Dne 26.6.2017 v 11:08 Eric Ren napsal(a):
When the primary mirror device fails, activating a mirrored
LV will crash the kernel. It can be reproduced 100% with the
scripts below:
"""
dd if=/dev/zero of=file.raw bs=1G count=1
loopdev=$(losetup -f)
losetup $loopdev file.raw
dmsetup create pv1 --table "0 102399 linear $loopdev 0"
dmsetup create pv2 --table "0 102399 linear $loopdev 102400"
vgcreate vgtest /dev/mapper/pv1 /dev/mapper/pv2
lvcreate -l1 --type mirror -m1 -n mirror12 --mirrorlog core \
vgtest /dev/mapper/pv1 /dev/mapper/pv2
vgchange -an vgtest
echo 0 10000000 error | dmsetup load /dev/mapper/pv1
dmsetup resume /dev/mapper/pv1
vgchange -ay vgtest
" > The call trace:
"""
[ 287.008629] device-mapper: raid1: Unable to read primary mirror during
recovery
[ 287.008632] device-mapper: raid1: Primary mirror (254:10) failed while
out-of-sync: Reads may fail.
...
[ 287.012480] BUG: unable to handle kernel NULL pointer dereference at
0000000000000019
[ 287.012515] IP: [<ffffffffa00d944f>] mirror_end_io+0x7f/0x130 [dm_mirror]
...
[ 291.994645] Call Trace:
[ 291.994671] [<ffffffffa007b215>] clone_endio+0x35/0xe0 [dm_mod]
[ 291.994675] [<ffffffffa0589ced>] do_reads+0x17d/0x1d0 [dm_mirror]
[ 291.994680] [<ffffffffa058af5c>] do_mirror+0xec/0x250 [dm_mirror]
[ 291.994687] [<ffffffff810958fe>] process_one_work+0x14e/0x410
[ 291.994691] [<ffffffff81096156>] worker_thread+0x116/0x490
[ 291.994694] [<ffffffff8109b627>] kthread+0xc7/0xe0
"""
Fixes it by setting "details.bi_bdev" to NULL in error path beforing
calling into mirror_end_io(), which will fail the IO properly.
Hi
Which kernel version is this ?
I'd thought we've already fixed this BZ for old mirrors:
https://bugzilla.redhat.com/show_bug.cgi?id=1382382
There similar BZ for md-raid based mirrors (--type raid1)
https://bugzilla.redhat.com/show_bug.cgi?id=1416099
Regards
Zdenej
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel