2 * Copyright (c) 2011-2014, Intel Corporation.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
17 #include <linux/nvme.h>
18 #include <linux/pci.h>
19 #include <linux/kref.h>
20 #include <linux/blk-mq.h>
21 #include <linux/lightnvm.h>
25 * Driver internal status code for commands that were cancelled due
26 * to timeouts or controller shutdown. The value is negative so
27 * that it a) doesn't overlap with the unsigned hardware error codes,
28 * and b) can easily be tested for.
30 NVME_SC_CANCELLED
= -EINTR
,
33 extern unsigned char nvme_io_timeout
;
34 #define NVME_IO_TIMEOUT (nvme_io_timeout * HZ)
36 extern unsigned char admin_timeout
;
37 #define ADMIN_TIMEOUT (admin_timeout * HZ)
39 extern unsigned char shutdown_timeout
;
40 #define SHUTDOWN_TIMEOUT (shutdown_timeout * HZ)
42 #define NVME_DEFAULT_KATO 5
43 #define NVME_KATO_GRACE 10
45 extern unsigned int nvme_max_retries
;
53 * List of workarounds for devices that required behavior not specified in
58 * Prefers I/O aligned to a stripe size specified in a vendor
59 * specific Identify field.
61 NVME_QUIRK_STRIPE_SIZE
= (1 << 0),
64 * The controller doesn't handle Identify value others than 0 or 1
67 NVME_QUIRK_IDENTIFY_CNS
= (1 << 1),
70 * The controller deterministically returns O's on reads to discarded
73 NVME_QUIRK_DISCARD_ZEROES
= (1 << 2),
76 * The controller needs a delay before starts checking the device
77 * readiness, which is done by reading the NVME_CSTS_RDY bit.
79 NVME_QUIRK_DELAY_BEFORE_CHK_RDY
= (1 << 3),
83 * Common request structure for NVMe passthrough. All drivers must have
84 * this structure as the first member of their request-private data.
87 struct nvme_command
*cmd
;
88 union nvme_result result
;
91 static inline struct nvme_request
*nvme_req(struct request
*req
)
93 return blk_mq_rq_to_pdu(req
);
96 /* The below value is the specific amount of delay needed before checking
97 * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
98 * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
101 #define NVME_QUIRK_DELAY_AMOUNT 2000
103 enum nvme_ctrl_state
{
107 NVME_CTRL_RECONNECTING
,
113 enum nvme_ctrl_state state
;
115 const struct nvme_ctrl_ops
*ops
;
116 struct request_queue
*admin_q
;
117 struct request_queue
*connect_q
;
121 struct blk_mq_tag_set
*tagset
;
122 struct list_head namespaces
;
123 struct mutex namespaces_mutex
;
124 struct device
*device
; /* char device */
125 struct list_head node
;
131 char firmware_rev
[8];
140 atomic_t abort_limit
;
148 unsigned long quirks
;
149 struct work_struct scan_work
;
150 struct work_struct async_event_work
;
151 struct delayed_work ka_work
;
159 struct nvmf_ctrl_options
*opts
;
163 * An NVM Express namespace is equivalent to a SCSI LUN
166 struct list_head list
;
168 struct nvme_ctrl
*ctrl
;
169 struct request_queue
*queue
;
170 struct gendisk
*disk
;
171 struct nvm_dev
*ndev
;
185 #define NVME_NS_REMOVING 0
186 #define NVME_NS_DEAD 1
188 u64 mode_select_num_blocks
;
189 u32 mode_select_block_len
;
192 struct nvme_ctrl_ops
{
194 struct module
*module
;
196 int (*reg_read32
)(struct nvme_ctrl
*ctrl
, u32 off
, u32
*val
);
197 int (*reg_write32
)(struct nvme_ctrl
*ctrl
, u32 off
, u32 val
);
198 int (*reg_read64
)(struct nvme_ctrl
*ctrl
, u32 off
, u64
*val
);
199 int (*reset_ctrl
)(struct nvme_ctrl
*ctrl
);
200 void (*free_ctrl
)(struct nvme_ctrl
*ctrl
);
201 void (*submit_async_event
)(struct nvme_ctrl
*ctrl
, int aer_idx
);
202 int (*delete_ctrl
)(struct nvme_ctrl
*ctrl
);
203 const char *(*get_subsysnqn
)(struct nvme_ctrl
*ctrl
);
204 int (*get_address
)(struct nvme_ctrl
*ctrl
, char *buf
, int size
);
207 static inline bool nvme_ctrl_ready(struct nvme_ctrl
*ctrl
)
211 if (ctrl
->ops
->reg_read32(ctrl
, NVME_REG_CSTS
, &val
))
213 return val
& NVME_CSTS_RDY
;
216 static inline int nvme_reset_subsystem(struct nvme_ctrl
*ctrl
)
218 if (!ctrl
->subsystem
)
220 return ctrl
->ops
->reg_write32(ctrl
, NVME_REG_NSSR
, 0x4E564D65);
223 static inline u64
nvme_block_nr(struct nvme_ns
*ns
, sector_t sector
)
225 return (sector
>> (ns
->lba_shift
- 9));
228 static inline void nvme_cleanup_cmd(struct request
*req
)
230 if (req
->rq_flags
& RQF_SPECIAL_PAYLOAD
) {
231 kfree(page_address(req
->special_vec
.bv_page
) +
232 req
->special_vec
.bv_offset
);
236 static inline int nvme_error_status(u16 status
)
238 switch (status
& 0x7ff) {
239 case NVME_SC_SUCCESS
:
241 case NVME_SC_CAP_EXCEEDED
:
248 static inline bool nvme_req_needs_retry(struct request
*req
, u16 status
)
250 return !(status
& NVME_SC_DNR
|| blk_noretry_request(req
)) &&
251 (jiffies
- req
->start_time
) < req
->timeout
&&
252 req
->retries
< nvme_max_retries
;
255 void nvme_cancel_request(struct request
*req
, void *data
, bool reserved
);
256 bool nvme_change_ctrl_state(struct nvme_ctrl
*ctrl
,
257 enum nvme_ctrl_state new_state
);
258 int nvme_disable_ctrl(struct nvme_ctrl
*ctrl
, u64 cap
);
259 int nvme_enable_ctrl(struct nvme_ctrl
*ctrl
, u64 cap
);
260 int nvme_shutdown_ctrl(struct nvme_ctrl
*ctrl
);
261 int nvme_init_ctrl(struct nvme_ctrl
*ctrl
, struct device
*dev
,
262 const struct nvme_ctrl_ops
*ops
, unsigned long quirks
);
263 void nvme_uninit_ctrl(struct nvme_ctrl
*ctrl
);
264 void nvme_put_ctrl(struct nvme_ctrl
*ctrl
);
265 int nvme_init_identify(struct nvme_ctrl
*ctrl
);
267 void nvme_queue_scan(struct nvme_ctrl
*ctrl
);
268 void nvme_remove_namespaces(struct nvme_ctrl
*ctrl
);
270 #define NVME_NR_AERS 1
271 void nvme_complete_async_event(struct nvme_ctrl
*ctrl
, __le16 status
,
272 union nvme_result
*res
);
273 void nvme_queue_async_events(struct nvme_ctrl
*ctrl
);
275 void nvme_stop_queues(struct nvme_ctrl
*ctrl
);
276 void nvme_start_queues(struct nvme_ctrl
*ctrl
);
277 void nvme_kill_queues(struct nvme_ctrl
*ctrl
);
279 #define NVME_QID_ANY -1
280 struct request
*nvme_alloc_request(struct request_queue
*q
,
281 struct nvme_command
*cmd
, unsigned int flags
, int qid
);
282 void nvme_requeue_req(struct request
*req
);
283 int nvme_setup_cmd(struct nvme_ns
*ns
, struct request
*req
,
284 struct nvme_command
*cmd
);
285 int nvme_submit_sync_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
286 void *buf
, unsigned bufflen
);
287 int __nvme_submit_sync_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
288 union nvme_result
*result
, void *buffer
, unsigned bufflen
,
289 unsigned timeout
, int qid
, int at_head
, int flags
);
290 int nvme_submit_user_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
291 void __user
*ubuffer
, unsigned bufflen
, u32
*result
,
293 int __nvme_submit_user_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
294 void __user
*ubuffer
, unsigned bufflen
,
295 void __user
*meta_buffer
, unsigned meta_len
, u32 meta_seed
,
296 u32
*result
, unsigned timeout
);
297 int nvme_identify_ctrl(struct nvme_ctrl
*dev
, struct nvme_id_ctrl
**id
);
298 int nvme_identify_ns(struct nvme_ctrl
*dev
, unsigned nsid
,
299 struct nvme_id_ns
**id
);
300 int nvme_get_log_page(struct nvme_ctrl
*dev
, struct nvme_smart_log
**log
);
301 int nvme_get_features(struct nvme_ctrl
*dev
, unsigned fid
, unsigned nsid
,
302 void *buffer
, size_t buflen
, u32
*result
);
303 int nvme_set_features(struct nvme_ctrl
*dev
, unsigned fid
, unsigned dword11
,
304 void *buffer
, size_t buflen
, u32
*result
);
305 int nvme_set_queue_count(struct nvme_ctrl
*ctrl
, int *count
);
306 void nvme_start_keep_alive(struct nvme_ctrl
*ctrl
);
307 void nvme_stop_keep_alive(struct nvme_ctrl
*ctrl
);
311 int nvme_sg_io(struct nvme_ns
*ns
, struct sg_io_hdr __user
*u_hdr
);
312 int nvme_sg_io32(struct nvme_ns
*ns
, unsigned long arg
);
313 int nvme_sg_get_version_num(int __user
*ip
);
316 int nvme_nvm_ns_supported(struct nvme_ns
*ns
, struct nvme_id_ns
*id
);
317 int nvme_nvm_register(struct nvme_ns
*ns
, char *disk_name
, int node
);
318 void nvme_nvm_unregister(struct nvme_ns
*ns
);
319 int nvme_nvm_register_sysfs(struct nvme_ns
*ns
);
320 void nvme_nvm_unregister_sysfs(struct nvme_ns
*ns
);
322 static inline int nvme_nvm_register(struct nvme_ns
*ns
, char *disk_name
,
328 static inline void nvme_nvm_unregister(struct nvme_ns
*ns
) {};
329 static inline int nvme_nvm_register_sysfs(struct nvme_ns
*ns
)
333 static inline void nvme_nvm_unregister_sysfs(struct nvme_ns
*ns
) {};
334 static inline int nvme_nvm_ns_supported(struct nvme_ns
*ns
, struct nvme_id_ns
*id
)
338 #endif /* CONFIG_NVM */
340 static inline struct nvme_ns
*nvme_get_ns_from_dev(struct device
*dev
)
342 return dev_to_disk(dev
)->private_data
;
345 int __init
nvme_core_init(void);
346 void nvme_core_exit(void);