The SCSI ioctl reset path is smart enough to set the
flag tmf_in_progress when a user-requested reset is
processed, but it does not wait for IO that is in
flight. This can result in lost IOs and hung
processes. We should wait for a reasonable amount
of time for either the IOs to complete or to fail
the request.
Signed-off-by: Lee Duncan <ldun...@suse.com>
---
drivers/scsi/scsi_error.c | 26 ++++++++++++++++++++++++++
1 file changed, 26 insertions(+)
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index 38942050b265..b964152611c3 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -57,6 +57,14 @@
#define BUS_RESET_SETTLE_TIME (10)
#define HOST_RESET_SETTLE_TIME (10)
+/*
+ * Time to wait for outstanding IOs when about to send
+ * a device reset, e.g. sg_reset. The msecs to wait must
+ * be an multiple of the msecs to wait per try.
+ */
+#define MSECS_PER_TRY_FOR_IO_ON_RESET 500
+#define MSECS_TO_WAIT_FOR_IO_ON_RESET (MSECS_PER_TRY_FOR_IO_ON_RESET * 10)
+
static int scsi_eh_try_stu(struct scsi_cmnd *scmd);
static int scsi_try_to_abort_cmd(struct scsi_host_template *,
struct scsi_cmnd *);
@@ -2269,6 +2277,7 @@ void scsi_report_device_reset(struct Scsi_Host *shost,
int channel, int target)
struct request *rq;
unsigned long flags;
int error = 0, rtn, val;
+ unsigned int msecs_to_wait = MSECS_TO_WAIT_FOR_IO_ON_RESET;
if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
return -EACCES;
@@ -2301,6 +2310,22 @@ void scsi_report_device_reset(struct Scsi_Host *shost,
int channel, int target)
spin_lock_irqsave(shost->host_lock, flags);
shost->tmf_in_progress = 1;
+
+ /* if any IOs in progress wait for them a while */
+ while ((atomic_read(&shost->host_busy) > 0) && (msecs_to_wait > 0)) {
+ spin_unlock_irqrestore(shost->host_lock, flags);
+ msleep(MSECS_PER_TRY_FOR_IO_ON_RESET);
+ msecs_to_wait -= MSECS_PER_TRY_FOR_IO_ON_RESET;
+ spin_lock_irqsave(shost->host_lock, flags);
+ }
+ if (atomic_read(&shost->host_busy)) {
+ shost->tmf_in_progress = 0;
+ spin_unlock_irqrestore(shost->host_lock, flags);
+ SCSI_LOG_ERROR_RECOVERY(3,
+ printk("%s: device reset failed: outstanding IO\n",
__func__));
+ goto out_put_scmd_and_free;
+ }
+
spin_unlock_irqrestore(shost->host_lock, flags);
switch (val & ~SG_SCSI_RESET_NO_ESCALATE) {
@@ -2349,6 +2374,7 @@ void scsi_report_device_reset(struct Scsi_Host *shost,
int channel, int target)
wake_up(&shost->host_wait);
scsi_run_host_queues(shost);
+out_put_scmd_and_free:
scsi_put_command(scmd);
kfree(rq);
--
1.8.5.6
