[PATCH] scsi: myrs: Put the SCSI device after logging a physical event

[PATCH] scsi: myrs: Put the SCSI device after logging a physical event

[Zhu Jun]

myrs_log_event() takes a SCSI device reference when processing a
physical-device event, but does not release it after logging the event
and updating the cached device state. Each matching event leaks a
device reference and a reference to the low-level driver module.

Put the device after processing the physical event, including when no
host-private information is available.

Fixes: 77266186397c ("scsi: myrs: Add Mylex RAID controller (SCSI interface)")
Signed-off-by: Zhu Jun <zhujun2@cmss.chinamobile.com>
---
 drivers/scsi/myrs.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/scsi/myrs.c b/drivers/scsi/myrs.c
index 1192b01fab8e..1f248501a611 100644
--- a/drivers/scsi/myrs.c
+++ b/drivers/scsi/myrs.c
@@ -877,6 +877,8 @@ static void myrs_log_event(struct myrs_hba *cs, struct myrs_event *ev)
 				break;
 			}
 		}
+		if (sdev)
+			scsi_device_put(sdev);
 		break;
 	case 'L':
 		shost_printk(KERN_INFO, shost,
-- 
2.51.0

Re: [PATCH] scsi: myrs: Put the SCSI device after logging a physical event

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found no new issues
in this patch, but noticed 1 related pre-existing issue it does not
introduce:
- [Medium] scsi: myrs: silent drop of physical device events in
  myrs_log_event() [1]
--

commit ac88587479e352e8cedb430c156919340b859e81
Author: Zhu Jun <zhujun2@cmss.chinamobile.com>

scsi: myrs: Put the SCSI device after logging a physical event

This commit calls scsi_device_put() in myrs_log_event() to release the
SCSI device reference acquired via scsi_device_lookup() when handling a
physical device event.

> diff --git a/drivers/scsi/myrs.c b/drivers/scsi/myrs.c
> index 1192b01fab8ec..1f248501a611f 100644
> --- a/drivers/scsi/myrs.c
> +++ b/drivers/scsi/myrs.c
> @@ -877,6 +877,8 @@ static void myrs_log_event(struct myrs_hba *cs, struct myrs_event *ev)
>  				break;
>  			}
>  		}
> +		if (sdev)
> +			scsi_device_put(sdev);
>  		break;
>  	case 'L':
>  		shost_printk(KERN_INFO, shost,

[Severity: Medium]
This problem wasn't introduced by this patch, but does myrs_log_event()
silently drop physical device events [1] when scsi_device_lookup()
returns NULL?

drivers/scsi/myrs.c:myrs_log_event() {
	...
	case 'P':
		sdev = scsi_device_lookup(shost, ev->channel,
					  ev->target, 0);
		sdev_printk(KERN_INFO, sdev, "event %d: Physical Device %s\n",
			    ev->ev_seq, ev_msg);
	...
}

When a physical drive is hidden by the controller, unconfigured, or
already removed, scsi_device_lookup() returns NULL. Because sdev_printk()
-> sdev_prefix_printk() returns immediately when sdev is NULL instead of
falling back to shost_printk(), drive failure and removal events are never
logged.

[1] https://sashiko.dev/#/bug/linux-5aadc823-d60c-4405-bf84-45e4ed87504d

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261112071752.154639-1-zhujun2@cmss.chinamobile.com?part=1