nvme_private.h revision 248763
1/*-
2 * Copyright (C) 2012 Intel Corporation
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: head/sys/dev/nvme/nvme_private.h 248763 2013-03-26 21:19:26Z jimharris $
27 */
28
29#ifndef __NVME_PRIVATE_H__
30#define __NVME_PRIVATE_H__
31
32#include <sys/param.h>
33#include <sys/kernel.h>
34#include <sys/lock.h>
35#include <sys/malloc.h>
36#include <sys/mutex.h>
37#include <sys/rman.h>
38#include <sys/systm.h>
39#include <sys/taskqueue.h>
40
41#include <vm/uma.h>
42
43#include <machine/bus.h>
44
45#include "nvme.h"
46
47#define DEVICE2SOFTC(dev) ((struct nvme_controller *) device_get_softc(dev))
48
49MALLOC_DECLARE(M_NVME);
50
51#define CHATHAM2
52
53#ifdef CHATHAM2
54#define CHATHAM_PCI_ID		0x20118086
55#define CHATHAM_CONTROL_BAR	0
56#endif
57
58#define IDT32_PCI_ID		0x80d0111d /* 32 channel board */
59#define IDT8_PCI_ID		0x80d2111d /* 8 channel board */
60
61#define NVME_MAX_PRP_LIST_ENTRIES	(32)
62
63/*
64 * For commands requiring more than 2 PRP entries, one PRP will be
65 *  embedded in the command (prp1), and the rest of the PRP entries
66 *  will be in a list pointed to by the command (prp2).  This means
67 *  that real max number of PRP entries we support is 32+1, which
68 *  results in a max xfer size of 32*PAGE_SIZE.
69 */
70#define NVME_MAX_XFER_SIZE	NVME_MAX_PRP_LIST_ENTRIES * PAGE_SIZE
71
72#define NVME_ADMIN_TRACKERS	(16)
73#define NVME_ADMIN_ENTRIES	(128)
74/* min and max are defined in admin queue attributes section of spec */
75#define NVME_MIN_ADMIN_ENTRIES	(2)
76#define NVME_MAX_ADMIN_ENTRIES	(4096)
77
78/*
79 * NVME_IO_ENTRIES defines the size of an I/O qpair's submission and completion
80 *  queues, while NVME_IO_TRACKERS defines the maximum number of I/O that we
81 *  will allow outstanding on an I/O qpair at any time.  The only advantage in
82 *  having IO_ENTRIES > IO_TRACKERS is for debugging purposes - when dumping
83 *  the contents of the submission and completion queues, it will show a longer
84 *  history of data.
85 */
86#define NVME_IO_ENTRIES		(256)
87#define NVME_IO_TRACKERS	(128)
88#define NVME_MIN_IO_TRACKERS	(16)
89#define NVME_MAX_IO_TRACKERS	(1024)
90
91/*
92 * NVME_MAX_IO_ENTRIES is not defined, since it is specified in CC.MQES
93 *  for each controller.
94 */
95
96#define NVME_INT_COAL_TIME	(0)	/* disabled */
97#define NVME_INT_COAL_THRESHOLD (0)	/* 0-based */
98
99#define NVME_MAX_NAMESPACES	(16)
100#define NVME_MAX_CONSUMERS	(2)
101#define NVME_MAX_ASYNC_EVENTS	(8)
102
103#define NVME_DEFAULT_TIMEOUT_PERIOD	(30)    /* in seconds */
104#define NVME_MIN_TIMEOUT_PERIOD		(5)
105#define NVME_MAX_TIMEOUT_PERIOD		(120)
106
107#define NVME_DEFAULT_RETRY_COUNT	(4)
108
109/* Maximum log page size to fetch for AERs. */
110#define NVME_MAX_AER_LOG_SIZE		(4096)
111
112#ifndef CACHE_LINE_SIZE
113#define CACHE_LINE_SIZE		(64)
114#endif
115
116extern uma_zone_t	nvme_request_zone;
117extern int32_t		nvme_retry_count;
118
119struct nvme_request {
120
121	struct nvme_command		cmd;
122	void				*payload;
123	uint32_t			payload_size;
124	boolean_t			timeout;
125	struct uio			*uio;
126	nvme_cb_fn_t			cb_fn;
127	void				*cb_arg;
128	int32_t				retries;
129	STAILQ_ENTRY(nvme_request)	stailq;
130};
131
132struct nvme_async_event_request {
133
134	struct nvme_controller		*ctrlr;
135	struct nvme_request		*req;
136	struct nvme_completion		cpl;
137	uint32_t			log_page_id;
138	uint32_t			log_page_size;
139	uint8_t				log_page_buffer[NVME_MAX_AER_LOG_SIZE];
140};
141
142struct nvme_tracker {
143
144	TAILQ_ENTRY(nvme_tracker)	tailq;
145	struct nvme_request		*req;
146	struct nvme_qpair		*qpair;
147	struct callout			timer;
148	bus_dmamap_t			payload_dma_map;
149	uint16_t			cid;
150
151	uint64_t			prp[NVME_MAX_PRP_LIST_ENTRIES];
152	bus_addr_t			prp_bus_addr;
153	bus_dmamap_t			prp_dma_map;
154};
155
156struct nvme_qpair {
157
158	struct nvme_controller	*ctrlr;
159	uint32_t		id;
160	uint32_t		phase;
161
162	uint16_t		vector;
163	int			rid;
164	struct resource		*res;
165	void 			*tag;
166
167	uint32_t		max_xfer_size;
168	uint32_t		num_entries;
169	uint32_t		num_trackers;
170	uint32_t		sq_tdbl_off;
171	uint32_t		cq_hdbl_off;
172
173	uint32_t		sq_head;
174	uint32_t		sq_tail;
175	uint32_t		cq_head;
176
177	int64_t			num_cmds;
178	int64_t			num_intr_handler_calls;
179
180	struct nvme_command	*cmd;
181	struct nvme_completion	*cpl;
182
183	bus_dma_tag_t		dma_tag;
184
185	bus_dmamap_t		cmd_dma_map;
186	uint64_t		cmd_bus_addr;
187
188	bus_dmamap_t		cpl_dma_map;
189	uint64_t		cpl_bus_addr;
190
191	TAILQ_HEAD(, nvme_tracker)	free_tr;
192	TAILQ_HEAD(, nvme_tracker)	outstanding_tr;
193	STAILQ_HEAD(, nvme_request)	queued_req;
194
195	struct nvme_tracker	**act_tr;
196
197	boolean_t		is_enabled;
198
199	struct mtx		lock __aligned(CACHE_LINE_SIZE);
200
201} __aligned(CACHE_LINE_SIZE);
202
203struct nvme_namespace {
204
205	struct nvme_controller		*ctrlr;
206	struct nvme_namespace_data	data;
207	uint16_t			id;
208	uint16_t			flags;
209	struct cdev			*cdev;
210	void				*cons_cookie[NVME_MAX_CONSUMERS];
211};
212
213/*
214 * One of these per allocated PCI device.
215 */
216struct nvme_controller {
217
218	device_t		dev;
219
220	uint32_t		ready_timeout_in_ms;
221
222	bus_space_tag_t		bus_tag;
223	bus_space_handle_t	bus_handle;
224	int			resource_id;
225	struct resource		*resource;
226
227	/*
228	 * The NVMe spec allows for the MSI-X table to be placed in BAR 4/5,
229	 *  separate from the control registers which are in BAR 0/1.  These
230	 *  members track the mapping of BAR 4/5 for that reason.
231	 */
232	int			bar4_resource_id;
233	struct resource		*bar4_resource;
234
235#ifdef CHATHAM2
236	bus_space_tag_t		chatham_bus_tag;
237	bus_space_handle_t	chatham_bus_handle;
238	int			chatham_resource_id;
239	struct resource		*chatham_resource;
240#endif
241
242	uint32_t		msix_enabled;
243	uint32_t		force_intx;
244	uint32_t		enable_aborts;
245
246	uint32_t		num_io_queues;
247	boolean_t		per_cpu_io_queues;
248
249	/* Fields for tracking progress during controller initialization. */
250	struct intr_config_hook	config_hook;
251	uint32_t		ns_identified;
252	uint32_t		queues_created;
253	struct task		reset_task;
254	struct taskqueue	*taskqueue;
255
256	/* For shared legacy interrupt. */
257	int			rid;
258	struct resource		*res;
259	void			*tag;
260
261	bus_dma_tag_t		hw_desc_tag;
262	bus_dmamap_t		hw_desc_map;
263
264	/** maximum i/o size in bytes */
265	uint32_t		max_xfer_size;
266
267	/** minimum page size supported by this controller in bytes */
268	uint32_t		min_page_size;
269
270	/** interrupt coalescing time period (in microseconds) */
271	uint32_t		int_coal_time;
272
273	/** interrupt coalescing threshold */
274	uint32_t		int_coal_threshold;
275
276	/** timeout period in seconds */
277	uint32_t		timeout_period;
278
279	struct nvme_qpair	adminq;
280	struct nvme_qpair	*ioq;
281
282	struct nvme_registers		*regs;
283
284	struct nvme_controller_data	cdata;
285	struct nvme_namespace		ns[NVME_MAX_NAMESPACES];
286
287	struct cdev			*cdev;
288
289	uint32_t			num_aers;
290	struct nvme_async_event_request	aer[NVME_MAX_ASYNC_EVENTS];
291
292	void				*cons_cookie[NVME_MAX_CONSUMERS];
293
294	uint32_t		is_resetting;
295
296#ifdef CHATHAM2
297	uint64_t		chatham_size;
298	uint64_t		chatham_lbas;
299#endif
300};
301
302#define nvme_mmio_offsetof(reg)						       \
303	offsetof(struct nvme_registers, reg)
304
305#define nvme_mmio_read_4(sc, reg)					       \
306	bus_space_read_4((sc)->bus_tag, (sc)->bus_handle,		       \
307	    nvme_mmio_offsetof(reg))
308
309#define nvme_mmio_write_4(sc, reg, val)					       \
310	bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,		       \
311	    nvme_mmio_offsetof(reg), val)
312
313#define nvme_mmio_write_8(sc, reg, val) \
314	do {								       \
315		bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,	       \
316		    nvme_mmio_offsetof(reg), val & 0xFFFFFFFF); 	       \
317		bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,	       \
318		    nvme_mmio_offsetof(reg)+4,				       \
319		    (val & 0xFFFFFFFF00000000UL) >> 32);		       \
320	} while (0);
321
322#ifdef CHATHAM2
323#define chatham_read_4(softc, reg) \
324	bus_space_read_4((softc)->chatham_bus_tag,			       \
325	    (softc)->chatham_bus_handle, reg)
326
327#define chatham_write_8(sc, reg, val)					       \
328	do {								       \
329		bus_space_write_4((sc)->chatham_bus_tag,		       \
330		    (sc)->chatham_bus_handle, reg, val & 0xffffffff);	       \
331		bus_space_write_4((sc)->chatham_bus_tag,		       \
332		    (sc)->chatham_bus_handle, reg+4,			       \
333		    (val & 0xFFFFFFFF00000000UL) >> 32);		       \
334	} while (0);
335
336#endif /* CHATHAM2 */
337
338#if __FreeBSD_version < 800054
339#define wmb()	__asm volatile("sfence" ::: "memory")
340#define mb()	__asm volatile("mfence" ::: "memory")
341#endif
342
343void	nvme_ns_test(struct nvme_namespace *ns, u_long cmd, caddr_t arg);
344
345void	nvme_ctrlr_cmd_identify_controller(struct nvme_controller *ctrlr,
346					   void *payload,
347					   nvme_cb_fn_t cb_fn, void *cb_arg);
348void	nvme_ctrlr_cmd_identify_namespace(struct nvme_controller *ctrlr,
349					  uint16_t nsid, void *payload,
350					  nvme_cb_fn_t cb_fn, void *cb_arg);
351void	nvme_ctrlr_cmd_set_interrupt_coalescing(struct nvme_controller *ctrlr,
352						uint32_t microseconds,
353						uint32_t threshold,
354						nvme_cb_fn_t cb_fn,
355						void *cb_arg);
356void	nvme_ctrlr_cmd_get_error_page(struct nvme_controller *ctrlr,
357				      struct nvme_error_information_entry *payload,
358				      uint32_t num_entries, /* 0 = max */
359				      nvme_cb_fn_t cb_fn,
360				      void *cb_arg);
361void	nvme_ctrlr_cmd_get_health_information_page(struct nvme_controller *ctrlr,
362						   uint32_t nsid,
363						   struct nvme_health_information_page *payload,
364						   nvme_cb_fn_t cb_fn,
365						   void *cb_arg);
366void	nvme_ctrlr_cmd_get_firmware_page(struct nvme_controller *ctrlr,
367					 struct nvme_firmware_page *payload,
368					 nvme_cb_fn_t cb_fn,
369					 void *cb_arg);
370void	nvme_ctrlr_cmd_create_io_cq(struct nvme_controller *ctrlr,
371				    struct nvme_qpair *io_que, uint16_t vector,
372				    nvme_cb_fn_t cb_fn, void *cb_arg);
373void	nvme_ctrlr_cmd_create_io_sq(struct nvme_controller *ctrlr,
374				    struct nvme_qpair *io_que,
375				    nvme_cb_fn_t cb_fn, void *cb_arg);
376void	nvme_ctrlr_cmd_delete_io_cq(struct nvme_controller *ctrlr,
377				    struct nvme_qpair *io_que,
378				    nvme_cb_fn_t cb_fn, void *cb_arg);
379void	nvme_ctrlr_cmd_delete_io_sq(struct nvme_controller *ctrlr,
380				    struct nvme_qpair *io_que,
381				    nvme_cb_fn_t cb_fn, void *cb_arg);
382void	nvme_ctrlr_cmd_set_num_queues(struct nvme_controller *ctrlr,
383				      uint32_t num_queues, nvme_cb_fn_t cb_fn,
384				      void *cb_arg);
385void	nvme_ctrlr_cmd_set_async_event_config(struct nvme_controller *ctrlr,
386					      union nvme_critical_warning_state state,
387					      nvme_cb_fn_t cb_fn, void *cb_arg);
388void	nvme_ctrlr_cmd_abort(struct nvme_controller *ctrlr, uint16_t cid,
389			     uint16_t sqid, nvme_cb_fn_t cb_fn, void *cb_arg);
390
391void	nvme_payload_map(void *arg, bus_dma_segment_t *seg, int nseg,
392			 int error);
393void	nvme_payload_map_uio(void *arg, bus_dma_segment_t *seg, int nseg,
394			     bus_size_t mapsize, int error);
395
396int	nvme_ctrlr_construct(struct nvme_controller *ctrlr, device_t dev);
397void	nvme_ctrlr_destruct(struct nvme_controller *ctrlr, device_t dev);
398int	nvme_ctrlr_hw_reset(struct nvme_controller *ctrlr);
399void	nvme_ctrlr_reset(struct nvme_controller *ctrlr);
400/* ctrlr defined as void * to allow use with config_intrhook. */
401void	nvme_ctrlr_start_config_hook(void *ctrlr_arg);
402void	nvme_ctrlr_submit_admin_request(struct nvme_controller *ctrlr,
403					struct nvme_request *req);
404void	nvme_ctrlr_submit_io_request(struct nvme_controller *ctrlr,
405				     struct nvme_request *req);
406
407void	nvme_qpair_construct(struct nvme_qpair *qpair, uint32_t id,
408			     uint16_t vector, uint32_t num_entries,
409			     uint32_t num_trackers, uint32_t max_xfer_size,
410			     struct nvme_controller *ctrlr);
411void	nvme_qpair_submit_tracker(struct nvme_qpair *qpair,
412				  struct nvme_tracker *tr);
413void	nvme_qpair_process_completions(struct nvme_qpair *qpair);
414void	nvme_qpair_submit_request(struct nvme_qpair *qpair,
415				  struct nvme_request *req);
416void	nvme_qpair_reset(struct nvme_qpair *qpair);
417
418void	nvme_admin_qpair_enable(struct nvme_qpair *qpair);
419void	nvme_admin_qpair_disable(struct nvme_qpair *qpair);
420void	nvme_admin_qpair_destroy(struct nvme_qpair *qpair);
421
422void	nvme_io_qpair_enable(struct nvme_qpair *qpair);
423void	nvme_io_qpair_disable(struct nvme_qpair *qpair);
424void	nvme_io_qpair_destroy(struct nvme_qpair *qpair);
425
426int	nvme_ns_construct(struct nvme_namespace *ns, uint16_t id,
427			  struct nvme_controller *ctrlr);
428void	nvme_ns_destruct(struct nvme_namespace *ns);
429
430int	nvme_ns_physio(struct cdev *dev, struct uio *uio, int ioflag);
431
432void	nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr);
433
434void	nvme_dump_command(struct nvme_command *cmd);
435void	nvme_dump_completion(struct nvme_completion *cpl);
436
437static __inline void
438nvme_single_map(void *arg, bus_dma_segment_t *seg, int nseg, int error)
439{
440	uint64_t *bus_addr = (uint64_t *)arg;
441
442	*bus_addr = seg[0].ds_addr;
443}
444
445static __inline struct nvme_request *
446nvme_allocate_request(void *payload, uint32_t payload_size, nvme_cb_fn_t cb_fn,
447		      void *cb_arg)
448{
449	struct nvme_request *req;
450
451	req = uma_zalloc(nvme_request_zone, M_NOWAIT | M_ZERO);
452	if (req == NULL)
453		return (NULL);
454
455	req->payload = payload;
456	req->payload_size = payload_size;
457	req->cb_fn = cb_fn;
458	req->cb_arg = cb_arg;
459	req->timeout = TRUE;
460
461	return (req);
462}
463
464static __inline struct nvme_request *
465nvme_allocate_request_uio(struct uio *uio, nvme_cb_fn_t cb_fn, void *cb_arg)
466{
467	struct nvme_request *req;
468
469	req = uma_zalloc(nvme_request_zone, M_NOWAIT | M_ZERO);
470	if (req == NULL)
471		return (NULL);
472
473	req->uio = uio;
474	req->cb_fn = cb_fn;
475	req->cb_arg = cb_arg;
476	req->timeout = TRUE;
477
478	return (req);
479}
480
481#define nvme_free_request(req)	uma_zfree(nvme_request_zone, req)
482
483void	nvme_notify_async_consumers(struct nvme_controller *ctrlr,
484				    const struct nvme_completion *async_cpl,
485				    uint32_t log_page_id, void *log_page_buffer,
486				    uint32_t log_page_size);
487
488#endif /* __NVME_PRIVATE_H__ */
489