Regression in 6.11-rc1 in scsi/sd?

[Coelho, Luciano]

Hi,

We're getting some lockdep splats with 6.11-rc1 in some of our older CI
machines:

<4>[   25.357106] ======================================================
<4>[   25.358383] WARNING: possible circular locking dependency detected
<4>[   25.359636] 6.11.0-rc1-CI_DRM_15151-gb6f9528c7fff+ #1 Not tainted
<4>[   25.360902] ------------------------------------------------------
<4>[   25.362184] rc.local/609 is trying to acquire lock:
<4>[   25.363450] ffff888102358670 (&q->limits_lock){+.+.}-{3:3}, at: queue_max_discard_sectors_store+0x8e/0x110
<4>[   25.364798] 
                  but task is already holding lock:
<4>[   25.367410] ffff888102358550 (&q->sysfs_lock){+.+.}-{3:3}, at: queue_attr_store+0x45/0x90
<4>[   25.368773] 
                  which lock already depends on the new lock.

...during device probe.  You can find the full dmesg, for example, here:

https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_15151/fi-bsw-n3050/boot0.txt

The stackdumps seem to point to sd_probe() and sd_revalidate_disk().

Is this an known issue? Does anyone have any idea what is causing it?

--
Cheers,
Luca.