linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [RFC PATCH] scsi: smartpqi: create module parameters for LUN reset
@ 2021-01-21 17:03 Enzo Matsumiya
  2021-02-23 15:54 ` Enzo Matsumiya
  0 siblings, 1 reply; 2+ messages in thread
From: Enzo Matsumiya @ 2021-01-21 17:03 UTC (permalink / raw)
  To: linux-scsi
  Cc: Enzo Matsumiya, Don Brace, James E.J. Bottomley,
	Martin K. Petersen, storagedev, linux-kernel

Commit c2922f174fa0 ("scsi: smartpqi: fix LUN reset when fw bkgnd thread is hung")
added support for a timeout on LUN resets.

However, when there are 2 or more devices connected to the same
controller and you hot-remove one of them, I/O will stall on the
devices still online for PQI_LUN_RESET_RETRIES * PQI_LUN_RESET_RETRY_INTERVAL_MSECS
miliseconds.

This commit makes those values configurable via module parameters.

Changing the bail out condition on rc in _pqi_device_reset() might be possible,
but could also break the original purpose of commit c2922f174fa0.

Signed-off-by: Enzo Matsumiya <ematsumiya@suse.de>
---
 drivers/scsi/smartpqi/smartpqi_init.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/scsi/smartpqi/smartpqi_init.c b/drivers/scsi/smartpqi/smartpqi_init.c
index c53f456fbd09..9835b2e5b91a 100644
--- a/drivers/scsi/smartpqi/smartpqi_init.c
+++ b/drivers/scsi/smartpqi/smartpqi_init.c
@@ -157,6 +157,18 @@ module_param_named(hide_vsep,
 MODULE_PARM_DESC(hide_vsep,
 	"Hide the virtual SEP for direct attached drives.");
 
+static int pqi_lun_reset_retries = 3;
+module_param_named(lun_reset_retries,
+	pqi_lun_reset_retries, int, 0644);
+MODULE_PARM_DESC(lun_reset_retries,
+	"Number of retries when resetting a LUN");
+
+static int pqi_lun_reset_tmo_interval = 10000;
+module_param_named(lun_reset_tmo_interval,
+	pqi_lun_reset_tmo_interval, int, 0644);
+MODULE_PARM_DESC(lun_reset_tmo_interval,
+	"LUN reset timeout interval (in miliseconds)");
+
 static char *raid_levels[] = {
 	"RAID-0",
 	"RAID-4",
@@ -5687,8 +5699,6 @@ static int pqi_lun_reset(struct pqi_ctrl_info *ctrl_info,
 
 /* Performs a reset at the LUN level. */
 
-#define PQI_LUN_RESET_RETRIES			3
-#define PQI_LUN_RESET_RETRY_INTERVAL_MSECS	10000
 #define PQI_LUN_RESET_PENDING_IO_TIMEOUT_SECS	120
 
 static int _pqi_device_reset(struct pqi_ctrl_info *ctrl_info,
@@ -5700,9 +5710,9 @@ static int _pqi_device_reset(struct pqi_ctrl_info *ctrl_info,
 
 	for (retries = 0;;) {
 		rc = pqi_lun_reset(ctrl_info, device);
-		if (rc == 0 || ++retries > PQI_LUN_RESET_RETRIES)
+		if (rc == 0 || ++retries > pqi_lun_reset_retries)
 			break;
-		msleep(PQI_LUN_RESET_RETRY_INTERVAL_MSECS);
+		msleep(pqi_lun_reset_tmo_interval);
 	}
 
 	timeout_secs = rc ? PQI_LUN_RESET_PENDING_IO_TIMEOUT_SECS : NO_TIMEOUT;
-- 
2.30.0


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [RFC PATCH] scsi: smartpqi: create module parameters for LUN reset
  2021-01-21 17:03 [RFC PATCH] scsi: smartpqi: create module parameters for LUN reset Enzo Matsumiya
@ 2021-02-23 15:54 ` Enzo Matsumiya
  0 siblings, 0 replies; 2+ messages in thread
From: Enzo Matsumiya @ 2021-02-23 15:54 UTC (permalink / raw)
  To: linux-scsi
  Cc: Don Brace, James E.J. Bottomley, Martin K. Petersen, storagedev,
	linux-kernel

Hi,

On 01/21, Enzo Matsumiya wrote:
>Commit c2922f174fa0 ("scsi: smartpqi: fix LUN reset when fw bkgnd thread is hung")
>added support for a timeout on LUN resets.
>
>However, when there are 2 or more devices connected to the same
>controller and you hot-remove one of them, I/O will stall on the
>devices still online for PQI_LUN_RESET_RETRIES * PQI_LUN_RESET_RETRY_INTERVAL_MSECS
>miliseconds.
>
>This commit makes those values configurable via module parameters.
>
>Changing the bail out condition on rc in _pqi_device_reset() might be possible,
>but could also break the original purpose of commit c2922f174fa0.
>
>Signed-off-by: Enzo Matsumiya <ematsumiya@suse.de>
>---
> drivers/scsi/smartpqi/smartpqi_init.c | 18 ++++++++++++++----
> 1 file changed, 14 insertions(+), 4 deletions(-)
>
>diff --git a/drivers/scsi/smartpqi/smartpqi_init.c b/drivers/scsi/smartpqi/smartpqi_init.c
>index c53f456fbd09..9835b2e5b91a 100644
>--- a/drivers/scsi/smartpqi/smartpqi_init.c
>+++ b/drivers/scsi/smartpqi/smartpqi_init.c
>@@ -157,6 +157,18 @@ module_param_named(hide_vsep,
> MODULE_PARM_DESC(hide_vsep,
> 	"Hide the virtual SEP for direct attached drives.");
>
>+static int pqi_lun_reset_retries = 3;
>+module_param_named(lun_reset_retries,
>+	pqi_lun_reset_retries, int, 0644);
>+MODULE_PARM_DESC(lun_reset_retries,
>+	"Number of retries when resetting a LUN");
>+
>+static int pqi_lun_reset_tmo_interval = 10000;
>+module_param_named(lun_reset_tmo_interval,
>+	pqi_lun_reset_tmo_interval, int, 0644);
>+MODULE_PARM_DESC(lun_reset_tmo_interval,
>+	"LUN reset timeout interval (in miliseconds)");
>+
> static char *raid_levels[] = {
> 	"RAID-0",
> 	"RAID-4",
>@@ -5687,8 +5699,6 @@ static int pqi_lun_reset(struct pqi_ctrl_info *ctrl_info,
>
> /* Performs a reset at the LUN level. */
>
>-#define PQI_LUN_RESET_RETRIES			3
>-#define PQI_LUN_RESET_RETRY_INTERVAL_MSECS	10000
> #define PQI_LUN_RESET_PENDING_IO_TIMEOUT_SECS	120
>
> static int _pqi_device_reset(struct pqi_ctrl_info *ctrl_info,
>@@ -5700,9 +5710,9 @@ static int _pqi_device_reset(struct pqi_ctrl_info *ctrl_info,
>
> 	for (retries = 0;;) {
> 		rc = pqi_lun_reset(ctrl_info, device);
>-		if (rc == 0 || ++retries > PQI_LUN_RESET_RETRIES)
>+		if (rc == 0 || ++retries > pqi_lun_reset_retries)
> 			break;
>-		msleep(PQI_LUN_RESET_RETRY_INTERVAL_MSECS);
>+		msleep(pqi_lun_reset_tmo_interval);
> 	}
>
> 	timeout_secs = rc ? PQI_LUN_RESET_PENDING_IO_TIMEOUT_SECS : NO_TIMEOUT;
>-- 
>2.30.0
>

Can anyone give me some feedback on this please?


Cheers,

Enzo

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-02-23 15:55 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-21 17:03 [RFC PATCH] scsi: smartpqi: create module parameters for LUN reset Enzo Matsumiya
2021-02-23 15:54 ` Enzo Matsumiya

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).