2011-01-21 01:50:14 +08:00
|
|
|
/*
|
|
|
|
* NVM Express device driver
|
2014-03-24 22:11:22 +08:00
|
|
|
* Copyright (c) 2011-2014, Intel Corporation.
|
2011-01-21 01:50:14 +08:00
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify it
|
|
|
|
* under the terms and conditions of the GNU General Public License,
|
|
|
|
* version 2, as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope it will be useful, but WITHOUT
|
|
|
|
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
|
|
|
* more details.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/nvme.h>
|
|
|
|
#include <linux/bio.h>
|
2011-05-13 01:50:28 +08:00
|
|
|
#include <linux/bitops.h>
|
2011-01-21 01:50:14 +08:00
|
|
|
#include <linux/blkdev.h>
|
2014-03-25 00:46:25 +08:00
|
|
|
#include <linux/cpu.h>
|
2011-05-06 20:37:54 +08:00
|
|
|
#include <linux/delay.h>
|
2011-01-21 01:50:14 +08:00
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/genhd.h>
|
2014-04-03 05:45:37 +08:00
|
|
|
#include <linux/hdreg.h>
|
2011-05-06 20:45:47 +08:00
|
|
|
#include <linux/idr.h>
|
2011-01-21 01:50:14 +08:00
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/io.h>
|
|
|
|
#include <linux/kdev_t.h>
|
2011-03-03 07:37:18 +08:00
|
|
|
#include <linux/kthread.h>
|
2011-01-21 01:50:14 +08:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/moduleparam.h>
|
|
|
|
#include <linux/pci.h>
|
2014-03-25 00:46:25 +08:00
|
|
|
#include <linux/percpu.h>
|
2011-02-06 20:53:23 +08:00
|
|
|
#include <linux/poison.h>
|
2013-07-09 05:26:25 +08:00
|
|
|
#include <linux/ptrace.h>
|
2011-01-21 01:50:14 +08:00
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/types.h>
|
2013-03-05 09:40:58 +08:00
|
|
|
#include <scsi/sg.h>
|
2012-02-07 10:45:33 +08:00
|
|
|
#include <asm-generic/io-64-nonatomic-lo-hi.h>
|
|
|
|
|
2014-04-29 02:30:52 +08:00
|
|
|
#include <trace/events/block.h>
|
|
|
|
|
2014-05-14 01:42:02 +08:00
|
|
|
#define NVME_Q_DEPTH 1024
|
2011-01-21 01:50:14 +08:00
|
|
|
#define SQ_SIZE(depth) (depth * sizeof(struct nvme_command))
|
|
|
|
#define CQ_SIZE(depth) (depth * sizeof(struct nvme_completion))
|
2014-05-14 01:42:02 +08:00
|
|
|
#define ADMIN_TIMEOUT (admin_timeout * HZ)
|
|
|
|
#define IOD_TIMEOUT (retry_time * HZ)
|
|
|
|
|
|
|
|
static unsigned char admin_timeout = 60;
|
|
|
|
module_param(admin_timeout, byte, 0644);
|
|
|
|
MODULE_PARM_DESC(admin_timeout, "timeout in seconds for admin commands");
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-05 01:43:36 +08:00
|
|
|
unsigned char io_timeout = 30;
|
|
|
|
module_param(io_timeout, byte, 0644);
|
|
|
|
MODULE_PARM_DESC(io_timeout, "timeout in seconds for I/O");
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-05-14 01:42:01 +08:00
|
|
|
static unsigned char retry_time = 30;
|
|
|
|
module_param(retry_time, byte, 0644);
|
|
|
|
MODULE_PARM_DESC(retry_time, "time in seconds to retry failed I/O");
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
static int nvme_major;
|
|
|
|
module_param(nvme_major, int, 0);
|
|
|
|
|
2011-02-06 20:28:06 +08:00
|
|
|
static int use_threaded_interrupts;
|
|
|
|
module_param(use_threaded_interrupts, int, 0);
|
|
|
|
|
2011-03-03 07:37:18 +08:00
|
|
|
static DEFINE_SPINLOCK(dev_list_lock);
|
|
|
|
static LIST_HEAD(dev_list);
|
|
|
|
static struct task_struct *nvme_thread;
|
2013-12-11 04:10:36 +08:00
|
|
|
static struct workqueue_struct *nvme_workq;
|
2014-04-08 07:10:11 +08:00
|
|
|
static wait_queue_head_t nvme_kthread_wait;
|
2011-03-03 07:37:18 +08:00
|
|
|
|
2013-12-11 04:10:37 +08:00
|
|
|
static void nvme_reset_failed_dev(struct work_struct *ws);
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
struct async_cmd_info {
|
|
|
|
struct kthread_work work;
|
|
|
|
struct kthread_worker *worker;
|
|
|
|
u32 result;
|
|
|
|
int status;
|
|
|
|
void *ctx;
|
|
|
|
};
|
2011-03-03 07:37:18 +08:00
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
/*
|
|
|
|
* An NVM Express queue. Each device has at least two (one for admin
|
|
|
|
* commands and one for I/O commands).
|
|
|
|
*/
|
|
|
|
struct nvme_queue {
|
2014-02-22 05:13:44 +08:00
|
|
|
struct rcu_head r_head;
|
2011-01-21 01:50:14 +08:00
|
|
|
struct device *q_dmadev;
|
2011-02-10 22:56:01 +08:00
|
|
|
struct nvme_dev *dev;
|
2014-01-28 04:57:22 +08:00
|
|
|
char irqname[24]; /* nvme4294967295-65535\0 */
|
2011-01-21 01:50:14 +08:00
|
|
|
spinlock_t q_lock;
|
|
|
|
struct nvme_command *sq_cmds;
|
|
|
|
volatile struct nvme_completion *cqes;
|
|
|
|
dma_addr_t sq_dma_addr;
|
|
|
|
dma_addr_t cq_dma_addr;
|
|
|
|
wait_queue_head_t sq_full;
|
2011-03-03 07:37:18 +08:00
|
|
|
wait_queue_t sq_cong_wait;
|
2011-01-21 01:50:14 +08:00
|
|
|
struct bio_list sq_cong;
|
2014-04-04 06:45:23 +08:00
|
|
|
struct list_head iod_bio;
|
2011-01-21 01:50:14 +08:00
|
|
|
u32 __iomem *q_db;
|
|
|
|
u16 q_depth;
|
|
|
|
u16 cq_vector;
|
|
|
|
u16 sq_head;
|
|
|
|
u16 sq_tail;
|
|
|
|
u16 cq_head;
|
2013-12-11 04:10:38 +08:00
|
|
|
u16 qid;
|
2013-06-24 23:47:34 +08:00
|
|
|
u8 cq_phase;
|
|
|
|
u8 cqe_seen;
|
2013-07-16 05:02:20 +08:00
|
|
|
u8 q_suspended;
|
2014-03-25 00:46:25 +08:00
|
|
|
cpumask_var_t cpu_mask;
|
2013-12-11 04:10:40 +08:00
|
|
|
struct async_cmd_info cmdinfo;
|
2011-01-21 01:50:14 +08:00
|
|
|
unsigned long cmdid_data[];
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check we didin't inadvertently grow the command struct
|
|
|
|
*/
|
|
|
|
static inline void _nvme_check_size(void)
|
|
|
|
{
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_rw_command) != 64);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_create_cq) != 64);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_create_sq) != 64);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_delete_queue) != 64);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_features) != 64);
|
2013-03-27 19:13:41 +08:00
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_format_cmd) != 64);
|
2013-12-11 04:10:38 +08:00
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_abort_cmd) != 64);
|
2011-01-21 01:50:14 +08:00
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_command) != 64);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_id_ctrl) != 4096);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_id_ns) != 4096);
|
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_lba_range_type) != 64);
|
2012-09-27 02:49:27 +08:00
|
|
|
BUILD_BUG_ON(sizeof(struct nvme_smart_log) != 512);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
typedef void (*nvme_completion_fn)(struct nvme_queue *, void *,
|
2011-10-15 19:33:46 +08:00
|
|
|
struct nvme_completion *);
|
|
|
|
|
2011-02-07 07:30:16 +08:00
|
|
|
struct nvme_cmd_info {
|
2011-10-15 19:33:46 +08:00
|
|
|
nvme_completion_fn fn;
|
|
|
|
void *ctx;
|
2011-02-07 07:30:16 +08:00
|
|
|
unsigned long timeout;
|
2013-12-11 04:10:38 +08:00
|
|
|
int aborted;
|
2011-02-07 07:30:16 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static struct nvme_cmd_info *nvme_cmd_info(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
return (void *)&nvmeq->cmdid_data[BITS_TO_LONGS(nvmeq->q_depth)];
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
static unsigned nvme_queue_extra(int depth)
|
|
|
|
{
|
|
|
|
return DIV_ROUND_UP(depth, 8) + (depth * sizeof(struct nvme_cmd_info));
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
/**
|
2011-03-17 04:28:24 +08:00
|
|
|
* alloc_cmdid() - Allocate a Command ID
|
|
|
|
* @nvmeq: The queue that will be used for this command
|
|
|
|
* @ctx: A pointer that will be passed to the handler
|
2011-10-15 19:33:46 +08:00
|
|
|
* @handler: The function to call on completion
|
2011-01-21 01:50:14 +08:00
|
|
|
*
|
|
|
|
* Allocate a Command ID for a queue. The data passed in will
|
|
|
|
* be passed to the completion handler. This is implemented by using
|
|
|
|
* the bottom two bits of the ctx pointer to store the handler ID.
|
|
|
|
* Passing in a pointer that's not 4-byte aligned will cause a BUG.
|
|
|
|
* We can change this if it becomes a problem.
|
2011-05-12 09:36:38 +08:00
|
|
|
*
|
|
|
|
* May be called with local interrupts disabled and the q_lock held,
|
|
|
|
* or with interrupts enabled and no locks held.
|
2011-01-21 01:50:14 +08:00
|
|
|
*/
|
2011-10-15 19:33:46 +08:00
|
|
|
static int alloc_cmdid(struct nvme_queue *nvmeq, void *ctx,
|
|
|
|
nvme_completion_fn handler, unsigned timeout)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-02-24 21:49:41 +08:00
|
|
|
int depth = nvmeq->q_depth - 1;
|
2011-02-07 07:30:16 +08:00
|
|
|
struct nvme_cmd_info *info = nvme_cmd_info(nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
int cmdid;
|
|
|
|
|
|
|
|
do {
|
|
|
|
cmdid = find_first_zero_bit(nvmeq->cmdid_data, depth);
|
|
|
|
if (cmdid >= depth)
|
|
|
|
return -EBUSY;
|
|
|
|
} while (test_and_set_bit(cmdid, nvmeq->cmdid_data));
|
|
|
|
|
2011-10-15 19:33:46 +08:00
|
|
|
info[cmdid].fn = handler;
|
|
|
|
info[cmdid].ctx = ctx;
|
2011-02-07 07:30:16 +08:00
|
|
|
info[cmdid].timeout = jiffies + timeout;
|
2013-12-11 04:10:38 +08:00
|
|
|
info[cmdid].aborted = 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
return cmdid;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int alloc_cmdid_killable(struct nvme_queue *nvmeq, void *ctx,
|
2011-10-15 19:33:46 +08:00
|
|
|
nvme_completion_fn handler, unsigned timeout)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
|
|
|
int cmdid;
|
|
|
|
wait_event_killable(nvmeq->sq_full,
|
2011-02-07 07:30:16 +08:00
|
|
|
(cmdid = alloc_cmdid(nvmeq, ctx, handler, timeout)) >= 0);
|
2011-01-21 01:50:14 +08:00
|
|
|
return (cmdid < 0) ? -EINTR : cmdid;
|
|
|
|
}
|
|
|
|
|
2011-10-15 19:33:46 +08:00
|
|
|
/* Special values must be less than 0x1000 */
|
|
|
|
#define CMD_CTX_BASE ((void *)POISON_POINTER_DELTA)
|
2011-02-08 04:55:59 +08:00
|
|
|
#define CMD_CTX_CANCELLED (0x30C + CMD_CTX_BASE)
|
|
|
|
#define CMD_CTX_COMPLETED (0x310 + CMD_CTX_BASE)
|
|
|
|
#define CMD_CTX_INVALID (0x314 + CMD_CTX_BASE)
|
2014-04-30 01:41:29 +08:00
|
|
|
#define CMD_CTX_ABORT (0x318 + CMD_CTX_BASE)
|
2011-02-06 20:53:23 +08:00
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static void special_completion(struct nvme_queue *nvmeq, void *ctx,
|
2011-10-15 19:33:46 +08:00
|
|
|
struct nvme_completion *cqe)
|
|
|
|
{
|
|
|
|
if (ctx == CMD_CTX_CANCELLED)
|
|
|
|
return;
|
2013-12-11 04:10:38 +08:00
|
|
|
if (ctx == CMD_CTX_ABORT) {
|
2014-04-04 06:45:23 +08:00
|
|
|
++nvmeq->dev->abort_limit;
|
2013-12-11 04:10:38 +08:00
|
|
|
return;
|
|
|
|
}
|
2011-10-15 19:33:46 +08:00
|
|
|
if (ctx == CMD_CTX_COMPLETED) {
|
2014-04-04 06:45:23 +08:00
|
|
|
dev_warn(nvmeq->q_dmadev,
|
2011-10-15 19:33:46 +08:00
|
|
|
"completed id %d twice on queue %d\n",
|
|
|
|
cqe->command_id, le16_to_cpup(&cqe->sq_id));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (ctx == CMD_CTX_INVALID) {
|
2014-04-04 06:45:23 +08:00
|
|
|
dev_warn(nvmeq->q_dmadev,
|
2011-10-15 19:33:46 +08:00
|
|
|
"invalid id %d completed on queue %d\n",
|
|
|
|
cqe->command_id, le16_to_cpup(&cqe->sq_id));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
dev_warn(nvmeq->q_dmadev, "Unknown special completion %p\n", ctx);
|
2011-10-15 19:33:46 +08:00
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static void async_completion(struct nvme_queue *nvmeq, void *ctx,
|
2013-12-11 04:10:40 +08:00
|
|
|
struct nvme_completion *cqe)
|
|
|
|
{
|
|
|
|
struct async_cmd_info *cmdinfo = ctx;
|
|
|
|
cmdinfo->result = le32_to_cpup(&cqe->result);
|
|
|
|
cmdinfo->status = le16_to_cpup(&cqe->status) >> 1;
|
|
|
|
queue_kthread_work(cmdinfo->worker, &cmdinfo->work);
|
|
|
|
}
|
|
|
|
|
2011-05-12 09:36:38 +08:00
|
|
|
/*
|
|
|
|
* Called with local interrupts disabled and the q_lock held. May not sleep.
|
|
|
|
*/
|
2011-10-15 19:33:46 +08:00
|
|
|
static void *free_cmdid(struct nvme_queue *nvmeq, int cmdid,
|
|
|
|
nvme_completion_fn *fn)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-10-15 19:33:46 +08:00
|
|
|
void *ctx;
|
2011-02-07 07:30:16 +08:00
|
|
|
struct nvme_cmd_info *info = nvme_cmd_info(nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-25 08:53:50 +08:00
|
|
|
if (cmdid >= nvmeq->q_depth || !info[cmdid].fn) {
|
|
|
|
if (fn)
|
|
|
|
*fn = special_completion;
|
2011-02-06 21:51:15 +08:00
|
|
|
return CMD_CTX_INVALID;
|
2011-10-15 19:33:46 +08:00
|
|
|
}
|
2012-08-03 04:05:59 +08:00
|
|
|
if (fn)
|
|
|
|
*fn = info[cmdid].fn;
|
2011-10-15 19:33:46 +08:00
|
|
|
ctx = info[cmdid].ctx;
|
|
|
|
info[cmdid].fn = special_completion;
|
2011-02-07 07:30:16 +08:00
|
|
|
info[cmdid].ctx = CMD_CTX_COMPLETED;
|
2011-01-21 01:50:14 +08:00
|
|
|
clear_bit(cmdid, nvmeq->cmdid_data);
|
|
|
|
wake_up(&nvmeq->sq_full);
|
2011-10-15 19:33:46 +08:00
|
|
|
return ctx;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2011-10-15 19:33:46 +08:00
|
|
|
static void *cancel_cmdid(struct nvme_queue *nvmeq, int cmdid,
|
|
|
|
nvme_completion_fn *fn)
|
2011-02-05 05:03:56 +08:00
|
|
|
{
|
2011-10-15 19:33:46 +08:00
|
|
|
void *ctx;
|
2011-02-07 07:30:16 +08:00
|
|
|
struct nvme_cmd_info *info = nvme_cmd_info(nvmeq);
|
2011-10-15 19:33:46 +08:00
|
|
|
if (fn)
|
|
|
|
*fn = info[cmdid].fn;
|
|
|
|
ctx = info[cmdid].ctx;
|
|
|
|
info[cmdid].fn = special_completion;
|
2011-02-07 07:30:16 +08:00
|
|
|
info[cmdid].ctx = CMD_CTX_CANCELLED;
|
2011-10-15 19:33:46 +08:00
|
|
|
return ctx;
|
2011-02-05 05:03:56 +08:00
|
|
|
}
|
|
|
|
|
2014-02-22 05:13:44 +08:00
|
|
|
static struct nvme_queue *raw_nvmeq(struct nvme_dev *dev, int qid)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
return rcu_dereference_raw(dev->queues[qid]);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2014-03-04 07:39:13 +08:00
|
|
|
static struct nvme_queue *get_nvmeq(struct nvme_dev *dev) __acquires(RCU)
|
2014-02-22 05:13:44 +08:00
|
|
|
{
|
2014-05-14 00:32:46 +08:00
|
|
|
struct nvme_queue *nvmeq;
|
2014-03-25 00:46:25 +08:00
|
|
|
unsigned queue_id = get_cpu_var(*dev->io_queue);
|
2014-05-14 00:32:46 +08:00
|
|
|
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_read_lock();
|
2014-05-14 00:32:46 +08:00
|
|
|
nvmeq = rcu_dereference(dev->queues[queue_id]);
|
|
|
|
if (nvmeq)
|
|
|
|
return nvmeq;
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
put_cpu_var(*dev->io_queue);
|
|
|
|
return NULL;
|
2014-02-22 05:13:44 +08:00
|
|
|
}
|
|
|
|
|
2014-03-04 07:39:13 +08:00
|
|
|
static void put_nvmeq(struct nvme_queue *nvmeq) __releases(RCU)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_read_unlock();
|
2014-03-25 00:46:25 +08:00
|
|
|
put_cpu_var(nvmeq->dev->io_queue);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2014-03-04 07:39:13 +08:00
|
|
|
static struct nvme_queue *lock_nvmeq(struct nvme_dev *dev, int q_idx)
|
|
|
|
__acquires(RCU)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-05-14 00:32:46 +08:00
|
|
|
struct nvme_queue *nvmeq;
|
|
|
|
|
2014-03-04 07:39:13 +08:00
|
|
|
rcu_read_lock();
|
2014-05-14 00:32:46 +08:00
|
|
|
nvmeq = rcu_dereference(dev->queues[q_idx]);
|
|
|
|
if (nvmeq)
|
|
|
|
return nvmeq;
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
return NULL;
|
2014-03-04 07:39:13 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void unlock_nvmeq(struct nvme_queue *nvmeq) __releases(RCU)
|
|
|
|
{
|
|
|
|
rcu_read_unlock();
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2011-03-17 04:28:24 +08:00
|
|
|
* nvme_submit_cmd() - Copy a command into a queue and ring the doorbell
|
2011-01-21 01:50:14 +08:00
|
|
|
* @nvmeq: The queue to use
|
|
|
|
* @cmd: The command to send
|
|
|
|
*
|
|
|
|
* Safe to use from interrupt context
|
|
|
|
*/
|
|
|
|
static int nvme_submit_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
u16 tail;
|
|
|
|
spin_lock_irqsave(&nvmeq->q_lock, flags);
|
2014-03-04 07:39:13 +08:00
|
|
|
if (nvmeq->q_suspended) {
|
|
|
|
spin_unlock_irqrestore(&nvmeq->q_lock, flags);
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
tail = nvmeq->sq_tail;
|
|
|
|
memcpy(&nvmeq->sq_cmds[tail], cmd, sizeof(*cmd));
|
|
|
|
if (++tail == nvmeq->q_depth)
|
|
|
|
tail = 0;
|
2011-02-16 22:59:59 +08:00
|
|
|
writel(tail, nvmeq->q_db);
|
2011-01-21 01:50:14 +08:00
|
|
|
nvmeq->sq_tail = tail;
|
|
|
|
spin_unlock_irqrestore(&nvmeq->q_lock, flags);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
static __le64 **iod_list(struct nvme_iod *iod)
|
2011-02-10 21:51:24 +08:00
|
|
|
{
|
2011-12-21 02:34:52 +08:00
|
|
|
return ((void *)iod) + iod->offset;
|
2011-02-10 21:51:24 +08:00
|
|
|
}
|
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
/*
|
|
|
|
* Will slightly overestimate the number of pages needed. This is OK
|
|
|
|
* as it only leads to a small amount of wasted memory for the lifetime of
|
|
|
|
* the I/O.
|
|
|
|
*/
|
|
|
|
static int nvme_npages(unsigned size)
|
|
|
|
{
|
|
|
|
unsigned nprps = DIV_ROUND_UP(size + PAGE_SIZE, PAGE_SIZE);
|
|
|
|
return DIV_ROUND_UP(8 * nprps, PAGE_SIZE - 8);
|
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
static struct nvme_iod *
|
|
|
|
nvme_alloc_iod(unsigned nseg, unsigned nbytes, gfp_t gfp)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-12-21 02:34:52 +08:00
|
|
|
struct nvme_iod *iod = kmalloc(sizeof(struct nvme_iod) +
|
|
|
|
sizeof(__le64 *) * nvme_npages(nbytes) +
|
|
|
|
sizeof(struct scatterlist) * nseg, gfp);
|
|
|
|
|
|
|
|
if (iod) {
|
|
|
|
iod->offset = offsetof(struct nvme_iod, sg[nseg]);
|
|
|
|
iod->npages = -1;
|
|
|
|
iod->length = nbytes;
|
2012-11-07 02:59:23 +08:00
|
|
|
iod->nents = 0;
|
2014-04-04 06:45:23 +08:00
|
|
|
iod->first_dma = 0ULL;
|
2013-05-30 05:59:39 +08:00
|
|
|
iod->start_time = jiffies;
|
2011-12-21 02:34:52 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
return iod;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-12-21 02:34:52 +08:00
|
|
|
const int last_prp = PAGE_SIZE / 8 - 1;
|
|
|
|
int i;
|
|
|
|
__le64 **list = iod_list(iod);
|
|
|
|
dma_addr_t prp_dma = iod->first_dma;
|
|
|
|
|
|
|
|
if (iod->npages == 0)
|
|
|
|
dma_pool_free(dev->prp_small_pool, list[0], prp_dma);
|
|
|
|
for (i = 0; i < iod->npages; i++) {
|
|
|
|
__le64 *prp_list = list[i];
|
|
|
|
dma_addr_t next_prp_dma = le64_to_cpu(prp_list[last_prp]);
|
|
|
|
dma_pool_free(dev->prp_page_pool, prp_list, prp_dma);
|
|
|
|
prp_dma = next_prp_dma;
|
|
|
|
}
|
|
|
|
kfree(iod);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-05-30 05:59:39 +08:00
|
|
|
static void nvme_start_io_acct(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct gendisk *disk = bio->bi_bdev->bd_disk;
|
2014-05-10 04:27:07 +08:00
|
|
|
if (blk_queue_io_stat(disk->queue)) {
|
|
|
|
const int rw = bio_data_dir(bio);
|
|
|
|
int cpu = part_stat_lock();
|
|
|
|
part_round_stats(cpu, &disk->part0);
|
|
|
|
part_stat_inc(cpu, &disk->part0, ios[rw]);
|
|
|
|
part_stat_add(cpu, &disk->part0, sectors[rw],
|
|
|
|
bio_sectors(bio));
|
|
|
|
part_inc_in_flight(&disk->part0, rw);
|
|
|
|
part_stat_unlock();
|
|
|
|
}
|
2013-05-30 05:59:39 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_end_io_acct(struct bio *bio, unsigned long start_time)
|
|
|
|
{
|
|
|
|
struct gendisk *disk = bio->bi_bdev->bd_disk;
|
2014-05-10 04:27:07 +08:00
|
|
|
if (blk_queue_io_stat(disk->queue)) {
|
|
|
|
const int rw = bio_data_dir(bio);
|
|
|
|
unsigned long duration = jiffies - start_time;
|
|
|
|
int cpu = part_stat_lock();
|
|
|
|
part_stat_add(cpu, &disk->part0, ticks[rw], duration);
|
|
|
|
part_round_stats(cpu, &disk->part0);
|
|
|
|
part_dec_in_flight(&disk->part0, rw);
|
|
|
|
part_stat_unlock();
|
|
|
|
}
|
2013-05-30 05:59:39 +08:00
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static void bio_completion(struct nvme_queue *nvmeq, void *ctx,
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_completion *cqe)
|
|
|
|
{
|
2011-12-21 02:34:52 +08:00
|
|
|
struct nvme_iod *iod = ctx;
|
|
|
|
struct bio *bio = iod->private;
|
2011-01-21 01:50:14 +08:00
|
|
|
u16 status = le16_to_cpup(&cqe->status) >> 1;
|
2014-04-29 02:30:52 +08:00
|
|
|
int error = 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
if (unlikely(status)) {
|
|
|
|
if (!(status & NVME_SC_DNR ||
|
|
|
|
bio->bi_rw & REQ_FAILFAST_MASK) &&
|
|
|
|
(jiffies - iod->start_time) < IOD_TIMEOUT) {
|
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
|
|
|
add_wait_queue(&nvmeq->sq_full,
|
|
|
|
&nvmeq->sq_cong_wait);
|
|
|
|
list_add_tail(&iod->node, &nvmeq->iod_bio);
|
|
|
|
wake_up(&nvmeq->sq_full);
|
|
|
|
return;
|
|
|
|
}
|
2014-04-29 02:30:52 +08:00
|
|
|
error = -EIO;
|
2014-04-04 06:45:23 +08:00
|
|
|
}
|
2013-08-09 00:25:38 +08:00
|
|
|
if (iod->nents) {
|
2014-04-04 06:45:23 +08:00
|
|
|
dma_unmap_sg(nvmeq->q_dmadev, iod->sg, iod->nents,
|
2011-01-21 01:50:14 +08:00
|
|
|
bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
|
2013-08-09 00:25:38 +08:00
|
|
|
nvme_end_io_acct(bio, iod->start_time);
|
|
|
|
}
|
2014-04-04 06:45:23 +08:00
|
|
|
nvme_free_iod(nvmeq->dev, iod);
|
2014-04-29 02:30:52 +08:00
|
|
|
|
|
|
|
trace_block_bio_complete(bdev_get_queue(bio->bi_bdev), bio, error);
|
|
|
|
bio_endio(bio, error);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2011-05-12 09:36:38 +08:00
|
|
|
/* length is in bytes. gfp flags indicates whether we may sleep. */
|
2014-04-04 06:45:23 +08:00
|
|
|
int nvme_setup_prps(struct nvme_dev *dev, struct nvme_iod *iod, int total_len,
|
|
|
|
gfp_t gfp)
|
2011-01-26 23:02:29 +08:00
|
|
|
{
|
2011-02-10 23:30:34 +08:00
|
|
|
struct dma_pool *pool;
|
2011-12-21 02:34:52 +08:00
|
|
|
int length = total_len;
|
|
|
|
struct scatterlist *sg = iod->sg;
|
2011-01-26 23:02:29 +08:00
|
|
|
int dma_len = sg_dma_len(sg);
|
|
|
|
u64 dma_addr = sg_dma_address(sg);
|
|
|
|
int offset = offset_in_page(dma_addr);
|
2011-02-10 21:51:24 +08:00
|
|
|
__le64 *prp_list;
|
2011-12-21 02:34:52 +08:00
|
|
|
__le64 **list = iod_list(iod);
|
2011-02-10 21:51:24 +08:00
|
|
|
dma_addr_t prp_dma;
|
2011-12-21 02:34:52 +08:00
|
|
|
int nprps, i;
|
2011-01-26 23:02:29 +08:00
|
|
|
|
|
|
|
length -= (PAGE_SIZE - offset);
|
|
|
|
if (length <= 0)
|
2011-12-21 02:34:52 +08:00
|
|
|
return total_len;
|
2011-01-26 23:02:29 +08:00
|
|
|
|
|
|
|
dma_len -= (PAGE_SIZE - offset);
|
|
|
|
if (dma_len) {
|
|
|
|
dma_addr += (PAGE_SIZE - offset);
|
|
|
|
} else {
|
|
|
|
sg = sg_next(sg);
|
|
|
|
dma_addr = sg_dma_address(sg);
|
|
|
|
dma_len = sg_dma_len(sg);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (length <= PAGE_SIZE) {
|
2014-04-04 06:45:23 +08:00
|
|
|
iod->first_dma = dma_addr;
|
2011-12-21 02:34:52 +08:00
|
|
|
return total_len;
|
2011-02-10 21:51:24 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
nprps = DIV_ROUND_UP(length, PAGE_SIZE);
|
2011-02-10 23:30:34 +08:00
|
|
|
if (nprps <= (256 / 8)) {
|
|
|
|
pool = dev->prp_small_pool;
|
2011-12-21 02:34:52 +08:00
|
|
|
iod->npages = 0;
|
2011-02-10 23:30:34 +08:00
|
|
|
} else {
|
|
|
|
pool = dev->prp_page_pool;
|
2011-12-21 02:34:52 +08:00
|
|
|
iod->npages = 1;
|
2011-02-10 23:30:34 +08:00
|
|
|
}
|
|
|
|
|
2011-05-13 01:51:41 +08:00
|
|
|
prp_list = dma_pool_alloc(pool, gfp, &prp_dma);
|
|
|
|
if (!prp_list) {
|
2014-04-04 06:45:23 +08:00
|
|
|
iod->first_dma = dma_addr;
|
2011-12-21 02:34:52 +08:00
|
|
|
iod->npages = -1;
|
|
|
|
return (total_len - length) + PAGE_SIZE;
|
2011-05-13 01:51:41 +08:00
|
|
|
}
|
2011-12-21 02:34:52 +08:00
|
|
|
list[0] = prp_list;
|
|
|
|
iod->first_dma = prp_dma;
|
2011-02-10 21:51:24 +08:00
|
|
|
i = 0;
|
|
|
|
for (;;) {
|
2011-03-17 04:43:40 +08:00
|
|
|
if (i == PAGE_SIZE / 8) {
|
2011-02-10 21:51:24 +08:00
|
|
|
__le64 *old_prp_list = prp_list;
|
2011-05-13 01:51:41 +08:00
|
|
|
prp_list = dma_pool_alloc(pool, gfp, &prp_dma);
|
2011-12-21 02:34:52 +08:00
|
|
|
if (!prp_list)
|
|
|
|
return total_len - length;
|
|
|
|
list[iod->npages++] = prp_list;
|
2011-03-17 04:43:40 +08:00
|
|
|
prp_list[0] = old_prp_list[i - 1];
|
|
|
|
old_prp_list[i - 1] = cpu_to_le64(prp_dma);
|
|
|
|
i = 1;
|
2011-02-10 21:51:24 +08:00
|
|
|
}
|
|
|
|
prp_list[i++] = cpu_to_le64(dma_addr);
|
|
|
|
dma_len -= PAGE_SIZE;
|
|
|
|
dma_addr += PAGE_SIZE;
|
|
|
|
length -= PAGE_SIZE;
|
|
|
|
if (length <= 0)
|
|
|
|
break;
|
|
|
|
if (dma_len > 0)
|
|
|
|
continue;
|
|
|
|
BUG_ON(dma_len < 0);
|
|
|
|
sg = sg_next(sg);
|
|
|
|
dma_addr = sg_dma_address(sg);
|
|
|
|
dma_len = sg_dma_len(sg);
|
2011-01-26 23:02:29 +08:00
|
|
|
}
|
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
return total_len;
|
2011-01-26 23:02:29 +08:00
|
|
|
}
|
|
|
|
|
2013-04-10 01:59:32 +08:00
|
|
|
static int nvme_split_and_submit(struct bio *bio, struct nvme_queue *nvmeq,
|
2013-11-24 10:21:01 +08:00
|
|
|
int len)
|
2013-04-10 01:59:32 +08:00
|
|
|
{
|
2013-11-24 10:21:01 +08:00
|
|
|
struct bio *split = bio_split(bio, len >> 9, GFP_ATOMIC, NULL);
|
|
|
|
if (!split)
|
2013-04-10 01:59:32 +08:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2014-04-29 02:30:52 +08:00
|
|
|
trace_block_split(bdev_get_queue(bio->bi_bdev), bio,
|
|
|
|
split->bi_iter.bi_sector);
|
2013-11-24 10:21:01 +08:00
|
|
|
bio_chain(split, bio);
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
2013-04-10 01:59:32 +08:00
|
|
|
add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
|
2013-11-24 10:21:01 +08:00
|
|
|
bio_list_add(&nvmeq->sq_cong, split);
|
|
|
|
bio_list_add(&nvmeq->sq_cong, bio);
|
2014-04-04 06:45:23 +08:00
|
|
|
wake_up(&nvmeq->sq_full);
|
2013-04-10 01:59:32 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-02-24 04:20:00 +08:00
|
|
|
/* NVMe scatterlists require no holes in the virtual address */
|
|
|
|
#define BIOVEC_NOT_VIRT_MERGEABLE(vec1, vec2) ((vec2)->bv_offset || \
|
|
|
|
(((vec1)->bv_offset + (vec1)->bv_len) % PAGE_SIZE))
|
|
|
|
|
2013-04-10 01:59:32 +08:00
|
|
|
static int nvme_map_bio(struct nvme_queue *nvmeq, struct nvme_iod *iod,
|
2011-01-21 01:50:14 +08:00
|
|
|
struct bio *bio, enum dma_data_direction dma_dir, int psegs)
|
|
|
|
{
|
2013-11-24 09:19:00 +08:00
|
|
|
struct bio_vec bvec, bvprv;
|
|
|
|
struct bvec_iter iter;
|
2011-02-11 02:55:39 +08:00
|
|
|
struct scatterlist *sg = NULL;
|
2013-11-24 09:19:00 +08:00
|
|
|
int length = 0, nsegs = 0, split_len = bio->bi_iter.bi_size;
|
|
|
|
int first = 1;
|
2013-04-10 07:13:20 +08:00
|
|
|
|
|
|
|
if (nvmeq->dev->stripe_size)
|
|
|
|
split_len = nvmeq->dev->stripe_size -
|
2013-10-12 06:44:27 +08:00
|
|
|
((bio->bi_iter.bi_sector << 9) &
|
|
|
|
(nvmeq->dev->stripe_size - 1));
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
sg_init_table(iod->sg, psegs);
|
2013-11-24 09:19:00 +08:00
|
|
|
bio_for_each_segment(bvec, bio, iter) {
|
|
|
|
if (!first && BIOVEC_PHYS_MERGEABLE(&bvprv, &bvec)) {
|
|
|
|
sg->length += bvec.bv_len;
|
2011-02-11 02:55:39 +08:00
|
|
|
} else {
|
2013-11-24 09:19:00 +08:00
|
|
|
if (!first && BIOVEC_NOT_VIRT_MERGEABLE(&bvprv, &bvec))
|
|
|
|
return nvme_split_and_submit(bio, nvmeq,
|
2013-11-24 10:21:01 +08:00
|
|
|
length);
|
2013-04-10 01:59:32 +08:00
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
sg = sg ? sg + 1 : iod->sg;
|
2013-11-24 09:19:00 +08:00
|
|
|
sg_set_page(sg, bvec.bv_page,
|
|
|
|
bvec.bv_len, bvec.bv_offset);
|
2011-02-11 02:55:39 +08:00
|
|
|
nsegs++;
|
|
|
|
}
|
2013-04-10 07:13:20 +08:00
|
|
|
|
2013-11-24 09:19:00 +08:00
|
|
|
if (split_len - length < bvec.bv_len)
|
2013-11-24 10:21:01 +08:00
|
|
|
return nvme_split_and_submit(bio, nvmeq, split_len);
|
2013-11-24 09:19:00 +08:00
|
|
|
length += bvec.bv_len;
|
2011-02-11 02:55:39 +08:00
|
|
|
bvprv = bvec;
|
2013-11-24 09:19:00 +08:00
|
|
|
first = 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
2011-12-21 02:34:52 +08:00
|
|
|
iod->nents = nsegs;
|
2011-02-11 02:55:39 +08:00
|
|
|
sg_mark_end(sg);
|
2013-04-10 01:59:32 +08:00
|
|
|
if (dma_map_sg(nvmeq->q_dmadev, iod->sg, iod->nents, dma_dir) == 0)
|
2011-02-24 04:20:00 +08:00
|
|
|
return -ENOMEM;
|
2013-04-10 01:59:32 +08:00
|
|
|
|
2013-10-12 06:44:27 +08:00
|
|
|
BUG_ON(length != bio->bi_iter.bi_size);
|
2011-02-24 04:20:00 +08:00
|
|
|
return length;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2012-11-10 07:33:05 +08:00
|
|
|
static int nvme_submit_discard(struct nvme_queue *nvmeq, struct nvme_ns *ns,
|
|
|
|
struct bio *bio, struct nvme_iod *iod, int cmdid)
|
|
|
|
{
|
2014-04-04 06:45:23 +08:00
|
|
|
struct nvme_dsm_range *range =
|
|
|
|
(struct nvme_dsm_range *)iod_list(iod)[0];
|
2012-11-10 07:33:05 +08:00
|
|
|
struct nvme_command *cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
|
|
|
|
|
|
|
|
range->cattr = cpu_to_le32(0);
|
2013-10-12 06:44:27 +08:00
|
|
|
range->nlb = cpu_to_le32(bio->bi_iter.bi_size >> ns->lba_shift);
|
|
|
|
range->slba = cpu_to_le64(nvme_block_nr(ns, bio->bi_iter.bi_sector));
|
2012-11-10 07:33:05 +08:00
|
|
|
|
|
|
|
memset(cmnd, 0, sizeof(*cmnd));
|
|
|
|
cmnd->dsm.opcode = nvme_cmd_dsm;
|
|
|
|
cmnd->dsm.command_id = cmdid;
|
|
|
|
cmnd->dsm.nsid = cpu_to_le32(ns->ns_id);
|
|
|
|
cmnd->dsm.prp1 = cpu_to_le64(iod->first_dma);
|
|
|
|
cmnd->dsm.nr = 0;
|
|
|
|
cmnd->dsm.attributes = cpu_to_le32(NVME_DSMGMT_AD);
|
|
|
|
|
|
|
|
if (++nvmeq->sq_tail == nvmeq->q_depth)
|
|
|
|
nvmeq->sq_tail = 0;
|
|
|
|
writel(nvmeq->sq_tail, nvmeq->q_db);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-02-23 03:18:30 +08:00
|
|
|
static int nvme_submit_flush(struct nvme_queue *nvmeq, struct nvme_ns *ns,
|
|
|
|
int cmdid)
|
|
|
|
{
|
|
|
|
struct nvme_command *cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
|
|
|
|
|
|
|
|
memset(cmnd, 0, sizeof(*cmnd));
|
|
|
|
cmnd->common.opcode = nvme_cmd_flush;
|
|
|
|
cmnd->common.command_id = cmdid;
|
|
|
|
cmnd->common.nsid = cpu_to_le32(ns->ns_id);
|
|
|
|
|
|
|
|
if (++nvmeq->sq_tail == nvmeq->q_depth)
|
|
|
|
nvmeq->sq_tail = 0;
|
|
|
|
writel(nvmeq->sq_tail, nvmeq->q_db);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-04-04 06:45:23 +08:00
|
|
|
struct bio *bio = iod->private;
|
|
|
|
struct nvme_ns *ns = bio->bi_bdev->bd_disk->private_data;
|
2011-01-26 23:02:29 +08:00
|
|
|
struct nvme_command *cmnd;
|
2014-04-04 06:45:23 +08:00
|
|
|
int cmdid;
|
2011-01-21 01:50:14 +08:00
|
|
|
u16 control;
|
|
|
|
u32 dsmgmt;
|
2011-02-23 03:18:30 +08:00
|
|
|
|
2011-12-21 02:53:01 +08:00
|
|
|
cmdid = alloc_cmdid(nvmeq, iod, bio_completion, NVME_IO_TIMEOUT);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (unlikely(cmdid < 0))
|
2014-04-04 06:45:23 +08:00
|
|
|
return cmdid;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
if (bio->bi_rw & REQ_DISCARD)
|
|
|
|
return nvme_submit_discard(nvmeq, ns, bio, iod, cmdid);
|
2014-04-30 01:41:29 +08:00
|
|
|
if (bio->bi_rw & REQ_FLUSH)
|
2011-02-23 03:18:30 +08:00
|
|
|
return nvme_submit_flush(nvmeq, ns, cmdid);
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
control = 0;
|
|
|
|
if (bio->bi_rw & REQ_FUA)
|
|
|
|
control |= NVME_RW_FUA;
|
|
|
|
if (bio->bi_rw & (REQ_FAILFAST_DEV | REQ_RAHEAD))
|
|
|
|
control |= NVME_RW_LR;
|
|
|
|
|
|
|
|
dsmgmt = 0;
|
|
|
|
if (bio->bi_rw & REQ_RAHEAD)
|
|
|
|
dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
|
|
|
|
|
2011-01-26 23:02:29 +08:00
|
|
|
cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
|
2011-01-26 23:08:25 +08:00
|
|
|
memset(cmnd, 0, sizeof(*cmnd));
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
cmnd->rw.opcode = bio_data_dir(bio) ? nvme_cmd_write : nvme_cmd_read;
|
2011-01-26 23:02:29 +08:00
|
|
|
cmnd->rw.command_id = cmdid;
|
|
|
|
cmnd->rw.nsid = cpu_to_le32(ns->ns_id);
|
2014-04-04 06:45:23 +08:00
|
|
|
cmnd->rw.prp1 = cpu_to_le64(sg_dma_address(iod->sg));
|
|
|
|
cmnd->rw.prp2 = cpu_to_le64(iod->first_dma);
|
2013-10-12 06:44:27 +08:00
|
|
|
cmnd->rw.slba = cpu_to_le64(nvme_block_nr(ns, bio->bi_iter.bi_sector));
|
2014-04-04 06:45:23 +08:00
|
|
|
cmnd->rw.length =
|
|
|
|
cpu_to_le16((bio->bi_iter.bi_size >> ns->lba_shift) - 1);
|
2011-01-26 23:02:29 +08:00
|
|
|
cmnd->rw.control = cpu_to_le16(control);
|
|
|
|
cmnd->rw.dsmgmt = cpu_to_le32(dsmgmt);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
if (++nvmeq->sq_tail == nvmeq->q_depth)
|
|
|
|
nvmeq->sq_tail = 0;
|
2011-02-16 22:59:59 +08:00
|
|
|
writel(nvmeq->sq_tail, nvmeq->q_db);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-02-11 01:01:09 +08:00
|
|
|
return 0;
|
2014-04-04 06:45:23 +08:00
|
|
|
}
|
|
|
|
|
2014-04-30 01:41:29 +08:00
|
|
|
static int nvme_split_flush_data(struct nvme_queue *nvmeq, struct bio *bio)
|
|
|
|
{
|
|
|
|
struct bio *split = bio_clone(bio, GFP_ATOMIC);
|
|
|
|
if (!split)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
split->bi_iter.bi_size = 0;
|
|
|
|
split->bi_phys_segments = 0;
|
|
|
|
bio->bi_rw &= ~REQ_FLUSH;
|
|
|
|
bio_chain(split, bio);
|
|
|
|
|
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
|
|
|
add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
|
|
|
|
bio_list_add(&nvmeq->sq_cong, split);
|
|
|
|
bio_list_add(&nvmeq->sq_cong, bio);
|
|
|
|
wake_up_process(nvme_thread);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
/*
|
|
|
|
* Called with local interrupts disabled and the q_lock held. May not sleep.
|
|
|
|
*/
|
|
|
|
static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
|
|
|
|
struct bio *bio)
|
|
|
|
{
|
|
|
|
struct nvme_iod *iod;
|
|
|
|
int psegs = bio_phys_segments(ns->queue, bio);
|
|
|
|
int result;
|
|
|
|
|
2014-04-30 01:41:29 +08:00
|
|
|
if ((bio->bi_rw & REQ_FLUSH) && psegs)
|
|
|
|
return nvme_split_flush_data(nvmeq, bio);
|
2014-04-04 06:45:23 +08:00
|
|
|
|
|
|
|
iod = nvme_alloc_iod(psegs, bio->bi_iter.bi_size, GFP_ATOMIC);
|
|
|
|
if (!iod)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
iod->private = bio;
|
|
|
|
if (bio->bi_rw & REQ_DISCARD) {
|
|
|
|
void *range;
|
|
|
|
/*
|
|
|
|
* We reuse the small pool to allocate the 16-byte range here
|
|
|
|
* as it is not worth having a special pool for these or
|
|
|
|
* additional cases to handle freeing the iod.
|
|
|
|
*/
|
|
|
|
range = dma_pool_alloc(nvmeq->dev->prp_small_pool,
|
|
|
|
GFP_ATOMIC,
|
|
|
|
&iod->first_dma);
|
|
|
|
if (!range) {
|
|
|
|
result = -ENOMEM;
|
|
|
|
goto free_iod;
|
|
|
|
}
|
|
|
|
iod_list(iod)[0] = (__le64 *)range;
|
|
|
|
iod->npages = 0;
|
|
|
|
} else if (psegs) {
|
|
|
|
result = nvme_map_bio(nvmeq, iod, bio,
|
|
|
|
bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
|
|
|
|
psegs);
|
|
|
|
if (result <= 0)
|
|
|
|
goto free_iod;
|
|
|
|
if (nvme_setup_prps(nvmeq->dev, iod, result, GFP_ATOMIC) !=
|
|
|
|
result) {
|
|
|
|
result = -ENOMEM;
|
|
|
|
goto free_iod;
|
|
|
|
}
|
|
|
|
nvme_start_io_acct(bio);
|
|
|
|
}
|
|
|
|
if (unlikely(nvme_submit_iod(nvmeq, iod))) {
|
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
|
|
|
add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
|
|
|
|
list_add_tail(&iod->node, &nvmeq->iod_bio);
|
|
|
|
}
|
|
|
|
return 0;
|
2011-02-11 01:01:09 +08:00
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
free_iod:
|
|
|
|
nvme_free_iod(nvmeq->dev, iod);
|
2011-02-15 04:55:33 +08:00
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-06-24 23:47:34 +08:00
|
|
|
static int nvme_process_cq(struct nvme_queue *nvmeq)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-01-21 02:24:06 +08:00
|
|
|
u16 head, phase;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
head = nvmeq->cq_head;
|
2011-01-21 02:24:06 +08:00
|
|
|
phase = nvmeq->cq_phase;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
for (;;) {
|
2011-10-15 19:33:46 +08:00
|
|
|
void *ctx;
|
|
|
|
nvme_completion_fn fn;
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_completion cqe = nvmeq->cqes[head];
|
2011-01-21 02:24:06 +08:00
|
|
|
if ((le16_to_cpu(cqe.status) & 1) != phase)
|
2011-01-21 01:50:14 +08:00
|
|
|
break;
|
|
|
|
nvmeq->sq_head = le16_to_cpu(cqe.sq_head);
|
|
|
|
if (++head == nvmeq->q_depth) {
|
|
|
|
head = 0;
|
2011-01-21 02:24:06 +08:00
|
|
|
phase = !phase;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2011-10-15 19:33:46 +08:00
|
|
|
ctx = free_cmdid(nvmeq, cqe.command_id, &fn);
|
2014-04-04 06:45:23 +08:00
|
|
|
fn(nvmeq, ctx, &cqe);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* If the controller ignores the cq head doorbell and continuously
|
|
|
|
* writes to the queue, it is theoretically possible to wrap around
|
|
|
|
* the queue twice and mistakenly return IRQ_NONE. Linux only
|
|
|
|
* requires that 0.1% of your interrupts are handled, so this isn't
|
|
|
|
* a big problem.
|
|
|
|
*/
|
2011-01-21 02:24:06 +08:00
|
|
|
if (head == nvmeq->cq_head && phase == nvmeq->cq_phase)
|
2013-06-24 23:47:34 +08:00
|
|
|
return 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-09-10 11:25:37 +08:00
|
|
|
writel(head, nvmeq->q_db + nvmeq->dev->db_stride);
|
2011-01-21 01:50:14 +08:00
|
|
|
nvmeq->cq_head = head;
|
2011-01-21 02:24:06 +08:00
|
|
|
nvmeq->cq_phase = phase;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-06-24 23:47:34 +08:00
|
|
|
nvmeq->cqe_seen = 1;
|
|
|
|
return 1;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-06-25 00:03:57 +08:00
|
|
|
static void nvme_make_request(struct request_queue *q, struct bio *bio)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns = q->queuedata;
|
|
|
|
struct nvme_queue *nvmeq = get_nvmeq(ns->dev);
|
|
|
|
int result = -EBUSY;
|
|
|
|
|
2013-07-16 05:02:23 +08:00
|
|
|
if (!nvmeq) {
|
|
|
|
bio_endio(bio, -EIO);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-06-25 00:03:57 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
if (!nvmeq->q_suspended && bio_list_empty(&nvmeq->sq_cong))
|
2013-06-25 00:03:57 +08:00
|
|
|
result = nvme_submit_bio_queue(nvmeq, ns, bio);
|
|
|
|
if (unlikely(result)) {
|
2014-04-04 06:45:23 +08:00
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
2013-06-25 00:03:57 +08:00
|
|
|
add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
|
|
|
|
bio_list_add(&nvmeq->sq_cong, bio);
|
|
|
|
}
|
|
|
|
|
|
|
|
nvme_process_cq(nvmeq);
|
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
|
|
|
put_nvmeq(nvmeq);
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
static irqreturn_t nvme_irq(int irq, void *data)
|
2011-02-06 20:28:06 +08:00
|
|
|
{
|
|
|
|
irqreturn_t result;
|
|
|
|
struct nvme_queue *nvmeq = data;
|
|
|
|
spin_lock(&nvmeq->q_lock);
|
2013-06-24 23:47:34 +08:00
|
|
|
nvme_process_cq(nvmeq);
|
|
|
|
result = nvmeq->cqe_seen ? IRQ_HANDLED : IRQ_NONE;
|
|
|
|
nvmeq->cqe_seen = 0;
|
2011-02-06 20:28:06 +08:00
|
|
|
spin_unlock(&nvmeq->q_lock);
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
static irqreturn_t nvme_irq_check(int irq, void *data)
|
|
|
|
{
|
|
|
|
struct nvme_queue *nvmeq = data;
|
|
|
|
struct nvme_completion cqe = nvmeq->cqes[nvmeq->cq_head];
|
|
|
|
if ((le16_to_cpu(cqe.status) & 1) != nvmeq->cq_phase)
|
|
|
|
return IRQ_NONE;
|
|
|
|
return IRQ_WAKE_THREAD;
|
|
|
|
}
|
|
|
|
|
2011-02-05 05:03:56 +08:00
|
|
|
static void nvme_abort_command(struct nvme_queue *nvmeq, int cmdid)
|
|
|
|
{
|
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2011-10-15 19:33:46 +08:00
|
|
|
cancel_cmdid(nvmeq, cmdid, NULL);
|
2011-02-05 05:03:56 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
|
|
|
}
|
|
|
|
|
2011-10-15 19:33:46 +08:00
|
|
|
struct sync_cmd_info {
|
|
|
|
struct task_struct *task;
|
|
|
|
u32 result;
|
|
|
|
int status;
|
|
|
|
};
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static void sync_completion(struct nvme_queue *nvmeq, void *ctx,
|
2011-10-15 19:33:46 +08:00
|
|
|
struct nvme_completion *cqe)
|
|
|
|
{
|
|
|
|
struct sync_cmd_info *cmdinfo = ctx;
|
|
|
|
cmdinfo->result = le32_to_cpup(&cqe->result);
|
|
|
|
cmdinfo->status = le16_to_cpup(&cqe->status) >> 1;
|
|
|
|
wake_up_process(cmdinfo->task);
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
/*
|
|
|
|
* Returns 0 on success. If the result is negative, it's a Linux error code;
|
|
|
|
* if the result is positive, it's an NVM Express status code
|
|
|
|
*/
|
2014-03-04 07:39:13 +08:00
|
|
|
static int nvme_submit_sync_cmd(struct nvme_dev *dev, int q_idx,
|
|
|
|
struct nvme_command *cmd,
|
2013-03-05 09:40:58 +08:00
|
|
|
u32 *result, unsigned timeout)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-03-04 07:39:13 +08:00
|
|
|
int cmdid, ret;
|
2011-01-21 01:50:14 +08:00
|
|
|
struct sync_cmd_info cmdinfo;
|
2014-03-04 07:39:13 +08:00
|
|
|
struct nvme_queue *nvmeq;
|
|
|
|
|
|
|
|
nvmeq = lock_nvmeq(dev, q_idx);
|
2014-05-14 00:32:46 +08:00
|
|
|
if (!nvmeq)
|
2014-03-04 07:39:13 +08:00
|
|
|
return -ENODEV;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
cmdinfo.task = current;
|
|
|
|
cmdinfo.status = -EINTR;
|
|
|
|
|
2014-03-04 07:39:13 +08:00
|
|
|
cmdid = alloc_cmdid(nvmeq, &cmdinfo, sync_completion, timeout);
|
|
|
|
if (cmdid < 0) {
|
|
|
|
unlock_nvmeq(nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
return cmdid;
|
2014-03-04 07:39:13 +08:00
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
cmd->common.command_id = cmdid;
|
|
|
|
|
2011-02-05 05:03:56 +08:00
|
|
|
set_current_state(TASK_KILLABLE);
|
2014-03-04 07:39:13 +08:00
|
|
|
ret = nvme_submit_cmd(nvmeq, cmd);
|
|
|
|
if (ret) {
|
|
|
|
free_cmdid(nvmeq, cmdid, NULL);
|
|
|
|
unlock_nvmeq(nvmeq);
|
|
|
|
set_current_state(TASK_RUNNING);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
unlock_nvmeq(nvmeq);
|
2013-04-20 04:11:06 +08:00
|
|
|
schedule_timeout(timeout);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-02-05 05:03:56 +08:00
|
|
|
if (cmdinfo.status == -EINTR) {
|
2014-03-04 07:39:13 +08:00
|
|
|
nvmeq = lock_nvmeq(dev, q_idx);
|
2014-05-14 00:32:46 +08:00
|
|
|
if (nvmeq) {
|
2014-03-04 07:39:13 +08:00
|
|
|
nvme_abort_command(nvmeq, cmdid);
|
2014-05-14 00:32:46 +08:00
|
|
|
unlock_nvmeq(nvmeq);
|
|
|
|
}
|
2011-02-05 05:03:56 +08:00
|
|
|
return -EINTR;
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
if (result)
|
|
|
|
*result = cmdinfo.result;
|
|
|
|
|
|
|
|
return cmdinfo.status;
|
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
static int nvme_submit_async_cmd(struct nvme_queue *nvmeq,
|
|
|
|
struct nvme_command *cmd,
|
|
|
|
struct async_cmd_info *cmdinfo, unsigned timeout)
|
|
|
|
{
|
|
|
|
int cmdid;
|
|
|
|
|
|
|
|
cmdid = alloc_cmdid_killable(nvmeq, cmdinfo, async_completion, timeout);
|
|
|
|
if (cmdid < 0)
|
|
|
|
return cmdid;
|
|
|
|
cmdinfo->status = -EINTR;
|
|
|
|
cmd->common.command_id = cmdid;
|
2014-03-04 07:39:13 +08:00
|
|
|
return nvme_submit_cmd(nvmeq, cmd);
|
2013-12-11 04:10:40 +08:00
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
int nvme_submit_admin_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
|
2011-01-21 01:50:14 +08:00
|
|
|
u32 *result)
|
|
|
|
{
|
2014-03-04 07:39:13 +08:00
|
|
|
return nvme_submit_sync_cmd(dev, 0, cmd, result, ADMIN_TIMEOUT);
|
|
|
|
}
|
|
|
|
|
|
|
|
int nvme_submit_io_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
|
|
|
|
u32 *result)
|
|
|
|
{
|
|
|
|
return nvme_submit_sync_cmd(dev, smp_processor_id() + 1, cmd, result,
|
|
|
|
NVME_IO_TIMEOUT);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
static int nvme_submit_admin_cmd_async(struct nvme_dev *dev,
|
|
|
|
struct nvme_command *cmd, struct async_cmd_info *cmdinfo)
|
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
return nvme_submit_async_cmd(raw_nvmeq(dev, 0), cmd, cmdinfo,
|
2013-12-11 04:10:40 +08:00
|
|
|
ADMIN_TIMEOUT);
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
static int adapter_delete_queue(struct nvme_dev *dev, u8 opcode, u16 id)
|
|
|
|
{
|
|
|
|
int status;
|
|
|
|
struct nvme_command c;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.delete_queue.opcode = opcode;
|
|
|
|
c.delete_queue.qid = cpu_to_le16(id);
|
|
|
|
|
|
|
|
status = nvme_submit_admin_cmd(dev, &c, NULL);
|
|
|
|
if (status)
|
|
|
|
return -EIO;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int adapter_alloc_cq(struct nvme_dev *dev, u16 qid,
|
|
|
|
struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
int status;
|
|
|
|
struct nvme_command c;
|
|
|
|
int flags = NVME_QUEUE_PHYS_CONTIG | NVME_CQ_IRQ_ENABLED;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.create_cq.opcode = nvme_admin_create_cq;
|
|
|
|
c.create_cq.prp1 = cpu_to_le64(nvmeq->cq_dma_addr);
|
|
|
|
c.create_cq.cqid = cpu_to_le16(qid);
|
|
|
|
c.create_cq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
|
|
|
|
c.create_cq.cq_flags = cpu_to_le16(flags);
|
|
|
|
c.create_cq.irq_vector = cpu_to_le16(nvmeq->cq_vector);
|
|
|
|
|
|
|
|
status = nvme_submit_admin_cmd(dev, &c, NULL);
|
|
|
|
if (status)
|
|
|
|
return -EIO;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int adapter_alloc_sq(struct nvme_dev *dev, u16 qid,
|
|
|
|
struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
int status;
|
|
|
|
struct nvme_command c;
|
|
|
|
int flags = NVME_QUEUE_PHYS_CONTIG | NVME_SQ_PRIO_MEDIUM;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.create_sq.opcode = nvme_admin_create_sq;
|
|
|
|
c.create_sq.prp1 = cpu_to_le64(nvmeq->sq_dma_addr);
|
|
|
|
c.create_sq.sqid = cpu_to_le16(qid);
|
|
|
|
c.create_sq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
|
|
|
|
c.create_sq.sq_flags = cpu_to_le16(flags);
|
|
|
|
c.create_sq.cqid = cpu_to_le16(qid);
|
|
|
|
|
|
|
|
status = nvme_submit_admin_cmd(dev, &c, NULL);
|
|
|
|
if (status)
|
|
|
|
return -EIO;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int adapter_delete_cq(struct nvme_dev *dev, u16 cqid)
|
|
|
|
{
|
|
|
|
return adapter_delete_queue(dev, nvme_admin_delete_cq, cqid);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int adapter_delete_sq(struct nvme_dev *dev, u16 sqid)
|
|
|
|
{
|
|
|
|
return adapter_delete_queue(dev, nvme_admin_delete_sq, sqid);
|
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
int nvme_identify(struct nvme_dev *dev, unsigned nsid, unsigned cns,
|
2011-09-20 05:08:14 +08:00
|
|
|
dma_addr_t dma_addr)
|
|
|
|
{
|
|
|
|
struct nvme_command c;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.identify.opcode = nvme_admin_identify;
|
|
|
|
c.identify.nsid = cpu_to_le32(nsid);
|
|
|
|
c.identify.prp1 = cpu_to_le64(dma_addr);
|
|
|
|
c.identify.cns = cpu_to_le32(cns);
|
|
|
|
|
|
|
|
return nvme_submit_admin_cmd(dev, &c, NULL);
|
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
|
2012-09-22 00:52:13 +08:00
|
|
|
dma_addr_t dma_addr, u32 *result)
|
2011-09-20 05:08:14 +08:00
|
|
|
{
|
|
|
|
struct nvme_command c;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.features.opcode = nvme_admin_get_features;
|
2012-07-26 06:06:38 +08:00
|
|
|
c.features.nsid = cpu_to_le32(nsid);
|
2011-09-20 05:08:14 +08:00
|
|
|
c.features.prp1 = cpu_to_le64(dma_addr);
|
|
|
|
c.features.fid = cpu_to_le32(fid);
|
|
|
|
|
2012-09-22 00:52:13 +08:00
|
|
|
return nvme_submit_admin_cmd(dev, &c, result);
|
2012-01-11 22:29:56 +08:00
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11,
|
|
|
|
dma_addr_t dma_addr, u32 *result)
|
2012-01-11 22:29:56 +08:00
|
|
|
{
|
|
|
|
struct nvme_command c;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.features.opcode = nvme_admin_set_features;
|
|
|
|
c.features.prp1 = cpu_to_le64(dma_addr);
|
|
|
|
c.features.fid = cpu_to_le32(fid);
|
|
|
|
c.features.dword11 = cpu_to_le32(dword11);
|
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
return nvme_submit_admin_cmd(dev, &c, result);
|
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:38 +08:00
|
|
|
/**
|
|
|
|
* nvme_abort_cmd - Attempt aborting a command
|
|
|
|
* @cmdid: Command id of a timed out IO
|
|
|
|
* @queue: The queue with timed out IO
|
|
|
|
*
|
|
|
|
* Schedule controller reset if the command was already aborted once before and
|
|
|
|
* still hasn't been returned to the driver, or if this is the admin queue.
|
|
|
|
*/
|
|
|
|
static void nvme_abort_cmd(int cmdid, struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
int a_cmdid;
|
|
|
|
struct nvme_command cmd;
|
|
|
|
struct nvme_dev *dev = nvmeq->dev;
|
|
|
|
struct nvme_cmd_info *info = nvme_cmd_info(nvmeq);
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *adminq;
|
2013-12-11 04:10:38 +08:00
|
|
|
|
|
|
|
if (!nvmeq->qid || info[cmdid].aborted) {
|
|
|
|
if (work_busy(&dev->reset_work))
|
|
|
|
return;
|
|
|
|
list_del_init(&dev->node);
|
|
|
|
dev_warn(&dev->pci_dev->dev,
|
|
|
|
"I/O %d QID %d timeout, reset controller\n", cmdid,
|
|
|
|
nvmeq->qid);
|
2014-03-07 23:24:49 +08:00
|
|
|
dev->reset_workfn = nvme_reset_failed_dev;
|
2013-12-11 04:10:38 +08:00
|
|
|
queue_work(nvme_workq, &dev->reset_work);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!dev->abort_limit)
|
|
|
|
return;
|
|
|
|
|
2014-02-22 05:13:44 +08:00
|
|
|
adminq = rcu_dereference(dev->queues[0]);
|
|
|
|
a_cmdid = alloc_cmdid(adminq, CMD_CTX_ABORT, special_completion,
|
2013-12-11 04:10:38 +08:00
|
|
|
ADMIN_TIMEOUT);
|
|
|
|
if (a_cmdid < 0)
|
|
|
|
return;
|
|
|
|
|
|
|
|
memset(&cmd, 0, sizeof(cmd));
|
|
|
|
cmd.abort.opcode = nvme_admin_abort_cmd;
|
|
|
|
cmd.abort.cid = cmdid;
|
|
|
|
cmd.abort.sqid = cpu_to_le16(nvmeq->qid);
|
|
|
|
cmd.abort.command_id = a_cmdid;
|
|
|
|
|
|
|
|
--dev->abort_limit;
|
|
|
|
info[cmdid].aborted = 1;
|
|
|
|
info[cmdid].timeout = jiffies + ADMIN_TIMEOUT;
|
|
|
|
|
|
|
|
dev_warn(nvmeq->q_dmadev, "Aborting I/O %d QID %d\n", cmdid,
|
|
|
|
nvmeq->qid);
|
2014-02-22 05:13:44 +08:00
|
|
|
nvme_submit_cmd(adminq, &cmd);
|
2013-12-11 04:10:38 +08:00
|
|
|
}
|
|
|
|
|
2012-08-08 03:56:23 +08:00
|
|
|
/**
|
|
|
|
* nvme_cancel_ios - Cancel outstanding I/Os
|
|
|
|
* @queue: The queue to cancel I/Os on
|
|
|
|
* @timeout: True to only cancel I/Os which have timed out
|
|
|
|
*/
|
|
|
|
static void nvme_cancel_ios(struct nvme_queue *nvmeq, bool timeout)
|
|
|
|
{
|
|
|
|
int depth = nvmeq->q_depth - 1;
|
|
|
|
struct nvme_cmd_info *info = nvme_cmd_info(nvmeq);
|
|
|
|
unsigned long now = jiffies;
|
|
|
|
int cmdid;
|
|
|
|
|
|
|
|
for_each_set_bit(cmdid, nvmeq->cmdid_data, depth) {
|
|
|
|
void *ctx;
|
|
|
|
nvme_completion_fn fn;
|
|
|
|
static struct nvme_completion cqe = {
|
2013-04-17 03:18:30 +08:00
|
|
|
.status = cpu_to_le16(NVME_SC_ABORT_REQ << 1),
|
2012-08-08 03:56:23 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
if (timeout && !time_after(now, info[cmdid].timeout))
|
|
|
|
continue;
|
2013-05-01 01:19:38 +08:00
|
|
|
if (info[cmdid].ctx == CMD_CTX_CANCELLED)
|
|
|
|
continue;
|
2013-12-11 04:10:38 +08:00
|
|
|
if (timeout && nvmeq->dev->initialized) {
|
|
|
|
nvme_abort_cmd(cmdid, nvmeq);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
dev_warn(nvmeq->q_dmadev, "Cancelling I/O %d QID %d\n", cmdid,
|
|
|
|
nvmeq->qid);
|
2012-08-08 03:56:23 +08:00
|
|
|
ctx = cancel_cmdid(nvmeq, cmdid, &fn);
|
2014-04-04 06:45:23 +08:00
|
|
|
fn(nvmeq, ctx, &cqe);
|
2012-08-08 03:56:23 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-22 05:13:44 +08:00
|
|
|
static void nvme_free_queue(struct rcu_head *r)
|
2012-08-04 01:55:56 +08:00
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq = container_of(r, struct nvme_queue, r_head);
|
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
|
|
|
while (bio_list_peek(&nvmeq->sq_cong)) {
|
|
|
|
struct bio *bio = bio_list_pop(&nvmeq->sq_cong);
|
|
|
|
bio_endio(bio, -EIO);
|
|
|
|
}
|
2014-04-04 06:45:23 +08:00
|
|
|
while (!list_empty(&nvmeq->iod_bio)) {
|
|
|
|
static struct nvme_completion cqe = {
|
|
|
|
.status = cpu_to_le16(
|
|
|
|
(NVME_SC_ABORT_REQ | NVME_SC_DNR) << 1),
|
|
|
|
};
|
|
|
|
struct nvme_iod *iod = list_first_entry(&nvmeq->iod_bio,
|
|
|
|
struct nvme_iod,
|
|
|
|
node);
|
|
|
|
list_del(&iod->node);
|
|
|
|
bio_completion(nvmeq, iod, &cqe);
|
|
|
|
}
|
2013-07-16 05:02:20 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
|
|
|
|
2012-08-04 01:55:56 +08:00
|
|
|
dma_free_coherent(nvmeq->q_dmadev, CQ_SIZE(nvmeq->q_depth),
|
|
|
|
(void *)nvmeq->cqes, nvmeq->cq_dma_addr);
|
|
|
|
dma_free_coherent(nvmeq->q_dmadev, SQ_SIZE(nvmeq->q_depth),
|
|
|
|
nvmeq->sq_cmds, nvmeq->sq_dma_addr);
|
2014-03-25 00:46:25 +08:00
|
|
|
if (nvmeq->qid)
|
|
|
|
free_cpumask_var(nvmeq->cpu_mask);
|
2012-08-04 01:55:56 +08:00
|
|
|
kfree(nvmeq);
|
|
|
|
}
|
|
|
|
|
2013-12-17 02:50:00 +08:00
|
|
|
static void nvme_free_queues(struct nvme_dev *dev, int lowest)
|
2013-07-16 05:02:20 +08:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2013-12-17 02:50:00 +08:00
|
|
|
for (i = dev->queue_count - 1; i >= lowest; i--) {
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
|
|
|
|
rcu_assign_pointer(dev->queues[i], NULL);
|
|
|
|
call_rcu(&nvmeq->r_head, nvme_free_queue);
|
2013-07-16 05:02:20 +08:00
|
|
|
dev->queue_count--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
/**
|
|
|
|
* nvme_suspend_queue - put queue into suspended state
|
|
|
|
* @nvmeq - queue to suspend
|
|
|
|
*
|
|
|
|
* Returns 1 if already suspended, 0 otherwise.
|
|
|
|
*/
|
|
|
|
static int nvme_suspend_queue(struct nvme_queue *nvmeq)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-12-11 04:10:40 +08:00
|
|
|
int vector = nvmeq->dev->entry[nvmeq->cq_vector].vector;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2012-08-08 03:56:23 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
if (nvmeq->q_suspended) {
|
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
2013-12-11 04:10:40 +08:00
|
|
|
return 1;
|
2012-08-21 04:57:49 +08:00
|
|
|
}
|
2013-07-16 05:02:20 +08:00
|
|
|
nvmeq->q_suspended = 1;
|
2014-03-25 00:46:25 +08:00
|
|
|
nvmeq->dev->online_queues--;
|
2012-08-08 03:56:23 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
|
|
|
|
2011-03-27 20:52:06 +08:00
|
|
|
irq_set_affinity_hint(vector, NULL);
|
|
|
|
free_irq(vector, nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
return 0;
|
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
static void nvme_clear_queue(struct nvme_queue *nvmeq)
|
|
|
|
{
|
2013-07-16 05:02:20 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
|
|
|
nvme_process_cq(nvmeq);
|
|
|
|
nvme_cancel_ios(nvmeq, false);
|
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
static void nvme_disable_queue(struct nvme_dev *dev, int qid)
|
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq = raw_nvmeq(dev, qid);
|
2013-12-11 04:10:40 +08:00
|
|
|
|
|
|
|
if (!nvmeq)
|
|
|
|
return;
|
|
|
|
if (nvme_suspend_queue(nvmeq))
|
|
|
|
return;
|
|
|
|
|
2013-12-11 04:10:39 +08:00
|
|
|
/* Don't tell the adapter to delete the admin queue.
|
|
|
|
* Don't tell a removed adapter to delete IO queues. */
|
|
|
|
if (qid && readl(&dev->bar->csts) != -1) {
|
2011-01-21 01:50:14 +08:00
|
|
|
adapter_delete_sq(dev, qid);
|
|
|
|
adapter_delete_cq(dev, qid);
|
|
|
|
}
|
2013-12-11 04:10:40 +08:00
|
|
|
nvme_clear_queue(nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid,
|
|
|
|
int depth, int vector)
|
|
|
|
{
|
|
|
|
struct device *dmadev = &dev->pci_dev->dev;
|
2013-07-16 05:02:20 +08:00
|
|
|
unsigned extra = nvme_queue_extra(depth);
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_queue *nvmeq = kzalloc(sizeof(*nvmeq) + extra, GFP_KERNEL);
|
|
|
|
if (!nvmeq)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
nvmeq->cqes = dma_alloc_coherent(dmadev, CQ_SIZE(depth),
|
|
|
|
&nvmeq->cq_dma_addr, GFP_KERNEL);
|
|
|
|
if (!nvmeq->cqes)
|
|
|
|
goto free_nvmeq;
|
|
|
|
memset((void *)nvmeq->cqes, 0, CQ_SIZE(depth));
|
|
|
|
|
|
|
|
nvmeq->sq_cmds = dma_alloc_coherent(dmadev, SQ_SIZE(depth),
|
|
|
|
&nvmeq->sq_dma_addr, GFP_KERNEL);
|
|
|
|
if (!nvmeq->sq_cmds)
|
|
|
|
goto free_cqdma;
|
|
|
|
|
2014-03-25 00:46:25 +08:00
|
|
|
if (qid && !zalloc_cpumask_var(&nvmeq->cpu_mask, GFP_KERNEL))
|
|
|
|
goto free_sqdma;
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
nvmeq->q_dmadev = dmadev;
|
2011-02-10 22:56:01 +08:00
|
|
|
nvmeq->dev = dev;
|
2014-01-28 04:57:22 +08:00
|
|
|
snprintf(nvmeq->irqname, sizeof(nvmeq->irqname), "nvme%dq%d",
|
|
|
|
dev->instance, qid);
|
2011-01-21 01:50:14 +08:00
|
|
|
spin_lock_init(&nvmeq->q_lock);
|
|
|
|
nvmeq->cq_head = 0;
|
2011-01-21 02:24:06 +08:00
|
|
|
nvmeq->cq_phase = 1;
|
2011-01-21 01:50:14 +08:00
|
|
|
init_waitqueue_head(&nvmeq->sq_full);
|
2011-03-03 07:37:18 +08:00
|
|
|
init_waitqueue_entry(&nvmeq->sq_cong_wait, nvme_thread);
|
2011-01-21 01:50:14 +08:00
|
|
|
bio_list_init(&nvmeq->sq_cong);
|
2014-04-04 06:45:23 +08:00
|
|
|
INIT_LIST_HEAD(&nvmeq->iod_bio);
|
2013-09-10 11:25:37 +08:00
|
|
|
nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
|
2011-01-21 01:50:14 +08:00
|
|
|
nvmeq->q_depth = depth;
|
|
|
|
nvmeq->cq_vector = vector;
|
2013-12-11 04:10:38 +08:00
|
|
|
nvmeq->qid = qid;
|
2013-07-16 05:02:20 +08:00
|
|
|
nvmeq->q_suspended = 1;
|
|
|
|
dev->queue_count++;
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_assign_pointer(dev->queues[qid], nvmeq);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
return nvmeq;
|
|
|
|
|
2014-03-25 00:46:25 +08:00
|
|
|
free_sqdma:
|
|
|
|
dma_free_coherent(dmadev, SQ_SIZE(depth), (void *)nvmeq->sq_cmds,
|
|
|
|
nvmeq->sq_dma_addr);
|
2011-01-21 01:50:14 +08:00
|
|
|
free_cqdma:
|
2013-05-02 03:07:47 +08:00
|
|
|
dma_free_coherent(dmadev, CQ_SIZE(depth), (void *)nvmeq->cqes,
|
2011-01-21 01:50:14 +08:00
|
|
|
nvmeq->cq_dma_addr);
|
|
|
|
free_nvmeq:
|
|
|
|
kfree(nvmeq);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2011-01-20 22:10:15 +08:00
|
|
|
static int queue_request_irq(struct nvme_dev *dev, struct nvme_queue *nvmeq,
|
|
|
|
const char *name)
|
|
|
|
{
|
2011-02-06 20:28:06 +08:00
|
|
|
if (use_threaded_interrupts)
|
|
|
|
return request_threaded_irq(dev->entry[nvmeq->cq_vector].vector,
|
2013-10-12 12:23:29 +08:00
|
|
|
nvme_irq_check, nvme_irq, IRQF_SHARED,
|
2011-02-06 20:28:06 +08:00
|
|
|
name, nvmeq);
|
2011-01-20 22:10:15 +08:00
|
|
|
return request_irq(dev->entry[nvmeq->cq_vector].vector, nvme_irq,
|
2013-10-12 12:23:29 +08:00
|
|
|
IRQF_SHARED, name, nvmeq);
|
2011-01-20 22:10:15 +08:00
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-07-16 05:02:20 +08:00
|
|
|
struct nvme_dev *dev = nvmeq->dev;
|
|
|
|
unsigned extra = nvme_queue_extra(nvmeq->q_depth);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
nvmeq->sq_tail = 0;
|
|
|
|
nvmeq->cq_head = 0;
|
|
|
|
nvmeq->cq_phase = 1;
|
2013-09-10 11:25:37 +08:00
|
|
|
nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
|
2013-07-16 05:02:20 +08:00
|
|
|
memset(nvmeq->cmdid_data, 0, extra);
|
|
|
|
memset((void *)nvmeq->cqes, 0, CQ_SIZE(nvmeq->q_depth));
|
|
|
|
nvme_cancel_ios(nvmeq, false);
|
|
|
|
nvmeq->q_suspended = 0;
|
2014-03-25 00:46:25 +08:00
|
|
|
dev->online_queues++;
|
2013-07-16 05:02:20 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_create_queue(struct nvme_queue *nvmeq, int qid)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = nvmeq->dev;
|
|
|
|
int result;
|
2011-02-01 21:39:04 +08:00
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
result = adapter_alloc_cq(dev, qid, nvmeq);
|
|
|
|
if (result < 0)
|
2013-07-16 05:02:20 +08:00
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
result = adapter_alloc_sq(dev, qid, nvmeq);
|
|
|
|
if (result < 0)
|
|
|
|
goto release_cq;
|
|
|
|
|
2014-01-28 04:57:22 +08:00
|
|
|
result = queue_request_irq(dev, nvmeq, nvmeq->irqname);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (result < 0)
|
|
|
|
goto release_sq;
|
|
|
|
|
2013-10-16 03:01:10 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
nvme_init_queue(nvmeq, qid);
|
2013-10-16 03:01:10 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
|
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
release_sq:
|
|
|
|
adapter_delete_sq(dev, qid);
|
|
|
|
release_cq:
|
|
|
|
adapter_delete_cq(dev, qid);
|
2013-07-16 05:02:20 +08:00
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-05-04 18:43:16 +08:00
|
|
|
static int nvme_wait_ready(struct nvme_dev *dev, u64 cap, bool enabled)
|
|
|
|
{
|
|
|
|
unsigned long timeout;
|
|
|
|
u32 bit = enabled ? NVME_CSTS_RDY : 0;
|
|
|
|
|
|
|
|
timeout = ((NVME_CAP_TIMEOUT(cap) + 1) * HZ / 2) + jiffies;
|
|
|
|
|
|
|
|
while ((readl(&dev->bar->csts) & NVME_CSTS_RDY) != bit) {
|
|
|
|
msleep(100);
|
|
|
|
if (fatal_signal_pending(current))
|
|
|
|
return -EINTR;
|
|
|
|
if (time_after(jiffies, timeout)) {
|
|
|
|
dev_err(&dev->pci_dev->dev,
|
2014-04-11 23:58:45 +08:00
|
|
|
"Device not ready; aborting %s\n", enabled ?
|
|
|
|
"initialisation" : "reset");
|
2013-05-04 18:43:16 +08:00
|
|
|
return -ENODEV;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the device has been passed off to us in an enabled state, just clear
|
|
|
|
* the enabled bit. The spec says we should set the 'shutdown notification
|
|
|
|
* bits', but doing so may cause the device to complete commands to the
|
|
|
|
* admin queue ... and we don't know what memory that might be pointing at!
|
|
|
|
*/
|
|
|
|
static int nvme_disable_ctrl(struct nvme_dev *dev, u64 cap)
|
|
|
|
{
|
2013-05-04 18:43:17 +08:00
|
|
|
u32 cc = readl(&dev->bar->cc);
|
|
|
|
|
|
|
|
if (cc & NVME_CC_ENABLE)
|
|
|
|
writel(cc & ~NVME_CC_ENABLE, &dev->bar->cc);
|
2013-05-04 18:43:16 +08:00
|
|
|
return nvme_wait_ready(dev, cap, false);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_enable_ctrl(struct nvme_dev *dev, u64 cap)
|
|
|
|
{
|
|
|
|
return nvme_wait_ready(dev, cap, true);
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:22 +08:00
|
|
|
static int nvme_shutdown_ctrl(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
unsigned long timeout;
|
|
|
|
u32 cc;
|
|
|
|
|
|
|
|
cc = (readl(&dev->bar->cc) & ~NVME_CC_SHN_MASK) | NVME_CC_SHN_NORMAL;
|
|
|
|
writel(cc, &dev->bar->cc);
|
|
|
|
|
|
|
|
timeout = 2 * HZ + jiffies;
|
|
|
|
while ((readl(&dev->bar->csts) & NVME_CSTS_SHST_MASK) !=
|
|
|
|
NVME_CSTS_SHST_CMPLT) {
|
|
|
|
msleep(100);
|
|
|
|
if (fatal_signal_pending(current))
|
|
|
|
return -EINTR;
|
|
|
|
if (time_after(jiffies, timeout)) {
|
|
|
|
dev_err(&dev->pci_dev->dev,
|
|
|
|
"Device shutdown incomplete; abort shutdown\n");
|
|
|
|
return -ENODEV;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-12-22 07:13:49 +08:00
|
|
|
static int nvme_configure_admin_queue(struct nvme_dev *dev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-05-04 18:43:16 +08:00
|
|
|
int result;
|
2011-01-21 01:50:14 +08:00
|
|
|
u32 aqa;
|
2013-05-04 18:43:16 +08:00
|
|
|
u64 cap = readq(&dev->bar->cap);
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_queue *nvmeq;
|
|
|
|
|
2013-05-04 18:43:16 +08:00
|
|
|
result = nvme_disable_ctrl(dev, cap);
|
|
|
|
if (result < 0)
|
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-02-22 05:13:44 +08:00
|
|
|
nvmeq = raw_nvmeq(dev, 0);
|
2013-07-16 05:02:23 +08:00
|
|
|
if (!nvmeq) {
|
|
|
|
nvmeq = nvme_alloc_queue(dev, 0, 64, 0);
|
|
|
|
if (!nvmeq)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
aqa = nvmeq->q_depth - 1;
|
|
|
|
aqa |= aqa << 16;
|
|
|
|
|
|
|
|
dev->ctrl_config = NVME_CC_ENABLE | NVME_CC_CSS_NVM;
|
|
|
|
dev->ctrl_config |= (PAGE_SHIFT - 12) << NVME_CC_MPS_SHIFT;
|
|
|
|
dev->ctrl_config |= NVME_CC_ARB_RR | NVME_CC_SHN_NONE;
|
2011-03-23 03:55:45 +08:00
|
|
|
dev->ctrl_config |= NVME_CC_IOSQES | NVME_CC_IOCQES;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
writel(aqa, &dev->bar->aqa);
|
|
|
|
writeq(nvmeq->sq_dma_addr, &dev->bar->asq);
|
|
|
|
writeq(nvmeq->cq_dma_addr, &dev->bar->acq);
|
|
|
|
writel(dev->ctrl_config, &dev->bar->cc);
|
|
|
|
|
2013-05-04 18:43:16 +08:00
|
|
|
result = nvme_enable_ctrl(dev, cap);
|
2013-05-02 03:07:51 +08:00
|
|
|
if (result)
|
2013-07-16 05:02:23 +08:00
|
|
|
return result;
|
2012-08-04 01:55:56 +08:00
|
|
|
|
2014-01-28 04:57:22 +08:00
|
|
|
result = queue_request_irq(dev, nvmeq, nvmeq->irqname);
|
2013-05-02 03:07:51 +08:00
|
|
|
if (result)
|
2013-07-16 05:02:23 +08:00
|
|
|
return result;
|
2013-05-02 03:07:51 +08:00
|
|
|
|
2013-10-16 03:01:10 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
nvme_init_queue(nvmeq, 0);
|
2013-10-16 03:01:10 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
2011-01-21 01:50:14 +08:00
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
|
2011-12-21 02:34:52 +08:00
|
|
|
unsigned long addr, unsigned length)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2011-01-24 20:52:07 +08:00
|
|
|
int i, err, count, nents, offset;
|
2011-01-27 06:05:50 +08:00
|
|
|
struct scatterlist *sg;
|
|
|
|
struct page **pages;
|
2011-12-21 02:34:52 +08:00
|
|
|
struct nvme_iod *iod;
|
2011-01-24 20:52:07 +08:00
|
|
|
|
|
|
|
if (addr & 3)
|
2011-12-21 02:34:52 +08:00
|
|
|
return ERR_PTR(-EINVAL);
|
2013-05-13 22:59:50 +08:00
|
|
|
if (!length || length > INT_MAX - PAGE_SIZE)
|
2011-12-21 02:34:52 +08:00
|
|
|
return ERR_PTR(-EINVAL);
|
2011-01-27 06:05:50 +08:00
|
|
|
|
2011-01-24 20:52:07 +08:00
|
|
|
offset = offset_in_page(addr);
|
2011-01-27 06:05:50 +08:00
|
|
|
count = DIV_ROUND_UP(offset + length, PAGE_SIZE);
|
|
|
|
pages = kcalloc(count, sizeof(*pages), GFP_KERNEL);
|
2012-01-20 20:55:30 +08:00
|
|
|
if (!pages)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
2011-01-24 20:52:07 +08:00
|
|
|
|
|
|
|
err = get_user_pages_fast(addr, count, 1, pages);
|
|
|
|
if (err < count) {
|
|
|
|
count = err;
|
|
|
|
err = -EFAULT;
|
|
|
|
goto put_pages;
|
|
|
|
}
|
2011-01-27 06:05:50 +08:00
|
|
|
|
2014-05-29 12:31:52 +08:00
|
|
|
err = -ENOMEM;
|
2011-12-21 02:34:52 +08:00
|
|
|
iod = nvme_alloc_iod(count, length, GFP_KERNEL);
|
2014-05-29 12:31:52 +08:00
|
|
|
if (!iod)
|
|
|
|
goto put_pages;
|
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
sg = iod->sg;
|
2011-01-24 20:52:07 +08:00
|
|
|
sg_init_table(sg, count);
|
2011-09-14 05:01:39 +08:00
|
|
|
for (i = 0; i < count; i++) {
|
|
|
|
sg_set_page(&sg[i], pages[i],
|
2013-05-13 22:59:50 +08:00
|
|
|
min_t(unsigned, length, PAGE_SIZE - offset),
|
|
|
|
offset);
|
2011-09-14 05:01:39 +08:00
|
|
|
length -= (PAGE_SIZE - offset);
|
|
|
|
offset = 0;
|
2011-01-27 06:05:50 +08:00
|
|
|
}
|
2012-01-07 04:49:25 +08:00
|
|
|
sg_mark_end(&sg[i - 1]);
|
2012-01-07 04:52:56 +08:00
|
|
|
iod->nents = count;
|
2011-01-27 06:05:50 +08:00
|
|
|
|
|
|
|
nents = dma_map_sg(&dev->pci_dev->dev, sg, count,
|
|
|
|
write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
|
2011-01-24 20:52:07 +08:00
|
|
|
if (!nents)
|
2011-12-21 02:34:52 +08:00
|
|
|
goto free_iod;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-01-27 06:05:50 +08:00
|
|
|
kfree(pages);
|
2011-12-21 02:34:52 +08:00
|
|
|
return iod;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
free_iod:
|
|
|
|
kfree(iod);
|
2011-01-27 06:05:50 +08:00
|
|
|
put_pages:
|
|
|
|
for (i = 0; i < count; i++)
|
|
|
|
put_page(pages[i]);
|
|
|
|
kfree(pages);
|
2011-12-21 02:34:52 +08:00
|
|
|
return ERR_PTR(err);
|
2011-01-27 06:05:50 +08:00
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-03-05 09:40:58 +08:00
|
|
|
void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
|
2012-01-07 04:52:56 +08:00
|
|
|
struct nvme_iod *iod)
|
2011-01-27 06:05:50 +08:00
|
|
|
{
|
2012-01-07 04:52:56 +08:00
|
|
|
int i;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2012-01-07 04:52:56 +08:00
|
|
|
dma_unmap_sg(&dev->pci_dev->dev, iod->sg, iod->nents,
|
|
|
|
write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
|
2011-01-27 06:05:50 +08:00
|
|
|
|
2012-01-07 04:52:56 +08:00
|
|
|
for (i = 0; i < iod->nents; i++)
|
|
|
|
put_page(sg_page(&iod->sg[i]));
|
2011-01-27 06:05:50 +08:00
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-02-02 05:13:29 +08:00
|
|
|
static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = ns->dev;
|
|
|
|
struct nvme_user_io io;
|
|
|
|
struct nvme_command c;
|
2013-04-24 07:23:59 +08:00
|
|
|
unsigned length, meta_len;
|
|
|
|
int status, i;
|
|
|
|
struct nvme_iod *iod, *meta_iod = NULL;
|
|
|
|
dma_addr_t meta_dma_addr;
|
|
|
|
void *meta, *uninitialized_var(meta_mem);
|
2011-02-02 05:13:29 +08:00
|
|
|
|
|
|
|
if (copy_from_user(&io, uio, sizeof(io)))
|
|
|
|
return -EFAULT;
|
2011-03-21 21:48:57 +08:00
|
|
|
length = (io.nblocks + 1) << ns->lba_shift;
|
2013-04-24 07:23:59 +08:00
|
|
|
meta_len = (io.nblocks + 1) * ns->ms;
|
|
|
|
|
|
|
|
if (meta_len && ((io.metadata & 3) || !io.metadata))
|
|
|
|
return -EINVAL;
|
2011-03-21 21:48:57 +08:00
|
|
|
|
|
|
|
switch (io.opcode) {
|
|
|
|
case nvme_cmd_write:
|
|
|
|
case nvme_cmd_read:
|
2011-05-21 01:03:42 +08:00
|
|
|
case nvme_cmd_compare:
|
2011-12-21 02:34:52 +08:00
|
|
|
iod = nvme_map_user_pages(dev, io.opcode & 1, io.addr, length);
|
2011-08-10 00:56:37 +08:00
|
|
|
break;
|
2011-03-21 21:48:57 +08:00
|
|
|
default:
|
2011-05-21 01:03:42 +08:00
|
|
|
return -EINVAL;
|
2011-03-21 21:48:57 +08:00
|
|
|
}
|
|
|
|
|
2011-12-21 02:34:52 +08:00
|
|
|
if (IS_ERR(iod))
|
|
|
|
return PTR_ERR(iod);
|
2011-02-02 05:13:29 +08:00
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.rw.opcode = io.opcode;
|
|
|
|
c.rw.flags = io.flags;
|
2011-03-21 21:48:57 +08:00
|
|
|
c.rw.nsid = cpu_to_le32(ns->ns_id);
|
2011-02-02 05:13:29 +08:00
|
|
|
c.rw.slba = cpu_to_le64(io.slba);
|
2011-03-21 21:48:57 +08:00
|
|
|
c.rw.length = cpu_to_le16(io.nblocks);
|
2011-02-02 05:13:29 +08:00
|
|
|
c.rw.control = cpu_to_le16(io.control);
|
2013-04-17 03:21:06 +08:00
|
|
|
c.rw.dsmgmt = cpu_to_le32(io.dsmgmt);
|
|
|
|
c.rw.reftag = cpu_to_le32(io.reftag);
|
|
|
|
c.rw.apptag = cpu_to_le16(io.apptag);
|
|
|
|
c.rw.appmask = cpu_to_le16(io.appmask);
|
2013-04-24 07:23:59 +08:00
|
|
|
|
|
|
|
if (meta_len) {
|
2013-07-19 02:13:51 +08:00
|
|
|
meta_iod = nvme_map_user_pages(dev, io.opcode & 1, io.metadata,
|
|
|
|
meta_len);
|
2013-04-24 07:23:59 +08:00
|
|
|
if (IS_ERR(meta_iod)) {
|
|
|
|
status = PTR_ERR(meta_iod);
|
|
|
|
meta_iod = NULL;
|
|
|
|
goto unmap;
|
|
|
|
}
|
|
|
|
|
|
|
|
meta_mem = dma_alloc_coherent(&dev->pci_dev->dev, meta_len,
|
|
|
|
&meta_dma_addr, GFP_KERNEL);
|
|
|
|
if (!meta_mem) {
|
|
|
|
status = -ENOMEM;
|
|
|
|
goto unmap;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (io.opcode & 1) {
|
|
|
|
int meta_offset = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < meta_iod->nents; i++) {
|
|
|
|
meta = kmap_atomic(sg_page(&meta_iod->sg[i])) +
|
|
|
|
meta_iod->sg[i].offset;
|
|
|
|
memcpy(meta_mem + meta_offset, meta,
|
|
|
|
meta_iod->sg[i].length);
|
|
|
|
kunmap_atomic(meta);
|
|
|
|
meta_offset += meta_iod->sg[i].length;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
c.rw.metadata = cpu_to_le64(meta_dma_addr);
|
|
|
|
}
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
length = nvme_setup_prps(dev, iod, length, GFP_KERNEL);
|
|
|
|
c.rw.prp1 = cpu_to_le64(sg_dma_address(iod->sg));
|
|
|
|
c.rw.prp2 = cpu_to_le64(iod->first_dma);
|
2011-02-02 05:13:29 +08:00
|
|
|
|
2011-05-13 01:51:41 +08:00
|
|
|
if (length != (io.nblocks + 1) << ns->lba_shift)
|
|
|
|
status = -ENOMEM;
|
|
|
|
else
|
2014-03-04 07:39:13 +08:00
|
|
|
status = nvme_submit_io_cmd(dev, &c, NULL);
|
2011-02-02 05:13:29 +08:00
|
|
|
|
2013-04-24 07:23:59 +08:00
|
|
|
if (meta_len) {
|
|
|
|
if (status == NVME_SC_SUCCESS && !(io.opcode & 1)) {
|
|
|
|
int meta_offset = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < meta_iod->nents; i++) {
|
|
|
|
meta = kmap_atomic(sg_page(&meta_iod->sg[i])) +
|
|
|
|
meta_iod->sg[i].offset;
|
|
|
|
memcpy(meta, meta_mem + meta_offset,
|
|
|
|
meta_iod->sg[i].length);
|
|
|
|
kunmap_atomic(meta);
|
|
|
|
meta_offset += meta_iod->sg[i].length;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
dma_free_coherent(&dev->pci_dev->dev, meta_len, meta_mem,
|
|
|
|
meta_dma_addr);
|
|
|
|
}
|
|
|
|
|
|
|
|
unmap:
|
2012-01-07 04:52:56 +08:00
|
|
|
nvme_unmap_user_pages(dev, io.opcode & 1, iod);
|
2011-12-21 02:34:52 +08:00
|
|
|
nvme_free_iod(dev, iod);
|
2013-04-24 07:23:59 +08:00
|
|
|
|
|
|
|
if (meta_iod) {
|
|
|
|
nvme_unmap_user_pages(dev, io.opcode & 1, meta_iod);
|
|
|
|
nvme_free_iod(dev, meta_iod);
|
|
|
|
}
|
|
|
|
|
2011-02-02 05:13:29 +08:00
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2012-07-26 06:07:55 +08:00
|
|
|
static int nvme_user_admin_cmd(struct nvme_dev *dev,
|
2011-05-21 01:03:42 +08:00
|
|
|
struct nvme_admin_cmd __user *ucmd)
|
2011-02-03 23:58:26 +08:00
|
|
|
{
|
2011-05-21 01:03:42 +08:00
|
|
|
struct nvme_admin_cmd cmd;
|
2011-02-03 23:58:26 +08:00
|
|
|
struct nvme_command c;
|
2011-12-21 02:34:52 +08:00
|
|
|
int status, length;
|
2012-07-28 01:53:28 +08:00
|
|
|
struct nvme_iod *uninitialized_var(iod);
|
2013-05-10 04:01:38 +08:00
|
|
|
unsigned timeout;
|
2011-02-03 23:58:26 +08:00
|
|
|
|
2011-05-21 01:03:42 +08:00
|
|
|
if (!capable(CAP_SYS_ADMIN))
|
|
|
|
return -EACCES;
|
|
|
|
if (copy_from_user(&cmd, ucmd, sizeof(cmd)))
|
2011-02-03 23:58:26 +08:00
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
2011-05-21 01:03:42 +08:00
|
|
|
c.common.opcode = cmd.opcode;
|
|
|
|
c.common.flags = cmd.flags;
|
|
|
|
c.common.nsid = cpu_to_le32(cmd.nsid);
|
|
|
|
c.common.cdw2[0] = cpu_to_le32(cmd.cdw2);
|
|
|
|
c.common.cdw2[1] = cpu_to_le32(cmd.cdw3);
|
|
|
|
c.common.cdw10[0] = cpu_to_le32(cmd.cdw10);
|
|
|
|
c.common.cdw10[1] = cpu_to_le32(cmd.cdw11);
|
|
|
|
c.common.cdw10[2] = cpu_to_le32(cmd.cdw12);
|
|
|
|
c.common.cdw10[3] = cpu_to_le32(cmd.cdw13);
|
|
|
|
c.common.cdw10[4] = cpu_to_le32(cmd.cdw14);
|
|
|
|
c.common.cdw10[5] = cpu_to_le32(cmd.cdw15);
|
|
|
|
|
|
|
|
length = cmd.data_len;
|
|
|
|
if (cmd.data_len) {
|
2012-01-07 04:42:45 +08:00
|
|
|
iod = nvme_map_user_pages(dev, cmd.opcode & 1, cmd.addr,
|
|
|
|
length);
|
2011-12-21 02:34:52 +08:00
|
|
|
if (IS_ERR(iod))
|
|
|
|
return PTR_ERR(iod);
|
2014-04-04 06:45:23 +08:00
|
|
|
length = nvme_setup_prps(dev, iod, length, GFP_KERNEL);
|
|
|
|
c.common.prp1 = cpu_to_le64(sg_dma_address(iod->sg));
|
|
|
|
c.common.prp2 = cpu_to_le64(iod->first_dma);
|
2011-05-21 01:03:42 +08:00
|
|
|
}
|
|
|
|
|
2013-05-10 04:01:38 +08:00
|
|
|
timeout = cmd.timeout_ms ? msecs_to_jiffies(cmd.timeout_ms) :
|
|
|
|
ADMIN_TIMEOUT;
|
2011-05-21 01:03:42 +08:00
|
|
|
if (length != cmd.data_len)
|
2011-05-13 01:51:41 +08:00
|
|
|
status = -ENOMEM;
|
|
|
|
else
|
2014-03-04 07:39:13 +08:00
|
|
|
status = nvme_submit_sync_cmd(dev, 0, &c, &cmd.result, timeout);
|
2011-12-21 02:34:52 +08:00
|
|
|
|
2011-05-21 01:03:42 +08:00
|
|
|
if (cmd.data_len) {
|
2012-01-07 04:52:56 +08:00
|
|
|
nvme_unmap_user_pages(dev, cmd.opcode & 1, iod);
|
2011-12-21 02:34:52 +08:00
|
|
|
nvme_free_iod(dev, iod);
|
2011-05-21 01:03:42 +08:00
|
|
|
}
|
2012-09-22 00:49:05 +08:00
|
|
|
|
2013-05-23 06:34:49 +08:00
|
|
|
if ((status >= 0) && copy_to_user(&ucmd->result, &cmd.result,
|
2012-09-22 00:49:05 +08:00
|
|
|
sizeof(cmd.result)))
|
|
|
|
status = -EFAULT;
|
|
|
|
|
2011-02-03 23:58:26 +08:00
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
static int nvme_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd,
|
|
|
|
unsigned long arg)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns = bdev->bd_disk->private_data;
|
|
|
|
|
|
|
|
switch (cmd) {
|
2011-05-21 01:03:42 +08:00
|
|
|
case NVME_IOCTL_ID:
|
2013-07-09 05:26:25 +08:00
|
|
|
force_successful_syscall_return();
|
2011-05-21 01:03:42 +08:00
|
|
|
return ns->ns_id;
|
|
|
|
case NVME_IOCTL_ADMIN_CMD:
|
2012-07-26 06:07:55 +08:00
|
|
|
return nvme_user_admin_cmd(ns->dev, (void __user *)arg);
|
2011-02-02 05:13:29 +08:00
|
|
|
case NVME_IOCTL_SUBMIT_IO:
|
|
|
|
return nvme_submit_io(ns, (void __user *)arg);
|
2013-03-05 09:40:58 +08:00
|
|
|
case SG_GET_VERSION_NUM:
|
|
|
|
return nvme_sg_get_version_num((void __user *)arg);
|
|
|
|
case SG_IO:
|
|
|
|
return nvme_sg_io(ns, (void __user *)arg);
|
2011-01-21 01:50:14 +08:00
|
|
|
default:
|
|
|
|
return -ENOTTY;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-10-24 03:07:34 +08:00
|
|
|
#ifdef CONFIG_COMPAT
|
|
|
|
static int nvme_compat_ioctl(struct block_device *bdev, fmode_t mode,
|
|
|
|
unsigned int cmd, unsigned long arg)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns = bdev->bd_disk->private_data;
|
|
|
|
|
|
|
|
switch (cmd) {
|
|
|
|
case SG_IO:
|
|
|
|
return nvme_sg_io32(ns, arg);
|
|
|
|
}
|
|
|
|
return nvme_ioctl(bdev, mode, cmd, arg);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
#define nvme_compat_ioctl NULL
|
|
|
|
#endif
|
|
|
|
|
2014-02-01 07:53:39 +08:00
|
|
|
static int nvme_open(struct block_device *bdev, fmode_t mode)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns = bdev->bd_disk->private_data;
|
|
|
|
struct nvme_dev *dev = ns->dev;
|
|
|
|
|
|
|
|
kref_get(&dev->kref);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_free_dev(struct kref *kref);
|
|
|
|
|
|
|
|
static void nvme_release(struct gendisk *disk, fmode_t mode)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns = disk->private_data;
|
|
|
|
struct nvme_dev *dev = ns->dev;
|
|
|
|
|
|
|
|
kref_put(&dev->kref, nvme_free_dev);
|
|
|
|
}
|
|
|
|
|
2014-04-03 05:45:37 +08:00
|
|
|
static int nvme_getgeo(struct block_device *bd, struct hd_geometry *geo)
|
|
|
|
{
|
|
|
|
/* some standard values */
|
|
|
|
geo->heads = 1 << 6;
|
|
|
|
geo->sectors = 1 << 5;
|
|
|
|
geo->cylinders = get_capacity(bd->bd_disk) >> 11;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
static const struct block_device_operations nvme_fops = {
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.ioctl = nvme_ioctl,
|
2013-10-24 03:07:34 +08:00
|
|
|
.compat_ioctl = nvme_compat_ioctl,
|
2014-02-01 07:53:39 +08:00
|
|
|
.open = nvme_open,
|
|
|
|
.release = nvme_release,
|
2014-04-03 05:45:37 +08:00
|
|
|
.getgeo = nvme_getgeo,
|
2011-01-21 01:50:14 +08:00
|
|
|
};
|
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
static void nvme_resubmit_iods(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
struct nvme_iod *iod, *next;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(iod, next, &nvmeq->iod_bio, node) {
|
|
|
|
if (unlikely(nvme_submit_iod(nvmeq, iod)))
|
|
|
|
break;
|
|
|
|
list_del(&iod->node);
|
|
|
|
if (bio_list_empty(&nvmeq->sq_cong) &&
|
|
|
|
list_empty(&nvmeq->iod_bio))
|
|
|
|
remove_wait_queue(&nvmeq->sq_full,
|
|
|
|
&nvmeq->sq_cong_wait);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-03-03 07:37:18 +08:00
|
|
|
static void nvme_resubmit_bios(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
while (bio_list_peek(&nvmeq->sq_cong)) {
|
|
|
|
struct bio *bio = bio_list_pop(&nvmeq->sq_cong);
|
|
|
|
struct nvme_ns *ns = bio->bi_bdev->bd_disk->private_data;
|
2013-04-10 01:59:32 +08:00
|
|
|
|
2014-04-04 06:45:23 +08:00
|
|
|
if (bio_list_empty(&nvmeq->sq_cong) &&
|
|
|
|
list_empty(&nvmeq->iod_bio))
|
2013-04-10 01:59:32 +08:00
|
|
|
remove_wait_queue(&nvmeq->sq_full,
|
|
|
|
&nvmeq->sq_cong_wait);
|
2011-03-03 07:37:18 +08:00
|
|
|
if (nvme_submit_bio_queue(nvmeq, ns, bio)) {
|
2014-04-04 06:45:23 +08:00
|
|
|
if (!waitqueue_active(&nvmeq->sq_full))
|
2013-04-10 01:59:32 +08:00
|
|
|
add_wait_queue(&nvmeq->sq_full,
|
|
|
|
&nvmeq->sq_cong_wait);
|
2011-03-03 07:37:18 +08:00
|
|
|
bio_list_add_head(&nvmeq->sq_cong, bio);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_kthread(void *data)
|
|
|
|
{
|
2013-12-11 04:10:37 +08:00
|
|
|
struct nvme_dev *dev, *next;
|
2011-03-03 07:37:18 +08:00
|
|
|
|
|
|
|
while (!kthread_should_stop()) {
|
2013-05-02 04:38:23 +08:00
|
|
|
set_current_state(TASK_INTERRUPTIBLE);
|
2011-03-03 07:37:18 +08:00
|
|
|
spin_lock(&dev_list_lock);
|
2013-12-11 04:10:37 +08:00
|
|
|
list_for_each_entry_safe(dev, next, &dev_list, node) {
|
2011-03-03 07:37:18 +08:00
|
|
|
int i;
|
2013-12-11 04:10:37 +08:00
|
|
|
if (readl(&dev->bar->csts) & NVME_CSTS_CFS &&
|
|
|
|
dev->initialized) {
|
|
|
|
if (work_busy(&dev->reset_work))
|
|
|
|
continue;
|
|
|
|
list_del_init(&dev->node);
|
|
|
|
dev_warn(&dev->pci_dev->dev,
|
|
|
|
"Failed status, reset controller\n");
|
2014-03-07 23:24:49 +08:00
|
|
|
dev->reset_workfn = nvme_reset_failed_dev;
|
2013-12-11 04:10:37 +08:00
|
|
|
queue_work(nvme_workq, &dev->reset_work);
|
|
|
|
continue;
|
|
|
|
}
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_read_lock();
|
2011-03-03 07:37:18 +08:00
|
|
|
for (i = 0; i < dev->queue_count; i++) {
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq =
|
|
|
|
rcu_dereference(dev->queues[i]);
|
2011-02-16 05:28:20 +08:00
|
|
|
if (!nvmeq)
|
|
|
|
continue;
|
2011-03-03 07:37:18 +08:00
|
|
|
spin_lock_irq(&nvmeq->q_lock);
|
2013-07-16 05:02:20 +08:00
|
|
|
if (nvmeq->q_suspended)
|
|
|
|
goto unlock;
|
2013-06-24 23:56:42 +08:00
|
|
|
nvme_process_cq(nvmeq);
|
2012-08-08 03:56:23 +08:00
|
|
|
nvme_cancel_ios(nvmeq, true);
|
2011-03-03 07:37:18 +08:00
|
|
|
nvme_resubmit_bios(nvmeq);
|
2014-04-04 06:45:23 +08:00
|
|
|
nvme_resubmit_iods(nvmeq);
|
2013-07-16 05:02:20 +08:00
|
|
|
unlock:
|
2011-03-03 07:37:18 +08:00
|
|
|
spin_unlock_irq(&nvmeq->q_lock);
|
|
|
|
}
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_read_unlock();
|
2011-03-03 07:37:18 +08:00
|
|
|
}
|
|
|
|
spin_unlock(&dev_list_lock);
|
2013-02-05 06:44:33 +08:00
|
|
|
schedule_timeout(round_jiffies_relative(HZ));
|
2011-03-03 07:37:18 +08:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-11-10 07:33:05 +08:00
|
|
|
static void nvme_config_discard(struct nvme_ns *ns)
|
|
|
|
{
|
|
|
|
u32 logical_block_size = queue_logical_block_size(ns->queue);
|
|
|
|
ns->queue->limits.discard_zeroes_data = 0;
|
|
|
|
ns->queue->limits.discard_alignment = logical_block_size;
|
|
|
|
ns->queue->limits.discard_granularity = logical_block_size;
|
|
|
|
ns->queue->limits.max_discard_sectors = 0xffffffff;
|
|
|
|
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue);
|
|
|
|
}
|
|
|
|
|
2013-07-09 05:26:25 +08:00
|
|
|
static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, unsigned nsid,
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_id_ns *id, struct nvme_lba_range_type *rt)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns;
|
|
|
|
struct gendisk *disk;
|
|
|
|
int lbaf;
|
|
|
|
|
|
|
|
if (rt->attributes & NVME_LBART_ATTRIB_HIDE)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ns = kzalloc(sizeof(*ns), GFP_KERNEL);
|
|
|
|
if (!ns)
|
|
|
|
return NULL;
|
|
|
|
ns->queue = blk_alloc_queue(GFP_KERNEL);
|
|
|
|
if (!ns->queue)
|
|
|
|
goto out_free_ns;
|
2012-01-11 05:35:08 +08:00
|
|
|
ns->queue->queue_flags = QUEUE_FLAG_DEFAULT;
|
|
|
|
queue_flag_set_unlocked(QUEUE_FLAG_NOMERGES, ns->queue);
|
|
|
|
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, ns->queue);
|
2011-01-21 01:50:14 +08:00
|
|
|
blk_queue_make_request(ns->queue, nvme_make_request);
|
|
|
|
ns->dev = dev;
|
|
|
|
ns->queue->queuedata = ns;
|
|
|
|
|
2013-12-10 01:58:46 +08:00
|
|
|
disk = alloc_disk(0);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (!disk)
|
|
|
|
goto out_free_queue;
|
2011-05-06 20:45:47 +08:00
|
|
|
ns->ns_id = nsid;
|
2011-01-21 01:50:14 +08:00
|
|
|
ns->disk = disk;
|
|
|
|
lbaf = id->flbas & 0xf;
|
|
|
|
ns->lba_shift = id->lbaf[lbaf].ds;
|
2013-04-24 07:23:59 +08:00
|
|
|
ns->ms = le16_to_cpu(id->lbaf[lbaf].ms);
|
2012-07-25 05:01:04 +08:00
|
|
|
blk_queue_logical_block_size(ns->queue, 1 << ns->lba_shift);
|
2012-07-27 01:29:57 +08:00
|
|
|
if (dev->max_hw_sectors)
|
|
|
|
blk_queue_max_hw_sectors(ns->queue, dev->max_hw_sectors);
|
2014-04-30 01:41:28 +08:00
|
|
|
if (dev->vwc & NVME_CTRL_VWC_PRESENT)
|
|
|
|
blk_queue_flush(ns->queue, REQ_FLUSH | REQ_FUA);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
disk->major = nvme_major;
|
2013-12-10 01:58:46 +08:00
|
|
|
disk->first_minor = 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
disk->fops = &nvme_fops;
|
|
|
|
disk->private_data = ns;
|
|
|
|
disk->queue = ns->queue;
|
2011-02-02 01:49:38 +08:00
|
|
|
disk->driverfs_dev = &dev->pci_dev->dev;
|
2013-12-10 01:58:46 +08:00
|
|
|
disk->flags = GENHD_FL_EXT_DEVT;
|
2011-05-06 20:45:47 +08:00
|
|
|
sprintf(disk->disk_name, "nvme%dn%d", dev->instance, nsid);
|
2011-01-21 01:50:14 +08:00
|
|
|
set_capacity(disk, le64_to_cpup(&id->nsze) << (ns->lba_shift - 9));
|
|
|
|
|
2012-11-10 07:33:05 +08:00
|
|
|
if (dev->oncs & NVME_CTRL_ONCS_DSM)
|
|
|
|
nvme_config_discard(ns);
|
|
|
|
|
2011-01-21 01:50:14 +08:00
|
|
|
return ns;
|
|
|
|
|
|
|
|
out_free_queue:
|
|
|
|
blk_cleanup_queue(ns->queue);
|
|
|
|
out_free_ns:
|
|
|
|
kfree(ns);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-03-25 00:46:25 +08:00
|
|
|
static int nvme_find_closest_node(int node)
|
|
|
|
{
|
|
|
|
int n, val, min_val = INT_MAX, best_node = node;
|
|
|
|
|
|
|
|
for_each_online_node(n) {
|
|
|
|
if (n == node)
|
|
|
|
continue;
|
|
|
|
val = node_distance(node, n);
|
|
|
|
if (val < min_val) {
|
|
|
|
min_val = val;
|
|
|
|
best_node = n;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return best_node;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_set_queue_cpus(cpumask_t *qmask, struct nvme_queue *nvmeq,
|
|
|
|
int count)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
for_each_cpu(cpu, qmask) {
|
|
|
|
if (cpumask_weight(nvmeq->cpu_mask) >= count)
|
|
|
|
break;
|
|
|
|
if (!cpumask_test_and_set_cpu(cpu, nvmeq->cpu_mask))
|
|
|
|
*per_cpu_ptr(nvmeq->dev->io_queue, cpu) = nvmeq->qid;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_add_cpus(cpumask_t *mask, const cpumask_t *unassigned_cpus,
|
|
|
|
const cpumask_t *new_mask, struct nvme_queue *nvmeq, int cpus_per_queue)
|
|
|
|
{
|
|
|
|
int next_cpu;
|
|
|
|
for_each_cpu(next_cpu, new_mask) {
|
|
|
|
cpumask_or(mask, mask, get_cpu_mask(next_cpu));
|
|
|
|
cpumask_or(mask, mask, topology_thread_cpumask(next_cpu));
|
|
|
|
cpumask_and(mask, mask, unassigned_cpus);
|
|
|
|
nvme_set_queue_cpus(mask, nvmeq, cpus_per_queue);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_create_io_queues(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
unsigned i, max;
|
|
|
|
|
|
|
|
max = min(dev->max_qid, num_online_cpus());
|
|
|
|
for (i = dev->queue_count; i <= max; i++)
|
|
|
|
if (!nvme_alloc_queue(dev, i, dev->q_depth, i - 1))
|
|
|
|
break;
|
|
|
|
|
|
|
|
max = min(dev->queue_count - 1, num_online_cpus());
|
|
|
|
for (i = dev->online_queues; i <= max; i++)
|
|
|
|
if (nvme_create_queue(raw_nvmeq(dev, i), i))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there are fewer queues than online cpus, this will try to optimally
|
|
|
|
* assign a queue to multiple cpus by grouping cpus that are "close" together:
|
|
|
|
* thread siblings, core, socket, closest node, then whatever else is
|
|
|
|
* available.
|
|
|
|
*/
|
|
|
|
static void nvme_assign_io_queues(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
unsigned cpu, cpus_per_queue, queues, remainder, i;
|
|
|
|
cpumask_var_t unassigned_cpus;
|
|
|
|
|
|
|
|
nvme_create_io_queues(dev);
|
|
|
|
|
|
|
|
queues = min(dev->online_queues - 1, num_online_cpus());
|
|
|
|
if (!queues)
|
|
|
|
return;
|
|
|
|
|
|
|
|
cpus_per_queue = num_online_cpus() / queues;
|
|
|
|
remainder = queues - (num_online_cpus() - queues * cpus_per_queue);
|
|
|
|
|
|
|
|
if (!alloc_cpumask_var(&unassigned_cpus, GFP_KERNEL))
|
|
|
|
return;
|
|
|
|
|
|
|
|
cpumask_copy(unassigned_cpus, cpu_online_mask);
|
|
|
|
cpu = cpumask_first(unassigned_cpus);
|
|
|
|
for (i = 1; i <= queues; i++) {
|
|
|
|
struct nvme_queue *nvmeq = lock_nvmeq(dev, i);
|
|
|
|
cpumask_t mask;
|
|
|
|
|
|
|
|
cpumask_clear(nvmeq->cpu_mask);
|
|
|
|
if (!cpumask_weight(unassigned_cpus)) {
|
|
|
|
unlock_nvmeq(nvmeq);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
mask = *get_cpu_mask(cpu);
|
|
|
|
nvme_set_queue_cpus(&mask, nvmeq, cpus_per_queue);
|
|
|
|
if (cpus_weight(mask) < cpus_per_queue)
|
|
|
|
nvme_add_cpus(&mask, unassigned_cpus,
|
|
|
|
topology_thread_cpumask(cpu),
|
|
|
|
nvmeq, cpus_per_queue);
|
|
|
|
if (cpus_weight(mask) < cpus_per_queue)
|
|
|
|
nvme_add_cpus(&mask, unassigned_cpus,
|
|
|
|
topology_core_cpumask(cpu),
|
|
|
|
nvmeq, cpus_per_queue);
|
|
|
|
if (cpus_weight(mask) < cpus_per_queue)
|
|
|
|
nvme_add_cpus(&mask, unassigned_cpus,
|
|
|
|
cpumask_of_node(cpu_to_node(cpu)),
|
|
|
|
nvmeq, cpus_per_queue);
|
|
|
|
if (cpus_weight(mask) < cpus_per_queue)
|
|
|
|
nvme_add_cpus(&mask, unassigned_cpus,
|
|
|
|
cpumask_of_node(
|
|
|
|
nvme_find_closest_node(
|
|
|
|
cpu_to_node(cpu))),
|
|
|
|
nvmeq, cpus_per_queue);
|
|
|
|
if (cpus_weight(mask) < cpus_per_queue)
|
|
|
|
nvme_add_cpus(&mask, unassigned_cpus,
|
|
|
|
unassigned_cpus,
|
|
|
|
nvmeq, cpus_per_queue);
|
|
|
|
|
|
|
|
WARN(cpumask_weight(nvmeq->cpu_mask) != cpus_per_queue,
|
|
|
|
"nvme%d qid:%d mis-matched queue-to-cpu assignment\n",
|
|
|
|
dev->instance, i);
|
|
|
|
|
|
|
|
irq_set_affinity_hint(dev->entry[nvmeq->cq_vector].vector,
|
|
|
|
nvmeq->cpu_mask);
|
|
|
|
cpumask_andnot(unassigned_cpus, unassigned_cpus,
|
|
|
|
nvmeq->cpu_mask);
|
|
|
|
cpu = cpumask_next(cpu, unassigned_cpus);
|
|
|
|
if (remainder && !--remainder)
|
|
|
|
cpus_per_queue++;
|
|
|
|
unlock_nvmeq(nvmeq);
|
|
|
|
}
|
|
|
|
WARN(cpumask_weight(unassigned_cpus), "nvme%d unassigned online cpus\n",
|
|
|
|
dev->instance);
|
|
|
|
i = 0;
|
|
|
|
cpumask_andnot(unassigned_cpus, cpu_possible_mask, cpu_online_mask);
|
|
|
|
for_each_cpu(cpu, unassigned_cpus)
|
|
|
|
*per_cpu_ptr(dev->io_queue, cpu) = (i++ % queues) + 1;
|
|
|
|
free_cpumask_var(unassigned_cpus);
|
|
|
|
}
|
|
|
|
|
2011-01-20 22:14:34 +08:00
|
|
|
static int set_queue_count(struct nvme_dev *dev, int count)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
|
|
|
int status;
|
|
|
|
u32 result;
|
2011-01-20 22:14:34 +08:00
|
|
|
u32 q_count = (count - 1) | ((count - 1) << 16);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2012-01-11 22:29:56 +08:00
|
|
|
status = nvme_set_features(dev, NVME_FEAT_NUM_QUEUES, q_count, 0,
|
2011-09-20 05:08:14 +08:00
|
|
|
&result);
|
2014-04-11 23:58:45 +08:00
|
|
|
if (status < 0)
|
|
|
|
return status;
|
|
|
|
if (status > 0) {
|
|
|
|
dev_err(&dev->pci_dev->dev, "Could not set queue count (%d)\n",
|
|
|
|
status);
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
return min(result & 0xffff, result >> 16) + 1;
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:24 +08:00
|
|
|
static size_t db_bar_size(struct nvme_dev *dev, unsigned nr_io_queues)
|
|
|
|
{
|
2013-09-10 11:25:37 +08:00
|
|
|
return 4096 + ((nr_io_queues + 1) * 8 * dev->db_stride);
|
2013-07-16 05:02:24 +08:00
|
|
|
}
|
|
|
|
|
2014-03-25 00:46:26 +08:00
|
|
|
static int nvme_cpu_notify(struct notifier_block *self,
|
|
|
|
unsigned long action, void *hcpu)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(self, struct nvme_dev, nb);
|
|
|
|
switch (action) {
|
|
|
|
case CPU_ONLINE:
|
|
|
|
case CPU_DEAD:
|
|
|
|
nvme_assign_io_queues(dev);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return NOTIFY_OK;
|
|
|
|
}
|
|
|
|
|
2012-12-22 07:13:49 +08:00
|
|
|
static int nvme_setup_io_queues(struct nvme_dev *dev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *adminq = raw_nvmeq(dev, 0);
|
2013-05-12 06:19:31 +08:00
|
|
|
struct pci_dev *pdev = dev->pci_dev;
|
2014-03-25 00:46:25 +08:00
|
|
|
int result, i, vecs, nr_io_queues, size;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-03-25 00:46:25 +08:00
|
|
|
nr_io_queues = num_possible_cpus();
|
2011-02-16 05:16:02 +08:00
|
|
|
result = set_queue_count(dev, nr_io_queues);
|
2011-01-21 02:01:49 +08:00
|
|
|
if (result < 0)
|
|
|
|
return result;
|
2011-02-16 05:16:02 +08:00
|
|
|
if (result < nr_io_queues)
|
|
|
|
nr_io_queues = result;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-07-16 05:02:24 +08:00
|
|
|
size = db_bar_size(dev, nr_io_queues);
|
|
|
|
if (size > 8192) {
|
2011-10-21 05:00:41 +08:00
|
|
|
iounmap(dev->bar);
|
2013-07-16 05:02:24 +08:00
|
|
|
do {
|
|
|
|
dev->bar = ioremap(pci_resource_start(pdev, 0), size);
|
|
|
|
if (dev->bar)
|
|
|
|
break;
|
|
|
|
if (!--nr_io_queues)
|
|
|
|
return -ENOMEM;
|
|
|
|
size = db_bar_size(dev, nr_io_queues);
|
|
|
|
} while (1);
|
2011-10-21 05:00:41 +08:00
|
|
|
dev->dbs = ((void __iomem *)dev->bar) + 4096;
|
2014-02-22 05:13:44 +08:00
|
|
|
adminq->q_db = dev->dbs;
|
2011-10-21 05:00:41 +08:00
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:24 +08:00
|
|
|
/* Deregister the admin queue's interrupt */
|
2014-01-28 04:57:22 +08:00
|
|
|
free_irq(dev->entry[0].vector, adminq);
|
2013-07-16 05:02:24 +08:00
|
|
|
|
2014-03-04 23:22:00 +08:00
|
|
|
for (i = 0; i < nr_io_queues; i++)
|
2011-01-21 02:01:49 +08:00
|
|
|
dev->entry[i].entry = i;
|
2014-03-04 23:22:00 +08:00
|
|
|
vecs = pci_enable_msix_range(pdev, dev->entry, 1, nr_io_queues);
|
|
|
|
if (vecs < 0) {
|
|
|
|
vecs = pci_enable_msi_range(pdev, 1, min(nr_io_queues, 32));
|
|
|
|
if (vecs < 0) {
|
|
|
|
vecs = 1;
|
|
|
|
} else {
|
|
|
|
for (i = 0; i < vecs; i++)
|
|
|
|
dev->entry[i].vector = i + pdev->irq;
|
2013-05-12 06:19:31 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-06-20 22:53:48 +08:00
|
|
|
/*
|
|
|
|
* Should investigate if there's a performance win from allocating
|
|
|
|
* more queues than interrupt vectors; it might allow the submission
|
|
|
|
* path to scale better, even if the receive path is limited by the
|
|
|
|
* number of interrupts.
|
|
|
|
*/
|
|
|
|
nr_io_queues = vecs;
|
2014-03-25 00:46:25 +08:00
|
|
|
dev->max_qid = nr_io_queues;
|
2013-06-20 22:53:48 +08:00
|
|
|
|
2014-01-28 04:57:22 +08:00
|
|
|
result = queue_request_irq(dev, adminq, adminq->irqname);
|
2013-07-16 05:02:24 +08:00
|
|
|
if (result) {
|
2014-01-28 04:57:22 +08:00
|
|
|
adminq->q_suspended = 1;
|
2013-07-16 05:02:20 +08:00
|
|
|
goto free_queues;
|
2013-07-16 05:02:24 +08:00
|
|
|
}
|
2011-01-21 02:01:49 +08:00
|
|
|
|
2013-07-16 05:02:23 +08:00
|
|
|
/* Free previously allocated queues that are no longer usable */
|
2014-03-25 00:46:25 +08:00
|
|
|
nvme_free_queues(dev, nr_io_queues + 1);
|
|
|
|
nvme_assign_io_queues(dev);
|
2011-03-17 04:52:19 +08:00
|
|
|
|
2014-03-25 00:46:26 +08:00
|
|
|
dev->nb.notifier_call = &nvme_cpu_notify;
|
|
|
|
result = register_hotcpu_notifier(&dev->nb);
|
|
|
|
if (result)
|
|
|
|
goto free_queues;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
return 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-07-16 05:02:20 +08:00
|
|
|
free_queues:
|
2013-12-17 02:50:00 +08:00
|
|
|
nvme_free_queues(dev, 1);
|
2013-07-16 05:02:20 +08:00
|
|
|
return result;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2013-04-16 23:22:36 +08:00
|
|
|
/*
|
|
|
|
* Return: error value if an error occurred setting up the queues or calling
|
|
|
|
* Identify Device. 0 if these succeeded, even if adding some of the
|
|
|
|
* namespaces failed. At the moment, these failures are silent. TBD which
|
|
|
|
* failures should be reported.
|
|
|
|
*/
|
2012-12-22 07:13:49 +08:00
|
|
|
static int nvme_dev_add(struct nvme_dev *dev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-06-22 02:36:34 +08:00
|
|
|
struct pci_dev *pdev = dev->pci_dev;
|
2013-07-09 05:26:25 +08:00
|
|
|
int res;
|
|
|
|
unsigned nn, i;
|
2013-05-02 03:07:49 +08:00
|
|
|
struct nvme_ns *ns;
|
2011-02-02 05:18:08 +08:00
|
|
|
struct nvme_id_ctrl *ctrl;
|
2011-09-20 05:08:14 +08:00
|
|
|
struct nvme_id_ns *id_ns;
|
|
|
|
void *mem;
|
2011-01-21 01:50:14 +08:00
|
|
|
dma_addr_t dma_addr;
|
2013-04-10 07:13:20 +08:00
|
|
|
int shift = NVME_CAP_MPSMIN(readq(&dev->bar->cap)) + 12;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-06-22 02:36:34 +08:00
|
|
|
mem = dma_alloc_coherent(&pdev->dev, 8192, &dma_addr, GFP_KERNEL);
|
2013-05-02 03:07:48 +08:00
|
|
|
if (!mem)
|
|
|
|
return -ENOMEM;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
res = nvme_identify(dev, 0, 1, dma_addr);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (res) {
|
2014-04-11 23:58:45 +08:00
|
|
|
dev_err(&pdev->dev, "Identify Controller failed (%d)\n", res);
|
2011-01-21 01:50:14 +08:00
|
|
|
res = -EIO;
|
2013-05-02 03:07:49 +08:00
|
|
|
goto out;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
ctrl = mem;
|
2011-02-02 05:18:08 +08:00
|
|
|
nn = le32_to_cpup(&ctrl->nn);
|
2012-11-10 07:33:05 +08:00
|
|
|
dev->oncs = le16_to_cpup(&ctrl->oncs);
|
2013-12-11 04:10:38 +08:00
|
|
|
dev->abort_limit = ctrl->acl + 1;
|
2014-04-30 01:41:28 +08:00
|
|
|
dev->vwc = ctrl->vwc;
|
2011-02-02 05:18:08 +08:00
|
|
|
memcpy(dev->serial, ctrl->sn, sizeof(ctrl->sn));
|
|
|
|
memcpy(dev->model, ctrl->mn, sizeof(ctrl->mn));
|
|
|
|
memcpy(dev->firmware_rev, ctrl->fr, sizeof(ctrl->fr));
|
2013-04-10 07:13:20 +08:00
|
|
|
if (ctrl->mdts)
|
2012-07-27 01:29:57 +08:00
|
|
|
dev->max_hw_sectors = 1 << (ctrl->mdts + shift - 9);
|
2013-06-22 02:36:34 +08:00
|
|
|
if ((pdev->vendor == PCI_VENDOR_ID_INTEL) &&
|
|
|
|
(pdev->device == 0x0953) && ctrl->vs[3])
|
2013-04-10 07:13:20 +08:00
|
|
|
dev->stripe_size = 1 << (ctrl->vs[3] + shift);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
id_ns = mem;
|
2011-10-08 01:10:13 +08:00
|
|
|
for (i = 1; i <= nn; i++) {
|
2011-09-20 05:08:14 +08:00
|
|
|
res = nvme_identify(dev, i, 0, dma_addr);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (res)
|
|
|
|
continue;
|
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
if (id_ns->ncap == 0)
|
2011-01-21 01:50:14 +08:00
|
|
|
continue;
|
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
res = nvme_get_features(dev, NVME_FEAT_LBA_RANGE, i,
|
2012-09-22 00:52:13 +08:00
|
|
|
dma_addr + 4096, NULL);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (res)
|
2013-02-01 05:40:38 +08:00
|
|
|
memset(mem + 4096, 0, 4096);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
ns = nvme_alloc_ns(dev, i, mem, mem + 4096);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (ns)
|
|
|
|
list_add_tail(&ns->list, &dev->namespaces);
|
|
|
|
}
|
|
|
|
list_for_each_entry(ns, &dev->namespaces, list)
|
|
|
|
add_disk(ns->disk);
|
2013-04-16 23:22:36 +08:00
|
|
|
res = 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-09-20 05:08:14 +08:00
|
|
|
out:
|
2011-09-20 05:14:53 +08:00
|
|
|
dma_free_coherent(&dev->pci_dev->dev, 8192, mem, dma_addr);
|
2011-01-21 01:50:14 +08:00
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:19 +08:00
|
|
|
static int nvme_dev_map(struct nvme_dev *dev)
|
|
|
|
{
|
2014-03-25 00:46:25 +08:00
|
|
|
u64 cap;
|
2013-07-16 05:02:19 +08:00
|
|
|
int bars, result = -ENOMEM;
|
|
|
|
struct pci_dev *pdev = dev->pci_dev;
|
|
|
|
|
|
|
|
if (pci_enable_device_mem(pdev))
|
|
|
|
return result;
|
|
|
|
|
|
|
|
dev->entry[0].vector = pdev->irq;
|
|
|
|
pci_set_master(pdev);
|
|
|
|
bars = pci_select_bars(pdev, IORESOURCE_MEM);
|
|
|
|
if (pci_request_selected_regions(pdev, bars, "nvme"))
|
|
|
|
goto disable_pci;
|
|
|
|
|
2013-06-27 06:49:11 +08:00
|
|
|
if (dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64)) &&
|
|
|
|
dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)))
|
|
|
|
goto disable;
|
2013-07-16 05:02:19 +08:00
|
|
|
|
|
|
|
dev->bar = ioremap(pci_resource_start(pdev, 0), 8192);
|
|
|
|
if (!dev->bar)
|
|
|
|
goto disable;
|
2013-12-11 04:10:39 +08:00
|
|
|
if (readl(&dev->bar->csts) == -1) {
|
|
|
|
result = -ENODEV;
|
|
|
|
goto unmap;
|
|
|
|
}
|
2014-03-25 00:46:25 +08:00
|
|
|
cap = readq(&dev->bar->cap);
|
|
|
|
dev->q_depth = min_t(int, NVME_CAP_MQES(cap) + 1, NVME_Q_DEPTH);
|
|
|
|
dev->db_stride = 1 << NVME_CAP_STRIDE(cap);
|
2013-07-16 05:02:19 +08:00
|
|
|
dev->dbs = ((void __iomem *)dev->bar) + 4096;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
2013-12-11 04:10:39 +08:00
|
|
|
unmap:
|
|
|
|
iounmap(dev->bar);
|
|
|
|
dev->bar = NULL;
|
2013-07-16 05:02:19 +08:00
|
|
|
disable:
|
|
|
|
pci_release_regions(pdev);
|
|
|
|
disable_pci:
|
|
|
|
pci_disable_device(pdev);
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_dev_unmap(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
if (dev->pci_dev->msi_enabled)
|
|
|
|
pci_disable_msi(dev->pci_dev);
|
|
|
|
else if (dev->pci_dev->msix_enabled)
|
|
|
|
pci_disable_msix(dev->pci_dev);
|
|
|
|
|
|
|
|
if (dev->bar) {
|
|
|
|
iounmap(dev->bar);
|
|
|
|
dev->bar = NULL;
|
2013-12-11 04:10:36 +08:00
|
|
|
pci_release_regions(dev->pci_dev);
|
2013-07-16 05:02:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (pci_is_enabled(dev->pci_dev))
|
|
|
|
pci_disable_device(dev->pci_dev);
|
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
struct nvme_delq_ctx {
|
|
|
|
struct task_struct *waiter;
|
|
|
|
struct kthread_worker *worker;
|
|
|
|
atomic_t refcount;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void nvme_wait_dq(struct nvme_delq_ctx *dq, struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
dq->waiter = current;
|
|
|
|
mb();
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
set_current_state(TASK_KILLABLE);
|
|
|
|
if (!atomic_read(&dq->refcount))
|
|
|
|
break;
|
|
|
|
if (!schedule_timeout(ADMIN_TIMEOUT) ||
|
|
|
|
fatal_signal_pending(current)) {
|
|
|
|
set_current_state(TASK_RUNNING);
|
|
|
|
|
|
|
|
nvme_disable_ctrl(dev, readq(&dev->bar->cap));
|
|
|
|
nvme_disable_queue(dev, 0);
|
|
|
|
|
|
|
|
send_sig(SIGKILL, dq->worker->task, 1);
|
|
|
|
flush_kthread_worker(dq->worker);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
set_current_state(TASK_RUNNING);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_put_dq(struct nvme_delq_ctx *dq)
|
|
|
|
{
|
|
|
|
atomic_dec(&dq->refcount);
|
|
|
|
if (dq->waiter)
|
|
|
|
wake_up_process(dq->waiter);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct nvme_delq_ctx *nvme_get_dq(struct nvme_delq_ctx *dq)
|
|
|
|
{
|
|
|
|
atomic_inc(&dq->refcount);
|
|
|
|
return dq;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_del_queue_end(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
struct nvme_delq_ctx *dq = nvmeq->cmdinfo.ctx;
|
|
|
|
|
|
|
|
nvme_clear_queue(nvmeq);
|
|
|
|
nvme_put_dq(dq);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int adapter_async_del_queue(struct nvme_queue *nvmeq, u8 opcode,
|
|
|
|
kthread_work_func_t fn)
|
|
|
|
{
|
|
|
|
struct nvme_command c;
|
|
|
|
|
|
|
|
memset(&c, 0, sizeof(c));
|
|
|
|
c.delete_queue.opcode = opcode;
|
|
|
|
c.delete_queue.qid = cpu_to_le16(nvmeq->qid);
|
|
|
|
|
|
|
|
init_kthread_work(&nvmeq->cmdinfo.work, fn);
|
|
|
|
return nvme_submit_admin_cmd_async(nvmeq->dev, &c, &nvmeq->cmdinfo);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_del_cq_work_handler(struct kthread_work *work)
|
|
|
|
{
|
|
|
|
struct nvme_queue *nvmeq = container_of(work, struct nvme_queue,
|
|
|
|
cmdinfo.work);
|
|
|
|
nvme_del_queue_end(nvmeq);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_delete_cq(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
return adapter_async_del_queue(nvmeq, nvme_admin_delete_cq,
|
|
|
|
nvme_del_cq_work_handler);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_del_sq_work_handler(struct kthread_work *work)
|
|
|
|
{
|
|
|
|
struct nvme_queue *nvmeq = container_of(work, struct nvme_queue,
|
|
|
|
cmdinfo.work);
|
|
|
|
int status = nvmeq->cmdinfo.status;
|
|
|
|
|
|
|
|
if (!status)
|
|
|
|
status = nvme_delete_cq(nvmeq);
|
|
|
|
if (status)
|
|
|
|
nvme_del_queue_end(nvmeq);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_delete_sq(struct nvme_queue *nvmeq)
|
|
|
|
{
|
|
|
|
return adapter_async_del_queue(nvmeq, nvme_admin_delete_sq,
|
|
|
|
nvme_del_sq_work_handler);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_del_queue_start(struct kthread_work *work)
|
|
|
|
{
|
|
|
|
struct nvme_queue *nvmeq = container_of(work, struct nvme_queue,
|
|
|
|
cmdinfo.work);
|
|
|
|
allow_signal(SIGKILL);
|
|
|
|
if (nvme_delete_sq(nvmeq))
|
|
|
|
nvme_del_queue_end(nvmeq);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_disable_io_queues(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
DEFINE_KTHREAD_WORKER_ONSTACK(worker);
|
|
|
|
struct nvme_delq_ctx dq;
|
|
|
|
struct task_struct *kworker_task = kthread_run(kthread_worker_fn,
|
|
|
|
&worker, "nvme%d", dev->instance);
|
|
|
|
|
|
|
|
if (IS_ERR(kworker_task)) {
|
|
|
|
dev_err(&dev->pci_dev->dev,
|
|
|
|
"Failed to create queue del task\n");
|
|
|
|
for (i = dev->queue_count - 1; i > 0; i--)
|
|
|
|
nvme_disable_queue(dev, i);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
dq.waiter = NULL;
|
|
|
|
atomic_set(&dq.refcount, 0);
|
|
|
|
dq.worker = &worker;
|
|
|
|
for (i = dev->queue_count - 1; i > 0; i--) {
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
|
2013-12-11 04:10:40 +08:00
|
|
|
|
|
|
|
if (nvme_suspend_queue(nvmeq))
|
|
|
|
continue;
|
|
|
|
nvmeq->cmdinfo.ctx = nvme_get_dq(&dq);
|
|
|
|
nvmeq->cmdinfo.worker = dq.worker;
|
|
|
|
init_kthread_work(&nvmeq->cmdinfo.work, nvme_del_queue_start);
|
|
|
|
queue_kthread_work(dq.worker, &nvmeq->cmdinfo.work);
|
|
|
|
}
|
|
|
|
nvme_wait_dq(&dq, dev);
|
|
|
|
kthread_stop(kworker_task);
|
|
|
|
}
|
|
|
|
|
2014-04-08 07:10:11 +08:00
|
|
|
/*
|
|
|
|
* Remove the node from the device list and check
|
|
|
|
* for whether or not we need to stop the nvme_thread.
|
|
|
|
*/
|
|
|
|
static void nvme_dev_list_remove(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
struct task_struct *tmp = NULL;
|
|
|
|
|
|
|
|
spin_lock(&dev_list_lock);
|
|
|
|
list_del_init(&dev->node);
|
|
|
|
if (list_empty(&dev_list) && !IS_ERR_OR_NULL(nvme_thread)) {
|
|
|
|
tmp = nvme_thread;
|
|
|
|
nvme_thread = NULL;
|
|
|
|
}
|
|
|
|
spin_unlock(&dev_list_lock);
|
|
|
|
|
|
|
|
if (tmp)
|
|
|
|
kthread_stop(tmp);
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:21 +08:00
|
|
|
static void nvme_dev_shutdown(struct nvme_dev *dev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-07-16 05:02:20 +08:00
|
|
|
int i;
|
|
|
|
|
2013-12-11 04:10:37 +08:00
|
|
|
dev->initialized = 0;
|
2014-03-25 00:46:26 +08:00
|
|
|
unregister_hotcpu_notifier(&dev->nb);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2014-04-08 07:10:11 +08:00
|
|
|
nvme_dev_list_remove(dev);
|
2011-03-03 07:37:18 +08:00
|
|
|
|
2013-12-11 04:10:40 +08:00
|
|
|
if (!dev->bar || (dev->bar && readl(&dev->bar->csts) == -1)) {
|
|
|
|
for (i = dev->queue_count - 1; i >= 0; i--) {
|
2014-02-22 05:13:44 +08:00
|
|
|
struct nvme_queue *nvmeq = raw_nvmeq(dev, i);
|
2013-12-11 04:10:40 +08:00
|
|
|
nvme_suspend_queue(nvmeq);
|
|
|
|
nvme_clear_queue(nvmeq);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
nvme_disable_io_queues(dev);
|
2013-07-16 05:02:22 +08:00
|
|
|
nvme_shutdown_ctrl(dev);
|
2013-12-11 04:10:40 +08:00
|
|
|
nvme_disable_queue(dev, 0);
|
|
|
|
}
|
2013-07-16 05:02:21 +08:00
|
|
|
nvme_dev_unmap(dev);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_dev_remove(struct nvme_dev *dev)
|
|
|
|
{
|
2014-02-01 07:53:39 +08:00
|
|
|
struct nvme_ns *ns;
|
2013-07-16 05:02:21 +08:00
|
|
|
|
2014-02-01 07:53:39 +08:00
|
|
|
list_for_each_entry(ns, &dev->namespaces, list) {
|
|
|
|
if (ns->disk->flags & GENHD_FL_UP)
|
|
|
|
del_gendisk(ns->disk);
|
|
|
|
if (!blk_queue_dying(ns->queue))
|
|
|
|
blk_cleanup_queue(ns->queue);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-02-10 22:56:01 +08:00
|
|
|
static int nvme_setup_prp_pools(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
struct device *dmadev = &dev->pci_dev->dev;
|
|
|
|
dev->prp_page_pool = dma_pool_create("prp list page", dmadev,
|
|
|
|
PAGE_SIZE, PAGE_SIZE, 0);
|
|
|
|
if (!dev->prp_page_pool)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2011-02-10 23:30:34 +08:00
|
|
|
/* Optimisation for I/Os between 4k and 128k */
|
|
|
|
dev->prp_small_pool = dma_pool_create("prp list 256", dmadev,
|
|
|
|
256, 256, 0);
|
|
|
|
if (!dev->prp_small_pool) {
|
|
|
|
dma_pool_destroy(dev->prp_page_pool);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2011-02-10 22:56:01 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_release_prp_pools(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
dma_pool_destroy(dev->prp_page_pool);
|
2011-02-10 23:30:34 +08:00
|
|
|
dma_pool_destroy(dev->prp_small_pool);
|
2011-02-10 22:56:01 +08:00
|
|
|
}
|
|
|
|
|
2012-02-22 07:50:53 +08:00
|
|
|
static DEFINE_IDA(nvme_instance_ida);
|
|
|
|
|
|
|
|
static int nvme_set_instance(struct nvme_dev *dev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2012-02-22 07:50:53 +08:00
|
|
|
int instance, error;
|
|
|
|
|
|
|
|
do {
|
|
|
|
if (!ida_pre_get(&nvme_instance_ida, GFP_KERNEL))
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
spin_lock(&dev_list_lock);
|
|
|
|
error = ida_get_new(&nvme_instance_ida, &instance);
|
|
|
|
spin_unlock(&dev_list_lock);
|
|
|
|
} while (error == -EAGAIN);
|
|
|
|
|
|
|
|
if (error)
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
dev->instance = instance;
|
|
|
|
return 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_release_instance(struct nvme_dev *dev)
|
|
|
|
{
|
2012-02-22 07:50:53 +08:00
|
|
|
spin_lock(&dev_list_lock);
|
|
|
|
ida_remove(&nvme_instance_ida, dev->instance);
|
|
|
|
spin_unlock(&dev_list_lock);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
2014-02-01 07:53:39 +08:00
|
|
|
static void nvme_free_namespaces(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
struct nvme_ns *ns, *next;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(ns, next, &dev->namespaces, list) {
|
|
|
|
list_del(&ns->list);
|
|
|
|
put_disk(ns->disk);
|
|
|
|
kfree(ns);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-02-20 01:17:58 +08:00
|
|
|
static void nvme_free_dev(struct kref *kref)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(kref, struct nvme_dev, kref);
|
2014-02-01 07:53:39 +08:00
|
|
|
|
|
|
|
nvme_free_namespaces(dev);
|
2014-03-25 00:46:25 +08:00
|
|
|
free_percpu(dev->io_queue);
|
2013-02-20 01:17:58 +08:00
|
|
|
kfree(dev->queues);
|
|
|
|
kfree(dev->entry);
|
|
|
|
kfree(dev);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_dev_open(struct inode *inode, struct file *f)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(f->private_data, struct nvme_dev,
|
|
|
|
miscdev);
|
|
|
|
kref_get(&dev->kref);
|
|
|
|
f->private_data = dev;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_dev_release(struct inode *inode, struct file *f)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = f->private_data;
|
|
|
|
kref_put(&dev->kref, nvme_free_dev);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static long nvme_dev_ioctl(struct file *f, unsigned int cmd, unsigned long arg)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = f->private_data;
|
|
|
|
switch (cmd) {
|
|
|
|
case NVME_IOCTL_ADMIN_CMD:
|
|
|
|
return nvme_user_admin_cmd(dev, (void __user *)arg);
|
|
|
|
default:
|
|
|
|
return -ENOTTY;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct file_operations nvme_dev_fops = {
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.open = nvme_dev_open,
|
|
|
|
.release = nvme_dev_release,
|
|
|
|
.unlocked_ioctl = nvme_dev_ioctl,
|
|
|
|
.compat_ioctl = nvme_dev_ioctl,
|
|
|
|
};
|
|
|
|
|
2013-07-16 05:02:21 +08:00
|
|
|
static int nvme_dev_start(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
int result;
|
2014-04-08 07:10:11 +08:00
|
|
|
bool start_thread = false;
|
2013-07-16 05:02:21 +08:00
|
|
|
|
|
|
|
result = nvme_dev_map(dev);
|
|
|
|
if (result)
|
|
|
|
return result;
|
|
|
|
|
|
|
|
result = nvme_configure_admin_queue(dev);
|
|
|
|
if (result)
|
|
|
|
goto unmap;
|
|
|
|
|
|
|
|
spin_lock(&dev_list_lock);
|
2014-04-08 07:10:11 +08:00
|
|
|
if (list_empty(&dev_list) && IS_ERR_OR_NULL(nvme_thread)) {
|
|
|
|
start_thread = true;
|
|
|
|
nvme_thread = NULL;
|
|
|
|
}
|
2013-07-16 05:02:21 +08:00
|
|
|
list_add(&dev->node, &dev_list);
|
|
|
|
spin_unlock(&dev_list_lock);
|
|
|
|
|
2014-04-08 07:10:11 +08:00
|
|
|
if (start_thread) {
|
|
|
|
nvme_thread = kthread_run(nvme_kthread, NULL, "nvme");
|
|
|
|
wake_up(&nvme_kthread_wait);
|
|
|
|
} else
|
|
|
|
wait_event_killable(nvme_kthread_wait, nvme_thread);
|
|
|
|
|
|
|
|
if (IS_ERR_OR_NULL(nvme_thread)) {
|
|
|
|
result = nvme_thread ? PTR_ERR(nvme_thread) : -EINTR;
|
|
|
|
goto disable;
|
|
|
|
}
|
|
|
|
|
2013-07-16 05:02:21 +08:00
|
|
|
result = nvme_setup_io_queues(dev);
|
2013-09-06 04:45:07 +08:00
|
|
|
if (result && result != -EBUSY)
|
2013-07-16 05:02:21 +08:00
|
|
|
goto disable;
|
|
|
|
|
2013-09-06 04:45:07 +08:00
|
|
|
return result;
|
2013-07-16 05:02:21 +08:00
|
|
|
|
|
|
|
disable:
|
2013-12-17 02:50:00 +08:00
|
|
|
nvme_disable_queue(dev, 0);
|
2014-04-08 07:10:11 +08:00
|
|
|
nvme_dev_list_remove(dev);
|
2013-07-16 05:02:21 +08:00
|
|
|
unmap:
|
|
|
|
nvme_dev_unmap(dev);
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2013-12-11 04:10:36 +08:00
|
|
|
static int nvme_remove_dead_ctrl(void *arg)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = (struct nvme_dev *)arg;
|
|
|
|
struct pci_dev *pdev = dev->pci_dev;
|
|
|
|
|
|
|
|
if (pci_get_drvdata(pdev))
|
|
|
|
pci_stop_and_remove_bus_device(pdev);
|
|
|
|
kref_put(&dev->kref, nvme_free_dev);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_remove_disks(struct work_struct *ws)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(ws, struct nvme_dev, reset_work);
|
|
|
|
|
|
|
|
nvme_dev_remove(dev);
|
2014-02-22 05:13:44 +08:00
|
|
|
nvme_free_queues(dev, 1);
|
2013-12-11 04:10:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_dev_resume(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = nvme_dev_start(dev);
|
|
|
|
if (ret && ret != -EBUSY)
|
|
|
|
return ret;
|
|
|
|
if (ret == -EBUSY) {
|
|
|
|
spin_lock(&dev_list_lock);
|
2014-03-07 23:24:49 +08:00
|
|
|
dev->reset_workfn = nvme_remove_disks;
|
2013-12-11 04:10:36 +08:00
|
|
|
queue_work(nvme_workq, &dev->reset_work);
|
|
|
|
spin_unlock(&dev_list_lock);
|
|
|
|
}
|
2013-12-11 04:10:37 +08:00
|
|
|
dev->initialized = 1;
|
2013-12-11 04:10:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_dev_reset(struct nvme_dev *dev)
|
|
|
|
{
|
|
|
|
nvme_dev_shutdown(dev);
|
|
|
|
if (nvme_dev_resume(dev)) {
|
|
|
|
dev_err(&dev->pci_dev->dev, "Device failed to resume\n");
|
|
|
|
kref_get(&dev->kref);
|
|
|
|
if (IS_ERR(kthread_run(nvme_remove_dead_ctrl, dev, "nvme%d",
|
|
|
|
dev->instance))) {
|
|
|
|
dev_err(&dev->pci_dev->dev,
|
|
|
|
"Failed to start controller remove task\n");
|
|
|
|
kref_put(&dev->kref, nvme_free_dev);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nvme_reset_failed_dev(struct work_struct *ws)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(ws, struct nvme_dev, reset_work);
|
|
|
|
nvme_dev_reset(dev);
|
|
|
|
}
|
|
|
|
|
2014-03-07 23:24:49 +08:00
|
|
|
static void nvme_reset_workfn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = container_of(work, struct nvme_dev, reset_work);
|
|
|
|
dev->reset_workfn(work);
|
|
|
|
}
|
|
|
|
|
2012-12-22 07:13:49 +08:00
|
|
|
static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
2013-07-16 05:02:19 +08:00
|
|
|
int result = -ENOMEM;
|
2011-01-21 01:50:14 +08:00
|
|
|
struct nvme_dev *dev;
|
|
|
|
|
|
|
|
dev = kzalloc(sizeof(*dev), GFP_KERNEL);
|
|
|
|
if (!dev)
|
|
|
|
return -ENOMEM;
|
|
|
|
dev->entry = kcalloc(num_possible_cpus(), sizeof(*dev->entry),
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!dev->entry)
|
|
|
|
goto free;
|
2011-01-21 02:01:49 +08:00
|
|
|
dev->queues = kcalloc(num_possible_cpus() + 1, sizeof(void *),
|
|
|
|
GFP_KERNEL);
|
2011-01-21 01:50:14 +08:00
|
|
|
if (!dev->queues)
|
|
|
|
goto free;
|
2014-03-25 00:46:25 +08:00
|
|
|
dev->io_queue = alloc_percpu(unsigned short);
|
|
|
|
if (!dev->io_queue)
|
|
|
|
goto free;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&dev->namespaces);
|
2014-03-07 23:24:49 +08:00
|
|
|
dev->reset_workfn = nvme_reset_failed_dev;
|
|
|
|
INIT_WORK(&dev->reset_work, nvme_reset_workfn);
|
2011-01-21 01:50:14 +08:00
|
|
|
dev->pci_dev = pdev;
|
2013-12-11 04:10:36 +08:00
|
|
|
pci_set_drvdata(pdev, dev);
|
2012-02-22 07:50:53 +08:00
|
|
|
result = nvme_set_instance(dev);
|
|
|
|
if (result)
|
2013-07-16 05:02:19 +08:00
|
|
|
goto free;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-02-10 22:56:01 +08:00
|
|
|
result = nvme_setup_prp_pools(dev);
|
|
|
|
if (result)
|
2013-07-16 05:02:19 +08:00
|
|
|
goto release;
|
2011-02-10 22:56:01 +08:00
|
|
|
|
2014-03-04 02:09:47 +08:00
|
|
|
kref_init(&dev->kref);
|
2013-07-16 05:02:21 +08:00
|
|
|
result = nvme_dev_start(dev);
|
2013-09-06 04:45:07 +08:00
|
|
|
if (result) {
|
|
|
|
if (result == -EBUSY)
|
|
|
|
goto create_cdev;
|
2013-07-16 05:02:19 +08:00
|
|
|
goto release_pools;
|
2013-09-06 04:45:07 +08:00
|
|
|
}
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-02-16 05:28:20 +08:00
|
|
|
result = nvme_dev_add(dev);
|
2013-09-06 04:45:07 +08:00
|
|
|
if (result)
|
2013-07-16 05:02:21 +08:00
|
|
|
goto shutdown;
|
2011-02-16 05:28:20 +08:00
|
|
|
|
2013-09-06 04:45:07 +08:00
|
|
|
create_cdev:
|
2013-02-20 01:17:58 +08:00
|
|
|
scnprintf(dev->name, sizeof(dev->name), "nvme%d", dev->instance);
|
|
|
|
dev->miscdev.minor = MISC_DYNAMIC_MINOR;
|
|
|
|
dev->miscdev.parent = &pdev->dev;
|
|
|
|
dev->miscdev.name = dev->name;
|
|
|
|
dev->miscdev.fops = &nvme_dev_fops;
|
|
|
|
result = misc_register(&dev->miscdev);
|
|
|
|
if (result)
|
|
|
|
goto remove;
|
|
|
|
|
2013-12-11 04:10:37 +08:00
|
|
|
dev->initialized = 1;
|
2011-01-21 01:50:14 +08:00
|
|
|
return 0;
|
|
|
|
|
2013-02-20 01:17:58 +08:00
|
|
|
remove:
|
|
|
|
nvme_dev_remove(dev);
|
2014-02-01 07:53:39 +08:00
|
|
|
nvme_free_namespaces(dev);
|
2013-07-16 05:02:21 +08:00
|
|
|
shutdown:
|
|
|
|
nvme_dev_shutdown(dev);
|
2013-07-16 05:02:19 +08:00
|
|
|
release_pools:
|
2013-12-17 02:50:00 +08:00
|
|
|
nvme_free_queues(dev, 0);
|
2011-02-10 22:56:01 +08:00
|
|
|
nvme_release_prp_pools(dev);
|
2013-07-16 05:02:19 +08:00
|
|
|
release:
|
|
|
|
nvme_release_instance(dev);
|
2011-01-21 01:50:14 +08:00
|
|
|
free:
|
2014-03-25 00:46:25 +08:00
|
|
|
free_percpu(dev->io_queue);
|
2011-01-21 01:50:14 +08:00
|
|
|
kfree(dev->queues);
|
|
|
|
kfree(dev->entry);
|
|
|
|
kfree(dev);
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2014-01-28 00:29:40 +08:00
|
|
|
static void nvme_shutdown(struct pci_dev *pdev)
|
|
|
|
{
|
|
|
|
struct nvme_dev *dev = pci_get_drvdata(pdev);
|
|
|
|
nvme_dev_shutdown(dev);
|
|
|
|
}
|
|
|
|
|
2012-12-22 07:13:49 +08:00
|
|
|
static void nvme_remove(struct pci_dev *pdev)
|
2011-01-21 01:50:14 +08:00
|
|
|
{
|
|
|
|
struct nvme_dev *dev = pci_get_drvdata(pdev);
|
2013-12-11 04:10:36 +08:00
|
|
|
|
|
|
|
spin_lock(&dev_list_lock);
|
|
|
|
list_del_init(&dev->node);
|
|
|
|
spin_unlock(&dev_list_lock);
|
|
|
|
|
|
|
|
pci_set_drvdata(pdev, NULL);
|
|
|
|
flush_work(&dev->reset_work);
|
2013-02-20 01:17:58 +08:00
|
|
|
misc_deregister(&dev->miscdev);
|
2013-12-11 04:10:36 +08:00
|
|
|
nvme_dev_remove(dev);
|
|
|
|
nvme_dev_shutdown(dev);
|
2013-12-17 02:50:00 +08:00
|
|
|
nvme_free_queues(dev, 0);
|
2014-02-22 05:13:44 +08:00
|
|
|
rcu_barrier();
|
2013-12-11 04:10:36 +08:00
|
|
|
nvme_release_instance(dev);
|
|
|
|
nvme_release_prp_pools(dev);
|
2013-02-20 01:17:58 +08:00
|
|
|
kref_put(&dev->kref, nvme_free_dev);
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* These functions are yet to be implemented */
|
|
|
|
#define nvme_error_detected NULL
|
|
|
|
#define nvme_dump_registers NULL
|
|
|
|
#define nvme_link_reset NULL
|
|
|
|
#define nvme_slot_reset NULL
|
|
|
|
#define nvme_error_resume NULL
|
2013-07-16 05:02:23 +08:00
|
|
|
|
2014-02-13 10:19:14 +08:00
|
|
|
#ifdef CONFIG_PM_SLEEP
|
2013-07-16 05:02:23 +08:00
|
|
|
static int nvme_suspend(struct device *dev)
|
|
|
|
{
|
|
|
|
struct pci_dev *pdev = to_pci_dev(dev);
|
|
|
|
struct nvme_dev *ndev = pci_get_drvdata(pdev);
|
|
|
|
|
|
|
|
nvme_dev_shutdown(ndev);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nvme_resume(struct device *dev)
|
|
|
|
{
|
|
|
|
struct pci_dev *pdev = to_pci_dev(dev);
|
|
|
|
struct nvme_dev *ndev = pci_get_drvdata(pdev);
|
|
|
|
|
2013-12-11 04:10:36 +08:00
|
|
|
if (nvme_dev_resume(ndev) && !work_busy(&ndev->reset_work)) {
|
2014-03-07 23:24:49 +08:00
|
|
|
ndev->reset_workfn = nvme_reset_failed_dev;
|
2013-12-11 04:10:36 +08:00
|
|
|
queue_work(nvme_workq, &ndev->reset_work);
|
|
|
|
}
|
|
|
|
return 0;
|
2013-07-16 05:02:23 +08:00
|
|
|
}
|
2014-02-13 10:19:14 +08:00
|
|
|
#endif
|
2013-07-16 05:02:23 +08:00
|
|
|
|
|
|
|
static SIMPLE_DEV_PM_OPS(nvme_dev_pm_ops, nvme_suspend, nvme_resume);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2012-09-08 00:33:17 +08:00
|
|
|
static const struct pci_error_handlers nvme_err_handler = {
|
2011-01-21 01:50:14 +08:00
|
|
|
.error_detected = nvme_error_detected,
|
|
|
|
.mmio_enabled = nvme_dump_registers,
|
|
|
|
.link_reset = nvme_link_reset,
|
|
|
|
.slot_reset = nvme_slot_reset,
|
|
|
|
.resume = nvme_error_resume,
|
|
|
|
};
|
|
|
|
|
|
|
|
/* Move to pci_ids.h later */
|
|
|
|
#define PCI_CLASS_STORAGE_EXPRESS 0x010802
|
|
|
|
|
2014-03-24 22:11:22 +08:00
|
|
|
static const struct pci_device_id nvme_id_table[] = {
|
2011-01-21 01:50:14 +08:00
|
|
|
{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_EXPRESS, 0xffffff) },
|
|
|
|
{ 0, }
|
|
|
|
};
|
|
|
|
MODULE_DEVICE_TABLE(pci, nvme_id_table);
|
|
|
|
|
|
|
|
static struct pci_driver nvme_driver = {
|
|
|
|
.name = "nvme",
|
|
|
|
.id_table = nvme_id_table,
|
|
|
|
.probe = nvme_probe,
|
2012-12-22 07:13:49 +08:00
|
|
|
.remove = nvme_remove,
|
2014-01-28 00:29:40 +08:00
|
|
|
.shutdown = nvme_shutdown,
|
2013-07-16 05:02:23 +08:00
|
|
|
.driver = {
|
|
|
|
.pm = &nvme_dev_pm_ops,
|
|
|
|
},
|
2011-01-21 01:50:14 +08:00
|
|
|
.err_handler = &nvme_err_handler,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init nvme_init(void)
|
|
|
|
{
|
2012-08-01 01:31:15 +08:00
|
|
|
int result;
|
2011-03-03 07:37:18 +08:00
|
|
|
|
2014-04-08 07:10:11 +08:00
|
|
|
init_waitqueue_head(&nvme_kthread_wait);
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2013-12-11 04:10:36 +08:00
|
|
|
nvme_workq = create_singlethread_workqueue("nvme");
|
|
|
|
if (!nvme_workq)
|
2014-04-08 07:10:11 +08:00
|
|
|
return -ENOMEM;
|
2013-12-11 04:10:36 +08:00
|
|
|
|
2012-07-26 06:05:18 +08:00
|
|
|
result = register_blkdev(nvme_major, "nvme");
|
|
|
|
if (result < 0)
|
2013-12-11 04:10:36 +08:00
|
|
|
goto kill_workq;
|
2012-07-26 06:05:18 +08:00
|
|
|
else if (result > 0)
|
2012-08-01 01:31:15 +08:00
|
|
|
nvme_major = result;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
|
|
|
result = pci_register_driver(&nvme_driver);
|
2011-03-03 07:37:18 +08:00
|
|
|
if (result)
|
|
|
|
goto unregister_blkdev;
|
|
|
|
return 0;
|
2011-01-21 01:50:14 +08:00
|
|
|
|
2011-03-03 07:37:18 +08:00
|
|
|
unregister_blkdev:
|
2011-01-21 01:50:14 +08:00
|
|
|
unregister_blkdev(nvme_major, "nvme");
|
2013-12-11 04:10:36 +08:00
|
|
|
kill_workq:
|
|
|
|
destroy_workqueue(nvme_workq);
|
2011-01-21 01:50:14 +08:00
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit nvme_exit(void)
|
|
|
|
{
|
|
|
|
pci_unregister_driver(&nvme_driver);
|
|
|
|
unregister_blkdev(nvme_major, "nvme");
|
2013-12-11 04:10:36 +08:00
|
|
|
destroy_workqueue(nvme_workq);
|
2014-04-08 07:10:11 +08:00
|
|
|
BUG_ON(nvme_thread && !IS_ERR(nvme_thread));
|
2014-05-10 10:42:26 +08:00
|
|
|
_nvme_check_size();
|
2011-01-21 01:50:14 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
MODULE_AUTHOR("Matthew Wilcox <willy@linux.intel.com>");
|
|
|
|
MODULE_LICENSE("GPL");
|
2014-03-24 22:11:22 +08:00
|
|
|
MODULE_VERSION("0.9");
|
2011-01-21 01:50:14 +08:00
|
|
|
module_init(nvme_init);
|
|
|
|
module_exit(nvme_exit);
|