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>
22 #include <linux/sed-opal.h>
26 * Driver internal status code for commands that were cancelled due
27 * to timeouts or controller shutdown. The value is negative so
28 * that it a) doesn't overlap with the unsigned hardware error codes,
29 * and b) can easily be tested for.
31 NVME_SC_CANCELLED
= -EINTR
,
34 extern unsigned char nvme_io_timeout
;
35 #define NVME_IO_TIMEOUT (nvme_io_timeout * HZ)
37 extern unsigned char admin_timeout
;
38 #define ADMIN_TIMEOUT (admin_timeout * HZ)
40 extern unsigned char shutdown_timeout
;
41 #define SHUTDOWN_TIMEOUT (shutdown_timeout * HZ)
43 #define NVME_DEFAULT_KATO 5
44 #define NVME_KATO_GRACE 10
52 * List of workarounds for devices that required behavior not specified in
57 * Prefers I/O aligned to a stripe size specified in a vendor
58 * specific Identify field.
60 NVME_QUIRK_STRIPE_SIZE
= (1 << 0),
63 * The controller doesn't handle Identify value others than 0 or 1
66 NVME_QUIRK_IDENTIFY_CNS
= (1 << 1),
69 * The controller deterministically returns O's on reads to discarded
72 NVME_QUIRK_DISCARD_ZEROES
= (1 << 2),
75 * The controller needs a delay before starts checking the device
76 * readiness, which is done by reading the NVME_CSTS_RDY bit.
78 NVME_QUIRK_DELAY_BEFORE_CHK_RDY
= (1 << 3),
81 * APST should not be used.
83 NVME_QUIRK_NO_APST
= (1 << 4),
87 * Common request structure for NVMe passthrough. All drivers must have
88 * this structure as the first member of their request-private data.
91 struct nvme_command
*cmd
;
92 union nvme_result result
;
96 static inline struct nvme_request
*nvme_req(struct request
*req
)
98 return blk_mq_rq_to_pdu(req
);
101 /* The below value is the specific amount of delay needed before checking
102 * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
103 * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
106 #define NVME_QUIRK_DELAY_AMOUNT 2000
108 enum nvme_ctrl_state
{
112 NVME_CTRL_RECONNECTING
,
118 enum nvme_ctrl_state state
;
121 const struct nvme_ctrl_ops
*ops
;
122 struct request_queue
*admin_q
;
123 struct request_queue
*connect_q
;
127 struct blk_mq_tag_set
*tagset
;
128 struct list_head namespaces
;
129 struct mutex namespaces_mutex
;
130 struct device
*device
; /* char device */
131 struct list_head node
;
134 struct opal_dev
*opal_dev
;
139 char firmware_rev
[8];
149 atomic_t abort_limit
;
159 unsigned long quirks
;
160 struct nvme_id_power_state psd
[32];
161 struct work_struct scan_work
;
162 struct work_struct async_event_work
;
163 struct delayed_work ka_work
;
165 /* Power saving configuration */
166 u64 ps_max_latency_us
;
174 struct nvmf_ctrl_options
*opts
;
178 * An NVM Express namespace is equivalent to a SCSI LUN
181 struct list_head list
;
183 struct nvme_ctrl
*ctrl
;
184 struct request_queue
*queue
;
185 struct gendisk
*disk
;
186 struct nvm_dev
*ndev
;
200 #define NVME_NS_REMOVING 0
201 #define NVME_NS_DEAD 1
203 u64 mode_select_num_blocks
;
204 u32 mode_select_block_len
;
207 struct nvme_ctrl_ops
{
209 struct module
*module
;
211 int (*reg_read32
)(struct nvme_ctrl
*ctrl
, u32 off
, u32
*val
);
212 int (*reg_write32
)(struct nvme_ctrl
*ctrl
, u32 off
, u32 val
);
213 int (*reg_read64
)(struct nvme_ctrl
*ctrl
, u32 off
, u64
*val
);
214 int (*reset_ctrl
)(struct nvme_ctrl
*ctrl
);
215 void (*free_ctrl
)(struct nvme_ctrl
*ctrl
);
216 void (*submit_async_event
)(struct nvme_ctrl
*ctrl
, int aer_idx
);
217 int (*delete_ctrl
)(struct nvme_ctrl
*ctrl
);
218 const char *(*get_subsysnqn
)(struct nvme_ctrl
*ctrl
);
219 int (*get_address
)(struct nvme_ctrl
*ctrl
, char *buf
, int size
);
222 static inline bool nvme_ctrl_ready(struct nvme_ctrl
*ctrl
)
226 if (ctrl
->ops
->reg_read32(ctrl
, NVME_REG_CSTS
, &val
))
228 return val
& NVME_CSTS_RDY
;
231 static inline int nvme_reset_subsystem(struct nvme_ctrl
*ctrl
)
233 if (!ctrl
->subsystem
)
235 return ctrl
->ops
->reg_write32(ctrl
, NVME_REG_NSSR
, 0x4E564D65);
238 static inline u64
nvme_block_nr(struct nvme_ns
*ns
, sector_t sector
)
240 return (sector
>> (ns
->lba_shift
- 9));
243 static inline void nvme_cleanup_cmd(struct request
*req
)
245 if (req
->rq_flags
& RQF_SPECIAL_PAYLOAD
) {
246 kfree(page_address(req
->special_vec
.bv_page
) +
247 req
->special_vec
.bv_offset
);
251 static inline int nvme_error_status(u16 status
)
253 switch (status
& 0x7ff) {
254 case NVME_SC_SUCCESS
:
256 case NVME_SC_CAP_EXCEEDED
:
263 void nvme_complete_rq(struct request
*req
);
264 void nvme_cancel_request(struct request
*req
, void *data
, bool reserved
);
265 bool nvme_change_ctrl_state(struct nvme_ctrl
*ctrl
,
266 enum nvme_ctrl_state new_state
);
267 int nvme_disable_ctrl(struct nvme_ctrl
*ctrl
, u64 cap
);
268 int nvme_enable_ctrl(struct nvme_ctrl
*ctrl
, u64 cap
);
269 int nvme_shutdown_ctrl(struct nvme_ctrl
*ctrl
);
270 int nvme_init_ctrl(struct nvme_ctrl
*ctrl
, struct device
*dev
,
271 const struct nvme_ctrl_ops
*ops
, unsigned long quirks
);
272 void nvme_uninit_ctrl(struct nvme_ctrl
*ctrl
);
273 void nvme_put_ctrl(struct nvme_ctrl
*ctrl
);
274 int nvme_init_identify(struct nvme_ctrl
*ctrl
);
276 void nvme_queue_scan(struct nvme_ctrl
*ctrl
);
277 void nvme_remove_namespaces(struct nvme_ctrl
*ctrl
);
279 int nvme_sec_submit(void *data
, u16 spsp
, u8 secp
, void *buffer
, size_t len
,
282 #define NVME_NR_AERS 1
283 void nvme_complete_async_event(struct nvme_ctrl
*ctrl
, __le16 status
,
284 union nvme_result
*res
);
285 void nvme_queue_async_events(struct nvme_ctrl
*ctrl
);
287 void nvme_stop_queues(struct nvme_ctrl
*ctrl
);
288 void nvme_start_queues(struct nvme_ctrl
*ctrl
);
289 void nvme_kill_queues(struct nvme_ctrl
*ctrl
);
290 void nvme_unfreeze(struct nvme_ctrl
*ctrl
);
291 void nvme_wait_freeze(struct nvme_ctrl
*ctrl
);
292 void nvme_wait_freeze_timeout(struct nvme_ctrl
*ctrl
, long timeout
);
293 void nvme_start_freeze(struct nvme_ctrl
*ctrl
);
295 #define NVME_QID_ANY -1
296 struct request
*nvme_alloc_request(struct request_queue
*q
,
297 struct nvme_command
*cmd
, unsigned int flags
, int qid
);
298 int nvme_setup_cmd(struct nvme_ns
*ns
, struct request
*req
,
299 struct nvme_command
*cmd
);
300 int nvme_submit_sync_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
301 void *buf
, unsigned bufflen
);
302 int __nvme_submit_sync_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
303 union nvme_result
*result
, void *buffer
, unsigned bufflen
,
304 unsigned timeout
, int qid
, int at_head
, int flags
);
305 int nvme_submit_user_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
306 void __user
*ubuffer
, unsigned bufflen
, u32
*result
,
308 int __nvme_submit_user_cmd(struct request_queue
*q
, struct nvme_command
*cmd
,
309 void __user
*ubuffer
, unsigned bufflen
,
310 void __user
*meta_buffer
, unsigned meta_len
, u32 meta_seed
,
311 u32
*result
, unsigned timeout
);
312 int nvme_identify_ctrl(struct nvme_ctrl
*dev
, struct nvme_id_ctrl
**id
);
313 int nvme_identify_ns(struct nvme_ctrl
*dev
, unsigned nsid
,
314 struct nvme_id_ns
**id
);
315 int nvme_get_log_page(struct nvme_ctrl
*dev
, struct nvme_smart_log
**log
);
316 int nvme_get_features(struct nvme_ctrl
*dev
, unsigned fid
, unsigned nsid
,
317 void *buffer
, size_t buflen
, u32
*result
);
318 int nvme_set_features(struct nvme_ctrl
*dev
, unsigned fid
, unsigned dword11
,
319 void *buffer
, size_t buflen
, u32
*result
);
320 int nvme_set_queue_count(struct nvme_ctrl
*ctrl
, int *count
);
321 void nvme_start_keep_alive(struct nvme_ctrl
*ctrl
);
322 void nvme_stop_keep_alive(struct nvme_ctrl
*ctrl
);
326 int nvme_sg_io(struct nvme_ns
*ns
, struct sg_io_hdr __user
*u_hdr
);
327 int nvme_sg_io32(struct nvme_ns
*ns
, unsigned long arg
);
328 int nvme_sg_get_version_num(int __user
*ip
);
331 int nvme_nvm_ns_supported(struct nvme_ns
*ns
, struct nvme_id_ns
*id
);
332 int nvme_nvm_register(struct nvme_ns
*ns
, char *disk_name
, int node
);
333 void nvme_nvm_unregister(struct nvme_ns
*ns
);
334 int nvme_nvm_register_sysfs(struct nvme_ns
*ns
);
335 void nvme_nvm_unregister_sysfs(struct nvme_ns
*ns
);
336 int nvme_nvm_ioctl(struct nvme_ns
*ns
, unsigned int cmd
, unsigned long arg
);
338 static inline int nvme_nvm_register(struct nvme_ns
*ns
, char *disk_name
,
344 static inline void nvme_nvm_unregister(struct nvme_ns
*ns
) {};
345 static inline int nvme_nvm_register_sysfs(struct nvme_ns
*ns
)
349 static inline void nvme_nvm_unregister_sysfs(struct nvme_ns
*ns
) {};
350 static inline int nvme_nvm_ns_supported(struct nvme_ns
*ns
, struct nvme_id_ns
*id
)
354 static inline int nvme_nvm_ioctl(struct nvme_ns
*ns
, unsigned int cmd
,
359 #endif /* CONFIG_NVM */
361 static inline struct nvme_ns
*nvme_get_ns_from_dev(struct device
*dev
)
363 return dev_to_disk(dev
)->private_data
;
366 int __init
nvme_core_init(void);
367 void nvme_core_exit(void);