scsi: cxlflash: Fence EEH during probe

An EEH during probe can lead to a crash as the recovery thread races with the
probe thread. To avoid this issue, introduce new states to fence out EEH
recovery until probe has completed. Also ensure the reset wait queue is
flushed during device removal to avoid orphaned threads.

Signed-off-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Signed-off-by: Uma Krishnan <ukrishn@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

1 files changed, 21 insertions, 4 deletions

diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c
index 568cd63..ebba3c9 100644
--- a/drivers/scsi/cxlflash/main.c
+++ b/drivers/scsi/cxlflash/main.c
@@ -470,6 +470,8 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp)
 	spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags);
 
 	switch (cfg->state) {
+	case STATE_PROBING:
+	case STATE_PROBED:
 	case STATE_RESET:
 		dev_dbg_ratelimited(dev, "%s: device is in reset\n", __func__);
 		rc = SCSI_MLQUEUE_HOST_BUSY;
@@ -719,7 +721,8 @@ static void notify_shutdown(struct cxlflash_cfg *cfg, bool wait)
  * cxlflash_remove() - PCI entry point to tear down host
  * @pdev:	PCI device associated with the host.
  *
- * Safe to use as a cleanup in partially allocated/initialized state.
+ * Safe to use as a cleanup in partially allocated/initialized state. Note that
+ * the reset_waitq is flushed as part of the stop/termination of user contexts.
  */
 static void cxlflash_remove(struct pci_dev *pdev)
 {
@@ -752,7 +755,6 @@ static void cxlflash_remove(struct pci_dev *pdev)
 	case INIT_STATE_SCSI:
 		cxlflash_term_local_luns(cfg);
 		scsi_remove_host(cfg->host);
-		/* fall through */
 	case INIT_STATE_AFU:
 		term_afu(cfg);
 	case INIT_STATE_PCI:
@@ -2624,6 +2626,15 @@ static void cxlflash_worker_thread(struct work_struct *work)
  * @pdev:	PCI device associated with the host.
  * @dev_id:	PCI device id associated with device.
  *
+ * The device will initially start out in a 'probing' state and
+ * transition to the 'normal' state at the end of a successful
+ * probe. Should an EEH event occur during probe, the notification
+ * thread (error_detected()) will wait until the probe handler
+ * is nearly complete. At that time, the device will be moved to
+ * a 'probed' state and the EEH thread woken up to drive the slot
+ * reset and recovery (device moves to 'normal' state). Meanwhile,
+ * the probe will be allowed to exit successfully.
+ *
  * Return: 0 on success, -errno on failure
  */
 static int cxlflash_probe(struct pci_dev *pdev,
@@ -2707,7 +2718,7 @@ static int cxlflash_probe(struct pci_dev *pdev,
 	cfg->init_state = INIT_STATE_PCI;
 
 	rc = init_afu(cfg);
-	if (rc) {
+	if (rc && !wq_has_sleeper(&cfg->reset_waitq)) {
 		dev_err(dev, "%s: init_afu failed rc=%d\n", __func__, rc);
 		goto out_remove;
 	}
@@ -2720,6 +2731,11 @@ static int cxlflash_probe(struct pci_dev *pdev,
 	}
 	cfg->init_state = INIT_STATE_SCSI;
 
+	if (wq_has_sleeper(&cfg->reset_waitq)) {
+		cfg->state = STATE_PROBED;
+		wake_up_all(&cfg->reset_waitq);
+	} else
+		cfg->state = STATE_NORMAL;
 out:
 	dev_dbg(dev, "%s: returning rc=%d\n", __func__, rc);
 	return rc;
@@ -2750,7 +2766,8 @@ static pci_ers_result_t cxlflash_pci_error_detected(struct pci_dev *pdev,
 
 	switch (state) {
 	case pci_channel_io_frozen:
-		wait_event(cfg->reset_waitq, cfg->state != STATE_RESET);
+		wait_event(cfg->reset_waitq, cfg->state != STATE_RESET &&
+					     cfg->state != STATE_PROBING);
 		if (cfg->state == STATE_FAILTERM)
 			return PCI_ERS_RESULT_DISCONNECT;