Merge tag 'for-linus-20180413' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Followup fixes for this merge window. This contains:

   - Series from Ming, fixing corner cases in our CPU <-> queue mapping.

     This triggered repeated warnings on especially s390, but I also hit
     it in cpu hot plug/unplug testing while doing IO on NVMe on x86-64.

   - Another fix from Ming, ensuring that we always order budget and
     driver tag identically, avoiding a deadlock on QD=1 devices.

   - Loop locking regression fix from this merge window, from Omar.

   - Another loop locking fix, this time missing an unlock, from Tetsuo
     Handa.

   - Fix for racing IO submission with device removal from Bart.

   - sr reference fix from me, fixing a case where disk change or
     getevents can race with device removal.

   - Set of nvme fixes by way of Keith, from various contributors"

* tag 'for-linus-20180413' of git://git.kernel.dk/linux-block: (28 commits)
  nvme: expand nvmf_check_if_ready checks
  nvme: Use admin command effects for admin commands
  nvmet: fix space padding in serial number
  nvme: check return value of init_srcu_struct function
  nvmet: Fix nvmet_execute_write_zeroes sector count
  nvme-pci: Separate IO and admin queue IRQ vectors
  nvme-pci: Remove unused queue parameter
  nvme-pci: Skip queue deletion if there are no queues
  nvme: target: fix buffer overflow
  nvme: don't send keep-alives to the discovery controller
  nvme: unexport nvme_start_keep_alive
  nvme-loop: fix kernel oops in case of unhandled command
  nvme: enforce 64bit offset for nvme_get_log_ext fn
  sr: get/drop reference to device in revalidate and check_events
  blk-mq: Revert "blk-mq: reimplement blk_mq_hw_queue_mapped"
  blk-mq: Avoid that submitting a bio concurrently with device removal triggers a crash
  backing: silence compiler warning using __printf
  blk-mq: remove code for dealing with remapping queue
  blk-mq: reimplement blk_mq_hw_queue_mapped
  blk-mq: don't check queue mapped in __blk_mq_delay_run_hw_queue()
  ...

1 files changed, 82 insertions, 1 deletions

diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c
index 8f0f34d06d46..124c458806df 100644
--- a/drivers/nvme/host/fabrics.c
+++ b/drivers/nvme/host/fabrics.c
@@ -536,6 +536,85 @@ static struct nvmf_transport_ops *nvmf_lookup_transport(
 	return NULL;
 }
 
+blk_status_t nvmf_check_if_ready(struct nvme_ctrl *ctrl, struct request *rq,
+		bool queue_live, bool is_connected)
+{
+	struct nvme_command *cmd = nvme_req(rq)->cmd;
+
+	if (likely(ctrl->state == NVME_CTRL_LIVE && is_connected))
+		return BLK_STS_OK;
+
+	switch (ctrl->state) {
+	case NVME_CTRL_DELETING:
+		goto reject_io;
+
+	case NVME_CTRL_NEW:
+	case NVME_CTRL_CONNECTING:
+		if (!is_connected)
+			/*
+			 * This is the case of starting a new
+			 * association but connectivity was lost
+			 * before it was fully created. We need to
+			 * error the commands used to initialize the
+			 * controller so the reconnect can go into a
+			 * retry attempt. The commands should all be
+			 * marked REQ_FAILFAST_DRIVER, which will hit
+			 * the reject path below. Anything else will
+			 * be queued while the state settles.
+			 */
+			goto reject_or_queue_io;
+
+		if ((queue_live &&
+		     !(nvme_req(rq)->flags & NVME_REQ_USERCMD)) ||
+		    (!queue_live && blk_rq_is_passthrough(rq) &&
+		     cmd->common.opcode == nvme_fabrics_command &&
+		     cmd->fabrics.fctype == nvme_fabrics_type_connect))
+			/*
+			 * If queue is live, allow only commands that
+			 * are internally generated pass through. These
+			 * are commands on the admin queue to initialize
+			 * the controller. This will reject any ioctl
+			 * admin cmds received while initializing.
+			 *
+			 * If the queue is not live, allow only a
+			 * connect command. This will reject any ioctl
+			 * admin cmd as well as initialization commands
+			 * if the controller reverted the queue to non-live.
+			 */
+			return BLK_STS_OK;
+
+		/*
+		 * fall-thru to the reject_or_queue_io clause
+		 */
+		break;
+
+	/* these cases fall-thru
+	 * case NVME_CTRL_LIVE:
+	 * case NVME_CTRL_RESETTING:
+	 */
+	default:
+		break;
+	}
+
+reject_or_queue_io:
+	/*
+	 * Any other new io is something we're not in a state to send
+	 * to the device. Default action is to busy it and retry it
+	 * after the controller state is recovered. However, anything
+	 * marked for failfast or nvme multipath is immediately failed.
+	 * Note: commands used to initialize the controller will be
+	 *  marked for failfast.
+	 * Note: nvme cli/ioctl commands are marked for failfast.
+	 */
+	if (!blk_noretry_request(rq) && !(rq->cmd_flags & REQ_NVME_MPATH))
+		return BLK_STS_RESOURCE;
+
+reject_io:
+	nvme_req(rq)->status = NVME_SC_ABORT_REQ;
+	return BLK_STS_IOERR;
+}
+EXPORT_SYMBOL_GPL(nvmf_check_if_ready);
+
 static const match_table_t opt_tokens = {
 	{ NVMF_OPT_TRANSPORT,		"transport=%s"		},
 	{ NVMF_OPT_TRADDR,		"traddr=%s"		},
@@ -608,8 +687,10 @@ static int nvmf_parse_options(struct nvmf_ctrl_options *opts,
 			opts->discovery_nqn =
 				!(strcmp(opts->subsysnqn,
 					 NVME_DISC_SUBSYS_NAME));
-			if (opts->discovery_nqn)
+			if (opts->discovery_nqn) {
+				opts->kato = 0;
 				opts->nr_io_queues = 0;
+			}
 			break;
 		case NVMF_OPT_TRADDR:
 			p = match_strdup(args);