On 2017年11月13日 11:37, Chris Murphy wrote: > On Sun, Nov 12, 2017 at 7:25 PM, Su Yue <suy.f...@cn.fujitsu.com> wrote: > >> >> I have some patches for lowmem mode. But they are not fully tested and >> unstable. >> It will be so nice if you can provide some images. > > Sorry, Qu got the image offline, I wasn't paying attention that btrfs > list wasn't cc'd. The -ss option was taking too long and during the > imaging a systemd unit mounted the file system so I figured the image > might be useless or contain extra confusion. > > Here is the -s image I send to Qu. > > https://drive.google.com/open?id=1cIAANqMicIMp9NfgnvN_SA90kjM0kP0C
Patches for backref lost false warning Cced to you. Since the reason is clear and test case added, I think it should solve the "backref lost" false alerts. However for BUG_ON(), it's tree walk code who can't handle tree reloc tree, this may be a little complex compared to backref lost. But since it's a different story, I'll send the fix as another patch(set). The final bug about backref counts mismatch, without the image I'm not really sure what's going on. (But extent tree verification is really buggy though) Thanks, Qu
signature.asc
Description: OpenPGP digital signature
