nvme_private.h revision 248735
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 248735 2013-03-26 18:31:46Z 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
40#include <vm/uma.h>
41
42#include <machine/bus.h>
43
44#include "nvme.h"
45
46#define DEVICE2SOFTC(dev) ((struct nvme_controller *) device_get_softc(dev))
47
48MALLOC_DECLARE(M_NVME);
49
50#define CHATHAM2
51
52#ifdef CHATHAM2
53#define CHATHAM_PCI_ID		0x20118086
54#define CHATHAM_CONTROL_BAR	0
55#endif
56
57#define IDT32_PCI_ID		0x80d0111d /* 32 channel board */
58#define IDT8_PCI_ID		0x80d2111d /* 8 channel board */
59
60#define NVME_MAX_PRP_LIST_ENTRIES	(32)
61
62/*
63 * For commands requiring more than 2 PRP entries, one PRP will be
64 *  embedded in the command (prp1), and the rest of the PRP entries
65 *  will be in a list pointed to by the command (prp2).  This means
66 *  that real max number of PRP entries we support is 32+1, which
67 *  results in a max xfer size of 32*PAGE_SIZE.
68 */
69#define NVME_MAX_XFER_SIZE	NVME_MAX_PRP_LIST_ENTRIES * PAGE_SIZE
70
71#define NVME_ADMIN_TRACKERS	(16)
72#define NVME_ADMIN_ENTRIES	(128)
73/* min and max are defined in admin queue attributes section of spec */
74#define NVME_MIN_ADMIN_ENTRIES	(2)
75#define NVME_MAX_ADMIN_ENTRIES	(4096)
76
77/*
78 * NVME_IO_ENTRIES defines the size of an I/O qpair's submission and completion
79 *  queues, while NVME_IO_TRACKERS defines the maximum number of I/O that we
80 *  will allow outstanding on an I/O qpair at any time.  The only advantage in
81 *  having IO_ENTRIES > IO_TRACKERS is for debugging purposes - when dumping
82 *  the contents of the submission and completion queues, it will show a longer
83 *  history of data.
84 */
85#define NVME_IO_ENTRIES		(256)
86#define NVME_IO_TRACKERS	(128)
87#define NVME_MIN_IO_TRACKERS	(16)
88#define NVME_MAX_IO_TRACKERS	(1024)
89
90/*
91 * NVME_MAX_IO_ENTRIES is not defined, since it is specified in CC.MQES
92 *  for each controller.
93 */
94
95#define NVME_INT_COAL_TIME	(0)	/* disabled */
96#define NVME_INT_COAL_THRESHOLD (0)	/* 0-based */
97
98#define NVME_MAX_NAMESPACES	(16)
99#define NVME_MAX_CONSUMERS	(2)
100#define NVME_MAX_ASYNC_EVENTS	(4)
101
102#define NVME_TIMEOUT_IN_SEC	(30)
103
104#ifndef CACHE_LINE_SIZE
105#define CACHE_LINE_SIZE		(64)
106#endif
107
108extern uma_zone_t nvme_request_zone;
109
110struct nvme_request {
111
112	struct nvme_command		cmd;
113	void				*payload;
114	uint32_t			payload_size;
115	uint32_t			timeout;
116	struct uio			*uio;
117	nvme_cb_fn_t			cb_fn;
118	void				*cb_arg;
119	STAILQ_ENTRY(nvme_request)	stailq;
120};
121
122struct nvme_tracker {
123
124	SLIST_ENTRY(nvme_tracker)	slist;
125	struct nvme_request		*req;
126	struct nvme_qpair		*qpair;
127	struct callout			timer;
128	bus_dmamap_t			payload_dma_map;
129	uint16_t			cid;
130
131	uint64_t			prp[NVME_MAX_PRP_LIST_ENTRIES];
132	bus_addr_t			prp_bus_addr;
133	bus_dmamap_t			prp_dma_map;
134};
135
136struct nvme_qpair {
137
138	struct nvme_controller	*ctrlr;
139	uint32_t		id;
140	uint32_t		phase;
141
142	uint16_t		vector;
143	int			rid;
144	struct resource		*res;
145	void 			*tag;
146
147	uint32_t		max_xfer_size;
148	uint32_t		num_entries;
149	uint32_t		num_trackers;
150	uint32_t		sq_tdbl_off;
151	uint32_t		cq_hdbl_off;
152
153	uint32_t		sq_head;
154	uint32_t		sq_tail;
155	uint32_t		cq_head;
156
157	int64_t			num_cmds;
158	int64_t			num_intr_handler_calls;
159
160	struct nvme_command	*cmd;
161	struct nvme_completion	*cpl;
162
163	bus_dma_tag_t		dma_tag;
164
165	bus_dmamap_t		cmd_dma_map;
166	uint64_t		cmd_bus_addr;
167
168	bus_dmamap_t		cpl_dma_map;
169	uint64_t		cpl_bus_addr;
170
171	SLIST_HEAD(, nvme_tracker)	free_tr;
172	STAILQ_HEAD(, nvme_request)	queued_req;
173
174	struct nvme_tracker	**act_tr;
175
176	struct mtx		lock __aligned(CACHE_LINE_SIZE);
177
178} __aligned(CACHE_LINE_SIZE);
179
180struct nvme_namespace {
181
182	struct nvme_controller		*ctrlr;
183	struct nvme_namespace_data	data;
184	uint16_t			id;
185	uint16_t			flags;
186	struct cdev			*cdev;
187};
188
189/*
190 * One of these per allocated PCI device.
191 */
192struct nvme_controller {
193
194	device_t		dev;
195
196	uint32_t		ready_timeout_in_ms;
197
198	bus_space_tag_t		bus_tag;
199	bus_space_handle_t	bus_handle;
200	int			resource_id;
201	struct resource		*resource;
202
203	/*
204	 * The NVMe spec allows for the MSI-X table to be placed in BAR 4/5,
205	 *  separate from the control registers which are in BAR 0/1.  These
206	 *  members track the mapping of BAR 4/5 for that reason.
207	 */
208	int			bar4_resource_id;
209	struct resource		*bar4_resource;
210
211#ifdef CHATHAM2
212	bus_space_tag_t		chatham_bus_tag;
213	bus_space_handle_t	chatham_bus_handle;
214	int			chatham_resource_id;
215	struct resource		*chatham_resource;
216#endif
217
218	uint32_t		msix_enabled;
219	uint32_t		force_intx;
220
221	uint32_t		num_io_queues;
222	boolean_t		per_cpu_io_queues;
223
224	/* Fields for tracking progress during controller initialization. */
225	struct intr_config_hook	config_hook;
226	uint32_t		ns_identified;
227	uint32_t		queues_created;
228
229	/* For shared legacy interrupt. */
230	int			rid;
231	struct resource		*res;
232	void			*tag;
233
234	bus_dma_tag_t		hw_desc_tag;
235	bus_dmamap_t		hw_desc_map;
236
237	/** maximum i/o size in bytes */
238	uint32_t		max_xfer_size;
239
240	/** interrupt coalescing time period (in microseconds) */
241	uint32_t		int_coal_time;
242
243	/** interrupt coalescing threshold */
244	uint32_t		int_coal_threshold;
245
246	struct nvme_qpair	adminq;
247	struct nvme_qpair	*ioq;
248
249	struct nvme_registers		*regs;
250
251	struct nvme_controller_data	cdata;
252	struct nvme_namespace		ns[NVME_MAX_NAMESPACES];
253
254	struct cdev			*cdev;
255
256	boolean_t			is_started;
257
258#ifdef CHATHAM2
259	uint64_t		chatham_size;
260	uint64_t		chatham_lbas;
261#endif
262};
263
264#define nvme_mmio_offsetof(reg)						       \
265	offsetof(struct nvme_registers, reg)
266
267#define nvme_mmio_read_4(sc, reg)					       \
268	bus_space_read_4((sc)->bus_tag, (sc)->bus_handle,		       \
269	    nvme_mmio_offsetof(reg))
270
271#define nvme_mmio_write_4(sc, reg, val)					       \
272	bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,		       \
273	    nvme_mmio_offsetof(reg), val)
274
275#define nvme_mmio_write_8(sc, reg, val) \
276	do {								       \
277		bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,	       \
278		    nvme_mmio_offsetof(reg), val & 0xFFFFFFFF); 	       \
279		bus_space_write_4((sc)->bus_tag, (sc)->bus_handle,	       \
280		    nvme_mmio_offsetof(reg)+4,				       \
281		    (val & 0xFFFFFFFF00000000UL) >> 32);		       \
282	} while (0);
283
284#ifdef CHATHAM2
285#define chatham_read_4(softc, reg) \
286	bus_space_read_4((softc)->chatham_bus_tag,			       \
287	    (softc)->chatham_bus_handle, reg)
288
289#define chatham_write_8(sc, reg, val)					       \
290	do {								       \
291		bus_space_write_4((sc)->chatham_bus_tag,		       \
292		    (sc)->chatham_bus_handle, reg, val & 0xffffffff);	       \
293		bus_space_write_4((sc)->chatham_bus_tag,		       \
294		    (sc)->chatham_bus_handle, reg+4,			       \
295		    (val & 0xFFFFFFFF00000000UL) >> 32);		       \
296	} while (0);
297
298#endif /* CHATHAM2 */
299
300#if __FreeBSD_version < 800054
301#define wmb()	__asm volatile("sfence" ::: "memory")
302#define mb()	__asm volatile("mfence" ::: "memory")
303#endif
304
305void	nvme_ns_test(struct nvme_namespace *ns, u_long cmd, caddr_t arg);
306
307void	nvme_ctrlr_cmd_set_feature(struct nvme_controller *ctrlr,
308				   uint8_t feature, uint32_t cdw11,
309				   void *payload, uint32_t payload_size,
310				   nvme_cb_fn_t cb_fn, void *cb_arg);
311void	nvme_ctrlr_cmd_get_feature(struct nvme_controller *ctrlr,
312				   uint8_t feature, uint32_t cdw11,
313				   void *payload, uint32_t payload_size,
314				   nvme_cb_fn_t cb_fn, void *cb_arg);
315void	nvme_ctrlr_cmd_identify_controller(struct nvme_controller *ctrlr,
316					   void *payload,
317					   nvme_cb_fn_t cb_fn, void *cb_arg);
318void	nvme_ctrlr_cmd_identify_namespace(struct nvme_controller *ctrlr,
319					  uint16_t nsid, void *payload,
320					  nvme_cb_fn_t cb_fn, void *cb_arg);
321void	nvme_ctrlr_cmd_set_interrupt_coalescing(struct nvme_controller *ctrlr,
322						uint32_t microseconds,
323						uint32_t threshold,
324						nvme_cb_fn_t cb_fn,
325						void *cb_arg);
326void	nvme_ctrlr_cmd_get_health_information_page(struct nvme_controller *ctrlr,
327						   uint32_t nsid,
328						   struct nvme_health_information_page *payload,
329						   nvme_cb_fn_t cb_fn,
330						   void *cb_arg);
331void	nvme_ctrlr_cmd_create_io_cq(struct nvme_controller *ctrlr,
332				    struct nvme_qpair *io_que, uint16_t vector,
333				    nvme_cb_fn_t cb_fn, void *cb_arg);
334void	nvme_ctrlr_cmd_create_io_sq(struct nvme_controller *ctrlr,
335				    struct nvme_qpair *io_que,
336				    nvme_cb_fn_t cb_fn, void *cb_arg);
337void	nvme_ctrlr_cmd_delete_io_cq(struct nvme_controller *ctrlr,
338				    struct nvme_qpair *io_que,
339				    nvme_cb_fn_t cb_fn, void *cb_arg);
340void	nvme_ctrlr_cmd_delete_io_sq(struct nvme_controller *ctrlr,
341				    struct nvme_qpair *io_que,
342				    nvme_cb_fn_t cb_fn, void *cb_arg);
343void	nvme_ctrlr_cmd_set_num_queues(struct nvme_controller *ctrlr,
344				      uint32_t num_queues, nvme_cb_fn_t cb_fn,
345				      void *cb_arg);
346void	nvme_ctrlr_cmd_set_asynchronous_event_config(struct nvme_controller *ctrlr,
347					   union nvme_critical_warning_state state,
348					   nvme_cb_fn_t cb_fn, void *cb_arg);
349void	nvme_ctrlr_cmd_asynchronous_event_request(struct nvme_controller *ctrlr,
350						  nvme_cb_fn_t cb_fn,
351						  void *cb_arg);
352void	nvme_ctrlr_cmd_abort(struct nvme_controller *ctrlr, uint16_t cid,
353			     uint16_t sqid, nvme_cb_fn_t cb_fn, void *cb_arg);
354
355void	nvme_payload_map(void *arg, bus_dma_segment_t *seg, int nseg,
356			 int error);
357void	nvme_payload_map_uio(void *arg, bus_dma_segment_t *seg, int nseg,
358			     bus_size_t mapsize, int error);
359
360int	nvme_ctrlr_construct(struct nvme_controller *ctrlr, device_t dev);
361int	nvme_ctrlr_reset(struct nvme_controller *ctrlr);
362/* ctrlr defined as void * to allow use with config_intrhook. */
363void	nvme_ctrlr_start(void *ctrlr_arg);
364void	nvme_ctrlr_submit_admin_request(struct nvme_controller *ctrlr,
365					struct nvme_request *req);
366void	nvme_ctrlr_submit_io_request(struct nvme_controller *ctrlr,
367				     struct nvme_request *req);
368
369void	nvme_qpair_construct(struct nvme_qpair *qpair, uint32_t id,
370			     uint16_t vector, uint32_t num_entries,
371			     uint32_t num_trackers, uint32_t max_xfer_size,
372			     struct nvme_controller *ctrlr);
373void	nvme_qpair_submit_cmd(struct nvme_qpair *qpair,
374			      struct nvme_tracker *tr);
375void	nvme_qpair_process_completions(struct nvme_qpair *qpair);
376void	nvme_qpair_submit_request(struct nvme_qpair *qpair,
377				  struct nvme_request *req);
378
379void	nvme_admin_qpair_destroy(struct nvme_qpair *qpair);
380
381void	nvme_io_qpair_destroy(struct nvme_qpair *qpair);
382
383int	nvme_ns_construct(struct nvme_namespace *ns, uint16_t id,
384			  struct nvme_controller *ctrlr);
385
386int	nvme_ns_physio(struct cdev *dev, struct uio *uio, int ioflag);
387
388void	nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr);
389
390void	nvme_dump_command(struct nvme_command *cmd);
391void	nvme_dump_completion(struct nvme_completion *cpl);
392
393static __inline void
394nvme_single_map(void *arg, bus_dma_segment_t *seg, int nseg, int error)
395{
396	uint64_t *bus_addr = (uint64_t *)arg;
397
398	*bus_addr = seg[0].ds_addr;
399}
400
401static __inline struct nvme_request *
402nvme_allocate_request(void *payload, uint32_t payload_size, nvme_cb_fn_t cb_fn,
403		      void *cb_arg)
404{
405	struct nvme_request *req;
406
407	req = uma_zalloc(nvme_request_zone, M_NOWAIT | M_ZERO);
408	if (req == NULL)
409		return (NULL);
410
411	req->payload = payload;
412	req->payload_size = payload_size;
413	req->cb_fn = cb_fn;
414	req->cb_arg = cb_arg;
415	req->timeout = NVME_TIMEOUT_IN_SEC;
416
417	return (req);
418}
419
420static __inline struct nvme_request *
421nvme_allocate_request_uio(struct uio *uio, nvme_cb_fn_t cb_fn, void *cb_arg)
422{
423	struct nvme_request *req;
424
425	req = uma_zalloc(nvme_request_zone, M_NOWAIT | M_ZERO);
426	if (req == NULL)
427		return (NULL);
428
429	req->uio = uio;
430	req->cb_fn = cb_fn;
431	req->cb_arg = cb_arg;
432	req->timeout = NVME_TIMEOUT_IN_SEC;
433
434	return (req);
435}
436
437#define nvme_free_request(req)	uma_zfree(nvme_request_zone, req)
438
439#endif /* __NVME_PRIVATE_H__ */
440