On 2017年11月02日 13:50, Lakshmipathi.G wrote: >> >> I'll try to reproduce it with FS_CHECK_INTEGRITY enabled. >> > > Okay thanks for the details. Here's the kernel config file which hits > this issue: > https://github.com/Lakshmipathi/btrfsqa/blob/master/setup/config/kernel.config#L3906
Although I didn't reproduced your problem, I found fstests can easily trigger the same problem. The problem turns out to be that setup_items_for_insert() will call btrfs_mark_buffer_dirty(), however that's not done at correct timing. If FS_CHECK_INTEGRITY is enabled, every time we call btrfs_mark_buffer_dirty(), we must ensure have setup the leaf correctly. However for setup_items_for_insert(), it doesn't really insert the data but only keys, leaving item data uninitialized, and triggering the bug. (That's to say, we should not do leaf validation check at btrfs_mark_buffer_dirty() time). I'll try to move it to leaf write time to fix the problem. For now, please disable FS_CHECK_INTEGRITY. Thanks, Qu > > thanks! > ---- > Cheers, > Lakshmipathi.G > http://www.giis.co.in http://www.webminal.org >
signature.asc
Description: OpenPGP digital signature
