Obsoleted large bucket layout message once

Obsoleted large bucket layout message once

[Constantin Runge]

Hi everybody,

when registering and mounting my bcache devices (unortunately my root 
filesystem), the kernel informed me about `Detected obsoleted large 
bucket layout, all attached bcache device will be read-only'.

Subsequently, my root fs was, in fact, read-only. Thus, I have no logs 
for you unfortunately. As far as I remember, bcache-super-show showed 
the same bucket and block sizes as always (except that the backing 
device showed a clean cache state):

 > bcache-super-show /dev/sda6
sb.magic		ok
sb.first_sector		8 [match]
sb.csum			C5EB3BA8CA523D61 [match]
sb.version		4 [backing device]

dev.label		(empty)
dev.uuid		264d1acc-8461-4ddc-a825-f0acdeea3ede
dev.sectors_per_block	1
dev.sectors_per_bucket	1024
dev.data.first_sector	2048
dev.data.cache_mode	1 [writeback]
dev.data.cache_state	2 [dirty]

cset.uuid		755901be-a101-4581-b1d0-4d21d1619f27
 > bcache-super-show /dev/sdb3
sb.magic		ok
sb.first_sector		8 [match]
sb.csum			BDA92F1E72DBA095 [match]
sb.version		3 [cache device]

dev.label		(empty)
dev.uuid		25151c1a-904e-492e-b99b-6d07e13b047a
dev.sectors_per_block	1
dev.sectors_per_bucket	1024
dev.cache.first_sector	1024
dev.cache.cache_sectors	907564032
dev.cache.total_sectors	907565056
dev.cache.ordered	yes
dev.cache.discard	no
dev.cache.pos		0
dev.cache.replacement	0 [lru]

cset.uuid		755901be-a101-4581-b1d0-4d21d1619f27


For debugging purposes I tried to mount and chroot into my backing 
device directly, using a loopback interface. I was not able to find an 
offset for `losetup -f -o <offset> /dev/sda6', so that `mount -o loop 
/dev/loop0 /mnt/mnt0' would not tell me that there was no filesystem.
Also, `for i in {0.2000}; do dd if=/dev/sda6 skip=$i | file -; done' 
from [1] showed `data' for every offset.

After getting a little desperate, I booted into an older kernel 
(although I didn't update my kernel or my bcache-tools since the last 
boot) and everything worked. I then booted back into my current system 
and everything worked again.

Is this a bug with bcache or do I need to change my setup to prevent 
this from happening again? My bcache devices were created sometime 
around 2015 or 2016.

System info:
uname -a: `Linux Horus 5.10.10-gentoo #1 SMP PREEMPT Mon Feb 15 15:00:49 
CET 2021 x86_64 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz GenuineIntel 
GNU/Linux'
bcache-tools version: 1.1
The older kernel is a 5.8.11-gentoo with voluntary preempt.

Best regards and thanks,
Constantin


   [1] 
https://stackoverflow.com/questions/22820492/how-to-revert-bcache-device-to-regular-device

Re: Obsoleted large bucket layout message once

[Coly Li]

On 2/16/21 11:06 PM, Constantin Runge wrote:
> Hi everybody,
> 
> when registering and mounting my bcache devices (unortunately my root
> filesystem), the kernel informed me about `Detected obsoleted large
> bucket layout, all attached bcache device will be read-only'.
> 
> Subsequently, my root fs was, in fact, read-only. Thus, I have no logs
> for you unfortunately. As far as I remember, bcache-super-show showed
> the same bucket and block sizes as always (except that the backing
> device showed a clean cache state):
> 
>> bcache-super-show /dev/sda6
> sb.magic        ok
> sb.first_sector        8 [match]
> sb.csum            C5EB3BA8CA523D61 [match]
> sb.version        4 [backing device]
> 
> dev.label        (empty)
> dev.uuid        264d1acc-8461-4ddc-a825-f0acdeea3ede
> dev.sectors_per_block    1
> dev.sectors_per_bucket    1024
> dev.data.first_sector    2048
> dev.data.cache_mode    1 [writeback]
> dev.data.cache_state    2 [dirty]
> 
> cset.uuid        755901be-a101-4581-b1d0-4d21d1619f27
>> bcache-super-show /dev/sdb3
> sb.magic        ok
> sb.first_sector        8 [match]
> sb.csum            BDA92F1E72DBA095 [match]
> sb.version        3 [cache device]
> 
> dev.label        (empty)
> dev.uuid        25151c1a-904e-492e-b99b-6d07e13b047a
> dev.sectors_per_block    1
> dev.sectors_per_bucket    1024
> dev.cache.first_sector    1024
> dev.cache.cache_sectors    907564032
> dev.cache.total_sectors    907565056
> dev.cache.ordered    yes
> dev.cache.discard    no
> dev.cache.pos        0
> dev.cache.replacement    0 [lru]
> 
> cset.uuid        755901be-a101-4581-b1d0-4d21d1619f27
> 
> 
> For debugging purposes I tried to mount and chroot into my backing
> device directly, using a loopback interface. I was not able to find an
> offset for `losetup -f -o <offset> /dev/sda6', so that `mount -o loop
> /dev/loop0 /mnt/mnt0' would not tell me that there was no filesystem.
> Also, `for i in {0.2000}; do dd if=/dev/sda6 skip=$i | file -; done'
> from [1] showed `data' for every offset.
> 
> After getting a little desperate, I booted into an older kernel
> (although I didn't update my kernel or my bcache-tools since the last
> boot) and everything worked. I then booted back into my current system
> and everything worked again.
> 
> Is this a bug with bcache or do I need to change my setup to prevent
> this from happening again? My bcache devices were created sometime
> around 2015 or 2016.
> 
> System info:
> uname -a: `Linux Horus 5.10.10-gentoo #1 SMP PREEMPT Mon Feb 15 15:00:49
> CET 2021 x86_64 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz GenuineIntel
> GNU/Linux'
> bcache-tools version: 1.1
> The older kernel is a 5.8.11-gentoo with voluntary preempt.
> 
> Best regards and thanks,
> Constantin
> 
> 
>   [1]
> https://stackoverflow.com/questions/22820492/how-to-revert-bcache-device-to-regular-device
> 

This is a regression and fixed in 5.11-rc6 by commit 0df28cad06eb
("bcache: only check feature sets when sb->version >=
BCACHE_SB_VERSION_CDEV_WITH_FEATURES"). This patch has been in stable
kernel for a while, the distribution maintainer should have it in time.

Thanks.

Coly Li