Re: Fwd: Read time tree block corruption detected

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

[-- Attachment #1.1: Type: text/plain, Size: 19171 bytes --]



On 2020/8/23 上午9:15, Tyler Richmond wrote:
> Is my best bet just to downgrade the kernel and then try to delete the
> broken files? Or should I rebuild from scratch? Just don't know
> whether it's worth the time to try and figure this out or if the
> problems stem from the FS being too old and it's beyond trying to
> repair.

All invalid inode generations, should be able to be repaired by latest
btrfs-check.

If not, please provide the btrfs-image dump for us to determine what's
going wrong.

Thanks,
Qu
> 
> On Tue, Aug 18, 2020 at 8:18 AM Tyler Richmond <t.d.richmond@gmail.com> wrote:
>>
>> I didn't check dmesg during the btrfs check, but that was the only
>> output during the rm -f before it was forced readonly. I just checked
>> dmesg for inode generation values, and there are a lot of them.
>>
>> https://pastebin.com/stZdN0ta
>> The dmesg output had 990 lines containing inode generation.
>>
>> However, these were at least later. I tried to do a btrfs balance
>> -mconvert raid1 and it failed with an I/O error. That is probably what
>> generated these specific errors, but maybe they were also happening
>> during the btrfs repair.
>>
>> The FS is ~45TB, but the btrfs-image -c9 failed anway with:
>> ERROR: either extent tree is corrupted or deprecated extent ref format
>> ERROR: create failed: -5
>>
>>
>> On Tue, Aug 18, 2020 at 2:07 AM Qu Wenruo <quwenruo.btrfs@gmx.com> wrote:
>>>
>>>
>>>
>>> On 2020/8/18 上午11:35, Tyler Richmond wrote:
>>>> Qu,
>>>>
>>>> Sorry to resurrect this thread, but I just ran into something that I
>>>> can't really just ignore. I've found a folder that is full of files
>>>> which I guess have been broken somehow. I found a backup and restored
>>>> them, but I want to delete this folder of broken files. But whenever I
>>>> try, the fs is forced into readonly mode again. I just finished another
>>>> btrfs check --repair but it didn't fix the problem.
>>>>
>>>> https://pastebin.com/eTV3s3fr
>>>
>>> Is that the full output?
>>>
>>> No inode generation bugs?
>>>>
>>>>  I'm already on btrfs-progs v5.7. Any new suggestions?
>>>
>>> Strange.
>>>
>>> The detection and repair should have been merged into v5.5.
>>>
>>> If your fs is small enough, would you please provide the "btrfs-image
>>> -c9" dump?
>>>
>>> It would contain the filenames and directories names, but doesn't
>>> contain file contents.
>>>
>>> Thanks,
>>> Qu
>>>>
>>>> On Fri, May 8, 2020 at 9:52 AM Tyler Richmond <t.d.richmond@gmail.com
>>>> <mailto:t.d.richmond@gmail.com>> wrote:
>>>>
>>>>     5.6.1 also failed the same way. Here's the usage output. This is the
>>>>     part where you see I've been using RAID5 haha
>>>>
>>>>     WARNING: RAID56 detected, not implemented
>>>>     Overall:
>>>>         Device size:                  60.03TiB
>>>>         Device allocated:             98.06GiB
>>>>         Device unallocated:           59.93TiB
>>>>         Device missing:                  0.00B
>>>>         Used:                         92.56GiB
>>>>         Free (estimated):                0.00B      (min: 8.00EiB)
>>>>         Data ratio:                       0.00
>>>>         Metadata ratio:                   2.00
>>>>         Global reserve:              512.00MiB      (used: 0.00B)
>>>>         Multiple profiles:                  no
>>>>
>>>>     Data,RAID5: Size:40.35TiB, Used:40.12TiB (99.42%)
>>>>        /dev/sdh        8.07TiB
>>>>        /dev/sdf        8.07TiB
>>>>        /dev/sdg        8.07TiB
>>>>        /dev/sdd        8.07TiB
>>>>        /dev/sdc        8.07TiB
>>>>        /dev/sde        8.07TiB
>>>>
>>>>     Metadata,RAID1: Size:49.00GiB, Used:46.28GiB (94.44%)
>>>>        /dev/sdh       34.00GiB
>>>>        /dev/sdf       32.00GiB
>>>>        /dev/sdg       32.00GiB
>>>>
>>>>     System,RAID1: Size:32.00MiB, Used:2.20MiB (6.87%)
>>>>        /dev/sdf       32.00MiB
>>>>        /dev/sdg       32.00MiB
>>>>
>>>>     Unallocated:
>>>>        /dev/sdh        2.81TiB
>>>>        /dev/sdf        2.81TiB
>>>>        /dev/sdg        2.81TiB
>>>>        /dev/sdd        1.03TiB
>>>>        /dev/sdc        1.03TiB
>>>>        /dev/sde        1.03TiB
>>>>
>>>>     On Fri, May 8, 2020 at 1:47 AM Qu Wenruo <quwenruo.btrfs@gmx.com
>>>>     <mailto:quwenruo.btrfs@gmx.com>> wrote:
>>>>     >
>>>>     >
>>>>     >
>>>>     > On 2020/5/8 下午1:12, Tyler Richmond wrote:
>>>>     > > If this is saying there's no extra space for metadata, is that why
>>>>     > > adding more files often makes the system hang for 30-90s? Is there
>>>>     > > anything I should do about that?
>>>>     >
>>>>     > I'm not sure about the hang though.
>>>>     >
>>>>     > It would be nice to give more info to diagnosis.
>>>>     > The output of 'btrfs fi usage' is useful for space usage problem.
>>>>     >
>>>>     > But the common idea is, to keep at 1~2 Gi unallocated (not avaiable
>>>>     > space in vanilla df command) space for btrfs.
>>>>     >
>>>>     > Thanks,
>>>>     > Qu
>>>>     >
>>>>     > >
>>>>     > > Thank you so much for all of your help. I love how flexible BTRFS is
>>>>     > > but when things go wrong it's very hard for me to troubleshoot.
>>>>     > >
>>>>     > > On Fri, May 8, 2020 at 1:07 AM Qu Wenruo <quwenruo.btrfs@gmx.com
>>>>     <mailto:quwenruo.btrfs@gmx.com>> wrote:
>>>>     > >>
>>>>     > >>
>>>>     > >>
>>>>     > >> On 2020/5/8 下午12:23, Tyler Richmond wrote:
>>>>     > >>> Something went wrong:
>>>>     > >>>
>>>>     > >>> Reinitialize checksum tree
>>>>     > >>> Unable to find block group for 0
>>>>     > >>> Unable to find block group for 0
>>>>     > >>> Unable to find block group for 0
>>>>     > >>> ctree.c:2272: split_leaf: BUG_ON `1` triggered, value 1
>>>>     > >>> btrfs(+0x6dd94)[0x55a933af7d94]
>>>>     > >>> btrfs(+0x71b94)[0x55a933afbb94]
>>>>     > >>> btrfs(btrfs_search_slot+0x11f0)[0x55a933afd6c8]
>>>>     > >>> btrfs(btrfs_csum_file_block+0x432)[0x55a933b19d09]
>>>>     > >>> btrfs(+0x360b2)[0x55a933ac00b2]
>>>>     > >>> btrfs(+0x46a3e)[0x55a933ad0a3e]
>>>>     > >>> btrfs(main+0x98)[0x55a933a9fe88]
>>>>     > >>>
>>>>     /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf3)[0x7f263ed550b3]
>>>>     > >>> btrfs(_start+0x2e)[0x55a933a9fa0e]
>>>>     > >>> Aborted
>>>>     > >>
>>>>     > >> This means no space for extra metadata...
>>>>     > >>
>>>>     > >> Anyway the csum tree problem shouldn't be a big thing, you
>>>>     could leave
>>>>     > >> it and call it a day.
>>>>     > >>
>>>>     > >> BTW, as long as btrfs check reports no extra problem for the inode
>>>>     > >> generation, it should be pretty safe to use the fs.
>>>>     > >>
>>>>     > >> Thanks,
>>>>     > >> Qu
>>>>     > >>>
>>>>     > >>> I just noticed I have btrfs-progs 5.6 installed and 5.6.1 is
>>>>     > >>> available. I'll let that try overnight?
>>>>     > >>>
>>>>     > >>> On Thu, May 7, 2020 at 8:11 PM Qu Wenruo
>>>>     <quwenruo.btrfs@gmx.com <mailto:quwenruo.btrfs@gmx.com>> wrote:
>>>>     > >>>>
>>>>     > >>>>
>>>>     > >>>>
>>>>     > >>>> On 2020/5/7 下午11:52, Tyler Richmond wrote:
>>>>     > >>>>> Thank you for helping. The end result of the scan was:
>>>>     > >>>>>
>>>>     > >>>>>
>>>>     > >>>>> [1/7] checking root items
>>>>     > >>>>> [2/7] checking extents
>>>>     > >>>>> [3/7] checking free space cache
>>>>     > >>>>> [4/7] checking fs roots
>>>>     > >>>>
>>>>     > >>>> Good news is, your fs is still mostly fine.
>>>>     > >>>>
>>>>     > >>>>> [5/7] checking only csums items (without verifying data)
>>>>     > >>>>> there are no extents for csum range 0-69632
>>>>     > >>>>> csum exists for 0-69632 but there is no extent record
>>>>     > >>>>> ...
>>>>     > >>>>> ...
>>>>     > >>>>> there are no extents for csum range 946692096-946827264
>>>>     > >>>>> csum exists for 946692096-946827264 but there is no extent
>>>>     record
>>>>     > >>>>> there are no extents for csum range 946831360-947912704
>>>>     > >>>>> csum exists for 946831360-947912704 but there is no extent
>>>>     record
>>>>     > >>>>> ERROR: errors found in csum tree
>>>>     > >>>>
>>>>     > >>>> Only extent tree is corrupted.
>>>>     > >>>>
>>>>     > >>>> Normally btrfs check --init-csum-tree should be able to
>>>>     handle it.
>>>>     > >>>>
>>>>     > >>>> But still, please be sure you're using the latest btrfs-progs
>>>>     to fix it.
>>>>     > >>>>
>>>>     > >>>> Thanks,
>>>>     > >>>> Qu
>>>>     > >>>>
>>>>     > >>>>> [6/7] checking root refs
>>>>     > >>>>> [7/7] checking quota groups skipped (not enabled on this FS)
>>>>     > >>>>> found 44157956026368 bytes used, error(s) found
>>>>     > >>>>> total csum bytes: 42038602716
>>>>     > >>>>> total tree bytes: 49688616960
>>>>     > >>>>> total fs tree bytes: 1256427520
>>>>     > >>>>> total extent tree bytes: 1709105152
>>>>     > >>>>> btree space waste bytes: 3172727316
>>>>     > >>>>> file data blocks allocated: 261625653436416
>>>>     > >>>>>  referenced 47477768499200
>>>>     > >>>>>
>>>>     > >>>>> What do I need to do to fix all of this?
>>>>     > >>>>>
>>>>     > >>>>> On Thu, May 7, 2020 at 1:52 AM Qu Wenruo
>>>>     <quwenruo.btrfs@gmx.com <mailto:quwenruo.btrfs@gmx.com>> wrote:
>>>>     > >>>>>>
>>>>     > >>>>>>
>>>>     > >>>>>>
>>>>     > >>>>>> On 2020/5/7 下午1:43, Tyler Richmond wrote:
>>>>     > >>>>>>> Well, the repair doesn't look terribly successful.
>>>>     > >>>>>>>
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>
>>>>     > >>>>>> This means there are more problems, not only the hash name
>>>>     mismatch.
>>>>     > >>>>>>
>>>>     > >>>>>> This means the fs is already corrupted, the name hash is
>>>>     just one
>>>>     > >>>>>> unrelated symptom.
>>>>     > >>>>>>
>>>>     > >>>>>> The only good news is, btrfs-progs abort the transaction,
>>>>     thus no
>>>>     > >>>>>> further damage to the fs.
>>>>     > >>>>>>
>>>>     > >>>>>> Please run a plain btrfs-check to show what's the problem
>>>>     first.
>>>>     > >>>>>>
>>>>     > >>>>>> Thanks,
>>>>     > >>>>>> Qu
>>>>     > >>>>>>
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> parent transid verify failed on 218620880703488 wanted
>>>>     6875841 found 6876224
>>>>     > >>>>>>> Ignoring transid failure
>>>>     > >>>>>>> ERROR: child eb corrupted: parent bytenr=225049956061184
>>>>     item=84
>>>>     > >>>>>>> parent level=1
>>>>     > >>>>>>>                                             child level=4
>>>>     > >>>>>>> ERROR: failed to zero log tree: -17
>>>>     > >>>>>>> ERROR: attempt to start transaction over already running one
>>>>     > >>>>>>> WARNING: reserved space leaked, flag=0x4 bytes_reserved=4096
>>>>     > >>>>>>> extent buffer leak: start 225049066086400 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066086400 len 4096
>>>>     > >>>>>>> WARNING: dirty eb leak (aborted trans): start
>>>>     225049066086400 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066094592 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066094592 len 4096
>>>>     > >>>>>>> WARNING: dirty eb leak (aborted trans): start
>>>>     225049066094592 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066102784 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066102784 len 4096
>>>>     > >>>>>>> WARNING: dirty eb leak (aborted trans): start
>>>>     225049066102784 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066131456 len 4096
>>>>     > >>>>>>> extent buffer leak: start 225049066131456 len 4096
>>>>     > >>>>>>> WARNING: dirty eb leak (aborted trans): start
>>>>     225049066131456 len 4096
>>>>     > >>>>>>>
>>>>     > >>>>>>> What is going on?
>>>>     > >>>>>>>
>>>>     > >>>>>>> On Wed, May 6, 2020 at 9:30 PM Tyler Richmond
>>>>     <t.d.richmond@gmail.com <mailto:t.d.richmond@gmail.com>> wrote:
>>>>     > >>>>>>>>
>>>>     > >>>>>>>> Chris, I had used the correct mountpoint in the command.
>>>>     I just edited
>>>>     > >>>>>>>> it in the email to be /mountpoint for consistency.
>>>>     > >>>>>>>>
>>>>     > >>>>>>>> Qu, I'll try the repair. Fingers crossed!
>>>>     > >>>>>>>>
>>>>     > >>>>>>>> On Wed, May 6, 2020 at 9:13 PM Qu Wenruo
>>>>     <quwenruo.btrfs@gmx.com <mailto:quwenruo.btrfs@gmx.com>> wrote:
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>> On 2020/5/7 上午5:54, Tyler Richmond wrote:
>>>>     > >>>>>>>>>> Hello,
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>> I looked up this error and it basically says ask a
>>>>     developer to
>>>>     > >>>>>>>>>> determine if it's a false error or not. I just started
>>>>     getting some
>>>>     > >>>>>>>>>> slow response times, and looked at the dmesg log to
>>>>     find a ton of
>>>>     > >>>>>>>>>> these errors.
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>> [192088.446299] BTRFS critical (device sdh): corrupt
>>>>     leaf: root=5
>>>>     > >>>>>>>>>> block=203510940835840 slot=4 ino=1311670, invalid inode
>>>>     generation:
>>>>     > >>>>>>>>>> has 18446744073709551492 expect [0, 6875827]
>>>>     > >>>>>>>>>> [192088.449823] BTRFS error (device sdh):
>>>>     block=203510940835840 read
>>>>     > >>>>>>>>>> time tree block corruption detected
>>>>     > >>>>>>>>>> [192088.459238] BTRFS critical (device sdh): corrupt
>>>>     leaf: root=5
>>>>     > >>>>>>>>>> block=203510940835840 slot=4 ino=1311670, invalid inode
>>>>     generation:
>>>>     > >>>>>>>>>> has 18446744073709551492 expect [0, 6875827]
>>>>     > >>>>>>>>>> [192088.462773] BTRFS error (device sdh):
>>>>     block=203510940835840 read
>>>>     > >>>>>>>>>> time tree block corruption detected
>>>>     > >>>>>>>>>> [192088.464711] BTRFS critical (device sdh): corrupt
>>>>     leaf: root=5
>>>>     > >>>>>>>>>> block=203510940835840 slot=4 ino=1311670, invalid inode
>>>>     generation:
>>>>     > >>>>>>>>>> has 18446744073709551492 expect [0, 6875827]
>>>>     > >>>>>>>>>> [192088.468457] BTRFS error (device sdh):
>>>>     block=203510940835840 read
>>>>     > >>>>>>>>>> time tree block corruption detected
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>> btrfs device stats, however, doesn't show any errors.
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>> Is there anything I should do about this, or should I
>>>>     just continue
>>>>     > >>>>>>>>>> using my array as normal?
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>> This is caused by older kernel underflow inode generation.
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>> Latest btrfs-progs can fix it, using btrfs check --repair.
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>> Or you can go safer, by manually locating the inode
>>>>     using its inode
>>>>     > >>>>>>>>> number (1311670), and copy it to some new location using
>>>>     previous
>>>>     > >>>>>>>>> working kernel, then delete the old file, copy the new
>>>>     one back to fix it.
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>> Thanks,
>>>>     > >>>>>>>>> Qu
>>>>     > >>>>>>>>>
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>> Thank you!
>>>>     > >>>>>>>>>>
>>>>     > >>>>>>>>>
>>>>     > >>>>>>
>>>>     > >>>>
>>>>     > >>
>>>>     >
>>>>
>>>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]