unit-wipe-test: Wait a bit so scsi_debug has a chance to fully initialize.

On my test system `tests/unit-wipe-test` fails (as root) due to a race
condition in add_device():

	root@host:~# ./unit-wipe-test
	[1] Wipe full file [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[2] Wipe blocks in file [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[  105.828258] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
	[  105.830450] scsi host2: scsi_debug: version 0191 [20210520]
	[  105.830450]   dev_size_mb=8, opts=0x0, submit_queues=1, statistics=0
	[  105.832924] scsi 2:0:0:0: Direct-Access     Linux    scsi_debug       0191 PQ: 0 ANSI: 7
	[  105.835417] scsi 2:0:0:0: Attached scsi generic sg1 type 0
	FAIL Cannot find /dev/.
	FAILED backtrace:
	48 ./unit-wipe-test
	144 main ./unit-wipe-test
	[  105.875131] sd 2:0:0:0: Power-on or device reset occurred
	[  105.876069] sd 2:0:0:0: [sda] 16384 512-byte logical blocks: (8.39 MB/8.00 MiB)
	[  105.877190] sd 2:0:0:0: [sda] Write Protect is off
	[  105.878002] sd 2:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
	[  105.879619] sd 2:0:0:0: [sda] Optimal transfer size 524288 bytes
	[  105.914222] sd 2:0:0:0: [sda] Attached SCSI disk
	[  106.866296] sd 2:0:0:0: [sda] Synchronizing SCSI cache

Observe how the “Power-on or device reset occurred” event occurs only
after add_device() has returned.  Interestingly, for subsequent runs the
delay appears to be much shorter and doesn't trigger the race condition:

	root@host:~# ./unit-wipe-test
	[1] Wipe full file [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[2] Wipe blocks in file [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[  130.639855] scsi_debug:sdebug_driver_probe: scsi_debug: trim poll_queues to 0. poll_q/nr_hw = (0/1)
	[  130.641463] scsi host2: scsi_debug: version 0191 [20210520]
	[  130.641463]   dev_size_mb=8, opts=0x0, submit_queues=1, statistics=0
	[  130.643809] scsi 2:0:0:0: Direct-Access     Linux    scsi_debug       0191 PQ: 0 ANSI: 7
	[  130.645342] sd 2:0:0:0: Power-on or device reset occurred
	[  130.646364] sd 2:0:0:0: [sda] 16384 512-byte logical blocks: (8.39 MB/8.00 MiB)
	[  130.647585] sd 2:0:0:0: [sda] Write Protect is off
	[  130.648428] sd 2:0:0:0: Attached scsi generic sg1 type 0
	[  130.649339] sd 2:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
	[  130.650763] sd 2:0:0:0: [sda] Optimal transfer size 524288 bytes
	[  130.682223] sd 2:0:0:0: [sda] Attached SCSI disk
	[3] Wipe full block device [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[4] Wipe blocks in block device [0/DIO][0][1048576/DIO][1048576][4194304/DIO][4194304][OK]
	[  137.858283] sd 2:0:0:0: [sda] Synchronizing SCSI cache

This commit adds an optional 2s delay if scsi_debug hasn't shown up in
sysfs after the modprobe call.
This commit is contained in:
Guilhem Moulin
2022-07-15 18:10:20 +02:00
committed by Milan Broz
parent 3106b4e2c1
commit a76c96d361

View File

@@ -43,6 +43,7 @@ function add_device()
echo "This kernel seems to not support proper scsi_debug module, test skipped." echo "This kernel seems to not support proper scsi_debug module, test skipped."
exit 77 exit 77
fi fi
grep -q scsi_debug /sys/block/*/device/model || sleep 2
DEV=$(grep -l -e scsi_debug /sys/block/*/device/model | cut -f4 -d /) DEV=$(grep -l -e scsi_debug /sys/block/*/device/model | cut -f4 -d /)
DEV="/dev/$DEV" DEV="/dev/$DEV"
[ -b $DEV ] || fail "Cannot find $DEV." [ -b $DEV ] || fail "Cannot find $DEV."