From b80d5ccca3a012e91ca64a2a0b13049163a6a698 Mon Sep 17 00:00:00 2001 From: Haiyan Hu Date: Tue, 10 Sep 2013 11:25:37 +0800 Subject: NVMe: Avoid shift operation when writing cq head doorbell Changes the type of dev->db_stride to unsigned and changes the value stored there to be 1 << the current value. Then there is less calculation to be done at completion time. Signed-off-by: Haiyan Hu Signed-off-by: Matthew Wilcox --- include/linux/nvme.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/linux') diff --git a/include/linux/nvme.h b/include/linux/nvme.h index 26ebcf41c213..8119a476cb29 100644 --- a/include/linux/nvme.h +++ b/include/linux/nvme.h @@ -80,7 +80,7 @@ struct nvme_dev { struct dma_pool *prp_small_pool; int instance; int queue_count; - int db_stride; + u32 db_stride; u32 ctrl_config; struct msix_entry *entry; struct nvme_bar __iomem *bar; -- cgit v1.2.3 From 320a382746e0ab1304476ea7e986a8d416ab99db Mon Sep 17 00:00:00 2001 From: Keith Busch Date: Wed, 23 Oct 2013 13:07:34 -0600 Subject: NVMe: compat SG_IO ioctl For 32-bit versions of sg3-utils running on a 64-bit system. This is mostly a copy from the relevent portions of fs/compat_ioctl.c, with slight modifications for going through block_device_operations. Signed-off-by: Keith Busch Reviewed-by: Vishal Verma [fixed up CONFIG_COMPAT=n build problems] Signed-off-by: Matthew Wilcox --- drivers/block/nvme-core.c | 18 +++++- drivers/block/nvme-scsi.c | 147 ++++++++++++++++++++++++++++++++++++++++++++++ include/linux/nvme.h | 1 + 3 files changed, 165 insertions(+), 1 deletion(-) (limited to 'include/linux') diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c index df3daeb6227f..e44350de8a21 100644 --- a/drivers/block/nvme-core.c +++ b/drivers/block/nvme-core.c @@ -1568,10 +1568,26 @@ static int nvme_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd, } } +#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 + static const struct block_device_operations nvme_fops = { .owner = THIS_MODULE, .ioctl = nvme_ioctl, - .compat_ioctl = nvme_ioctl, + .compat_ioctl = nvme_compat_ioctl, }; static void nvme_resubmit_bios(struct nvme_queue *nvmeq) diff --git a/drivers/block/nvme-scsi.c b/drivers/block/nvme-scsi.c index 4a4ff4eb8e23..4a0ceb64e269 100644 --- a/drivers/block/nvme-scsi.c +++ b/drivers/block/nvme-scsi.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -3038,6 +3039,152 @@ int nvme_sg_io(struct nvme_ns *ns, struct sg_io_hdr __user *u_hdr) return retcode; } +#ifdef CONFIG_COMPAT +typedef struct sg_io_hdr32 { + compat_int_t interface_id; /* [i] 'S' for SCSI generic (required) */ + compat_int_t dxfer_direction; /* [i] data transfer direction */ + unsigned char cmd_len; /* [i] SCSI command length ( <= 16 bytes) */ + unsigned char mx_sb_len; /* [i] max length to write to sbp */ + unsigned short iovec_count; /* [i] 0 implies no scatter gather */ + compat_uint_t dxfer_len; /* [i] byte count of data transfer */ + compat_uint_t dxferp; /* [i], [*io] points to data transfer memory + or scatter gather list */ + compat_uptr_t cmdp; /* [i], [*i] points to command to perform */ + compat_uptr_t sbp; /* [i], [*o] points to sense_buffer memory */ + compat_uint_t timeout; /* [i] MAX_UINT->no timeout (unit: millisec) */ + compat_uint_t flags; /* [i] 0 -> default, see SG_FLAG... */ + compat_int_t pack_id; /* [i->o] unused internally (normally) */ + compat_uptr_t usr_ptr; /* [i->o] unused internally */ + unsigned char status; /* [o] scsi status */ + unsigned char masked_status; /* [o] shifted, masked scsi status */ + unsigned char msg_status; /* [o] messaging level data (optional) */ + unsigned char sb_len_wr; /* [o] byte count actually written to sbp */ + unsigned short host_status; /* [o] errors from host adapter */ + unsigned short driver_status; /* [o] errors from software driver */ + compat_int_t resid; /* [o] dxfer_len - actual_transferred */ + compat_uint_t duration; /* [o] time taken by cmd (unit: millisec) */ + compat_uint_t info; /* [o] auxiliary information */ +} sg_io_hdr32_t; /* 64 bytes long (on sparc32) */ + +typedef struct sg_iovec32 { + compat_uint_t iov_base; + compat_uint_t iov_len; +} sg_iovec32_t; + +static int sg_build_iovec(sg_io_hdr_t __user *sgio, void __user *dxferp, u16 iovec_count) +{ + sg_iovec_t __user *iov = (sg_iovec_t __user *) (sgio + 1); + sg_iovec32_t __user *iov32 = dxferp; + int i; + + for (i = 0; i < iovec_count; i++) { + u32 base, len; + + if (get_user(base, &iov32[i].iov_base) || + get_user(len, &iov32[i].iov_len) || + put_user(compat_ptr(base), &iov[i].iov_base) || + put_user(len, &iov[i].iov_len)) + return -EFAULT; + } + + if (put_user(iov, &sgio->dxferp)) + return -EFAULT; + return 0; +} + +int nvme_sg_io32(struct nvme_ns *ns, unsigned long arg) +{ + sg_io_hdr32_t __user *sgio32 = (sg_io_hdr32_t __user *)arg; + sg_io_hdr_t __user *sgio; + u16 iovec_count; + u32 data; + void __user *dxferp; + int err; + int interface_id; + + if (get_user(interface_id, &sgio32->interface_id)) + return -EFAULT; + if (interface_id != 'S') + return -EINVAL; + + if (get_user(iovec_count, &sgio32->iovec_count)) + return -EFAULT; + + { + void __user *top = compat_alloc_user_space(0); + void __user *new = compat_alloc_user_space(sizeof(sg_io_hdr_t) + + (iovec_count * sizeof(sg_iovec_t))); + if (new > top) + return -EINVAL; + + sgio = new; + } + + /* Ok, now construct. */ + if (copy_in_user(&sgio->interface_id, &sgio32->interface_id, + (2 * sizeof(int)) + + (2 * sizeof(unsigned char)) + + (1 * sizeof(unsigned short)) + + (1 * sizeof(unsigned int)))) + return -EFAULT; + + if (get_user(data, &sgio32->dxferp)) + return -EFAULT; + dxferp = compat_ptr(data); + if (iovec_count) { + if (sg_build_iovec(sgio, dxferp, iovec_count)) + return -EFAULT; + } else { + if (put_user(dxferp, &sgio->dxferp)) + return -EFAULT; + } + + { + unsigned char __user *cmdp; + unsigned char __user *sbp; + + if (get_user(data, &sgio32->cmdp)) + return -EFAULT; + cmdp = compat_ptr(data); + + if (get_user(data, &sgio32->sbp)) + return -EFAULT; + sbp = compat_ptr(data); + + if (put_user(cmdp, &sgio->cmdp) || + put_user(sbp, &sgio->sbp)) + return -EFAULT; + } + + if (copy_in_user(&sgio->timeout, &sgio32->timeout, + 3 * sizeof(int))) + return -EFAULT; + + if (get_user(data, &sgio32->usr_ptr)) + return -EFAULT; + if (put_user(compat_ptr(data), &sgio->usr_ptr)) + return -EFAULT; + + err = nvme_sg_io(ns, sgio); + if (err >= 0) { + void __user *datap; + + if (copy_in_user(&sgio32->pack_id, &sgio->pack_id, + sizeof(int)) || + get_user(datap, &sgio->usr_ptr) || + put_user((u32)(unsigned long)datap, + &sgio32->usr_ptr) || + copy_in_user(&sgio32->status, &sgio->status, + (4 * sizeof(unsigned char)) + + (2 * sizeof(unsigned short)) + + (3 * sizeof(int)))) + err = -EFAULT; + } + + return err; +} +#endif + int nvme_sg_get_version_num(int __user *ip) { return put_user(sg_version_num, ip); diff --git a/include/linux/nvme.h b/include/linux/nvme.h index 8119a476cb29..5bc29197d90e 100644 --- a/include/linux/nvme.h +++ b/include/linux/nvme.h @@ -165,6 +165,7 @@ int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11, struct sg_io_hdr; int nvme_sg_io(struct nvme_ns *ns, struct sg_io_hdr __user *u_hdr); +int nvme_sg_io32(struct nvme_ns *ns, unsigned long arg); int nvme_sg_get_version_num(int __user *ip); #endif /* _LINUX_NVME_H */ -- cgit v1.2.3 From 9a6b94584de1a0467d85b435df9c744c5c45a270 Mon Sep 17 00:00:00 2001 From: Keith Busch Date: Tue, 10 Dec 2013 13:10:36 -0700 Subject: NVMe: Device resume error handling Adds controller error handling on resume power management. If the device fails to initialize, the device is queued for a reset. If the reset fails, a thread is spawned to remove the pci device. If the device resumes as "busy", the device is responding to admin commands but will not create IO queues. In this case, we need to remove the gendisks and free the IO queues since they can't be used and may be holding bios in their lists. From testing, the dma pools require a pci device so this had to change the pci driver 'remove' to release the dma resources in line with that call instead of after all references to the device are released. Signed-off-by: Keith Busch Signed-off-by: Matthew Wilcox --- drivers/block/nvme-core.c | 108 +++++++++++++++++++++++++++++++++++++++------- include/linux/nvme.h | 1 + 2 files changed, 94 insertions(+), 15 deletions(-) (limited to 'include/linux') diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c index 300766973d76..000bca43c23b 100644 --- a/drivers/block/nvme-core.c +++ b/drivers/block/nvme-core.c @@ -58,6 +58,7 @@ module_param(use_threaded_interrupts, int, 0); static DEFINE_SPINLOCK(dev_list_lock); static LIST_HEAD(dev_list); static struct task_struct *nvme_thread; +static struct workqueue_struct *nvme_workq; /* * An NVM Express queue. Each device has at least two (one for admin @@ -1968,7 +1969,6 @@ static int nvme_dev_map(struct nvme_dev *dev) dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32))) goto disable; - pci_set_drvdata(pdev, dev); dev->bar = ioremap(pci_resource_start(pdev, 0), 8192); if (!dev->bar) goto disable; @@ -1995,9 +1995,9 @@ static void nvme_dev_unmap(struct nvme_dev *dev) if (dev->bar) { iounmap(dev->bar); dev->bar = NULL; + pci_release_regions(dev->pci_dev); } - pci_release_regions(dev->pci_dev); if (pci_is_enabled(dev->pci_dev)) pci_disable_device(dev->pci_dev); } @@ -2085,11 +2085,6 @@ static void nvme_release_instance(struct nvme_dev *dev) static void nvme_free_dev(struct kref *kref) { struct nvme_dev *dev = container_of(kref, struct nvme_dev, kref); - nvme_dev_remove(dev); - nvme_dev_shutdown(dev); - nvme_free_queues(dev); - nvme_release_instance(dev); - nvme_release_prp_pools(dev); kfree(dev->queues); kfree(dev->entry); kfree(dev); @@ -2161,6 +2156,70 @@ static int nvme_dev_start(struct nvme_dev *dev) return result; } +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) +{ + int i; + struct nvme_dev *dev = container_of(ws, struct nvme_dev, reset_work); + + nvme_dev_remove(dev); + spin_lock(&dev_list_lock); + for (i = dev->queue_count - 1; i > 0; i--) { + BUG_ON(!dev->queues[i] || !dev->queues[i]->q_suspended); + nvme_free_queue(dev->queues[i]); + dev->queue_count--; + dev->queues[i] = NULL; + } + spin_unlock(&dev_list_lock); +} + +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); + INIT_WORK(&dev->reset_work, nvme_remove_disks); + queue_work(nvme_workq, &dev->reset_work); + spin_unlock(&dev_list_lock); + } + 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); +} + static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id) { int result = -ENOMEM; @@ -2180,7 +2239,7 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id) INIT_LIST_HEAD(&dev->namespaces); dev->pci_dev = pdev; - + pci_set_drvdata(pdev, dev); result = nvme_set_instance(dev); if (result) goto free; @@ -2232,7 +2291,19 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id) static void nvme_remove(struct pci_dev *pdev) { struct nvme_dev *dev = pci_get_drvdata(pdev); + + 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); misc_deregister(&dev->miscdev); + nvme_dev_remove(dev); + nvme_dev_shutdown(dev); + nvme_free_queues(dev); + nvme_release_instance(dev); + nvme_release_prp_pools(dev); kref_put(&dev->kref, nvme_free_dev); } @@ -2256,13 +2327,12 @@ static int nvme_resume(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); struct nvme_dev *ndev = pci_get_drvdata(pdev); - int ret; - ret = nvme_dev_start(ndev); - /* XXX: should remove gendisks if resume fails */ - if (ret) - nvme_free_queues(ndev); - return ret; + if (nvme_dev_resume(ndev) && !work_busy(&ndev->reset_work)) { + INIT_WORK(&ndev->reset_work, nvme_reset_failed_dev); + queue_work(nvme_workq, &ndev->reset_work); + } + return 0; } static SIMPLE_DEV_PM_OPS(nvme_dev_pm_ops, nvme_suspend, nvme_resume); @@ -2303,9 +2373,14 @@ static int __init nvme_init(void) if (IS_ERR(nvme_thread)) return PTR_ERR(nvme_thread); + result = -ENOMEM; + nvme_workq = create_singlethread_workqueue("nvme"); + if (!nvme_workq) + goto kill_kthread; + result = register_blkdev(nvme_major, "nvme"); if (result < 0) - goto kill_kthread; + goto kill_workq; else if (result > 0) nvme_major = result; @@ -2316,6 +2391,8 @@ static int __init nvme_init(void) unregister_blkdev: unregister_blkdev(nvme_major, "nvme"); + kill_workq: + destroy_workqueue(nvme_workq); kill_kthread: kthread_stop(nvme_thread); return result; @@ -2325,6 +2402,7 @@ static void __exit nvme_exit(void) { pci_unregister_driver(&nvme_driver); unregister_blkdev(nvme_major, "nvme"); + destroy_workqueue(nvme_workq); kthread_stop(nvme_thread); } diff --git a/include/linux/nvme.h b/include/linux/nvme.h index 5bc29197d90e..eed81cc56d7e 100644 --- a/include/linux/nvme.h +++ b/include/linux/nvme.h @@ -87,6 +87,7 @@ struct nvme_dev { struct list_head namespaces; struct kref kref; struct miscdevice miscdev; + struct work_struct reset_work; char name[12]; char serial[20]; char model[40]; -- cgit v1.2.3 From d4b4ff8e28b474fac0fbfa9cfc40f88b9e41e380 Mon Sep 17 00:00:00 2001 From: Keith Busch Date: Tue, 10 Dec 2013 13:10:37 -0700 Subject: NVMe: Schedule reset for failed controllers Schedules a controller reset when it indicates it has a failed status. If the device does not become ready after a reset, the pci device will be scheduled for removal. Signed-off-by: Keith Busch [fixed checkpatch issue] Signed-off-by: Matthew Wilcox --- drivers/block/nvme-core.c | 21 +++++++++++++++++++-- include/linux/nvme.h | 1 + 2 files changed, 20 insertions(+), 2 deletions(-) (limited to 'include/linux') diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c index 000bca43c23b..2f5b9f5f5a21 100644 --- a/drivers/block/nvme-core.c +++ b/drivers/block/nvme-core.c @@ -60,6 +60,8 @@ static LIST_HEAD(dev_list); static struct task_struct *nvme_thread; static struct workqueue_struct *nvme_workq; +static void nvme_reset_failed_dev(struct work_struct *ws); + /* * An NVM Express queue. Each device has at least two (one for admin * commands and one for I/O commands). @@ -1612,13 +1614,25 @@ static void nvme_resubmit_bios(struct nvme_queue *nvmeq) static int nvme_kthread(void *data) { - struct nvme_dev *dev; + struct nvme_dev *dev, *next; while (!kthread_should_stop()) { set_current_state(TASK_INTERRUPTIBLE); spin_lock(&dev_list_lock); - list_for_each_entry(dev, &dev_list, node) { + list_for_each_entry_safe(dev, next, &dev_list, node) { int i; + 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"); + INIT_WORK(&dev->reset_work, + nvme_reset_failed_dev); + queue_work(nvme_workq, &dev->reset_work); + continue; + } for (i = 0; i < dev->queue_count; i++) { struct nvme_queue *nvmeq = dev->queues[i]; if (!nvmeq) @@ -2006,6 +2020,7 @@ static void nvme_dev_shutdown(struct nvme_dev *dev) { int i; + dev->initialized = 0; for (i = dev->queue_count - 1; i >= 0; i--) nvme_disable_queue(dev, i); @@ -2196,6 +2211,7 @@ static int nvme_dev_resume(struct nvme_dev *dev) queue_work(nvme_workq, &dev->reset_work); spin_unlock(&dev_list_lock); } + dev->initialized = 1; return 0; } @@ -2269,6 +2285,7 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id) if (result) goto remove; + dev->initialized = 1; kref_init(&dev->kref); return 0; diff --git a/include/linux/nvme.h b/include/linux/nvme.h index eed81cc56d7e..117d877e8be5 100644 --- a/include/linux/nvme.h +++ b/include/linux/nvme.h @@ -95,6 +95,7 @@ struct nvme_dev { u32 max_hw_sectors; u32 stripe_size; u16 oncs; + u8 initialized; }; /* -- cgit v1.2.3 From c30341dc3c436cf43508cd44cdfbb3810c38c195 Mon Sep 17 00:00:00 2001 From: Keith Busch Date: Tue, 10 Dec 2013 13:10:38 -0700 Subject: NVMe: Abort timed out commands Send nvme abort command to io requests that have timed out on an initialized device. If the command is not returned after another timeout, schedule the controller for reset. Signed-off-by: Keith Busch [fix endianness issues] Signed-off-by: Matthew Wilcox --- drivers/block/nvme-core.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++- include/linux/nvme.h | 1 + include/uapi/linux/nvme.h | 11 ++++++++ 3 files changed, 79 insertions(+), 1 deletion(-) (limited to 'include/linux') diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c index 2f5b9f5f5a21..75049532df96 100644 --- a/drivers/block/nvme-core.c +++ b/drivers/block/nvme-core.c @@ -83,6 +83,7 @@ struct nvme_queue { u16 sq_head; u16 sq_tail; u16 cq_head; + u16 qid; u8 cq_phase; u8 cqe_seen; u8 q_suspended; @@ -100,6 +101,7 @@ static inline void _nvme_check_size(void) BUILD_BUG_ON(sizeof(struct nvme_delete_queue) != 64); BUILD_BUG_ON(sizeof(struct nvme_features) != 64); BUILD_BUG_ON(sizeof(struct nvme_format_cmd) != 64); + BUILD_BUG_ON(sizeof(struct nvme_abort_cmd) != 64); 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); @@ -114,6 +116,7 @@ struct nvme_cmd_info { nvme_completion_fn fn; void *ctx; unsigned long timeout; + int aborted; }; static struct nvme_cmd_info *nvme_cmd_info(struct nvme_queue *nvmeq) @@ -157,6 +160,7 @@ static int alloc_cmdid(struct nvme_queue *nvmeq, void *ctx, info[cmdid].fn = handler; info[cmdid].ctx = ctx; info[cmdid].timeout = jiffies + timeout; + info[cmdid].aborted = 0; return cmdid; } @@ -175,6 +179,7 @@ static int alloc_cmdid_killable(struct nvme_queue *nvmeq, void *ctx, #define CMD_CTX_COMPLETED (0x310 + CMD_CTX_BASE) #define CMD_CTX_INVALID (0x314 + CMD_CTX_BASE) #define CMD_CTX_FLUSH (0x318 + CMD_CTX_BASE) +#define CMD_CTX_ABORT (0x31C + CMD_CTX_BASE) static void special_completion(struct nvme_dev *dev, void *ctx, struct nvme_completion *cqe) @@ -183,6 +188,10 @@ static void special_completion(struct nvme_dev *dev, void *ctx, return; if (ctx == CMD_CTX_FLUSH) return; + if (ctx == CMD_CTX_ABORT) { + ++dev->abort_limit; + return; + } if (ctx == CMD_CTX_COMPLETED) { dev_warn(&dev->pci_dev->dev, "completed id %d twice on queue %d\n", @@ -1004,6 +1013,56 @@ int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11, return nvme_submit_admin_cmd(dev, &c, result); } +/** + * 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); + + 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); + INIT_WORK(&dev->reset_work, nvme_reset_failed_dev); + queue_work(nvme_workq, &dev->reset_work); + return; + } + + if (!dev->abort_limit) + return; + + a_cmdid = alloc_cmdid(dev->queues[0], CMD_CTX_ABORT, special_completion, + 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); + nvme_submit_cmd(dev->queues[0], &cmd); +} + /** * nvme_cancel_ios - Cancel outstanding I/Os * @queue: The queue to cancel I/Os on @@ -1027,7 +1086,12 @@ static void nvme_cancel_ios(struct nvme_queue *nvmeq, bool timeout) continue; if (info[cmdid].ctx == CMD_CTX_CANCELLED) continue; - dev_warn(nvmeq->q_dmadev, "Cancelling I/O %d\n", cmdid); + 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); ctx = cancel_cmdid(nvmeq, cmdid, &fn); fn(nvmeq->dev, ctx, &cqe); } @@ -1119,6 +1183,7 @@ static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid, nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride]; nvmeq->q_depth = depth; nvmeq->cq_vector = vector; + nvmeq->qid = qid; nvmeq->q_suspended = 1; dev->queue_count++; @@ -1929,6 +1994,7 @@ static int nvme_dev_add(struct nvme_dev *dev) ctrl = mem; nn = le32_to_cpup(&ctrl->nn); dev->oncs = le16_to_cpup(&ctrl->oncs); + dev->abort_limit = ctrl->acl + 1; 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)); diff --git a/include/linux/nvme.h b/include/linux/nvme.h index 117d877e8be5..69ae03f6eb15 100644 --- a/include/linux/nvme.h +++ b/include/linux/nvme.h @@ -95,6 +95,7 @@ struct nvme_dev { u32 max_hw_sectors; u32 stripe_size; u16 oncs; + u16 abort_limit; u8 initialized; }; diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h index 989c04e0c563..e5ab62201119 100644 --- a/include/uapi/linux/nvme.h +++ b/include/uapi/linux/nvme.h @@ -350,6 +350,16 @@ struct nvme_delete_queue { __u32 rsvd11[5]; }; +struct nvme_abort_cmd { + __u8 opcode; + __u8 flags; + __u16 command_id; + __u32 rsvd1[9]; + __le16 sqid; + __u16 cid; + __u32 rsvd11[5]; +}; + struct nvme_download_firmware { __u8 opcode; __u8 flags; @@ -384,6 +394,7 @@ struct nvme_command { struct nvme_download_firmware dlfw; struct nvme_format_cmd format; struct nvme_dsm_cmd dsm; + struct nvme_abort_cmd abort; }; }; -- cgit v1.2.3