the SERDES drive level is set
4245	 * to 1.2V.
4246	 */
4247	if (sc->bge_asicrev == BGE_ASICREV_BCM5704 &&
4248	    sc->bge_flags & BGE_FLAG_TBI) {
4249		val = CSR_READ_4(sc, BGE_SERDES_CFG);
4250		val = (val & ~0xFFF) | 0x880;
4251		CSR_WRITE_4(sc, BGE_SERDES_CFG, val);
4252	}
4253
4254	/* XXX: Broadcom Linux driver. */
4255	if (sc->bge_flags & BGE_FLAG_PCIE &&
4256	    !BGE_IS_5717_PLUS(sc) &&
4257	    sc->bge_chipid != BGE_CHIPID_BCM5750_A0 &&
4258	    sc->bge_asicrev != BGE_ASICREV_BCM5785) {
4259		/* Enable Data FIFO protection. */
4260		val = CSR_READ_4(sc, 0x7C00);
4261		CSR_WRITE_4(sc, 0x7C00, val | (1 << 25));
4262	}
4263
4264	if (sc->bge_asicrev == BGE_ASICREV_BCM5720)
4265		BGE_CLRBIT(sc, BGE_CPMU_CLCK_ORIDE,
4266		    CPMU_CLCK_ORIDE_MAC_ORIDE_EN);
4267
4268	return (0);
4269}
4270
4271static __inline void
4272bge_rxreuse_std(struct bge_softc *sc, int i)
4273{
4274	struct bge_rx_bd *r;
4275
4276	r = &sc->bge_ldata.bge_rx_std_ring[sc->bge_std];
4277	r->bge_flags = BGE_RXBDFLAG_END;
4278	r->bge_len = sc->bge_cdata.bge_rx_std_seglen[i];
4279	r->bge_idx = i;
4280	BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
4281}
4282
4283static __inline void
4284bge_rxreuse_jumbo(struct bge_softc *sc, int i)
4285{
4286	struct bge_extrx_bd *r;
4287
4288	r = &sc->bge_ldata.bge_rx_jumbo_ring[sc->bge_jumbo];
4289	r->bge_flags = BGE_RXBDFLAG_JUMBO_RING | BGE_RXBDFLAG_END;
4290	r->bge_len0 = sc->bge_cdata.bge_rx_jumbo_seglen[i][0];
4291	r->bge_len1 = sc->bge_cdata.bge_rx_jumbo_seglen[i][1];
4292	r->bge_len2 = sc->bge_cdata.bge_rx_jumbo_seglen[i][2];
4293	r->bge_len3 = sc->bge_cdata.bge_rx_jumbo_seglen[i][3];
4294	r->bge_idx = i;
4295	BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_RING_CNT);
4296}
4297
4298/*
4299 * Frame reception handling. This is called if there's a frame
4300 * on the receive return list.
4301 *
4302 * Note: we have to be able to handle two possibilities here:
4303 * 1) the frame is from the jumbo receive ring
4304 * 2) the frame is from the standard receive ring
4305 */
4306
4307static int
4308bge_rxeof(struct bge_softc *sc, uint16_t rx_prod, int holdlck)
4309{
4310	struct ifnet *ifp;
4311	int rx_npkts = 0, stdcnt = 0, jumbocnt = 0;
4312	uint16_t rx_cons;
4313
4314	rx_cons = sc->bge_rx_saved_considx;
4315
4316	/* Nothing to do. */
4317	if (rx_cons == rx_prod)
4318		return (rx_npkts);
4319
4320	ifp = sc->bge_ifp;
4321
4322	bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
4323	    sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_POSTREAD);
4324	bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
4325	    sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_POSTWRITE);
4326	if (BGE_IS_JUMBO_CAPABLE(sc) &&
4327	    ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
4328	    (MCLBYTES - ETHER_ALIGN))
4329		bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
4330		    sc->bge_cdata.bge_rx_jumbo_ring_map, BUS_DMASYNC_POSTWRITE);
4331
4332	while (rx_cons != rx_prod) {
4333		struct bge_rx_bd	*cur_rx;
4334		uint32_t		rxidx;
4335		struct mbuf		*m = NULL;
4336		uint16_t		vlan_tag = 0;
4337		int			have_tag = 0;
4338
4339#ifdef DEVICE_POLLING
4340		if (ifp->if_capenable & IFCAP_POLLING) {
4341			if (sc->rxcycles <= 0)
4342				break;
4343			sc->rxcycles--;
4344		}
4345#endif
4346
4347		cur_rx = &sc->bge_ldata.bge_rx_return_ring[rx_cons];
4348
4349		rxidx = cur_rx->bge_idx;
4350		BGE_INC(rx_cons, sc->bge_return_ring_cnt);
4351
4352		if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING &&
4353		    cur_rx->bge_flags & BGE_RXBDFLAG_VLAN_TAG) {
4354			have_tag = 1;
4355			vlan_tag = cur_rx->bge_vlan_tag;
4356		}
4357
4358		if (cur_rx->bge_flags & BGE_RXBDFLAG_JUMBO_RING) {
4359			jumbocnt++;
4360			m = sc->bge_cdata.bge_rx_jumbo_chain[rxidx];
4361			if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
4362				bge_rxreuse_jumbo(sc, rxidx);
4363				continue;
4364			}
4365			if (bge_newbuf_jumbo(sc, rxidx) != 0) {
4366				bge_rxreuse_jumbo(sc, rxidx);
4367				ifp->if_iqdrops++;
4368				continue;
4369			}
4370			BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_RING_CNT);
4371		} else {
4372			stdcnt++;
4373			m = sc->bge_cdata.bge_rx_std_chain[rxidx];
4374			if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
4375				bge_rxreuse_std(sc, rxidx);
4376				continue;
4377			}
4378			if (bge_newbuf_std(sc, rxidx) != 0) {
4379				bge_rxreuse_std(sc, rxidx);
4380				ifp->if_iqdrops++;
4381				continue;
4382			}
4383			BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
4384		}
4385
4386		ifp->if_ipackets++;
4387#ifndef __NO_STRICT_ALIGNMENT
4388		/*
4389		 * For architectures with strict alignment we must make sure
4390		 * the payload is aligned.
4391		 */
4392		if (sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) {
4393			bcopy(m->m_data, m->m_data + ETHER_ALIGN,
4394			    cur_rx->bge_len);
4395			m->m_data += ETHER_ALIGN;
4396		}
4397#endif
4398		m->m_pkthdr.len = m->m_len = cur_rx->bge_len - ETHER_CRC_LEN;
4399		m->m_pkthdr.rcvif = ifp;
4400
4401		if (ifp->if_capenable & IFCAP_RXCSUM)
4402			bge_rxcsum(sc, cur_rx, m);
4403
4404		/*
4405		 * If we received a packet with a vlan tag,
4406		 * attach that information to the packet.
4407		 */
4408		if (have_tag) {
4409			m->m_pkthdr.ether_vtag = vlan_tag;
4410			m->m_flags |= M_VLANTAG;
4411		}
4412
4413		if (holdlck != 0) {
4414			BGE_UNLOCK(sc);
4415			(*ifp->if_input)(ifp, m);
4416			BGE_LOCK(sc);
4417		} else
4418			(*ifp->if_input)(ifp, m);
4419		rx_npkts++;
4420
4421		if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
4422			return (rx_npkts);
4423	}
4424
4425	bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
4426	    sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_PREREAD);
4427	if (stdcnt > 0)
4428		bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
4429		    sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
4430
4431	if (jumbocnt > 0)
4432		bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
4433		    sc->bge_cdata.bge_rx_jumbo_ring_map, BUS_DMASYNC_PREWRITE);
4434
4435	sc->bge_rx_saved_considx = rx_cons;
4436	bge_writembx(sc, BGE_MBX_RX_CONS0_LO, sc->bge_rx_saved_considx);
4437	if (stdcnt)
4438		bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, (sc->bge_std +
4439		    BGE_STD_RX_RING_CNT - 1) % BGE_STD_RX_RING_CNT);
4440	if (jumbocnt)
4441		bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, (sc->bge_jumbo +
4442		    BGE_JUMBO_RX_RING_CNT - 1) % BGE_JUMBO_RX_RING_CNT);
4443#ifdef notyet
4444	/*
4445	 * This register wraps very quickly under heavy packet drops.
4446	 * If you need correct statistics, you can enable this check.
4447	 */
4448	if (BGE_IS_5705_PLUS(sc))
4449		ifp->if_ierrors += CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
4450#endif
4451	return (rx_npkts);
4452}
4453
4454static void
4455bge_rxcsum(struct bge_softc *sc, struct bge_rx_bd *cur_rx, struct mbuf *m)
4456{
4457
4458	if (BGE_IS_5717_PLUS(sc)) {
4459		if ((cur_rx->bge_flags & BGE_RXBDFLAG_IPV6) == 0) {
4460			if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
4461				m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
4462				if ((cur_rx->bge_error_flag &
4463				    BGE_RXERRFLAG_IP_CSUM_NOK) == 0)
4464					m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
4465			}
4466			if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM) {
4467				m->m_pkthdr.csum_data =
4468				    cur_rx->bge_tcp_udp_csum;
4469				m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
4470				    CSUM_PSEUDO_HDR;
4471			}
4472		}
4473	} else {
4474		if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
4475			m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
4476			if ((cur_rx->bge_ip_csum ^ 0xFFFF) == 0)
4477				m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
4478		}
4479		if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM &&
4480		    m->m_pkthdr.len >= ETHER_MIN_NOPAD) {
4481			m->m_pkthdr.csum_data =
4482			    cur_rx->bge_tcp_udp_csum;
4483			m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
4484			    CSUM_PSEUDO_HDR;
4485		}
4486	}
4487}
4488
4489static void
4490bge_txeof(struct bge_softc *sc, uint16_t tx_cons)
4491{
4492	struct bge_tx_bd *cur_tx;
4493	struct ifnet *ifp;
4494
4495	BGE_LOCK_ASSERT(sc);
4496
4497	/* Nothing to do. */
4498	if (sc->bge_tx_saved_considx == tx_cons)
4499		return;
4500
4501	ifp = sc->bge_ifp;
4502
4503	bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
4504	    sc->bge_cdata.bge_tx_ring_map, BUS_DMASYNC_POSTWRITE);
4505	/*
4506	 * Go through our tx ring and free mbufs for those
4507	 * frames that have been sent.
4508	 */
4509	while (sc->bge_tx_saved_considx != tx_cons) {
4510		uint32_t		idx;
4511
4512		idx = sc->bge_tx_saved_considx;
4513		cur_tx = &sc->bge_ldata.bge_tx_ring[idx];
4514		if (cur_tx->bge_flags & BGE_TXBDFLAG_END)
4515			ifp->if_opackets++;
4516		if (sc->bge_cdata.bge_tx_chain[idx] != NULL) {
4517			bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag,
4518			    sc->bge_cdata.bge_tx_dmamap[idx],
4519			    BUS_DMASYNC_POSTWRITE);
4520			bus_dmamap_unload(sc->bge_cdata.bge_tx_mtag,
4521			    sc->bge_cdata.bge_tx_dmamap[idx]);
4522			m_freem(sc->bge_cdata.bge_tx_chain[idx]);
4523			sc->bge_cdata.bge_tx_chain[idx] = NULL;
4524		}
4525		sc->bge_txcnt--;
4526		BGE_INC(sc->bge_tx_saved_considx, BGE_TX_RING_CNT);
4527	}
4528
4529	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
4530	if (sc->bge_txcnt == 0)
4531		sc->bge_timer = 0;
4532}
4533
4534#ifdef DEVICE_POLLING
4535static int
4536bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
4537{
4538	struct bge_softc *sc = ifp->if_softc;
4539	uint16_t rx_prod, tx_cons;
4540	uint32_t statusword;
4541	int rx_npkts = 0;
4542
4543	BGE_LOCK(sc);
4544	if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
4545		BGE_UNLOCK(sc);
4546		return (rx_npkts);
4547	}
4548
4549	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4550	    sc->bge_cdata.bge_status_map,
4551	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
4552	/* Fetch updates from the status block. */
4553	rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
4554	tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
4555
4556	statusword = sc->bge_ldata.bge_status_block->bge_status;
4557	/* Clear the status so the next pass only sees the changes. */
4558	sc->bge_ldata.bge_status_block->bge_status = 0;
4559
4560	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4561	    sc->bge_cdata.bge_status_map,
4562	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
4563
4564	/* Note link event. It will be processed by POLL_AND_CHECK_STATUS. */
4565	if (statusword & BGE_STATFLAG_LINKSTATE_CHANGED)
4566		sc->bge_link_evt++;
4567
4568	if (cmd == POLL_AND_CHECK_STATUS)
4569		if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
4570		    sc->bge_chipid != BGE_CHIPID_BCM5700_B2) ||
4571		    sc->bge_link_evt || (sc->bge_flags & BGE_FLAG_TBI))
4572			bge_link_upd(sc);
4573
4574	sc->rxcycles = count;
4575	rx_npkts = bge_rxeof(sc, rx_prod, 1);
4576	if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
4577		BGE_UNLOCK(sc);
4578		return (rx_npkts);
4579	}
4580	bge_txeof(sc, tx_cons);
4581	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
4582		bge_start_locked(ifp);
4583
4584	BGE_UNLOCK(sc);
4585	return (rx_npkts);
4586}
4587#endif /* DEVICE_POLLING */
4588
4589static int
4590bge_msi_intr(void *arg)
4591{
4592	struct bge_softc *sc;
4593
4594	sc = (struct bge_softc *)arg;
4595	/*
4596	 * This interrupt is not shared and controller already
4597	 * disabled further interrupt.
4598	 */
4599	taskqueue_enqueue(sc->bge_tq, &sc->bge_intr_task);
4600	return (FILTER_HANDLED);
4601}
4602
4603static void
4604bge_intr_task(void *arg, int pending)
4605{
4606	struct bge_softc *sc;
4607	struct ifnet *ifp;
4608	uint32_t status, status_tag;
4609	uint16_t rx_prod, tx_cons;
4610
4611	sc = (struct bge_softc *)arg;
4612	ifp = sc->bge_ifp;
4613
4614	BGE_LOCK(sc);
4615	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
4616		BGE_UNLOCK(sc);
4617		return;
4618	}
4619
4620	/* Get updated status block. */
4621	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4622	    sc->bge_cdata.bge_status_map,
4623	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
4624
4625	/* Save producer/consumer indices. */
4626	rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
4627	tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
4628	status = sc->bge_ldata.bge_status_block->bge_status;
4629	status_tag = sc->bge_ldata.bge_status_block->bge_status_tag << 24;
4630	/* Dirty the status flag. */
4631	sc->bge_ldata.bge_status_block->bge_status = 0;
4632	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4633	    sc->bge_cdata.bge_status_map,
4634	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
4635	if ((sc->bge_flags & BGE_FLAG_TAGGED_STATUS) == 0)
4636		status_tag = 0;
4637
4638	if ((status & BGE_STATFLAG_LINKSTATE_CHANGED) != 0)
4639		bge_link_upd(sc);
4640
4641	/* Let controller work. */
4642	bge_writembx(sc, BGE_MBX_IRQ0_LO, status_tag);
4643
4644	if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
4645	    sc->bge_rx_saved_considx != rx_prod) {
4646		/* Check RX return ring producer/consumer. */
4647		BGE_UNLOCK(sc);
4648		bge_rxeof(sc, rx_prod, 0);
4649		BGE_LOCK(sc);
4650	}
4651	if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
4652		/* Check TX ring producer/consumer. */
4653		bge_txeof(sc, tx_cons);
4654		if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
4655			bge_start_locked(ifp);
4656	}
4657	BGE_UNLOCK(sc);
4658}
4659
4660static void
4661bge_intr(void *xsc)
4662{
4663	struct bge_softc *sc;
4664	struct ifnet *ifp;
4665	uint32_t statusword;
4666	uint16_t rx_prod, tx_cons;
4667
4668	sc = xsc;
4669
4670	BGE_LOCK(sc);
4671
4672	ifp = sc->bge_ifp;
4673
4674#ifdef DEVICE_POLLING
4675	if (ifp->if_capenable & IFCAP_POLLING) {
4676		BGE_UNLOCK(sc);
4677		return;
4678	}
4679#endif
4680
4681	/*
4682	 * Ack the interrupt by writing something to BGE_MBX_IRQ0_LO.  Don't
4683	 * disable interrupts by writing nonzero like we used to, since with
4684	 * our current organization this just gives complications and
4685	 * pessimizations for re-enabling interrupts.  We used to have races
4686	 * instead of the necessary complications.  Disabling interrupts
4687	 * would just reduce the chance of a status update while we are
4688	 * running (by switching to the interrupt-mode coalescence
4689	 * parameters), but this chance is already very low so it is more
4690	 * efficient to get another interrupt than prevent it.
4691	 *
4692	 * We do the ack first to ensure another interrupt if there is a
4693	 * status update after the ack.  We don't check for the status
4694	 * changing later because it is more efficient to get another
4695	 * interrupt than prevent it, not quite as above (not checking is
4696	 * a smaller optimization than not toggling the interrupt enable,
4697	 * since checking doesn't involve PCI accesses and toggling require
4698	 * the status check).  So toggling would probably be a pessimization
4699	 * even with MSI.  It would only be needed for using a task queue.
4700	 */
4701	bge_writembx(sc, BGE_MBX_IRQ0_LO, 0);
4702
4703	/*
4704	 * Do the mandatory PCI flush as well as get the link status.
4705	 */
4706	statusword = CSR_READ_4(sc, BGE_MAC_STS) & BGE_MACSTAT_LINK_CHANGED;
4707
4708	/* Make sure the descriptor ring indexes are coherent. */
4709	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4710	    sc->bge_cdata.bge_status_map,
4711	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
4712	rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
4713	tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
4714	sc->bge_ldata.bge_status_block->bge_status = 0;
4715	bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
4716	    sc->bge_cdata.bge_status_map,
4717	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
4718
4719	if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
4720	    sc->bge_chipid != BGE_CHIPID_BCM5700_B2) ||
4721	    statusword || sc->bge_link_evt)
4722		bge_link_upd(sc);
4723
4724	if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
4725		/* Check RX return ring producer/consumer. */
4726		bge_rxeof(sc, rx_prod, 1);
4727	}
4728
4729	if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
4730		/* Check TX ring producer/consumer. */
4731		bge_txeof(sc, tx_cons);
4732	}
4733
4734	if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
4735	    !IFQ_DRV_IS_EMPTY(&ifp->if_snd))
4736		bge_start_locked(ifp);
4737
4738	BGE_UNLOCK(sc);
4739}
4740
4741static void
4742bge_asf_driver_up(struct bge_softc *sc)
4743{
4744	if (sc->bge_asf_mode & ASF_STACKUP) {
4745		/* Send ASF heartbeat aprox. every 2s */
4746		if (sc->bge_asf_count)
4747			sc->bge_asf_count --;
4748		else {
4749			sc->bge_asf_count = 2;
4750			bge_writemem_ind(sc, BGE_SRAM_FW_CMD_MB,
4751			    BGE_FW_CMD_DRV_ALIVE);
4752			bge_writemem_ind(sc, BGE_SRAM_FW_CMD_LEN_MB, 4);
4753			bge_writemem_ind(sc, BGE_SRAM_FW_CMD_DATA_MB,
4754			    BGE_FW_HB_TIMEOUT_SEC);
4755			CSR_WRITE_4(sc, BGE_RX_CPU_EVENT,
4756			    CSR_READ_4(sc, BGE_RX_CPU_EVENT) |
4757			    BGE_RX_CPU_DRV_EVENT);
4758		}
4759	}
4760}
4761
4762static void
4763bge_tick(void *xsc)
4764{
4765	struct bge_softc *sc = xsc;
4766	struct mii_data *mii = NULL;
4767
4768	BGE_LOCK_ASSERT(sc);
4769
4770	/* Synchronize with possible callout reset/stop. */
4771	if (callout_pending(&sc->bge_stat_ch) ||
4772	    !callout_active(&sc->bge_stat_ch))
4773		return;
4774
4775	if (BGE_IS_5705_PLUS(sc))
4776		bge_stats_update_regs(sc);
4777	else
4778		bge_stats_update(sc);
4779
4780	/* XXX Add APE heartbeat check here? */
4781
4782	if ((sc->bge_flags & BGE_FLAG_TBI) == 0) {
4783		mii = device_get_softc(sc->bge_miibus);
4784		/*
4785		 * Do not touch PHY if we have link up. This could break
4786		 * IPMI/ASF mode or produce extra input errors
4787		 * (extra errors was reported for bcm5701 & bcm5704).
4788		 */
4789		if (!sc->bge_link)
4790			mii_tick(mii);
4791	} else {
4792		/*
4793		 * Since in TBI mode auto-polling can't be used we should poll
4794		 * link status manually. Here we register pending link event
4795		 * and trigger interrupt.
4796		 */
4797#ifdef DEVICE_POLLING
4798		/* In polling mode we poll link state in bge_poll(). */
4799		if (!(sc->bge_ifp->if_capenable & IFCAP_POLLING))
4800#endif
4801		{
4802		sc->bge_link_evt++;
4803		if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
4804		    sc->bge_flags & BGE_FLAG_5788)
4805			BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
4806		else
4807			BGE_SETBIT(sc, BGE_HCC_MODE, BGE_HCCMODE_COAL_NOW);
4808		}
4809	}
4810
4811	bge_asf_driver_up(sc);
4812	bge_watchdog(sc);
4813
4814	callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
4815}
4816
4817static void
4818bge_stats_update_regs(struct bge_softc *sc)
4819{
4820	struct ifnet *ifp;
4821	struct bge_mac_stats *stats;
4822	uint32_t val;
4823
4824	ifp = sc->bge_ifp;
4825	stats = &sc->bge_mac_stats;
4826
4827	stats->ifHCOutOctets +=
4828	    CSR_READ_4(sc, BGE_TX_MAC_STATS_OCTETS);
4829	stats->etherStatsCollisions +=
4830	    CSR_READ_4(sc, BGE_TX_MAC_STATS_COLLS);
4831	stats->outXonSent +=
4832	    CSR_READ_4(sc, BGE_TX_MAC_STATS_XON_SENT);
4833	stats->outXoffSent +=
4834	    CSR_READ_4(sc, BGE_TX_MAC_STATS_XOFF_SENT);
4835	stats->dot3StatsInternalMacTransmitErrors +=
4836	    CSR_READ_4(sc, BGE_TX_MAC_STATS_ERRORS);
4837	stats->dot3StatsSingleCollisionFrames +=
4838	    CSR_READ_4(sc, BGE_TX_MAC_STATS_SINGLE_COLL);
4839	stats->dot3StatsMultipleCollisionFrames +=
4840	    CSR_READ_4(sc, BGE_TX_MAC_STATS_MULTI_COLL);
4841	stats->dot3StatsDeferredTransmissions +=
4842	    CSR_READ_4(sc, BGE_TX_MAC_STATS_DEFERRED);
4843	stats->dot3StatsExcessiveCollisions +=
4844	    CSR_READ_4(sc, BGE_TX_MAC_STATS_EXCESS_COLL);
4845	stats->dot3StatsLateCollisions +=
4846	    CSR_READ_4(sc, BGE_TX_MAC_STATS_LATE_COLL);
4847	stats->ifHCOutUcastPkts +=
4848	    CSR_READ_4(sc, BGE_TX_MAC_STATS_UCAST);
4849	stats->ifHCOutMulticastPkts +=
4850	    CSR_READ_4(sc, BGE_TX_MAC_STATS_MCAST);
4851	stats->ifHCOutBroadcastPkts +=
4852	    CSR_READ_4(sc, BGE_TX_MAC_STATS_BCAST);
4853
4854	stats->ifHCInOctets +=
4855	    CSR_READ_4(sc, BGE_RX_MAC_STATS_OCTESTS);
4856	stats->etherStatsFragments +=
4857	    CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAGMENTS);
4858	stats->ifHCInUcastPkts +=
4859	    CSR_READ_4(sc, BGE_RX_MAC_STATS_UCAST);
4860	stats->ifHCInMulticastPkts +=
4861	    CSR_READ_4(sc, BGE_RX_MAC_STATS_MCAST);
4862	stats->ifHCInBroadcastPkts +=
4863	    CSR_READ_4(sc, BGE_RX_MAC_STATS_BCAST);
4864	stats->dot3StatsFCSErrors +=
4865	    CSR_READ_4(sc, BGE_RX_MAC_STATS_FCS_ERRORS);
4866	stats->dot3StatsAlignmentErrors +=
4867	    CSR_READ_4(sc, BGE_RX_MAC_STATS_ALGIN_ERRORS);
4868	stats->xonPauseFramesReceived +=
4869	    CSR_READ_4(sc, BGE_RX_MAC_STATS_XON_RCVD);
4870	stats->xoffPauseFramesReceived +=
4871	    CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_RCVD);
4872	stats->macControlFramesReceived +=
4873	    CSR_READ_4(sc, BGE_RX_MAC_STATS_CTRL_RCVD);
4874	stats->xoffStateEntered +=
4875	    CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_ENTERED);
4876	stats->dot3StatsFramesTooLong +=
4877	    CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAME_TOO_LONG);
4878	stats->etherStatsJabbers +=
4879	    CSR_READ_4(sc, BGE_RX_MAC_STATS_JABBERS);
4880	stats->etherStatsUndersizePkts +=
4881	    CSR_READ_4(sc, BGE_RX_MAC_STATS_UNDERSIZE);
4882
4883	stats->FramesDroppedDueToFilters +=
4884	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_FILTDROP);
4885	stats->DmaWriteQueueFull +=
4886	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_WRQ_FULL);
4887	stats->DmaWriteHighPriQueueFull +=
4888	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_HPWRQ_FULL);
4889	stats->NoMoreRxBDs +=
4890	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_OUT_OF_BDS);
4891	/*
4892	 * XXX
4893	 * Unlike other controllers, BGE_RXLP_LOCSTAT_IFIN_DROPS
4894	 * counter of BCM5717, BCM5718, BCM5719 A0 and BCM5720 A0
4895	 * includes number of unwanted multicast frames.  This comes
4896	 * from silicon bug and known workaround to get rough(not
4897	 * exact) counter is to enable interrupt on MBUF low water
4898	 * attention.  This can be accomplished by setting
4899	 * BGE_HCCMODE_ATTN bit of BGE_HCC_MODE,
4900	 * BGE_BMANMODE_LOMBUF_ATTN bit of BGE_BMAN_MODE and
4901	 * BGE_MODECTL_FLOWCTL_ATTN_INTR bit of BGE_MODE_CTL.
4902	 * However that change would generate more interrupts and
4903	 * there are still possibilities of losing multiple frames
4904	 * during BGE_MODECTL_FLOWCTL_ATTN_INTR interrupt handling.
4905	 * Given that the workaround still would not get correct
4906	 * counter I don't think it's worth to implement it.  So
4907	 * ignore reading the counter on controllers that have the
4908	 * silicon bug.
4909	 */
4910	if (sc->bge_asicrev != BGE_ASICREV_BCM5717 &&
4911	    sc->bge_chipid != BGE_CHIPID_BCM5719_A0 &&
4912	    sc->bge_chipid != BGE_CHIPID_BCM5720_A0)
4913		stats->InputDiscards +=
4914		    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
4915	stats->InputErrors +=
4916	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_ERRORS);
4917	stats->RecvThresholdHit +=
4918	    CSR_READ_4(sc, BGE_RXLP_LOCSTAT_RXTHRESH_HIT);
4919
4920	ifp->if_collisions = (u_long)stats->etherStatsCollisions;
4921	ifp->if_ierrors = (u_long)(stats->NoMoreRxBDs + stats->InputDiscards +
4922	    stats->InputErrors);
4923
4924	if (sc->bge_flags & BGE_FLAG_RDMA_BUG) {
4925		/*
4926		 * If controller transmitted more than BGE_NUM_RDMA_CHANNELS
4927		 * frames, it's safe to disable workaround for DMA engine's
4928		 * miscalculation of TXMBUF space.
4929		 */
4930		if (stats->ifHCOutUcastPkts + stats->ifHCOutMulticastPkts +
4931		    stats->ifHCOutBroadcastPkts > BGE_NUM_RDMA_CHANNELS) {
4932			val = CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL);
4933			if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
4934				val &= ~BGE_RDMA_TX_LENGTH_WA_5719;
4935			else
4936				val &= ~BGE_RDMA_TX_LENGTH_WA_5720;
4937			CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL, val);
4938			sc->bge_flags &= ~BGE_FLAG_RDMA_BUG;
4939		}
4940	}
4941}
4942
4943static void
4944bge_stats_clear_regs(struct bge_softc *sc)
4945{
4946
4947	CSR_READ_4(sc, BGE_TX_MAC_STATS_OCTETS);
4948	CSR_READ_4(sc, BGE_TX_MAC_STATS_COLLS);
4949	CSR_READ_4(sc, BGE_TX_MAC_STATS_XON_SENT);
4950	CSR_READ_4(sc, BGE_TX_MAC_STATS_XOFF_SENT);
4951	CSR_READ_4(sc, BGE_TX_MAC_STATS_ERRORS);
4952	CSR_READ_4(sc, BGE_TX_MAC_STATS_SINGLE_COLL);
4953	CSR_READ_4(sc, BGE_TX_MAC_STATS_MULTI_COLL);
4954	CSR_READ_4(sc, BGE_TX_MAC_STATS_DEFERRED);
4955	CSR_READ_4(sc, BGE_TX_MAC_STATS_EXCESS_COLL);
4956	CSR_READ_4(sc, BGE_TX_MAC_STATS_LATE_COLL);
4957	CSR_READ_4(sc, BGE_TX_MAC_STATS_UCAST);
4958	CSR_READ_4(sc, BGE_TX_MAC_STATS_MCAST);
4959	CSR_READ_4(sc, BGE_TX_MAC_STATS_BCAST);
4960
4961	CSR_READ_4(sc, BGE_RX_MAC_STATS_OCTESTS);
4962	CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAGMENTS);
4963	CSR_READ_4(sc, BGE_RX_MAC_STATS_UCAST);
4964	CSR_READ_4(sc, BGE_RX_MAC_STATS_MCAST);
4965	CSR_READ_4(sc, BGE_RX_MAC_STATS_BCAST);
4966	CSR_READ_4(sc, BGE_RX_MAC_STATS_FCS_ERRORS);
4967	CSR_READ_4(sc, BGE_RX_MAC_STATS_ALGIN_ERRORS);
4968	CSR_READ_4(sc, BGE_RX_MAC_STATS_XON_RCVD);
4969	CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_RCVD);
4970	CSR_READ_4(sc, BGE_RX_MAC_STATS_CTRL_RCVD);
4971	CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_ENTERED);
4972	CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAME_TOO_LONG);
4973	CSR_READ_4(sc, BGE_RX_MAC_STATS_JABBERS);
4974	CSR_READ_4(sc, BGE_RX_MAC_STATS_UNDERSIZE);
4975
4976	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_FILTDROP);
4977	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_WRQ_FULL);
4978	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_HPWRQ_FULL);
4979	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_OUT_OF_BDS);
4980	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
4981	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_ERRORS);
4982	CSR_READ_4(sc, BGE_RXLP_LOCSTAT_RXTHRESH_HIT);
4983}
4984
4985static void
4986bge_stats_update(struct bge_softc *sc)
4987{
4988	struct ifnet *ifp;
4989	bus_size_t stats;
4990	uint32_t cnt;	/* current register value */
4991
4992	ifp = sc->bge_ifp;
4993
4994	stats = BGE_MEMWIN_START + BGE_STATS_BLOCK;
4995
4996#define	READ_STAT(sc, stats, stat) \
4997	CSR_READ_4(sc, stats + offsetof(struct bge_stats, stat))
4998
4999	cnt = READ_STAT(sc, stats, txstats.etherStatsCollisions.bge_addr_lo);
5000	ifp->if_collisions += (uint32_t)(cnt - sc->bge_tx_collisions);
5001	sc->bge_tx_collisions = cnt;
5002
5003	cnt = READ_STAT(sc, stats, nicNoMoreRxBDs.bge_addr_lo);
5004	ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_nobds);
5005	sc->bge_rx_nobds = cnt;
5006	cnt = READ_STAT(sc, stats, ifInErrors.bge_addr_lo);
5007	ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_inerrs);
5008	sc->bge_rx_inerrs = cnt;
5009	cnt = READ_STAT(sc, stats, ifInDiscards.bge_addr_lo);
5010	ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_discards);
5011	sc->bge_rx_discards = cnt;
5012
5013	cnt = READ_STAT(sc, stats, txstats.ifOutDiscards.bge_addr_lo);
5014	ifp->if_oerrors += (uint32_t)(cnt - sc->bge_tx_discards);
5015	sc->bge_tx_discards = cnt;
5016
5017#undef	READ_STAT
5018}
5019
5020/*
5021 * Pad outbound frame to ETHER_MIN_NOPAD for an unusual reason.
5022 * The bge hardware will pad out Tx runts to ETHER_MIN_NOPAD,
5023 * but when such padded frames employ the bge IP/TCP checksum offload,
5024 * the hardware checksum assist gives incorrect results (possibly
5025 * from incorporating its own padding into the UDP/TCP checksum; who knows).
5026 * If we pad such runts with zeros, the onboard checksum comes out correct.
5027 */
5028static __inline int
5029bge_cksum_pad(struct mbuf *m)
5030{
5031	int padlen = ETHER_MIN_NOPAD - m->m_pkthdr.len;
5032	struct mbuf *last;
5033
5034	/* If there's only the packet-header and we can pad there, use it. */
5035	if (m->m_pkthdr.len == m->m_len && M_WRITABLE(m) &&
5036	    M_TRAILINGSPACE(m) >= padlen) {
5037		last = m;
5038	} else {
5039		/*
5040		 * Walk packet chain to find last mbuf. We will either
5041		 * pad there, or append a new mbuf and pad it.
5042		 */
5043		for (last = m; last->m_next != NULL; last = last->m_next);
5044		if (!(M_WRITABLE(last) && M_TRAILINGSPACE(last) >= padlen)) {
5045			/* Allocate new empty mbuf, pad it. Compact later. */
5046			struct mbuf *n;
5047
5048			MGET(n, M_NOWAIT, MT_DATA);
5049			if (n == NULL)
5050				return (ENOBUFS);
5051			n->m_len = 0;
5052			last->m_next = n;
5053			last = n;
5054		}
5055	}
5056
5057	/* Now zero the pad area, to avoid the bge cksum-assist bug. */
5058	memset(mtod(last, caddr_t) + last->m_len, 0, padlen);
5059	last->m_len += padlen;
5060	m->m_pkthdr.len += padlen;
5061
5062	return (0);
5063}
5064
5065static struct mbuf *
5066bge_check_short_dma(struct mbuf *m)
5067{
5068	struct mbuf *n;
5069	int found;
5070
5071	/*
5072	 * If device receive two back-to-back send BDs with less than
5073	 * or equal to 8 total bytes then the device may hang.  The two
5074	 * back-to-back send BDs must in the same frame for this failure
5075	 * to occur.  Scan mbuf chains and see whether two back-to-back
5076	 * send BDs are there. If this is the case, allocate new mbuf
5077	 * and copy the frame to workaround the silicon bug.
5078	 */
5079	for (n = m, found = 0; n != NULL; n = n->m_next) {
5080		if (n->m_len < 8) {
5081			found++;
5082			if (found > 1)
5083				break;
5084			continue;
5085		}
5086		found = 0;
5087	}
5088
5089	if (found > 1) {
5090		n = m_defrag(m, M_NOWAIT);
5091		if (n == NULL)
5092			m_freem(m);
5093	} else
5094		n = m;
5095	return (n);
5096}
5097
5098static struct mbuf *
5099bge_setup_tso(struct bge_softc *sc, struct mbuf *m, uint16_t *mss,
5100    uint16_t *flags)
5101{
5102	struct ip *ip;
5103	struct tcphdr *tcp;
5104	struct mbuf *n;
5105	uint16_t hlen;
5106	uint32_t poff;
5107
5108	if (M_WRITABLE(m) == 0) {
5109		/* Get a writable copy. */
5110		n = m_dup(m, M_NOWAIT);
5111		m_freem(m);
5112		if (n == NULL)
5113			return (NULL);
5114		m = n;
5115	}
5116	m = m_pullup(m, sizeof(struct ether_header) + sizeof(struct ip));
5117	if (m == NULL)
5118		return (NULL);
5119	ip = (struct ip *)(mtod(m, char *) + sizeof(struct ether_header));
5120	poff = sizeof(struct ether_header) + (ip->ip_hl << 2);
5121	m = m_pullup(m, poff + sizeof(struct tcphdr));
5122	if (m == NULL)
5123		return (NULL);
5124	tcp = (struct tcphdr *)(mtod(m, char *) + poff);
5125	m = m_pullup(m, poff + (tcp->th_off << 2));
5126	if (m == NULL)
5127		return (NULL);
5128	/*
5129	 * It seems controller doesn't modify IP length and TCP pseudo
5130	 * checksum. These checksum computed by upper stack should be 0.
5131	 */
5132	*mss = m->m_pkthdr.tso_segsz;
5133	ip = (struct ip *)(mtod(m, char *) + sizeof(struct ether_header));
5134	ip->ip_sum = 0;
5135	ip->ip_len = htons(*mss + (ip->ip_hl << 2) + (tcp->th_off << 2));
5136	/* Clear pseudo checksum computed by TCP stack. */
5137	tcp = (struct tcphdr *)(mtod(m, char *) + poff);
5138	tcp->th_sum = 0;
5139	/*
5140	 * Broadcom controllers uses different descriptor format for
5141	 * TSO depending on ASIC revision. Due to TSO-capable firmware
5142	 * license issue and lower performance of firmware based TSO
5143	 * we only support hardware based TSO.
5144	 */
5145	/* Calculate header length, incl. TCP/IP options, in 32 bit units. */
5146	hlen = ((ip->ip_hl << 2) + (tcp->th_off << 2)) >> 2;
5147	if (sc->bge_flags & BGE_FLAG_TSO3) {
5148		/*
5149		 * For BCM5717 and newer controllers, hardware based TSO
5150		 * uses the 14 lower bits of the bge_mss field to store the
5151		 * MSS and the upper 2 bits to store the lowest 2 bits of
5152		 * the IP/TCP header length.  The upper 6 bits of the header
5153		 * length are stored in the bge_flags[14:10,4] field.  Jumbo
5154		 * frames are supported.
5155		 */
5156		*mss |= ((hlen & 0x3) << 14);
5157		*flags |= ((hlen & 0xF8) << 7) | ((hlen & 0x4) << 2);
5158	} else {
5159		/*
5160		 * For BCM5755 and newer controllers, hardware based TSO uses
5161		 * the lower 11	bits to store the MSS and the upper 5 bits to
5162		 * store the IP/TCP header length. Jumbo frames are not
5163		 * supported.
5164		 */
5165		*mss |= (hlen << 11);
5166	}
5167	return (m);
5168}
5169
5170/*
5171 * Encapsulate an mbuf chain in the tx ring  by coupling the mbuf data
5172 * pointers to descriptors.
5173 */
5174static int
5175bge_encap(struct bge_softc *sc, struct mbuf **m_head, uint32_t *txidx)
5176{
5177	bus_dma_segment_t	segs[BGE_NSEG_NEW];
5178	bus_dmamap_t		map;
5179	struct bge_tx_bd	*d;
5180	struct mbuf		*m = *m_head;
5181	uint32_t		idx = *txidx;
5182	uint16_t		csum_flags, mss, vlan_tag;
5183	int			nsegs, i, error;
5184
5185	csum_flags = 0;
5186	mss = 0;
5187	vlan_tag = 0;
5188	if ((sc->bge_flags & BGE_FLAG_SHORT_DMA_BUG) != 0 &&
5189	    m->m_next != NULL) {
5190		*m_head = bge_check_short_dma(m);
5191		if (*m_head == NULL)
5192			return (ENOBUFS);
5193		m = *m_head;
5194	}
5195	if ((m->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
5196		*m_head = m = bge_setup_tso(sc, m, &mss, &csum_flags);
5197		if (*m_head == NULL)
5198			return (ENOBUFS);
5199		csum_flags |= BGE_TXBDFLAG_CPU_PRE_DMA |
5200		    BGE_TXBDFLAG_CPU_POST_DMA;
5201	} else if ((m->m_pkthdr.csum_flags & sc->bge_csum_features) != 0) {
5202		if (m->m_pkthdr.csum_flags & CSUM_IP)
5203			csum_flags |= BGE_TXBDFLAG_IP_CSUM;
5204		if (m->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP)) {
5205			csum_flags |= BGE_TXBDFLAG_TCP_UDP_CSUM;
5206			if (m->m_pkthdr.len < ETHER_MIN_NOPAD &&
5207			    (error = bge_cksum_pad(m)) != 0) {
5208				m_freem(m);
5209				*m_head = NULL;
5210				return (error);
5211			}
5212		}
5213	}
5214
5215	if ((m->m_pkthdr.csum_flags & CSUM_TSO) == 0) {
5216		if (sc->bge_flags & BGE_FLAG_JUMBO_FRAME &&
5217		    m->m_pkthdr.len > ETHER_MAX_LEN)
5218			csum_flags |= BGE_TXBDFLAG_JUMBO_FRAME;
5219		if (sc->bge_forced_collapse > 0 &&
5220		    (sc->bge_flags & BGE_FLAG_PCIE) != 0 && m->m_next != NULL) {
5221			/*
5222			 * Forcedly collapse mbuf chains to overcome hardware
5223			 * limitation which only support a single outstanding
5224			 * DMA read operation.
5225			 */
5226			if (sc->bge_forced_collapse == 1)
5227				m = m_defrag(m, M_NOWAIT);
5228			else
5229				m = m_collapse(m, M_NOWAIT,
5230				    sc->bge_forced_collapse);
5231			if (m == NULL)
5232				m = *m_head;
5233			*m_head = m;
5234		}
5235	}
5236
5237	map = sc->bge_cdata.bge_tx_dmamap[idx];
5238	error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_tx_mtag, map, m, segs,
5239	    &nsegs, BUS_DMA_NOWAIT);
5240	if (error == EFBIG) {
5241		m = m_collapse(m, M_NOWAIT, BGE_NSEG_NEW);
5242		if (m == NULL) {
5243			m_freem(*m_head);
5244			*m_head = NULL;
5245			return (ENOBUFS);
5246		}
5247		*m_head = m;
5248		error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_tx_mtag, map,
5249		    m, segs, &nsegs, BUS_DMA_NOWAIT);
5250		if (error) {
5251			m_freem(m);
5252			*m_head = NULL;
5253			return (error);
5254		}
5255	} else if (error != 0)
5256		return (error);
5257
5258	/* Check if we have enough free send BDs. */
5259	if (sc->bge_txcnt + nsegs >= BGE_TX_RING_CNT) {
5260		bus_dmamap_unload(sc->bge_cdata.bge_tx_mtag, map);
5261		return (ENOBUFS);
5262	}
5263
5264	bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag, map, BUS_DMASYNC_PREWRITE);
5265
5266	if (m->m_flags & M_VLANTAG) {
5267		csum_flags |= BGE_TXBDFLAG_VLAN_TAG;
5268		vlan_tag = m->m_pkthdr.ether_vtag;
5269	}
5270
5271	if (sc->bge_asicrev == BGE_ASICREV_BCM5762 &&
5272	    (m->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
5273		/*
5274		 * 5725 family of devices corrupts TSO packets when TSO DMA
5275		 * buffers cross into regions which are within MSS bytes of
5276		 * a 4GB boundary.  If we encounter the condition, drop the
5277		 * packet.
5278		 */
5279		for (i = 0; ; i++) {
5280			d = &sc->bge_ldata.bge_tx_ring[idx];
5281			d->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[i].ds_addr);
5282			d->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[i].ds_addr);
5283			d->bge_len = segs[i].ds_len;
5284			if (d->bge_addr.bge_addr_lo + segs[i].ds_len + mss <
5285			    d->bge_addr.bge_addr_lo)
5286				break;
5287			d->bge_flags = csum_flags;
5288			d->bge_vlan_tag = vlan_tag;
5289			d->bge_mss = mss;
5290			if (i == nsegs - 1)
5291				break;
5292			BGE_INC(idx, BGE_TX_RING_CNT);
5293		}
5294		if (i != nsegs - 1) {
5295			bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag, map,
5296			    BUS_DMASYNC_POSTWRITE);
5297			bus_dmamap_unload(sc->bge_cdata.bge_tx_mtag, map);
5298			m_freem(*m_head);
5299			*m_head = NULL;
5300			return (EIO);
5301		}
5302	} else {
5303		for (i = 0; ; i++) {
5304			d = &sc->bge_ldata.bge_tx_ring[idx];
5305			d->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[i].ds_addr);
5306			d->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[i].ds_addr);
5307			d->bge_len = segs[i].ds_len;
5308			d->bge_flags = csum_flags;
5309			d->bge_vlan_tag = vlan_tag;
5310			d->bge_mss = mss;
5311			if (i == nsegs - 1)
5312				break;
5313			BGE_INC(idx, BGE_TX_RING_CNT);
5314		}
5315	}
5316
5317	/* Mark the last segment as end of packet... */
5318	d->bge_flags |= BGE_TXBDFLAG_END;
5319
5320	/*
5321	 * Insure that the map for this transmission
5322	 * is placed at the array index of the last descriptor
5323	 * in this chain.
5324	 */
5325	sc->bge_cdata.bge_tx_dmamap[*txidx] = sc->bge_cdata.bge_tx_dmamap[idx];
5326	sc->bge_cdata.bge_tx_dmamap[idx] = map;
5327	sc->bge_cdata.bge_tx_chain[idx] = m;
5328	sc->bge_txcnt += nsegs;
5329
5330	BGE_INC(idx, BGE_TX_RING_CNT);
5331	*txidx = idx;
5332
5333	return (0);
5334}
5335
5336/*
5337 * Main transmit routine. To avoid having to do mbuf copies, we put pointers
5338 * to the mbuf data regions directly in the transmit descriptors.
5339 */
5340static void
5341bge_start_locked(struct ifnet *ifp)
5342{
5343	struct bge_softc *sc;
5344	struct mbuf *m_head;
5345	uint32_t prodidx;
5346	int count;
5347
5348	sc = ifp->if_softc;
5349	BGE_LOCK_ASSERT(sc);
5350
5351	if (!sc->bge_link ||
5352	    (ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
5353	    IFF_DRV_RUNNING)
5354		return;
5355
5356	prodidx = sc->bge_tx_prodidx;
5357
5358	for (count = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd);) {
5359		if (sc->bge_txcnt > BGE_TX_RING_CNT - 16) {
5360			ifp->if_drv_flags |= IFF_DRV_OACTIVE;
5361			break;
5362		}
5363		IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
5364		if (m_head == NULL)
5365			break;
5366
5367		/*
5368		 * Pack the data into the transmit ring. If we
5369		 * don't have room, set the OACTIVE flag and wait
5370		 * for the NIC to drain the ring.
5371		 */
5372		if (bge_encap(sc, &m_head, &prodidx)) {
5373			if (m_head == NULL)
5374				break;
5375			IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
5376			ifp->if_drv_flags |= IFF_DRV_OACTIVE;
5377			break;
5378		}
5379		++count;
5380
5381		/*
5382		 * If there's a BPF listener, bounce a copy of this frame
5383		 * to him.
5384		 */
5385#ifdef ETHER_BPF_MTAP
5386		ETHER_BPF_MTAP(ifp, m_head);
5387#else
5388		BPF_MTAP(ifp, m_head);
5389#endif
5390	}
5391
5392	if (count > 0) {
5393		bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
5394		    sc->bge_cdata.bge_tx_ring_map, BUS_DMASYNC_PREWRITE);
5395		/* Transmit. */
5396		bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
5397		/* 5700 b2 errata */
5398		if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
5399			bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
5400
5401		sc->bge_tx_prodidx = prodidx;
5402
5403		/*
5404		 * Set a timeout in case the chip goes out to lunch.
5405		 */
5406		sc->bge_timer = BGE_TX_TIMEOUT;
5407	}
5408}
5409
5410/*
5411 * Main transmit routine. To avoid having to do mbuf copies, we put pointers
5412 * to the mbuf data regions directly in the transmit descriptors.
5413 */
5414static void
5415bge_start(struct ifnet *ifp)
5416{
5417	struct bge_softc *sc;
5418
5419	sc = ifp->if_softc;
5420	BGE_LOCK(sc);
5421	bge_start_locked(ifp);
5422	BGE_UNLOCK(sc);
5423}
5424
5425static void
5426bge_init_locked(struct bge_softc *sc)
5427{
5428	struct ifnet *ifp;
5429	uint16_t *m;
5430	uint32_t mode;
5431
5432	BGE_LOCK_ASSERT(sc);
5433
5434	ifp = sc->bge_ifp;
5435
5436	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
5437		return;
5438
5439	/* Cancel pending I/O and flush buffers. */
5440	bge_stop(sc);
5441
5442	bge_stop_fw(sc);
5443	bge_sig_pre_reset(sc, BGE_RESET_START);
5444	bge_reset(sc);
5445	bge_sig_legacy(sc, BGE_RESET_START);
5446	bge_sig_post_reset(sc, BGE_RESET_START);
5447
5448	bge_chipinit(sc);
5449
5450	/*
5451	 * Init the various state machines, ring
5452	 * control blocks and firmware.
5453	 */
5454	if (bge_blockinit(sc)) {
5455		device_printf(sc->bge_dev, "initialization failure\n");
5456		return;
5457	}
5458
5459	ifp = sc->bge_ifp;
5460
5461	/* Specify MTU. */
5462	CSR_WRITE_4(sc, BGE_RX_MTU, ifp->if_mtu +
5463	    ETHER_HDR_LEN + ETHER_CRC_LEN +
5464	    (ifp->if_capenable & IFCAP_VLAN_MTU ? ETHER_VLAN_ENCAP_LEN : 0));
5465
5466	/* Load our MAC address. */
5467	m = (uint16_t *)IF_LLADDR(sc->bge_ifp);
5468	CSR_WRITE_4(sc, BGE_MAC_ADDR1_LO, htons(m[0]));
5469	CSR_WRITE_4(sc, BGE_MAC_ADDR1_HI, (htons(m[1]) << 16) | htons(m[2]));
5470
5471	/* Program promiscuous mode. */
5472	bge_setpromisc(sc);
5473
5474	/* Program multicast filter. */
5475	bge_setmulti(sc);
5476
5477	/* Program VLAN tag stripping. */
5478	bge_setvlan(sc);
5479
5480	/* Override UDP checksum offloading. */
5481	if (sc->bge_forced_udpcsum == 0)
5482		sc->bge_csum_features &= ~CSUM_UDP;
5483	else
5484		sc->bge_csum_features |= CSUM_UDP;
5485	if (ifp->if_capabilities & IFCAP_TXCSUM &&
5486	    ifp->if_capenable & IFCAP_TXCSUM) {
5487		ifp->if_hwassist &= ~(BGE_CSUM_FEATURES | CSUM_UDP);
5488		ifp->if_hwassist |= sc->bge_csum_features;
5489	}
5490
5491	/* Init RX ring. */
5492	if (bge_init_rx_ring_std(sc) != 0) {
5493		device_printf(sc->bge_dev, "no memory for std Rx buffers.\n");
5494		bge_stop(sc);
5495		return;
5496	}
5497
5498	/*
5499	 * Workaround for a bug in 5705 ASIC rev A0. Poll the NIC's
5500	 * memory to insure that the chip has in fact read the first
5501	 * entry of the ring.
5502	 */
5503	if (sc->bge_chipid == BGE_CHIPID_BCM5705_A0) {
5504		uint32_t		v, i;
5505		for (i = 0; i < 10; i++) {
5506			DELAY(20);
5507			v = bge_readmem_ind(sc, BGE_STD_RX_RINGS + 8);
5508			if (v == (MCLBYTES - ETHER_ALIGN))
5509				break;
5510		}
5511		if (i == 10)
5512			device_printf (sc->bge_dev,
5513			    "5705 A0 chip failed to load RX ring\n");
5514	}
5515
5516	/* Init jumbo RX ring. */
5517	if (BGE_IS_JUMBO_CAPABLE(sc) &&
5518	    ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
5519	    (MCLBYTES - ETHER_ALIGN)) {
5520		if (bge_init_rx_ring_jumbo(sc) != 0) {
5521			device_printf(sc->bge_dev,
5522			    "no memory for jumbo Rx buffers.\n");
5523			bge_stop(sc);
5524			return;
5525		}
5526	}
5527
5528	/* Init our RX return ring index. */
5529	sc->bge_rx_saved_considx = 0;
5530
5531	/* Init our RX/TX stat counters. */
5532	sc->bge_rx_discards = sc->bge_tx_discards = sc->bge_tx_collisions = 0;
5533
5534	/* Init TX ring. */
5535	bge_init_tx_ring(sc);
5536
5537	/* Enable TX MAC state machine lockup fix. */
5538	mode = CSR_READ_4(sc, BGE_TX_MODE);
5539	if (BGE_IS_5755_PLUS(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5906)
5540		mode |= BGE_TXMODE_MBUF_LOCKUP_FIX;
5541	if (sc->bge_asicrev == BGE_ASICREV_BCM5720 ||
5542	    sc->bge_asicrev == BGE_ASICREV_BCM5762) {
5543		mode &= ~(BGE_TXMODE_JMB_FRM_LEN | BGE_TXMODE_CNT_DN_MODE);
5544		mode |= CSR_READ_4(sc, BGE_TX_MODE) &
5545		    (BGE_TXMODE_JMB_FRM_LEN | BGE_TXMODE_CNT_DN_MODE);
5546	}
5547	/* Turn on transmitter. */
5548	CSR_WRITE_4(sc, BGE_TX_MODE, mode | BGE_TXMODE_ENABLE);
5549	DELAY(100);
5550
5551	/* Turn on receiver. */
5552	mode = CSR_READ_4(sc, BGE_RX_MODE);
5553	if (BGE_IS_5755_PLUS(sc))
5554		mode |= BGE_RXMODE_IPV6_ENABLE;
5555	if (sc->bge_asicrev == BGE_ASICREV_BCM5762)
5556		mode |= BGE_RXMODE_IPV4_FRAG_FIX;
5557	CSR_WRITE_4(sc,BGE_RX_MODE, mode | BGE_RXMODE_ENABLE);
5558	DELAY(10);
5559
5560	/*
5561	 * Set the number of good frames to receive after RX MBUF
5562	 * Low Watermark has been reached. After the RX MAC receives
5563	 * this number of frames, it will drop subsequent incoming
5564	 * frames until the MBUF High Watermark is reached.
5565	 */
5566	if (BGE_IS_57765_PLUS(sc))
5567		CSR_WRITE_4(sc, BGE_MAX_RX_FRAME_LOWAT, 1);
5568	else
5569		CSR_WRITE_4(sc, BGE_MAX_RX_FRAME_LOWAT, 2);
5570
5571	/* Clear MAC statistics. */
5572	if (BGE_IS_5705_PLUS(sc))
5573		bge_stats_clear_regs(sc);
5574
5575	/* Tell firmware we're alive. */
5576	BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
5577
5578#ifdef DEVICE_POLLING
5579	/* Disable interrupts if we are polling. */
5580	if (ifp->if_capenable & IFCAP_POLLING) {
5581		BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
5582		    BGE_PCIMISCCTL_MASK_PCI_INTR);
5583		bge_writembx(sc, BGE_MBX_IRQ0_LO, 1);
5584	} else
5585#endif
5586
5587	/* Enable host interrupts. */
5588	{
5589	BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_CLEAR_INTA);
5590	BGE_CLRBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
5591	bge_writembx(sc, BGE_MBX_IRQ0_LO, 0);
5592	}
5593
5594	ifp->if_drv_flags |= IFF_DRV_RUNNING;
5595	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
5596
5597	bge_ifmedia_upd_locked(ifp);
5598
5599	callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
5600}
5601
5602static void
5603bge_init(void *xsc)
5604{
5605	struct bge_softc *sc = xsc;
5606
5607	BGE_LOCK(sc);
5608	bge_init_locked(sc);
5609	BGE_UNLOCK(sc);
5610}
5611
5612/*
5613 * Set media options.
5614 */
5615static int
5616bge_ifmedia_upd(struct ifnet *ifp)
5617{
5618	struct bge_softc *sc = ifp->if_softc;
5619	int res;
5620
5621	BGE_LOCK(sc);
5622	res = bge_ifmedia_upd_locked(ifp);
5623	BGE_UNLOCK(sc);
5624
5625	return (res);
5626}
5627
5628static int
5629bge_ifmedia_upd_locked(struct ifnet *ifp)
5630{
5631	struct bge_softc *sc = ifp->if_softc;
5632	struct mii_data *mii;
5633	struct mii_softc *miisc;
5634	struct ifmedia *ifm;
5635
5636	BGE_LOCK_ASSERT(sc);
5637
5638	ifm = &sc->bge_ifmedia;
5639
5640	/* If this is a 1000baseX NIC, enable the TBI port. */
5641	if (sc->bge_flags & BGE_FLAG_TBI) {
5642		if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
5643			return (EINVAL);
5644		switch(IFM_SUBTYPE(ifm->ifm_media)) {
5645		case IFM_AUTO:
5646			/*
5647			 * The BCM5704 ASIC appears to have a special
5648			 * mechanism for programming the autoneg
5649			 * advertisement registers in TBI mode.
5650			 */
5651			if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
5652				uint32_t sgdig;
5653				sgdig = CSR_READ_4(sc, BGE_SGDIG_STS);
5654				if (sgdig & BGE_SGDIGSTS_DONE) {
5655					CSR_WRITE_4(sc, BGE_TX_TBI_AUTONEG, 0);
5656					sgdig = CSR_READ_4(sc, BGE_SGDIG_CFG);
5657					sgdig |= BGE_SGDIGCFG_AUTO |
5658					    BGE_SGDIGCFG_PAUSE_CAP |
5659					    BGE_SGDIGCFG_ASYM_PAUSE;
5660					CSR_WRITE_4(sc, BGE_SGDIG_CFG,
5661					    sgdig | BGE_SGDIGCFG_SEND);
5662					DELAY(5);
5663					CSR_WRITE_4(sc, BGE_SGDIG_CFG, sgdig);
5664				}
5665			}
5666			break;
5667		case IFM_1000_SX:
5668			if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
5669				BGE_CLRBIT(sc, BGE_MAC_MODE,
5670				    BGE_MACMODE_HALF_DUPLEX);
5671			} else {
5672				BGE_SETBIT(sc, BGE_MAC_MODE,
5673				    BGE_MACMODE_HALF_DUPLEX);
5674			}
5675			DELAY(40);
5676			break;
5677		default:
5678			return (EINVAL);
5679		}
5680		return (0);
5681	}
5682
5683	sc->bge_link_evt++;
5684	mii = device_get_softc(sc->bge_miibus);
5685	LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
5686		PHY_RESET(miisc);
5687	mii_mediachg(mii);
5688
5689	/*
5690	 * Force an interrupt so that we will call bge_link_upd
5691	 * if needed and clear any pending link state attention.
5692	 * Without this we are not getting any further interrupts
5693	 * for link state changes and thus will not UP the link and
5694	 * not be able to send in bge_start_locked. The only
5695	 * way to get things working was to receive a packet and
5696	 * get an RX intr.
5697	 * bge_tick should help for fiber cards and we might not
5698	 * need to do this here if BGE_FLAG_TBI is set but as
5699	 * we poll for fiber anyway it should not harm.
5700	 */
5701	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
5702	    sc->bge_flags & BGE_FLAG_5788)
5703		BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
5704	else
5705		BGE_SETBIT(sc, BGE_HCC_MODE, BGE_HCCMODE_COAL_NOW);
5706
5707	return (0);
5708}
5709
5710/*
5711 * Report current media status.
5712 */
5713static void
5714bge_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
5715{
5716	struct bge_softc *sc = ifp->if_softc;
5717	struct mii_data *mii;
5718
5719	BGE_LOCK(sc);
5720
5721	if ((ifp->if_flags & IFF_UP) == 0) {
5722		BGE_UNLOCK(sc);
5723		return;
5724	}
5725	if (sc->bge_flags & BGE_FLAG_TBI) {
5726		ifmr->ifm_status = IFM_AVALID;
5727		ifmr->ifm_active = IFM_ETHER;
5728		if (CSR_READ_4(sc, BGE_MAC_STS) &
5729		    BGE_MACSTAT_TBI_PCS_SYNCHED)
5730			ifmr->ifm_status |= IFM_ACTIVE;
5731		else {
5732			ifmr->ifm_active |= IFM_NONE;
5733			BGE_UNLOCK(sc);
5734			return;
5735		}
5736		ifmr->ifm_active |= IFM_1000_SX;
5737		if (CSR_READ_4(sc, BGE_MAC_MODE) & BGE_MACMODE_HALF_DUPLEX)
5738			ifmr->ifm_active |= IFM_HDX;
5739		else
5740			ifmr->ifm_active |= IFM_FDX;
5741		BGE_UNLOCK(sc);
5742		return;
5743	}
5744
5745	mii = device_get_softc(sc->bge_miibus);
5746	mii_pollstat(mii);
5747	ifmr->ifm_active = mii->mii_media_active;
5748	ifmr->ifm_status = mii->mii_media_status;
5749
5750	BGE_UNLOCK(sc);
5751}
5752
5753static int
5754bge_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
5755{
5756	struct bge_softc *sc = ifp->if_softc;
5757	struct ifreq *ifr = (struct ifreq *) data;
5758	struct mii_data *mii;
5759	int flags, mask, error = 0;
5760
5761	switch (command) {
5762	case SIOCSIFMTU:
5763		if (BGE_IS_JUMBO_CAPABLE(sc) ||
5764		    (sc->bge_flags & BGE_FLAG_JUMBO_STD)) {
5765			if (ifr->ifr_mtu < ETHERMIN ||
5766			    ifr->ifr_mtu > BGE_JUMBO_MTU) {
5767				error = EINVAL;
5768				break;
5769			}
5770		} else if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > ETHERMTU) {
5771			error = EINVAL;
5772			break;
5773		}
5774		BGE_LOCK(sc);
5775		if (ifp->if_mtu != ifr->ifr_mtu) {
5776			ifp->if_mtu = ifr->ifr_mtu;
5777			if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
5778				ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
5779				bge_init_locked(sc);
5780			}
5781		}
5782		BGE_UNLOCK(sc);
5783		break;
5784	case SIOCSIFFLAGS:
5785		BGE_LOCK(sc);
5786		if (ifp->if_flags & IFF_UP) {
5787			/*
5788			 * If only the state of the PROMISC flag changed,
5789			 * then just use the 'set promisc mode' command
5790			 * instead of reinitializing the entire NIC. Doing
5791			 * a full re-init means reloading the firmware and
5792			 * waiting for it to start up, which may take a
5793			 * second or two.  Similarly for ALLMULTI.
5794			 */
5795			if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
5796				flags = ifp->if_flags ^ sc->bge_if_flags;
5797				if (flags & IFF_PROMISC)
5798					bge_setpromisc(sc);
5799				if (flags & IFF_ALLMULTI)
5800					bge_setmulti(sc);
5801			} else
5802				bge_init_locked(sc);
5803		} else {
5804			if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
5805				bge_stop(sc);
5806			}
5807		}
5808		sc->bge_if_flags = ifp->if_flags;
5809		BGE_UNLOCK(sc);
5810		error = 0;
5811		break;
5812	case SIOCADDMULTI:
5813	case SIOCDELMULTI:
5814		if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
5815			BGE_LOCK(sc);
5816			bge_setmulti(sc);
5817			BGE_UNLOCK(sc);
5818			error = 0;
5819		}
5820		break;
5821	case SIOCSIFMEDIA:
5822	case SIOCGIFMEDIA:
5823		if (sc->bge_flags & BGE_FLAG_TBI) {
5824			error = ifmedia_ioctl(ifp, ifr,
5825			    &sc->bge_ifmedia, command);
5826		} else {
5827			mii = device_get_softc(sc->bge_miibus);
5828			error = ifmedia_ioctl(ifp, ifr,
5829			    &mii->mii_media, command);
5830		}
5831		break;
5832	case SIOCSIFCAP:
5833		mask = ifr->ifr_reqcap ^ ifp->if_capenable;
5834#ifdef DEVICE_POLLING
5835		if (mask & IFCAP_POLLING) {
5836			if (ifr->ifr_reqcap & IFCAP_POLLING) {
5837				error = ether_poll_register(bge_poll, ifp);
5838				if (error)
5839					return (error);
5840				BGE_LOCK(sc);
5841				BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
5842				    BGE_PCIMISCCTL_MASK_PCI_INTR);
5843				bge_writembx(sc, BGE_MBX_IRQ0_LO, 1);
5844				ifp->if_capenable |= IFCAP_POLLING;
5845				BGE_UNLOCK(sc);
5846			} else {
5847				error = ether_poll_deregister(ifp);
5848				/* Enable interrupt even in error case */
5849				BGE_LOCK(sc);
5850				BGE_CLRBIT(sc, BGE_PCI_MISC_CTL,
5851				    BGE_PCIMISCCTL_MASK_PCI_INTR);
5852				bge_writembx(sc, BGE_MBX_IRQ0_LO, 0);
5853				ifp->if_capenable &= ~IFCAP_POLLING;
5854				BGE_UNLOCK(sc);
5855			}
5856		}
5857#endif
5858		if ((mask & IFCAP_TXCSUM) != 0 &&
5859		    (ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
5860			ifp->if_capenable ^= IFCAP_TXCSUM;
5861			if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
5862				ifp->if_hwassist |= sc->bge_csum_features;
5863			else
5864				ifp->if_hwassist &= ~sc->bge_csum_features;
5865		}
5866
5867		if ((mask & IFCAP_RXCSUM) != 0 &&
5868		    (ifp->if_capabilities & IFCAP_RXCSUM) != 0)
5869			ifp->if_capenable ^= IFCAP_RXCSUM;
5870
5871		if ((mask & IFCAP_TSO4) != 0 &&
5872		    (ifp->if_capabilities & IFCAP_TSO4) != 0) {
5873			ifp->if_capenable ^= IFCAP_TSO4;
5874			if ((ifp->if_capenable & IFCAP_TSO4) != 0)
5875				ifp->if_hwassist |= CSUM_TSO;
5876			else
5877				ifp->if_hwassist &= ~CSUM_TSO;
5878		}
5879
5880		if (mask & IFCAP_VLAN_MTU) {
5881			ifp->if_capenable ^= IFCAP_VLAN_MTU;
5882			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
5883			bge_init(sc);
5884		}
5885
5886		if ((mask & IFCAP_VLAN_HWTSO) != 0 &&
5887		    (ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0)
5888			ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
5889		if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
5890		    (ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) {
5891			ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
5892			if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
5893				ifp->if_capenable &= ~IFCAP_VLAN_HWTSO;
5894			BGE_LOCK(sc);
5895			bge_setvlan(sc);
5896			BGE_UNLOCK(sc);
5897		}
5898#ifdef VLAN_CAPABILITIES
5899		VLAN_CAPABILITIES(ifp);
5900#endif
5901		break;
5902	default:
5903		error = ether_ioctl(ifp, command, data);
5904		break;
5905	}
5906
5907	return (error);
5908}
5909
5910static void
5911bge_watchdog(struct bge_softc *sc)
5912{
5913	struct ifnet *ifp;
5914	uint32_t status;
5915
5916	BGE_LOCK_ASSERT(sc);
5917
5918	if (sc->bge_timer == 0 || --sc->bge_timer)
5919		return;
5920
5921	/* If pause frames are active then don't reset the hardware. */
5922	if ((CSR_READ_4(sc, BGE_RX_MODE) & BGE_RXMODE_FLOWCTL_ENABLE) != 0) {
5923		status = CSR_READ_4(sc, BGE_RX_STS);
5924		if ((status & BGE_RXSTAT_REMOTE_XOFFED) != 0) {
5925			/*
5926			 * If link partner has us in XOFF state then wait for
5927			 * the condition to clear.
5928			 */
5929			CSR_WRITE_4(sc, BGE_RX_STS, status);
5930			sc->bge_timer = BGE_TX_TIMEOUT;
5931			return;
5932		} else if ((status & BGE_RXSTAT_RCVD_XOFF) != 0 &&
5933		    (status & BGE_RXSTAT_RCVD_XON) != 0) {
5934			/*
5935			 * If link partner has us in XOFF state then wait for
5936			 * the condition to clear.
5937			 */
5938			CSR_WRITE_4(sc, BGE_RX_STS, status);
5939			sc->bge_timer = BGE_TX_TIMEOUT;
5940			return;
5941		}
5942		/*
5943		 * Any other condition is unexpected and the controller
5944		 * should be reset.
5945		 */
5946	}
5947
5948	ifp = sc->bge_ifp;
5949
5950	if_printf(ifp, "watchdog timeout -- resetting\n");
5951
5952	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
5953	bge_init_locked(sc);
5954
5955	ifp->if_oerrors++;
5956}
5957
5958static void
5959bge_stop_block(struct bge_softc *sc, bus_size_t reg, uint32_t bit)
5960{
5961	int i;
5962
5963	BGE_CLRBIT(sc, reg, bit);
5964
5965	for (i = 0; i < BGE_TIMEOUT; i++) {
5966		if ((CSR_READ_4(sc, reg) & bit) == 0)
5967			return;
5968		DELAY(100);
5969        }
5970}
5971
5972/*
5973 * Stop the adapter and free any mbufs allocated to the
5974 * RX and TX lists.
5975 */
5976static void
5977bge_stop(struct bge_softc *sc)
5978{
5979	struct ifnet *ifp;
5980
5981	BGE_LOCK_ASSERT(sc);
5982
5983	ifp = sc->bge_ifp;
5984
5985	callout_stop(&sc->bge_stat_ch);
5986
5987	/* Disable host interrupts. */
5988	BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
5989	bge_writembx(sc, BGE_MBX_IRQ0_LO, 1);
5990
5991	/*
5992	 * Tell firmware we're shutting down.
5993	 */
5994	bge_stop_fw(sc);
5995	bge_sig_pre_reset(sc, BGE_RESET_SHUTDOWN);
5996
5997	/*
5998	 * Disable all of the receiver blocks.
5999	 */
6000	bge_stop_block(sc, BGE_RX_MODE, BGE_RXMODE_ENABLE);
6001	bge_stop_block(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
6002	bge_stop_block(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
6003	if (BGE_IS_5700_FAMILY(sc))
6004		bge_stop_block(sc, BGE_RXLS_MODE, BGE_RXLSMODE_ENABLE);
6005	bge_stop_block(sc, BGE_RDBDI_MODE, BGE_RBDIMODE_ENABLE);
6006	bge_stop_block(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
6007	bge_stop_block(sc, BGE_RBDC_MODE, BGE_RBDCMODE_ENABLE);
6008
6009	/*
6010	 * Disable all of the transmit blocks.
6011	 */
6012	bge_stop_block(sc, BGE_SRS_MODE, BGE_SRSMODE_ENABLE);
6013	bge_stop_block(sc, BGE_SBDI_MODE, BGE_SBDIMODE_ENABLE);
6014	bge_stop_block(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE);
6015	bge_stop_block(sc, BGE_RDMA_MODE, BGE_RDMAMODE_ENABLE);
6016	bge_stop_block(sc, BGE_SDC_MODE, BGE_SDCMODE_ENABLE);
6017	if (BGE_IS_5700_FAMILY(sc))
6018		bge_stop_block(sc, BGE_DMAC_MODE, BGE_DMACMODE_ENABLE);
6019	bge_stop_block(sc, BGE_SBDC_MODE, BGE_SBDCMODE_ENABLE);
6020
6021	/*
6022	 * Shut down all of the memory managers and related
6023	 * state machines.
6024	 */
6025	bge_stop_block(sc, BGE_HCC_MODE, BGE_HCCMODE_ENABLE);
6026	bge_stop_block(sc, BGE_WDMA_MODE, BGE_WDMAMODE_ENABLE);
6027	if (BGE_IS_5700_FAMILY(sc))
6028		bge_stop_block(sc, BGE_MBCF_MODE, BGE_MBCFMODE_ENABLE);
6029
6030	CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
6031	CSR_WRITE_4(sc, BGE_FTQ_RESET, 0);
6032	if (!(BGE_IS_5705_PLUS(sc))) {
6033		BGE_CLRBIT(sc, BGE_BMAN_MODE, BGE_BMANMODE_ENABLE);
6034		BGE_CLRBIT(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE);
6035	}
6036	/* Update MAC statistics. */
6037	if (BGE_IS_5705_PLUS(sc))
6038		bge_stats_update_regs(sc);
6039
6040	bge_reset(sc);
6041	bge_sig_legacy(sc, BGE_RESET_SHUTDOWN);
6042	bge_sig_post_reset(sc, BGE_RESET_SHUTDOWN);
6043
6044	/*
6045	 * Keep the ASF firmware running if up.
6046	 */
6047	if (sc->bge_asf_mode & ASF_STACKUP)
6048		BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
6049	else
6050		BGE_CLRBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
6051
6052	/* Free the RX lists. */
6053	bge_free_rx_ring_std(sc);
6054
6055	/* Free jumbo RX list. */
6056	if (BGE_IS_JUMBO_CAPABLE(sc))
6057		bge_free_rx_ring_jumbo(sc);
6058
6059	/* Free TX buffers. */
6060	bge_free_tx_ring(sc);
6061
6062	sc->bge_tx_saved_considx = BGE_TXCONS_UNSET;
6063
6064	/* Clear MAC's link state (PHY may still have link UP). */
6065	if (bootverbose && sc->bge_link)
6066		if_printf(sc->bge_ifp, "link DOWN\n");
6067	sc->bge_link = 0;
6068
6069	ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
6070}
6071
6072/*
6073 * Stop all chip I/O so that the kernel's probe routines don't
6074 * get confused by errant DMAs when rebooting.
6075 */
6076static int
6077bge_shutdown(device_t dev)
6078{
6079	struct bge_softc *sc;
6080
6081	sc = device_get_softc(dev);
6082	BGE_LOCK(sc);
6083	bge_stop(sc);
6084	BGE_UNLOCK(sc);
6085
6086	return (0);
6087}
6088
6089static int
6090bge_suspend(device_t dev)
6091{
6092	struct bge_softc *sc;
6093
6094	sc = device_get_softc(dev);
6095	BGE_LOCK(sc);
6096	bge_stop(sc);
6097	BGE_UNLOCK(sc);
6098
6099	return (0);
6100}
6101
6102static int
6103bge_resume(device_t dev)
6104{
6105	struct bge_softc *sc;
6106	struct ifnet *ifp;
6107
6108	sc = device_get_softc(dev);
6109	BGE_LOCK(sc);
6110	ifp = sc->bge_ifp;
6111	if (ifp->if_flags & IFF_UP) {
6112		bge_init_locked(sc);
6113		if (ifp->if_drv_flags & IFF_DRV_RUNNING)
6114			bge_start_locked(ifp);
6115	}
6116	BGE_UNLOCK(sc);
6117
6118	return (0);
6119}
6120
6121static void
6122bge_link_upd(struct bge_softc *sc)
6123{
6124	struct mii_data *mii;
6125	uint32_t link, status;
6126
6127	BGE_LOCK_ASSERT(sc);
6128
6129	/* Clear 'pending link event' flag. */
6130	sc->bge_link_evt = 0;
6131
6132	/*
6133	 * Process link state changes.
6134	 * Grrr. The link status word in the status block does
6135	 * not work correctly on the BCM5700 rev AX and BX chips,
6136	 * according to all available information. Hence, we have
6137	 * to enable MII interrupts in order to properly obtain
6138	 * async link changes. Unfortunately, this also means that
6139	 * we have to read the MAC status register to detect link
6140	 * changes, thereby adding an additional register access to
6141	 * the interrupt handler.
6142	 *
6143	 * XXX: perhaps link state detection procedure used for
6144	 * BGE_CHIPID_BCM5700_B2 can be used for others BCM5700 revisions.
6145	 */
6146
6147	if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
6148	    sc->bge_chipid != BGE_CHIPID_BCM5700_B2) {
6149		status = CSR_READ_4(sc, BGE_MAC_STS);
6150		if (status & BGE_MACSTAT_MI_INTERRUPT) {
6151			mii = device_get_softc(sc->bge_miibus);
6152			mii_pollstat(mii);
6153			if (!sc->bge_link &&
6154			    mii->mii_media_status & IFM_ACTIVE &&
6155			    IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
6156				sc->bge_link++;
6157				if (bootverbose)
6158					if_printf(sc->bge_ifp, "link UP\n");
6159			} else if (sc->bge_link &&
6160			    (!(mii->mii_media_status & IFM_ACTIVE) ||
6161			    IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
6162				sc->bge_link = 0;
6163				if (bootverbose)
6164					if_printf(sc->bge_ifp, "link DOWN\n");
6165			}
6166
6167			/* Clear the interrupt. */
6168			CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
6169			    BGE_EVTENB_MI_INTERRUPT);
6170			bge_miibus_readreg(sc->bge_dev, sc->bge_phy_addr,
6171			    BRGPHY_MII_ISR);
6172			bge_miibus_writereg(sc->bge_dev, sc->bge_phy_addr,
6173			    BRGPHY_MII_IMR, BRGPHY_INTRS);
6174		}
6175		return;
6176	}
6177
6178	if (sc->bge_flags & BGE_FLAG_TBI) {
6179		status = CSR_READ_4(sc, BGE_MAC_STS);
6180		if (status & BGE_MACSTAT_TBI_PCS_SYNCHED) {
6181			if (!sc->bge_link) {
6182				sc->bge_link++;
6183				if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
6184					BGE_CLRBIT(sc, BGE_MAC_MODE,
6185					    BGE_MACMODE_TBI_SEND_CFGS);
6186					DELAY(40);
6187				}
6188				CSR_WRITE_4(sc, BGE_MAC_STS, 0xFFFFFFFF);
6189				if (bootverbose)
6190					if_printf(sc->bge_ifp, "link UP\n");
6191				if_link_state_change(sc->bge_ifp,
6192				    LINK_STATE_UP);
6193			}
6194		} else if (sc->bge_link) {
6195			sc->bge_link = 0;
6196			if (bootverbose)
6197				if_printf(sc->bge_ifp, "link DOWN\n");
6198			if_link_state_change(sc->bge_ifp, LINK_STATE_DOWN);
6199		}
6200	} else if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
6201		/*
6202		 * Some broken BCM chips have BGE_STATFLAG_LINKSTATE_CHANGED bit
6203		 * in status word always set. Workaround this bug by reading
6204		 * PHY link status directly.
6205		 */
6206		link = (CSR_READ_4(sc, BGE_MI_STS) & BGE_MISTS_LINK) ? 1 : 0;
6207
6208		if (link != sc->bge_link ||
6209		    sc->bge_asicrev == BGE_ASICREV_BCM5700) {
6210			mii = device_get_softc(sc->bge_miibus);
6211			mii_pollstat(mii);
6212			if (!sc->bge_link &&
6213			    mii->mii_media_status & IFM_ACTIVE &&
6214			    IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
6215				sc->bge_link++;
6216				if (bootverbose)
6217					if_printf(sc->bge_ifp, "link UP\n");
6218			} else if (sc->bge_link &&
6219			    (!(mii->mii_media_status & IFM_ACTIVE) ||
6220			    IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
6221				sc->bge_link = 0;
6222				if (bootverbose)
6223					if_printf(sc->bge_ifp, "link DOWN\n");
6224			}
6225		}
6226	} else {
6227		/*
6228		 * For controllers that call mii_tick, we have to poll
6229		 * link status.
6230		 */
6231		mii = device_get_softc(sc->bge_miibus);
6232		mii_pollstat(mii);
6233		bge_miibus_statchg(sc->bge_dev);
6234	}
6235
6236	/* Disable MAC attention when link is up. */
6237	CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
6238	    BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
6239	    BGE_MACSTAT_LINK_CHANGED);
6240}
6241
6242static void
6243bge_add_sysctls(struct bge_softc *sc)
6244{
6245	struct sysctl_ctx_list *ctx;
6246	struct sysctl_oid_list *children;
6247	char tn[32];
6248	int unit;
6249
6250	ctx = device_get_sysctl_ctx(sc->bge_dev);
6251	children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->bge_dev));
6252
6253#ifdef BGE_REGISTER_DEBUG
6254	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "debug_info",
6255	    CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_debug_info, "I",
6256	    "Debug Information");
6257
6258	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "reg_read",
6259	    CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_reg_read, "I",
6260	    "MAC Register Read");
6261
6262	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ape_read",
6263	    CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_ape_read, "I",
6264	    "APE Register Read");
6265
6266	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "mem_read",
6267	    CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_mem_read, "I",
6268	    "Memory Read");
6269
6270#endif
6271
6272	unit = device_get_unit(sc->bge_dev);
6273	/*
6274	 * A common design characteristic for many Broadcom client controllers
6275	 * is that they only support a single outstanding DMA read operation
6276	 * on the PCIe bus. This means that it will take twice as long to fetch
6277	 * a TX frame that is split into header and payload buffers as it does
6278	 * to fetch a single, contiguous TX frame (2 reads vs. 1 read). For
6279	 * these controllers, coalescing buffers to reduce the number of memory
6280	 * reads is effective way to get maximum performance(about 940Mbps).
6281	 * Without collapsing TX buffers the maximum TCP bulk transfer
6282	 * performance is about 850Mbps. However forcing coalescing mbufs
6283	 * consumes a lot of CPU cycles, so leave it off by default.
6284	 */
6285	sc->bge_forced_collapse = 0;
6286	snprintf(tn, sizeof(tn), "dev.bge.%d.forced_collapse", unit);
6287	TUNABLE_INT_FETCH(tn, &sc->bge_forced_collapse);
6288	SYSCTL_ADD_INT(ctx, children, OID_AUTO, "forced_collapse",
6289	    CTLFLAG_RW, &sc->bge_forced_collapse, 0,
6290	    "Number of fragmented TX buffers of a frame allowed before "
6291	    "forced collapsing");
6292
6293	sc->bge_msi = 1;
6294	snprintf(tn, sizeof(tn), "dev.bge.%d.msi", unit);
6295	TUNABLE_INT_FETCH(tn, &sc->bge_msi);
6296	SYSCTL_ADD_INT(ctx, children, OID_AUTO, "msi",
6297	    CTLFLAG_RD, &sc->bge_msi, 0, "Enable MSI");
6298
6299	/*
6300	 * It seems all Broadcom controllers have a bug that can generate UDP
6301	 * datagrams with checksum value 0 when TX UDP checksum offloading is
6302	 * enabled.  Generating UDP checksum value 0 is RFC 768 violation.
6303	 * Even though the probability of generating such UDP datagrams is
6304	 * low, I don't want to see FreeBSD boxes to inject such datagrams
6305	 * into network so disable UDP checksum offloading by default.  Users
6306	 * still override this behavior by setting a sysctl variable,
6307	 * dev.bge.0.forced_udpcsum.
6308	 */
6309	sc->bge_forced_udpcsum = 0;
6310	snprintf(tn, sizeof(tn), "dev.bge.%d.bge_forced_udpcsum", unit);
6311	TUNABLE_INT_FETCH(tn, &sc->bge_forced_udpcsum);
6312	SYSCTL_ADD_INT(ctx, children, OID_AUTO, "forced_udpcsum",
6313	    CTLFLAG_RW, &sc->bge_forced_udpcsum, 0,
6314	    "Enable UDP checksum offloading even if controller can "
6315	    "generate UDP checksum value 0");
6316
6317	if (BGE_IS_5705_PLUS(sc))
6318		bge_add_sysctl_stats_regs(sc, ctx, children);
6319	else
6320		bge_add_sysctl_stats(sc, ctx, children);
6321}
6322
6323#define BGE_SYSCTL_STAT(sc, ctx, desc, parent, node, oid) \
6324	SYSCTL_ADD_PROC(ctx, parent, OID_AUTO, oid, CTLTYPE_UINT|CTLFLAG_RD, \
6325	    sc, offsetof(struct bge_stats, node), bge_sysctl_stats, "IU", \
6326	    desc)
6327
6328static void
6329bge_add_sysctl_stats(struct bge_softc *sc, struct sysctl_ctx_list *ctx,
6330    struct sysctl_oid_list *parent)
6331{
6332	struct sysctl_oid *tree;
6333	struct sysctl_oid_list *children, *schildren;
6334
6335	tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "stats", CTLFLAG_RD,
6336	    NULL, "BGE Statistics");
6337	schildren = children = SYSCTL_CHILDREN(tree);
6338	BGE_SYSCTL_STAT(sc, ctx, "Frames Dropped Due To Filters",
6339	    children, COSFramesDroppedDueToFilters,
6340	    "FramesDroppedDueToFilters");
6341	BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Write Queue Full",
6342	    children, nicDmaWriteQueueFull, "DmaWriteQueueFull");
6343	BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Write High Priority Queue Full",
6344	    children, nicDmaWriteHighPriQueueFull, "DmaWriteHighPriQueueFull");
6345	BGE_SYSCTL_STAT(sc, ctx, "NIC No More RX Buffer Descriptors",
6346	    children, nicNoMoreRxBDs, "NoMoreRxBDs");
6347	BGE_SYSCTL_STAT(sc, ctx, "Discarded Input Frames",
6348	    children, ifInDiscards, "InputDiscards");
6349	BGE_SYSCTL_STAT(sc, ctx, "Input Errors",
6350	    children, ifInErrors, "InputErrors");
6351	BGE_SYSCTL_STAT(sc, ctx, "NIC Recv Threshold Hit",
6352	    children, nicRecvThresholdHit, "RecvThresholdHit");
6353	BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Read Queue Full",
6354	    children, nicDmaReadQueueFull, "DmaReadQueueFull");
6355	BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Read High Priority Queue Full",
6356	    children, nicDmaReadHighPriQueueFull, "DmaReadHighPriQueueFull");
6357	BGE_SYSCTL_STAT(sc, ctx, "NIC Send Data Complete Queue Full",
6358	    children, nicSendDataCompQueueFull, "SendDataCompQueueFull");
6359	BGE_SYSCTL_STAT(sc, ctx, "NIC Ring Set Send Producer Index",
6360	    children, nicRingSetSendProdIndex, "RingSetSendProdIndex");
6361	BGE_SYSCTL_STAT(sc, ctx, "NIC Ring Status Update",
6362	    children, nicRingStatusUpdate, "RingStatusUpdate");
6363	BGE_SYSCTL_STAT(sc, ctx, "NIC Interrupts",
6364	    children, nicInterrupts, "Interrupts");
6365	BGE_SYSCTL_STAT(sc, ctx, "NIC Avoided Interrupts",
6366	    children, nicAvoidedInterrupts, "AvoidedInterrupts");
6367	BGE_SYSCTL_STAT(sc, ctx, "NIC Send Threshold Hit",
6368	    children, nicSendThresholdHit, "SendThresholdHit");
6369
6370	tree = SYSCTL_ADD_NODE(ctx, schildren, OID_AUTO, "rx", CTLFLAG_RD,
6371	    NULL, "BGE RX Statistics");
6372	children = SYSCTL_CHILDREN(tree);
6373	BGE_SYSCTL_STAT(sc, ctx, "Inbound Octets",
6374	    children, rxstats.ifHCInOctets, "ifHCInOctets");
6375	BGE_SYSCTL_STAT(sc, ctx, "Fragments",
6376	    children, rxstats.etherStatsFragments, "Fragments");
6377	BGE_SYSCTL_STAT(sc, ctx, "Inbound Unicast Packets",
6378	    children, rxstats.ifHCInUcastPkts, "UnicastPkts");
6379	BGE_SYSCTL_STAT(sc, ctx, "Inbound Multicast Packets",
6380	    children, rxstats.ifHCInMulticastPkts, "MulticastPkts");
6381	BGE_SYSCTL_STAT(sc, ctx, "FCS Errors",
6382	    children, rxstats.dot3StatsFCSErrors, "FCSErrors");
6383	BGE_SYSCTL_STAT(sc, ctx, "Alignment Errors",
6384	    children, rxstats.dot3StatsAlignmentErrors, "AlignmentErrors");
6385	BGE_SYSCTL_STAT(sc, ctx, "XON Pause Frames Received",
6386	    children, rxstats.xonPauseFramesReceived, "xonPauseFramesReceived");
6387	BGE_SYSCTL_STAT(sc, ctx, "XOFF Pause Frames Received",
6388	    children, rxstats.xoffPauseFramesReceived,
6389	    "xoffPauseFramesReceived");
6390	BGE_SYSCTL_STAT(sc, ctx, "MAC Control Frames Received",
6391	    children, rxstats.macControlFramesReceived,
6392	    "ControlFramesReceived");
6393	BGE_SYSCTL_STAT(sc, ctx, "XOFF State Entered",
6394	    children, rxstats.xoffStateEntered, "xoffStateEntered");
6395	BGE_SYSCTL_STAT(sc, ctx, "Frames Too Long",
6396	    children, rxstats.dot3StatsFramesTooLong, "FramesTooLong");
6397	BGE_SYSCTL_STAT(sc, ctx, "Jabbers",
6398	    children, rxstats.etherStatsJabbers, "Jabbers");
6399	BGE_SYSCTL_STAT(sc, ctx, "Undersized Packets",
6400	    children, rxstats.etherStatsUndersizePkts, "UndersizePkts");
6401	BGE_SYSCTL_STAT(sc, ctx, "Inbound Range Length Errors",
6402	    children, rxstats.inRangeLengthError, "inRangeLengthError");
6403	BGE_SYSCTL_STAT(sc, ctx, "Outbound Range Length Errors",
6404	    children, rxstats.outRangeLengthError, "outRangeLengthError");
6405
6406	tree = SYSCTL_ADD_NODE(ctx, schildren, OID_AUTO, "tx", CTLFLAG_RD,
6407	    NULL, "BGE TX Statistics");
6408	children = SYSCTL_CHILDREN(tree);
6409	BGE_SYSCTL_STAT(sc, ctx, "Outbound Octets",
6410	    children, txstats.ifHCOutOctets, "ifHCOutOctets");
6411	BGE_SYSCTL_STAT(sc, ctx, "TX Collisions",
6412	    children, txstats.etherStatsCollisions, "Collisions");
6413	BGE_SYSCTL_STAT(sc, ctx, "XON Sent",
6414	    children, txstats.outXonSent, "XonSent");
6415	BGE_SYSCTL_STAT(sc, ctx, "XOFF Sent",
6416	    children, txstats.outXoffSent, "XoffSent");
6417	BGE_SYSCTL_STAT(sc, ctx, "Flow Control Done",
6418	    children, txstats.flowControlDone, "flowControlDone");
6419	BGE_SYSCTL_STAT(sc, ctx, "Internal MAC TX errors",
6420	    children, txstats.dot3StatsInternalMacTransmitErrors,
6421	    "InternalMacTransmitErrors");
6422	BGE_SYSCTL_STAT(sc, ctx, "Single Collision Frames",
6423	    children, txstats.dot3StatsSingleCollisionFrames,
6424	    "SingleCollisionFrames");
6425	BGE_SYSCTL_STAT(sc, ctx, "Multiple Collision Frames",
6426	    children, txstats.dot3StatsMultipleCollisionFrames,
6427	    "MultipleCollisionFrames");
6428	BGE_SYSCTL_STAT(sc, ctx, "Deferred Transmissions",
6429	    children, txstats.dot3StatsDeferredTransmissions,
6430	    "DeferredTransmissions");
6431	BGE_SYSCTL_STAT(sc, ctx, "Excessive Collisions",
6432	    children, txstats.dot3StatsExcessiveCollisions,
6433	    "ExcessiveCollisions");
6434	BGE_SYSCTL_STAT(sc, ctx, "Late Collisions",
6435	    children, txstats.dot3StatsLateCollisions,
6436	    "LateCollisions");
6437	BGE_SYSCTL_STAT(sc, ctx, "Outbound Unicast Packets",
6438	    children, txstats.ifHCOutUcastPkts, "UnicastPkts");
6439	BGE_SYSCTL_STAT(sc, ctx, "Outbound Multicast Packets",
6440	    children, txstats.ifHCOutMulticastPkts, "MulticastPkts");
6441	BGE_SYSCTL_STAT(sc, ctx, "Outbound Broadcast Packets",
6442	    children, txstats.ifHCOutBroadcastPkts, "BroadcastPkts");
6443	BGE_SYSCTL_STAT(sc, ctx, "Carrier Sense Errors",
6444	    children, txstats.dot3StatsCarrierSenseErrors,
6445	    "CarrierSenseErrors");
6446	BGE_SYSCTL_STAT(sc, ctx, "Outbound Discards",
6447	    children, txstats.ifOutDiscards, "Discards");
6448	BGE_SYSCTL_STAT(sc, ctx, "Outbound Errors",
6449	    children, txstats.ifOutErrors, "Errors");
6450}
6451
6452#undef BGE_SYSCTL_STAT
6453
6454#define	BGE_SYSCTL_STAT_ADD64(c, h, n, p, d)	\
6455	    SYSCTL_ADD_UQUAD(c, h, OID_AUTO, n, CTLFLAG_RD, p, d)
6456
6457static void
6458bge_add_sysctl_stats_regs(struct bge_softc *sc, struct sysctl_ctx_list *ctx,
6459    struct sysctl_oid_list *parent)
6460{
6461	struct sysctl_oid *tree;
6462	struct sysctl_oid_list *child, *schild;
6463	struct bge_mac_stats *stats;
6464
6465	stats = &sc->bge_mac_stats;
6466	tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "stats", CTLFLAG_RD,
6467	    NULL, "BGE Statistics");
6468	schild = child = SYSCTL_CHILDREN(tree);
6469	BGE_SYSCTL_STAT_ADD64(ctx, child, "FramesDroppedDueToFilters",
6470	    &stats->FramesDroppedDueToFilters, "Frames Dropped Due to Filters");
6471	BGE_SYSCTL_STAT_ADD64(ctx, child, "DmaWriteQueueFull",
6472	    &stats->DmaWriteQueueFull, "NIC DMA Write Queue Full");
6473	BGE_SYSCTL_STAT_ADD64(ctx, child, "DmaWriteHighPriQueueFull",
6474	    &stats->DmaWriteHighPriQueueFull,
6475	    "NIC DMA Write High Priority Queue Full");
6476	BGE_SYSCTL_STAT_ADD64(ctx, child, "NoMoreRxBDs",
6477	    &stats->NoMoreRxBDs, "NIC No More RX Buffer Descriptors");
6478	BGE_SYSCTL_STAT_ADD64(ctx, child, "InputDiscards",
6479	    &stats->InputDiscards, "Discarded Input Frames");
6480	BGE_SYSCTL_STAT_ADD64(ctx, child, "InputErrors",
6481	    &stats->InputErrors, "Input Errors");
6482	BGE_SYSCTL_STAT_ADD64(ctx, child, "RecvThresholdHit",
6483	    &stats->RecvThresholdHit, "NIC Recv Threshold Hit");
6484
6485	tree = SYSCTL_ADD_NODE(ctx, schild, OID_AUTO, "rx", CTLFLAG_RD,
6486	    NULL, "BGE RX Statistics");
6487	child = SYSCTL_CHILDREN(tree);
6488	BGE_SYSCTL_STAT_ADD64(ctx, child, "ifHCInOctets",
6489	    &stats->ifHCInOctets, "Inbound Octets");
6490	BGE_SYSCTL_STAT_ADD64(ctx, child, "Fragments",
6491	    &stats->etherStatsFragments, "Fragments");
6492	BGE_SYSCTL_STAT_ADD64(ctx, child, "UnicastPkts",
6493	    &stats->ifHCInUcastPkts, "Inbound Unicast Packets");
6494	BGE_SYSCTL_STAT_ADD64(ctx, child, "MulticastPkts",
6495	    &stats->ifHCInMulticastPkts, "Inbound Multicast Packets");
6496	BGE_SYSCTL_STAT_ADD64(ctx, child, "BroadcastPkts",
6497	    &stats->ifHCInBroadcastPkts, "Inbound Broadcast Packets");
6498	BGE_SYSCTL_STAT_ADD64(ctx, child, "FCSErrors",
6499	    &stats->dot3StatsFCSErrors, "FCS Errors");
6500	BGE_SYSCTL_STAT_ADD64(ctx, child, "AlignmentErrors",
6501	    &stats->dot3StatsAlignmentErrors, "Alignment Errors");
6502	BGE_SYSCTL_STAT_ADD64(ctx, child, "xonPauseFramesReceived",
6503	    &stats->xonPauseFramesReceived, "XON Pause Frames Received");
6504	BGE_SYSCTL_STAT_ADD64(ctx, child, "xoffPauseFramesReceived",
6505	    &stats->xoffPauseFramesReceived, "XOFF Pause Frames Received");
6506	BGE_SYSCTL_STAT_ADD64(ctx, child, "ControlFramesReceived",
6507	    &stats->macControlFramesReceived, "MAC Control Frames Received");
6508	BGE_SYSCTL_STAT_ADD64(ctx, child, "xoffStateEntered",
6509	    &stats->xoffStateEntered, "XOFF State Entered");
6510	BGE_SYSCTL_STAT_ADD64(ctx, child, "FramesTooLong",
6511	    &stats->dot3StatsFramesTooLong, "Frames Too Long");
6512	BGE_SYSCTL_STAT_ADD64(ctx, child, "Jabbers",
6513	    &stats->etherStatsJabbers, "Jabbers");
6514	BGE_SYSCTL_STAT_ADD64(ctx, child, "UndersizePkts",
6515	    &stats->etherStatsUndersizePkts, "Undersized Packets");
6516
6517	tree = SYSCTL_ADD_NODE(ctx, schild, OID_AUTO, "tx", CTLFLAG_RD,
6518	    NULL, "BGE TX Statistics");
6519	child = SYSCTL_CHILDREN(tree);
6520	BGE_SYSCTL_STAT_ADD64(ctx, child, "ifHCOutOctets",
6521	    &stats->ifHCOutOctets, "Outbound Octets");
6522	BGE_SYSCTL_STAT_ADD64(ctx, child, "Collisions",
6523	    &stats->etherStatsCollisions, "TX Collisions");
6524	BGE_SYSCTL_STAT_ADD64(ctx, child, "XonSent",
6525	    &stats->outXonSent, "XON Sent");
6526	BGE_SYSCTL_STAT_ADD64(ctx, child, "XoffSent",
6527	    &stats->outXoffSent, "XOFF Sent");
6528	BGE_SYSCTL_STAT_ADD64(ctx, child, "InternalMacTransmitErrors",
6529	    &stats->dot3StatsInternalMacTransmitErrors,
6530	    "Internal MAC TX Errors");
6531	BGE_SYSCTL_STAT_ADD64(ctx, child, "SingleCollisionFrames",
6532	    &stats->dot3StatsSingleCollisionFrames, "Single Collision Frames");
6533	BGE_SYSCTL_STAT_ADD64(ctx, child, "MultipleCollisionFrames",
6534	    &stats->dot3StatsMultipleCollisionFrames,
6535	    "Multiple Collision Frames");
6536	BGE_SYSCTL_STAT_ADD64(ctx, child, "DeferredTransmissions",
6537	    &stats->dot3StatsDeferredTransmissions, "Deferred Transmissions");
6538	BGE_SYSCTL_STAT_ADD64(ctx, child, "ExcessiveCollisions",
6539	    &stats->dot3StatsExcessiveCollisions, "Excessive Collisions");
6540	BGE_SYSCTL_STAT_ADD64(ctx, child, "LateCollisions",
6541	    &stats->dot3StatsLateCollisions, "Late Collisions");
6542	BGE_SYSCTL_STAT_ADD64(ctx, child, "UnicastPkts",
6543	    &stats->ifHCOutUcastPkts, "Outbound Unicast Packets");
6544	BGE_SYSCTL_STAT_ADD64(ctx, child, "MulticastPkts",
6545	    &stats->ifHCOutMulticastPkts, "Outbound Multicast Packets");
6546	BGE_SYSCTL_STAT_ADD64(ctx, child, "BroadcastPkts",
6547	    &stats->ifHCOutBroadcastPkts, "Outbound Broadcast Packets");
6548}
6549
6550#undef	BGE_SYSCTL_STAT_ADD64
6551
6552static int
6553bge_sysctl_stats(SYSCTL_HANDLER_ARGS)
6554{
6555	struct bge_softc *sc;
6556	uint32_t result;
6557	int offset;
6558
6559	sc = (struct bge_softc *)arg1;
6560	offset = arg2;
6561	result = CSR_READ_4(sc, BGE_MEMWIN_START + BGE_STATS_BLOCK + offset +
6562	    offsetof(bge_hostaddr, bge_addr_lo));
6563	return (sysctl_handle_int(oidp, &result, 0, req));
6564}
6565
6566#ifdef BGE_REGISTER_DEBUG
6567static int
6568bge_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
6569{
6570	struct bge_softc *sc;
6571	uint16_t *sbdata;
6572	int error, result, sbsz;
6573	int i, j;
6574
6575	result = -1;
6576	error = sysctl_handle_int(oidp, &result, 0, req);
6577	if (error || (req->newptr == NULL))
6578		return (error);
6579
6580	if (result == 1) {
6581		sc = (struct bge_softc *)arg1;
6582
6583		if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
6584		    sc->bge_chipid != BGE_CHIPID_BCM5700_C0)
6585			sbsz = BGE_STATUS_BLK_SZ;
6586		else
6587			sbsz = 32;
6588		sbdata = (uint16_t *)sc->bge_ldata.bge_status_block;
6589		printf("Status Block:\n");
6590		BGE_LOCK(sc);
6591		bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
6592		    sc->bge_cdata.bge_status_map,
6593		    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
6594		for (i = 0x0; i < sbsz / sizeof(uint16_t); ) {
6595			printf("%06x:", i);
6596			for (j = 0; j < 8; j++)
6597				printf(" %04x", sbdata[i++]);
6598			printf("\n");
6599		}
6600
6601		printf("Registers:\n");
6602		for (i = 0x800; i < 0xA00; ) {
6603			printf("%06x:", i);
6604			for (j = 0; j < 8; j++) {
6605				printf(" %08x", CSR_READ_4(sc, i));
6606				i += 4;
6607			}
6608			printf("\n");
6609		}
6610		BGE_UNLOCK(sc);
6611
6612		printf("Hardware Flags:\n");
6613		if (BGE_IS_5717_PLUS(sc))
6614			printf(" - 5717 Plus\n");
6615		if (BGE_IS_5755_PLUS(sc))
6616			printf(" - 5755 Plus\n");
6617		if (BGE_IS_575X_PLUS(sc))
6618			printf(" - 575X Plus\n");
6619		if (BGE_IS_5705_PLUS(sc))
6620			printf(" - 5705 Plus\n");
6621		if (BGE_IS_5714_FAMILY(sc))
6622			printf(" - 5714 Family\n");
6623		if (BGE_IS_5700_FAMILY(sc))
6624			printf(" - 5700 Family\n");
6625		if (sc->bge_flags & BGE_FLAG_JUMBO)
6626			printf(" - Supports Jumbo Frames\n");
6627		if (sc->bge_flags & BGE_FLAG_PCIX)
6628			printf(" - PCI-X Bus\n");
6629		if (sc->bge_flags & BGE_FLAG_PCIE)
6630			printf(" - PCI Express Bus\n");
6631		if (sc->bge_phy_flags & BGE_PHY_NO_3LED)
6632			printf(" - No 3 LEDs\n");
6633		if (sc->bge_flags & BGE_FLAG_RX_ALIGNBUG)
6634			printf(" - RX Alignment Bug\n");
6635	}
6636
6637	return (error);
6638}
6639
6640static int
6641bge_sysctl_reg_read(SYSCTL_HANDLER_ARGS)
6642{
6643	struct bge_softc *sc;
6644	int error;
6645	uint16_t result;
6646	uint32_t val;
6647
6648	result = -1;
6649	error = sysctl_handle_int(oidp, &result, 0, req);
6650	if (error || (req->newptr == NULL))
6651		return (error);
6652
6653	if (result < 0x8000) {
6654		sc = (struct bge_softc *)arg1;
6655		val = CSR_READ_4(sc, result);
6656		printf("reg 0x%06X = 0x%08X\n", result, val);
6657	}
6658
6659	return (error);
6660}
6661
6662static int
6663bge_sysctl_ape_read(SYSCTL_HANDLER_ARGS)
6664{
6665	struct bge_softc *sc;
6666	int error;
6667	uint16_t result;
6668	uint32_t val;
6669
6670	result = -1;
6671	error = sysctl_handle_int(oidp, &result, 0, req);
6672	if (error || (req->newptr == NULL))
6673		return (error);
6674
6675	if (result < 0x8000) {
6676		sc = (struct bge_softc *)arg1;
6677		val = APE_READ_4(sc, result);
6678		printf("reg 0x%06X = 0x%08X\n", result, val);
6679	}
6680
6681	return (error);
6682}
6683
6684static int
6685bge_sysctl_mem_read(SYSCTL_HANDLER_ARGS)
6686{
6687	struct bge_softc *sc;
6688	int error;
6689	uint16_t result;
6690	uint32_t val;
6691
6692	result = -1;
6693	error = sysctl_handle_int(oidp, &result, 0, req);
6694	if (error || (req->newptr == NULL))
6695		return (error);
6696
6697	if (result < 0x8000) {
6698		sc = (struct bge_softc *)arg1;
6699		val = bge_readmem_ind(sc, result);
6700		printf("mem 0x%06X = 0x%08X\n", result, val);
6701	}
6702
6703	return (error);
6704}
6705#endif
6706
6707static int
6708bge_get_eaddr_fw(struct bge_softc *sc, uint8_t ether_addr[])
6709{
6710
6711	if (sc->bge_flags & BGE_FLAG_EADDR)
6712		return (1);
6713
6714#ifdef __sparc64__
6715	OF_getetheraddr(sc->bge_dev, ether_addr);
6716	return (0);
6717#endif
6718	return (1);
6719}
6720
6721static int
6722bge_get_eaddr_mem(struct bge_softc *sc, uint8_t ether_addr[])
6723{
6724	uint32_t mac_addr;
6725
6726	mac_addr = bge_readmem_ind(sc, BGE_SRAM_MAC_ADDR_HIGH_MB);
6727	if ((mac_addr >> 16) == 0x484b) {
6728		ether_addr[0] = (uint8_t)(mac_addr >> 8);
6729		ether_addr[1] = (uint8_t)mac_addr;
6730		mac_addr = bge_readmem_ind(sc, BGE_SRAM_MAC_ADDR_LOW_MB);
6731		ether_addr[2] = (uint8_t)(mac_addr >> 24);
6732		ether_addr[3] = (uint8_t)(mac_addr >> 16);
6733		ether_addr[4] = (uint8_t)(mac_addr >> 8);
6734		ether_addr[5] = (uint8_t)mac_addr;
6735		return (0);
6736	}
6737	return (1);
6738}
6739
6740static int
6741bge_get_eaddr_nvram(struct bge_softc *sc, uint8_t ether_addr[])
6742{
6743	int mac_offset = BGE_EE_MAC_OFFSET;
6744
6745	if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
6746		mac_offset = BGE_EE_MAC_OFFSET_5906;
6747
6748	return (bge_read_nvram(sc, ether_addr, mac_offset + 2,
6749	    ETHER_ADDR_LEN));
6750}
6751
6752static int
6753bge_get_eaddr_eeprom(struct bge_softc *sc, uint8_t ether_addr[])
6754{
6755
6756	if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
6757		return (1);
6758
6759	return (bge_read_eeprom(sc, ether_addr, BGE_EE_MAC_OFFSET + 2,
6760	   ETHER_ADDR_LEN));
6761}
6762
6763static int
6764bge_get_eaddr(struct bge_softc *sc, uint8_t eaddr[])
6765{
6766	static const bge_eaddr_fcn_t bge_eaddr_funcs[] = {
6767		/* NOTE: Order is critical */
6768		bge_get_eaddr_fw,
6769		bge_get_eaddr_mem,
6770		bge_get_eaddr_nvram,
6771		bge_get_eaddr_eeprom,
6772		NULL
6773	};
6774	const bge_eaddr_fcn_t *func;
6775
6776	for (func = bge_eaddr_funcs; *func != NULL; ++func) {
6777		if ((*func)(sc, eaddr) == 0)
6778			break;
6779	}
6780	return (*func == NULL ? ENXIO : 0);
6781}
6782