#define SQ_SIZE(depth) (depth * sizeof(struct nvme_command))
#define CQ_SIZE(depth) (depth * sizeof(struct nvme_completion))
#define NVME_MINORS 64
-#define IO_TIMEOUT (5 * HZ)
+#define NVME_IO_TIMEOUT (5 * HZ)
#define ADMIN_TIMEOUT (60 * HZ)
static int nvme_major;
char serial[20];
char model[40];
char firmware_rev[8];
+ u32 max_hw_sectors;
};
/*
BUILD_BUG_ON(sizeof(struct nvme_lba_range_type) != 64);
}
-typedef void (*nvme_completion_fn)(struct nvme_queue *, void *,
+typedef void (*nvme_completion_fn)(struct nvme_dev *, void *,
struct nvme_completion *);
struct nvme_cmd_info {
#define CMD_CTX_INVALID (0x314 + CMD_CTX_BASE)
#define CMD_CTX_FLUSH (0x318 + CMD_CTX_BASE)
-static void special_completion(struct nvme_queue *nvmeq, void *ctx,
+static void special_completion(struct nvme_dev *dev, void *ctx,
struct nvme_completion *cqe)
{
if (ctx == CMD_CTX_CANCELLED)
if (ctx == CMD_CTX_FLUSH)
return;
if (ctx == CMD_CTX_COMPLETED) {
- dev_warn(nvmeq->q_dmadev,
+ dev_warn(&dev->pci_dev->dev,
"completed id %d twice on queue %d\n",
cqe->command_id, le16_to_cpup(&cqe->sq_id));
return;
}
if (ctx == CMD_CTX_INVALID) {
- dev_warn(nvmeq->q_dmadev,
+ dev_warn(&dev->pci_dev->dev,
"invalid id %d completed on queue %d\n",
cqe->command_id, le16_to_cpup(&cqe->sq_id));
return;
}
- dev_warn(nvmeq->q_dmadev, "Unknown special completion %p\n", ctx);
+ dev_warn(&dev->pci_dev->dev, "Unknown special completion %p\n", ctx);
}
/*
return ctx;
}
-static struct nvme_queue *get_nvmeq(struct nvme_ns *ns)
+static struct nvme_queue *get_nvmeq(struct nvme_dev *dev)
{
- return ns->dev->queues[get_cpu() + 1];
+ return dev->queues[get_cpu() + 1];
}
static void put_nvmeq(struct nvme_queue *nvmeq)
return 0;
}
-struct nvme_prps {
- int npages; /* 0 means small pool in use */
+/*
+ * The nvme_iod describes the data in an I/O, including the list of PRP
+ * entries. You can't see it in this data structure because C doesn't let
+ * me express that. Use nvme_alloc_iod to ensure there's enough space
+ * allocated to store the PRP list.
+ */
+struct nvme_iod {
+ void *private; /* For the use of the submitter of the I/O */
+ int npages; /* In the PRP list. 0 means small pool in use */
+ int offset; /* Of PRP list */
+ int nents; /* Used in scatterlist */
+ int length; /* Of data, in bytes */
dma_addr_t first_dma;
- __le64 *list[0];
+ struct scatterlist sg[0];
};
-static void nvme_free_prps(struct nvme_dev *dev, struct nvme_prps *prps)
+static __le64 **iod_list(struct nvme_iod *iod)
{
- const int last_prp = PAGE_SIZE / 8 - 1;
- int i;
- dma_addr_t prp_dma;
+ return ((void *)iod) + iod->offset;
+}
- if (!prps)
- return;
+/*
+ * 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);
+}
- prp_dma = prps->first_dma;
+static struct nvme_iod *
+nvme_alloc_iod(unsigned nseg, unsigned nbytes, gfp_t gfp)
+{
+ struct nvme_iod *iod = kmalloc(sizeof(struct nvme_iod) +
+ sizeof(__le64 *) * nvme_npages(nbytes) +
+ sizeof(struct scatterlist) * nseg, gfp);
- if (prps->npages == 0)
- dma_pool_free(dev->prp_small_pool, prps->list[0], prp_dma);
- for (i = 0; i < prps->npages; i++) {
- __le64 *prp_list = prps->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;
+ if (iod) {
+ iod->offset = offsetof(struct nvme_iod, sg[nseg]);
+ iod->npages = -1;
+ iod->length = nbytes;
}
- kfree(prps);
-}
-struct nvme_bio {
- struct bio *bio;
- int nents;
- struct nvme_prps *prps;
- struct scatterlist sg[0];
-};
+ return iod;
+}
-/* XXX: use a mempool */
-static struct nvme_bio *alloc_nbio(unsigned nseg, gfp_t gfp)
+static void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
{
- return kzalloc(sizeof(struct nvme_bio) +
- sizeof(struct scatterlist) * nseg, gfp);
+ 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);
}
-static void free_nbio(struct nvme_queue *nvmeq, struct nvme_bio *nbio)
+static void requeue_bio(struct nvme_dev *dev, struct bio *bio)
{
- nvme_free_prps(nvmeq->dev, nbio->prps);
- kfree(nbio);
+ struct nvme_queue *nvmeq = get_nvmeq(dev);
+ if (bio_list_empty(&nvmeq->sq_cong))
+ add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
+ bio_list_add(&nvmeq->sq_cong, bio);
+ put_nvmeq(nvmeq);
+ wake_up_process(nvme_thread);
}
-static void bio_completion(struct nvme_queue *nvmeq, void *ctx,
+static void bio_completion(struct nvme_dev *dev, void *ctx,
struct nvme_completion *cqe)
{
- struct nvme_bio *nbio = ctx;
- struct bio *bio = nbio->bio;
+ struct nvme_iod *iod = ctx;
+ struct bio *bio = iod->private;
u16 status = le16_to_cpup(&cqe->status) >> 1;
- dma_unmap_sg(nvmeq->q_dmadev, nbio->sg, nbio->nents,
+ dma_unmap_sg(&dev->pci_dev->dev, iod->sg, iod->nents,
bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
- free_nbio(nvmeq, nbio);
+ nvme_free_iod(dev, iod);
if (status) {
bio_endio(bio, -EIO);
} else if (bio->bi_vcnt > bio->bi_idx) {
- if (bio_list_empty(&nvmeq->sq_cong))
- add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
- bio_list_add(&nvmeq->sq_cong, bio);
- wake_up_process(nvme_thread);
+ requeue_bio(dev, bio);
} else {
bio_endio(bio, 0);
}
}
/* length is in bytes. gfp flags indicates whether we may sleep. */
-static struct nvme_prps *nvme_setup_prps(struct nvme_dev *dev,
- struct nvme_common_command *cmd,
- struct scatterlist *sg, int *len,
- gfp_t gfp)
+static int nvme_setup_prps(struct nvme_dev *dev,
+ struct nvme_common_command *cmd, struct nvme_iod *iod,
+ int total_len, gfp_t gfp)
{
struct dma_pool *pool;
- int length = *len;
+ int length = total_len;
+ struct scatterlist *sg = iod->sg;
int dma_len = sg_dma_len(sg);
u64 dma_addr = sg_dma_address(sg);
int offset = offset_in_page(dma_addr);
__le64 *prp_list;
+ __le64 **list = iod_list(iod);
dma_addr_t prp_dma;
- int nprps, npages, i;
- struct nvme_prps *prps = NULL;
+ int nprps, i;
cmd->prp1 = cpu_to_le64(dma_addr);
length -= (PAGE_SIZE - offset);
if (length <= 0)
- return prps;
+ return total_len;
dma_len -= (PAGE_SIZE - offset);
if (dma_len) {
if (length <= PAGE_SIZE) {
cmd->prp2 = cpu_to_le64(dma_addr);
- return prps;
+ return total_len;
}
nprps = DIV_ROUND_UP(length, PAGE_SIZE);
- npages = DIV_ROUND_UP(8 * nprps, PAGE_SIZE - 8);
- prps = kmalloc(sizeof(*prps) + sizeof(__le64 *) * npages, gfp);
- if (!prps) {
- cmd->prp2 = cpu_to_le64(dma_addr);
- *len = (*len - length) + PAGE_SIZE;
- return prps;
- }
-
if (nprps <= (256 / 8)) {
pool = dev->prp_small_pool;
- prps->npages = 0;
+ iod->npages = 0;
} else {
pool = dev->prp_page_pool;
- prps->npages = 1;
+ iod->npages = 1;
}
prp_list = dma_pool_alloc(pool, gfp, &prp_dma);
if (!prp_list) {
cmd->prp2 = cpu_to_le64(dma_addr);
- *len = (*len - length) + PAGE_SIZE;
- kfree(prps);
- return NULL;
+ iod->npages = -1;
+ return (total_len - length) + PAGE_SIZE;
}
- prps->list[0] = prp_list;
- prps->first_dma = prp_dma;
+ list[0] = prp_list;
+ iod->first_dma = prp_dma;
cmd->prp2 = cpu_to_le64(prp_dma);
i = 0;
for (;;) {
if (i == PAGE_SIZE / 8) {
__le64 *old_prp_list = prp_list;
prp_list = dma_pool_alloc(pool, gfp, &prp_dma);
- if (!prp_list) {
- *len = (*len - length);
- return prps;
- }
- prps->list[prps->npages++] = prp_list;
+ if (!prp_list)
+ return total_len - length;
+ list[iod->npages++] = prp_list;
prp_list[0] = old_prp_list[i - 1];
old_prp_list[i - 1] = cpu_to_le64(prp_dma);
i = 1;
dma_len = sg_dma_len(sg);
}
- return prps;
+ return total_len;
}
/* 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))
-static int nvme_map_bio(struct device *dev, struct nvme_bio *nbio,
+static int nvme_map_bio(struct device *dev, struct nvme_iod *iod,
struct bio *bio, enum dma_data_direction dma_dir, int psegs)
{
struct bio_vec *bvec, *bvprv = NULL;
struct scatterlist *sg = NULL;
int i, old_idx, length = 0, nsegs = 0;
- sg_init_table(nbio->sg, psegs);
+ sg_init_table(iod->sg, psegs);
old_idx = bio->bi_idx;
bio_for_each_segment(bvec, bio, i) {
if (bvprv && BIOVEC_PHYS_MERGEABLE(bvprv, bvec)) {
} else {
if (bvprv && BIOVEC_NOT_VIRT_MERGEABLE(bvprv, bvec))
break;
- sg = sg ? sg + 1 : nbio->sg;
+ sg = sg ? sg + 1 : iod->sg;
sg_set_page(sg, bvec->bv_page, bvec->bv_len,
bvec->bv_offset);
nsegs++;
bvprv = bvec;
}
bio->bi_idx = i;
- nbio->nents = nsegs;
+ iod->nents = nsegs;
sg_mark_end(sg);
- if (dma_map_sg(dev, nbio->sg, nbio->nents, dma_dir) == 0) {
+ if (dma_map_sg(dev, iod->sg, iod->nents, dma_dir) == 0) {
bio->bi_idx = old_idx;
return -ENOMEM;
}
static int nvme_submit_flush_data(struct nvme_queue *nvmeq, struct nvme_ns *ns)
{
int cmdid = alloc_cmdid(nvmeq, (void *)CMD_CTX_FLUSH,
- special_completion, IO_TIMEOUT);
+ special_completion, NVME_IO_TIMEOUT);
if (unlikely(cmdid < 0))
return cmdid;
struct bio *bio)
{
struct nvme_command *cmnd;
- struct nvme_bio *nbio;
+ struct nvme_iod *iod;
enum dma_data_direction dma_dir;
int cmdid, length, result = -ENOMEM;
u16 control;
return result;
}
- nbio = alloc_nbio(psegs, GFP_ATOMIC);
- if (!nbio)
+ iod = nvme_alloc_iod(psegs, bio->bi_size, GFP_ATOMIC);
+ if (!iod)
goto nomem;
- nbio->bio = bio;
+ iod->private = bio;
result = -EBUSY;
- cmdid = alloc_cmdid(nvmeq, nbio, bio_completion, IO_TIMEOUT);
+ cmdid = alloc_cmdid(nvmeq, iod, bio_completion, NVME_IO_TIMEOUT);
if (unlikely(cmdid < 0))
- goto free_nbio;
+ goto free_iod;
if ((bio->bi_rw & REQ_FLUSH) && !psegs)
return nvme_submit_flush(nvmeq, ns, cmdid);
dma_dir = DMA_FROM_DEVICE;
}
- result = nvme_map_bio(nvmeq->q_dmadev, nbio, bio, dma_dir, psegs);
+ result = nvme_map_bio(nvmeq->q_dmadev, iod, bio, dma_dir, psegs);
if (result < 0)
- goto free_nbio;
+ goto free_iod;
length = result;
cmnd->rw.command_id = cmdid;
cmnd->rw.nsid = cpu_to_le32(ns->ns_id);
- nbio->prps = nvme_setup_prps(nvmeq->dev, &cmnd->common, nbio->sg,
- &length, GFP_ATOMIC);
+ length = nvme_setup_prps(nvmeq->dev, &cmnd->common, iod, length,
+ GFP_ATOMIC);
cmnd->rw.slba = cpu_to_le64(bio->bi_sector >> (ns->lba_shift - 9));
cmnd->rw.length = cpu_to_le16((length >> ns->lba_shift) - 1);
cmnd->rw.control = cpu_to_le16(control);
return 0;
- free_nbio:
- free_nbio(nvmeq, nbio);
+ free_iod:
+ nvme_free_iod(nvmeq->dev, iod);
nomem:
return result;
}
static int nvme_make_request(struct request_queue *q, struct bio *bio)
{
struct nvme_ns *ns = q->queuedata;
- struct nvme_queue *nvmeq = get_nvmeq(ns);
+ struct nvme_queue *nvmeq = get_nvmeq(ns->dev);
int result = -EBUSY;
spin_lock_irq(&nvmeq->q_lock);
}
ctx = free_cmdid(nvmeq, cqe.command_id, &fn);
- fn(nvmeq, ctx, &cqe);
+ fn(nvmeq->dev, ctx, &cqe);
}
/* If the controller ignores the cq head doorbell and continuously
int status;
};
-static void sync_completion(struct nvme_queue *nvmeq, void *ctx,
+static void sync_completion(struct nvme_dev *dev, void *ctx,
struct nvme_completion *cqe)
{
struct sync_cmd_info *cmdinfo = ctx;
}
static int nvme_get_features(struct nvme_dev *dev, unsigned fid,
- unsigned dword11, dma_addr_t dma_addr, u32 *result)
+ unsigned nsid, dma_addr_t dma_addr)
{
struct nvme_command c;
memset(&c, 0, sizeof(c));
c.features.opcode = nvme_admin_get_features;
+ c.features.nsid = cpu_to_le32(nsid);
+ c.features.prp1 = cpu_to_le64(dma_addr);
+ c.features.fid = cpu_to_le32(fid);
+
+ return nvme_submit_admin_cmd(dev, &c, NULL);
+}
+
+static int nvme_set_features(struct nvme_dev *dev, unsigned fid,
+ unsigned dword11, dma_addr_t dma_addr, u32 *result)
+{
+ 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);
return nvme_submit_admin_cmd(dev, &c, result);
}
+/**
+ * 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 = {
+ .status = cpu_to_le16(NVME_SC_ABORT_REQ) << 1,
+ };
+
+ if (timeout && !time_after(now, info[cmdid].timeout))
+ continue;
+ dev_warn(nvmeq->q_dmadev, "Cancelling I/O %d\n", cmdid);
+ ctx = cancel_cmdid(nvmeq, cmdid, &fn);
+ fn(nvmeq->dev, ctx, &cqe);
+ }
+}
+
+static void nvme_free_queue_mem(struct nvme_queue *nvmeq)
+{
+ 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);
+ kfree(nvmeq);
+}
+
static void nvme_free_queue(struct nvme_dev *dev, int qid)
{
struct nvme_queue *nvmeq = dev->queues[qid];
int vector = dev->entry[nvmeq->cq_vector].vector;
+ spin_lock_irq(&nvmeq->q_lock);
+ nvme_cancel_ios(nvmeq, false);
+ spin_unlock_irq(&nvmeq->q_lock);
+
irq_set_affinity_hint(vector, NULL);
free_irq(vector, nvmeq);
adapter_delete_cq(dev, qid);
}
- 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);
- kfree(nvmeq);
+ nvme_free_queue_mem(nvmeq);
}
static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid,
int depth, int vector)
{
struct device *dmadev = &dev->pci_dev->dev;
- unsigned extra = (depth / 8) + (depth * sizeof(struct nvme_cmd_info));
+ unsigned extra = DIV_ROUND_UP(depth, 8) + (depth *
+ sizeof(struct nvme_cmd_info));
struct nvme_queue *nvmeq = kzalloc(sizeof(*nvmeq) + extra, GFP_KERNEL);
if (!nvmeq)
return NULL;
static int __devinit nvme_configure_admin_queue(struct nvme_dev *dev)
{
- int result;
+ int result = 0;
u32 aqa;
u64 cap;
unsigned long timeout;
timeout = ((NVME_CAP_TIMEOUT(cap) + 1) * HZ / 2) + jiffies;
dev->db_stride = NVME_CAP_STRIDE(cap);
- while (!(readl(&dev->bar->csts) & NVME_CSTS_RDY)) {
+ while (!result && !(readl(&dev->bar->csts) & NVME_CSTS_RDY)) {
msleep(100);
if (fatal_signal_pending(current))
- return -EINTR;
+ result = -EINTR;
if (time_after(jiffies, timeout)) {
dev_err(&dev->pci_dev->dev,
"Device not ready; aborting initialisation\n");
- return -ENODEV;
+ result = -ENODEV;
}
}
+ if (result) {
+ nvme_free_queue_mem(nvmeq);
+ return result;
+ }
+
result = queue_request_irq(dev, nvmeq, "nvme admin");
dev->queues[0] = nvmeq;
return result;
}
-static int nvme_map_user_pages(struct nvme_dev *dev, int write,
- unsigned long addr, unsigned length,
- struct scatterlist **sgp)
+static struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
+ unsigned long addr, unsigned length)
{
int i, err, count, nents, offset;
struct scatterlist *sg;
struct page **pages;
+ struct nvme_iod *iod;
if (addr & 3)
- return -EINVAL;
+ return ERR_PTR(-EINVAL);
if (!length)
- return -EINVAL;
+ return ERR_PTR(-EINVAL);
offset = offset_in_page(addr);
count = DIV_ROUND_UP(offset + length, PAGE_SIZE);
pages = kcalloc(count, sizeof(*pages), GFP_KERNEL);
+ if (!pages)
+ return ERR_PTR(-ENOMEM);
err = get_user_pages_fast(addr, count, 1, pages);
if (err < count) {
goto put_pages;
}
- sg = kcalloc(count, sizeof(*sg), GFP_KERNEL);
+ iod = nvme_alloc_iod(count, length, GFP_KERNEL);
+ sg = iod->sg;
sg_init_table(sg, count);
for (i = 0; i < count; i++) {
sg_set_page(&sg[i], pages[i],
length -= (PAGE_SIZE - offset);
offset = 0;
}
+ sg_mark_end(&sg[i - 1]);
+ iod->nents = count;
err = -ENOMEM;
nents = dma_map_sg(&dev->pci_dev->dev, sg, count,
write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
if (!nents)
- goto put_pages;
+ goto free_iod;
kfree(pages);
- *sgp = sg;
- return nents;
+ return iod;
+ free_iod:
+ kfree(iod);
put_pages:
for (i = 0; i < count; i++)
put_page(pages[i]);
kfree(pages);
- return err;
+ return ERR_PTR(err);
}
static void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
- unsigned long addr, int length, struct scatterlist *sg)
+ struct nvme_iod *iod)
{
- int i, count;
+ int i;
- count = DIV_ROUND_UP(offset_in_page(addr) + length, PAGE_SIZE);
- dma_unmap_sg(&dev->pci_dev->dev, sg, count, DMA_FROM_DEVICE);
+ dma_unmap_sg(&dev->pci_dev->dev, iod->sg, iod->nents,
+ write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
- for (i = 0; i < count; i++)
- put_page(sg_page(&sg[i]));
+ for (i = 0; i < iod->nents; i++)
+ put_page(sg_page(&iod->sg[i]));
}
static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
struct nvme_user_io io;
struct nvme_command c;
unsigned length;
- int nents, status;
- struct scatterlist *sg;
- struct nvme_prps *prps;
+ int status;
+ struct nvme_iod *iod;
if (copy_from_user(&io, uio, sizeof(io)))
return -EFAULT;
case nvme_cmd_write:
case nvme_cmd_read:
case nvme_cmd_compare:
- nents = nvme_map_user_pages(dev, io.opcode & 1, io.addr,
- length, &sg);
+ iod = nvme_map_user_pages(dev, io.opcode & 1, io.addr, length);
break;
default:
return -EINVAL;
}
- if (nents < 0)
- return nents;
+ if (IS_ERR(iod))
+ return PTR_ERR(iod);
memset(&c, 0, sizeof(c));
c.rw.opcode = io.opcode;
c.rw.apptag = io.apptag;
c.rw.appmask = io.appmask;
/* XXX: metadata */
- prps = nvme_setup_prps(dev, &c.common, sg, &length, GFP_KERNEL);
+ length = nvme_setup_prps(dev, &c.common, iod, length, GFP_KERNEL);
- nvmeq = get_nvmeq(ns);
+ nvmeq = get_nvmeq(dev);
/*
* Since nvme_submit_sync_cmd sleeps, we can't keep preemption
* disabled. We may be preempted at any point, and be rescheduled
if (length != (io.nblocks + 1) << ns->lba_shift)
status = -ENOMEM;
else
- status = nvme_submit_sync_cmd(nvmeq, &c, NULL, IO_TIMEOUT);
+ status = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
- nvme_unmap_user_pages(dev, io.opcode & 1, io.addr, length, sg);
- nvme_free_prps(dev, prps);
+ nvme_unmap_user_pages(dev, io.opcode & 1, iod);
+ nvme_free_iod(dev, iod);
return status;
}
-static int nvme_user_admin_cmd(struct nvme_ns *ns,
+static int nvme_user_admin_cmd(struct nvme_dev *dev,
struct nvme_admin_cmd __user *ucmd)
{
- struct nvme_dev *dev = ns->dev;
struct nvme_admin_cmd cmd;
struct nvme_command c;
- int status, length, nents = 0;
- struct scatterlist *sg;
- struct nvme_prps *prps = NULL;
+ int status, length;
+ struct nvme_iod *uninitialized_var(iod);
if (!capable(CAP_SYS_ADMIN))
return -EACCES;
length = cmd.data_len;
if (cmd.data_len) {
- nents = nvme_map_user_pages(dev, 1, cmd.addr, length, &sg);
- if (nents < 0)
- return nents;
- prps = nvme_setup_prps(dev, &c.common, sg, &length, GFP_KERNEL);
+ iod = nvme_map_user_pages(dev, cmd.opcode & 1, cmd.addr,
+ length);
+ if (IS_ERR(iod))
+ return PTR_ERR(iod);
+ length = nvme_setup_prps(dev, &c.common, iod, length,
+ GFP_KERNEL);
}
if (length != cmd.data_len)
status = -ENOMEM;
else
status = nvme_submit_admin_cmd(dev, &c, NULL);
+
if (cmd.data_len) {
- nvme_unmap_user_pages(dev, 0, cmd.addr, cmd.data_len, sg);
- nvme_free_prps(dev, prps);
+ nvme_unmap_user_pages(dev, cmd.opcode & 1, iod);
+ nvme_free_iod(dev, iod);
}
return status;
}
case NVME_IOCTL_ID:
return ns->ns_id;
case NVME_IOCTL_ADMIN_CMD:
- return nvme_user_admin_cmd(ns, (void __user *)arg);
+ return nvme_user_admin_cmd(ns->dev, (void __user *)arg);
case NVME_IOCTL_SUBMIT_IO:
return nvme_submit_io(ns, (void __user *)arg);
default:
.compat_ioctl = nvme_ioctl,
};
-static void nvme_timeout_ios(struct nvme_queue *nvmeq)
-{
- 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 = { .status = cpu_to_le16(NVME_SC_ABORT_REQ) << 1, };
-
- if (!time_after(now, info[cmdid].timeout))
- continue;
- dev_warn(nvmeq->q_dmadev, "Timing out I/O %d\n", cmdid);
- ctx = cancel_cmdid(nvmeq, cmdid, &fn);
- fn(nvmeq, ctx, &cqe);
- }
-}
-
static void nvme_resubmit_bios(struct nvme_queue *nvmeq)
{
while (bio_list_peek(&nvmeq->sq_cong)) {
spin_lock_irq(&nvmeq->q_lock);
if (nvme_process_cq(nvmeq))
printk("process_cq did something\n");
- nvme_timeout_ios(nvmeq);
+ nvme_cancel_ios(nvmeq, true);
nvme_resubmit_bios(nvmeq);
spin_unlock_irq(&nvmeq->q_lock);
}
ns->queue = blk_alloc_queue(GFP_KERNEL);
if (!ns->queue)
goto out_free_ns;
- ns->queue->queue_flags = QUEUE_FLAG_DEFAULT | QUEUE_FLAG_NOMERGES |
- QUEUE_FLAG_NONROT | QUEUE_FLAG_DISCARD;
+ 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);
+/* queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue); */
blk_queue_make_request(ns->queue, nvme_make_request);
ns->dev = dev;
ns->queue->queuedata = ns;
ns->disk = disk;
lbaf = id->flbas & 0xf;
ns->lba_shift = id->lbaf[lbaf].ds;
+ blk_queue_logical_block_size(ns->queue, 1 << ns->lba_shift);
+ if (dev->max_hw_sectors)
+ blk_queue_max_hw_sectors(ns->queue, dev->max_hw_sectors);
disk->major = nvme_major;
disk->minors = NVME_MINORS;
u32 result;
u32 q_count = (count - 1) | ((count - 1) << 16);
- status = nvme_get_features(dev, NVME_FEAT_NUM_QUEUES, q_count, 0,
+ status = nvme_set_features(dev, NVME_FEAT_NUM_QUEUES, q_count, 0,
&result);
if (status)
return -EIO;
static int __devinit nvme_setup_io_queues(struct nvme_dev *dev)
{
- int result, cpu, i, nr_io_queues, db_bar_size;
+ int result, cpu, i, nr_io_queues, db_bar_size, q_depth;
nr_io_queues = num_online_cpus();
result = set_queue_count(dev, nr_io_queues);
cpu = cpumask_next(cpu, cpu_online_mask);
}
+ q_depth = min_t(int, NVME_CAP_MQES(readq(&dev->bar->cap)) + 1,
+ NVME_Q_DEPTH);
for (i = 0; i < nr_io_queues; i++) {
- dev->queues[i + 1] = nvme_create_queue(dev, i + 1,
- NVME_Q_DEPTH, i);
+ dev->queues[i + 1] = nvme_create_queue(dev, i + 1, q_depth, i);
if (IS_ERR(dev->queues[i + 1]))
return PTR_ERR(dev->queues[i + 1]);
dev->queue_count++;
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));
+ if (ctrl->mdts) {
+ int shift = NVME_CAP_MPSMIN(readq(&dev->bar->cap)) + 12;
+ dev->max_hw_sectors = 1 << (ctrl->mdts + shift - 9);
+ }
id_ns = mem;
for (i = 1; i <= nn; i++) {
continue;
res = nvme_get_features(dev, NVME_FEAT_LBA_RANGE, i,
- dma_addr + 4096, NULL);
+ dma_addr + 4096);
if (res)
continue;
list_del(&dev->node);
spin_unlock(&dev_list_lock);
- /* TODO: wait all I/O finished or cancel them */
-
list_for_each_entry_safe(ns, next, &dev->namespaces, list) {
list_del(&ns->list);
del_gendisk(ns->disk);
dma_pool_destroy(dev->prp_small_pool);
}
-/* XXX: Use an ida or something to let remove / add work correctly */
-static void nvme_set_instance(struct nvme_dev *dev)
+static DEFINE_IDA(nvme_instance_ida);
+
+static int nvme_set_instance(struct nvme_dev *dev)
{
- static int instance;
- dev->instance = instance++;
+ 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;
}
static void nvme_release_instance(struct nvme_dev *dev)
{
+ spin_lock(&dev_list_lock);
+ ida_remove(&nvme_instance_ida, dev->instance);
+ spin_unlock(&dev_list_lock);
}
static int __devinit nvme_probe(struct pci_dev *pdev,
pci_set_drvdata(pdev, dev);
dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
- nvme_set_instance(dev);
+ result = nvme_set_instance(dev);
+ if (result)
+ goto disable;
+
dev->entry[0].vector = pdev->irq;
result = nvme_setup_prp_pools(dev);
static int __init nvme_init(void)
{
- int result = -EBUSY;
+ int result;
nvme_thread = kthread_run(nvme_kthread, NULL, "nvme");
if (IS_ERR(nvme_thread))
return PTR_ERR(nvme_thread);
- nvme_major = register_blkdev(nvme_major, "nvme");
- if (nvme_major <= 0)
+ result = register_blkdev(nvme_major, "nvme");
+ if (result < 0)
goto kill_kthread;
+ else if (result > 0)
+ nvme_major = result;
result = pci_register_driver(&nvme_driver);
if (result)
MODULE_AUTHOR("Matthew Wilcox <willy@linux.intel.com>");
MODULE_LICENSE("GPL");
-MODULE_VERSION("0.7");
+MODULE_VERSION("0.8");
module_init(nvme_init);
module_exit(nvme_exit);