[BUG] BTRFS: error in btrfs_run_delayed_refs:2821: errno=-5 IO failure

[BUG] BTRFS: error in btrfs_run_delayed_refs:2821: errno=-5 IO failure

[Chris Murphy]

The short version: btrfs convert and subsequent balance will eat an
ext4 file system. I've reproduced this 6 for 6 times with both kernel
4.1 and 4.2rc1, but I get different back traces for the two kernels.

Kernel 4.1.0
Btrfs-progs 4.1

New ext4 file system populated with several thousand files (Fedora 22
installation) and a 1.5GB ISO.

Conversion succeeds with no user space or kernel errors. Btrfs check
comes up clean. Deletion of ext2_saved subvolume happens without
error, as does defragment. After this btrfs check is still clean. Then
major problems on btrfs balance and the file system is totally broken
afterward.


balance failure following ext4 to btrfs conversion, BTRFS: error in
btrfs_run_delayed_refs:2821: errno=-5 IO failure
https://bugzilla.kernel.org/show_bug.cgi?id=101191



-- 
Chris Murphy

Re: [BUG] BTRFS: error in btrfs_run_delayed_refs:2821: errno=-5 IO failure

[Noah Massey]

On Wed, Jul 8, 2015 at 10:01 AM, Chris Murphy <lists@colorremedies.com> wrote:
> The short version: btrfs convert and subsequent balance will eat an
> ext4 file system. I've reproduced this 6 for 6 times with both kernel
> 4.1 and 4.2rc1, but I get different back traces for the two kernels.
>
> Kernel 4.1.0
> Btrfs-progs 4.1
>
> New ext4 file system populated with several thousand files (Fedora 22
> installation) and a 1.5GB ISO.
>

Standard disclaimer: Not a developer, just a user.
The 1.5 GB ISO on a fresh ext4 system made me think of
http://thread.gmane.org/gmane.comp.file-systems.btrfs/36955
(ext4 extents over 1 GB causing ENOSPC).

Did the "btrfs filesystem balance" specify a "-t 900G" or similar?

Or has that issue been fixed?

Re: [BUG] BTRFS: error in btrfs_run_delayed_refs:2821: errno=-5 IO failure

[Николай Овчинников]

Looks like my case.
I also have some files over 1Gb ans skip defrag before balance.


2015-07-08 23:02 GMT+03:00 Noah Massey <noah.massey@gmail.com>:
> On Wed, Jul 8, 2015 at 10:01 AM, Chris Murphy <lists@colorremedies.com> wrote:
>> The short version: btrfs convert and subsequent balance will eat an
>> ext4 file system. I've reproduced this 6 for 6 times with both kernel
>> 4.1 and 4.2rc1, but I get different back traces for the two kernels.
>>
>> Kernel 4.1.0
>> Btrfs-progs 4.1
>>
>> New ext4 file system populated with several thousand files (Fedora 22
>> installation) and a 1.5GB ISO.
>>
>
> Standard disclaimer: Not a developer, just a user.
> The 1.5 GB ISO on a fresh ext4 system made me think of
> http://thread.gmane.org/gmane.comp.file-systems.btrfs/36955
> (ext4 extents over 1 GB causing ENOSPC).
>
> Did the "btrfs filesystem balance" specify a "-t 900G" or similar?
>
> Or has that issue been fixed?
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [BUG] BTRFS: error in btrfs_run_delayed_refs:2821: errno=-5 IO failure

[Duncan]

Noah Massey posted on Wed, 08 Jul 2015 16:02:04 -0400 as excerpted:

> Standard disclaimer: Not a developer, just a user.

Same here. =:^)

> The 1.5 GB ISO on a fresh ext4 system made me think of
> http://thread.gmane.org/gmane.comp.file-systems.btrfs/36955 (ext4
> extents over 1 GB causing ENOSPC).
> 
> Did the "btrfs filesystem balance" specify a "-t 900G" or similar?
> 
> Or has that issue been fixed?

Don't know about bug fixed, but that balance command syntax is definitely 
very deprecated, and there's no -t option documented for btrfs balance 
start, the current version.

Did you mean btrfs filesystem defrag?  It has a -t option that looks like 
it might have been what you were after.  However, until very recently (I 
think it's in integration but don't know if it's in an actual release 
yet), there was an integer overflow bug with the -t option, such that if 
given a value above 3G, (I believe the cutoff is actually 4G, so 
something like 3999M would work), it would behave as if -t0 had been 
given. =:^(  

But the effective high value is typically 1G anyway, so the workaround is 
simple enough, just use -t 1G or -t 2G.  That's what the patch 
effectively does anyway, reduce anything higher to 1G or 2G, IDR which.

So if you meant btrfs filesystem defrag -t, a 2G value would be 
preferable to a 900G value, with the latter very likely not working as 
intended.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman