W dniu 2016-09-04 02:17, Kent Overstreet napisał(a):
Hi!
On Sat, Sep 03, 2016 at 11:29:49PM +0200, Marcin wrote:
Hi!
Kernel at commit c820493652e830dc050e1418301e1bdec5691a1e
I createt to devices, fast has size
# blockdev --getsz /dev/sde1
20971520
and slower device:
# blockdev --getsz /dev/sdd1
2930209551
I was copying files from one disk to bcache, after some time I got:
BUG: soft lockup - CPU#0 stuck for 22s! [bch_copygc_read:5328]
Thanks for the report - can you run addr2line with your vmlinux file,
and the
RIP?
addr2line -i -e vmlinux ffffffffc028795b
It returned:
??:0
Probably due to I'm using bcache as module.
<long story>
As I mentioned before I wasn't sure which branch I used to test. So I
started to reproduce problem again and watch which branch is used. I
encountered strange behaviour, random BUGs in kernel, even random
reboots. It looked like a hardware problem but I never got such problems
using this box as backup & samba server. Memtest86+ didn't found any
problem, cpuburn didn't trigger anything suspicious. Gcc compiles
without any strange failures. Nothing suspicious happened when I didn't
use bcachefs. So i spent a couple of hours trying to find out what is
going on. I tried bcachefs without tiering. When I formated /dev/sde1
then everything was ok. But when I tried to use /dev/sdd1 something
happened:
# bcache format /dev/sdd1
/dev/sdd1 contains a bcache filesystem
Proceed anyway? (y,n) y
UUID: 78a90acf-0247-4edf-a56f-f4f82a51e13e
Set UUID: a0a6e9fa-d7bb-4cec-8c42-b7e4bbccb2d8
version: 6
nbuckets: 3815462
block_size: 1
bucket_size: 768
nr_in_set: 1
nr_this_dev: 0
first_bucket: 3
# LC_ALL=en_US.utf-8 mount -t bcache /dev/sdd1 /mnt/test/
mount: mount(2) failed: No such file or directory
Kernel throws:
zserwerek kernel: [ 5832.483306] bcache: bch_open_as_blockdevs()
register_cache_set err Bad bucket size
Hmmm. So I resized /dev/sdd1 from 1.4TB to 100GB. Formated again, and
mounted without problem. Hmm. I created tiered bcachefs with smaller
/dev/sdd1 and it works stable.
<conclucion>
Please look at line with "bucket size":
bucket_size: 768
If bucket size is higher than (probably) 512 then I can't mount simple
(without tiering) bcachefs filesystem. If I use such big device in
tiered bcachefs I'm expieriencing random problems with stability of box.
I think that bug in mail's subject is only random symptom of problem
when device is formated with bucket size >512.
What is going inside kernel in this case, is overwittem memory of other
processes?
Marcin
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
