NVMe: IOCTL path RCU protect queue access

This adds rcu protected access to a queue in the nvme IOCTL path
to fix potential races between a surprise removal and queue usage in
nvme_submit_sync_cmd. The fix holds the rcu_read_lock() here to prevent
the nvme_queue from freeing while this path is executing so it can't
sleep, and so this path will no longer wait for a available command
id should they all be in use at the time a passthrough IOCTL request
is received.
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>

drivers/block/nvme-core.c

@@ -268,18 +268,30 @@ static struct nvme_queue *raw_nvmeq(struct nvme_dev *dev, int qid)
 	return rcu_dereference_raw(dev->queues[qid]);
 }
 
-struct nvme_queue *get_nvmeq(struct nvme_dev *dev) __acquires(RCU)
+static struct nvme_queue *get_nvmeq(struct nvme_dev *dev) __acquires(RCU)
 {
 	rcu_read_lock();
 	return rcu_dereference(dev->queues[get_cpu() + 1]);
 }
 
-void put_nvmeq(struct nvme_queue *nvmeq) __releases(RCU)
+static void put_nvmeq(struct nvme_queue *nvmeq) __releases(RCU)
 {
 	put_cpu();
 	rcu_read_unlock();
 }
 
+static struct nvme_queue *lock_nvmeq(struct nvme_dev *dev, int q_idx)
+							__acquires(RCU)
+{
+	rcu_read_lock();
+	return rcu_dereference(dev->queues[q_idx]);
+}
+
+static void unlock_nvmeq(struct nvme_queue *nvmeq) __releases(RCU)
+{
+	rcu_read_unlock();
+}
+
 /**
  * nvme_submit_cmd() - Copy a command into a queue and ring the doorbell
  * @nvmeq: The queue to use
@@ -292,6 +304,10 @@ 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);
+	if (nvmeq->q_suspended) {
+		spin_unlock_irqrestore(&nvmeq->q_lock, flags);
+		return -EBUSY;
+	}
 	tail = nvmeq->sq_tail;
 	memcpy(&nvmeq->sq_cmds[tail], cmd, sizeof(*cmd));
 	if (++tail == nvmeq->q_depth)
@@ -812,27 +828,46 @@ static void sync_completion(struct nvme_dev *dev, void *ctx,
  * 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
  */
-int nvme_submit_sync_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd,
+static int nvme_submit_sync_cmd(struct nvme_dev *dev, int q_idx,
+						struct nvme_command *cmd,
 						u32 *result, unsigned timeout)
 {
-	int cmdid;
+	int cmdid, ret;
 	struct sync_cmd_info cmdinfo;
+	struct nvme_queue *nvmeq;
+
+	nvmeq = lock_nvmeq(dev, q_idx);
+	if (!nvmeq) {
+		unlock_nvmeq(nvmeq);
+		return -ENODEV;
+	}
 
 	cmdinfo.task = current;
 	cmdinfo.status = -EINTR;
 
-	cmdid = alloc_cmdid_killable(nvmeq, &cmdinfo, sync_completion,
-								timeout);
-	if (cmdid < 0)
+	cmdid = alloc_cmdid(nvmeq, &cmdinfo, sync_completion, timeout);
+	if (cmdid < 0) {
+		unlock_nvmeq(nvmeq);
 		return cmdid;
+	}
 	cmd->common.command_id = cmdid;
 
 	set_current_state(TASK_KILLABLE);
-	nvme_submit_cmd(nvmeq, cmd);
+	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);
 	schedule_timeout(timeout);
 
 	if (cmdinfo.status == -EINTR) {
+		nvmeq = lock_nvmeq(dev, q_idx);
+		if (nvmeq)
 			nvme_abort_command(nvmeq, cmdid);
+		unlock_nvmeq(nvmeq);
 		return -EINTR;
 	}
 
@@ -853,15 +888,20 @@ static int nvme_submit_async_cmd(struct nvme_queue *nvmeq,
 		return cmdid;
 	cmdinfo->status = -EINTR;
 	cmd->common.command_id = cmdid;
-	nvme_submit_cmd(nvmeq, cmd);
-	return 0;
+	return nvme_submit_cmd(nvmeq, cmd);
 }
 
 int nvme_submit_admin_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
 								u32 *result)
 {
-	return nvme_submit_sync_cmd(raw_nvmeq(dev, 0), cmd, result,
-								ADMIN_TIMEOUT);
+	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);
 }
 
 static int nvme_submit_admin_cmd_async(struct nvme_dev *dev,
@@ -1434,7 +1474,6 @@ void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
 static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 {
 	struct nvme_dev *dev = ns->dev;
-	struct nvme_queue *nvmeq;
 	struct nvme_user_io io;
 	struct nvme_command c;
 	unsigned length, meta_len;
@@ -1510,20 +1549,10 @@ static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 
 	length = nvme_setup_prps(dev, &c.common, iod, length, GFP_KERNEL);
 
-	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
-	 * to a different CPU.  That will cause cacheline bouncing, but no
-	 * additional races since q_lock already protects against other CPUs.
-	 */
-	put_nvmeq(nvmeq);
 	if (length != (io.nblocks + 1) << ns->lba_shift)
 		status = -ENOMEM;
-	else if (!nvmeq || nvmeq->q_suspended)
-		status = -EBUSY;
 	else
-		status = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
+		status = nvme_submit_io_cmd(dev, &c, NULL);
 
 	if (meta_len) {
 		if (status == NVME_SC_SUCCESS && !(io.opcode & 1)) {
@@ -1597,8 +1626,7 @@ static int nvme_user_admin_cmd(struct nvme_dev *dev,
 	if (length != cmd.data_len)
 		status = -ENOMEM;
 	else
-		status = nvme_submit_sync_cmd(raw_nvmeq(dev, 0), &c,
-							&cmd.result, timeout);
+		status = nvme_submit_sync_cmd(dev, 0, &c, &cmd.result, timeout);
 
 	if (cmd.data_len) {
 		nvme_unmap_user_pages(dev, cmd.opcode & 1, iod);

drivers/block/nvme-scsi.c

@@ -2033,7 +2033,6 @@ static int nvme_trans_do_nvme_io(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 	int res = SNTI_TRANSLATION_SUCCESS;
 	int nvme_sc;
 	struct nvme_dev *dev = ns->dev;
-	struct nvme_queue *nvmeq;
 	u32 num_cmds;
 	struct nvme_iod *iod;
 	u64 unit_len;
@@ -2106,18 +2105,7 @@ static int nvme_trans_do_nvme_io(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 
 		nvme_offset += unit_num_blocks;
 
-		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 to a different CPU.  That
-		 * will cause cacheline bouncing, but no additional
-		 * races since q_lock already protects against other
-		 * CPUs.
-		 */
-		put_nvmeq(nvmeq);
-		nvme_sc = nvme_submit_sync_cmd(nvmeq, &c, NULL,
-						NVME_IO_TIMEOUT);
+		nvme_sc = nvme_submit_io_cmd(dev, &c, NULL);
 		if (nvme_sc != NVME_SC_SUCCESS) {
 			nvme_unmap_user_pages(dev,
 				(is_write) ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
@@ -2644,7 +2632,6 @@ static int nvme_trans_start_stop(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 {
 	int res = SNTI_TRANSLATION_SUCCESS;
 	int nvme_sc;
-	struct nvme_queue *nvmeq;
 	struct nvme_command c;
 	u8 immed, pcmod, pc, no_flush, start;
 
@@ -2671,10 +2658,7 @@ static int nvme_trans_start_stop(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 			c.common.opcode = nvme_cmd_flush;
 			c.common.nsid = cpu_to_le32(ns->ns_id);
 
-			nvmeq = get_nvmeq(ns->dev);
-			put_nvmeq(nvmeq);
-			nvme_sc = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
-
+			nvme_sc = nvme_submit_io_cmd(ns->dev, &c, NULL);
 			res = nvme_trans_status_code(hdr, nvme_sc);
 			if (res)
 				goto out;
@@ -2697,15 +2681,12 @@ static int nvme_trans_synchronize_cache(struct nvme_ns *ns,
 	int res = SNTI_TRANSLATION_SUCCESS;
 	int nvme_sc;
 	struct nvme_command c;
-	struct nvme_queue *nvmeq;
 
 	memset(&c, 0, sizeof(c));
 	c.common.opcode = nvme_cmd_flush;
 	c.common.nsid = cpu_to_le32(ns->ns_id);
 
-	nvmeq = get_nvmeq(ns->dev);
-	put_nvmeq(nvmeq);
-	nvme_sc = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
+	nvme_sc = nvme_submit_io_cmd(ns->dev, &c, NULL);
 
 	res = nvme_trans_status_code(hdr, nvme_sc);
 	if (res)
@@ -2872,7 +2853,6 @@ static int nvme_trans_unmap(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 	struct nvme_dev *dev = ns->dev;
 	struct scsi_unmap_parm_list *plist;
 	struct nvme_dsm_range *range;
-	struct nvme_queue *nvmeq;
 	struct nvme_command c;
 	int i, nvme_sc, res = -ENOMEM;
 	u16 ndesc, list_len;
@@ -2914,10 +2894,7 @@ static int nvme_trans_unmap(struct nvme_ns *ns, struct sg_io_hdr *hdr,
 	c.dsm.nr = cpu_to_le32(ndesc - 1);
 	c.dsm.attributes = cpu_to_le32(NVME_DSMGMT_AD);
 
-	nvmeq = get_nvmeq(dev);
-	put_nvmeq(nvmeq);
-
-	nvme_sc = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
+	nvme_sc = nvme_submit_io_cmd(dev, &c, NULL);
 	res = nvme_trans_status_code(hdr, nvme_sc);
 
 	dma_free_coherent(&dev->pci_dev->dev, ndesc * sizeof(*range),

include/linux/nvme.h

@@ -151,10 +151,7 @@ struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
 				unsigned long addr, unsigned length);
 void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
 			struct nvme_iod *iod);
-struct nvme_queue *get_nvmeq(struct nvme_dev *dev);
-void put_nvmeq(struct nvme_queue *nvmeq);
-int nvme_submit_sync_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd,
-						u32 *result, unsigned timeout);
+int nvme_submit_io_cmd(struct nvme_dev *, struct nvme_command *, u32 *);
 int nvme_submit_flush_data(struct nvme_queue *nvmeq, struct nvme_ns *ns);
 int nvme_submit_admin_cmd(struct nvme_dev *, struct nvme_command *,
 							u32 *result);