On 2017年10月25日 14:54, Tom Hale wrote: > > > On 19/10/17 15:58, Qu Wenruo wrote: >> On 2017年10月19日 16:53, Tom Hale wrote: >>> In running btrfs check, I got the following message: >>> >>> warning line 4144 >>> >>> Could this be a little more descriptive? >>> >>> * Does it mean I should rebuild my FS from scratch? >>> * Is there anything I can do to remove this warning? >>> >> >> --repair is dangerous, use it unless you're sure the problem can be >> fixed by it. >> >> What's the output of "btrfs check" and "btrfs check --mode=lowmem" after >> doing the repair? > > Plain `btrfs check` gives me: > > checking extents > checking free space cache > checking fs roots > checking csums > checking root refs > Checking filesystem on /dev/mapper/vg_svelte-home > UUID: 93722fa7-7e8f-418a-a7ca-080aca8db94b > found 192318169092 bytes used, no error found > total csum bytes: 185482112 > total tree bytes: 1924104192 > total fs tree bytes: 1597767680 > total extent tree bytes: 102514688 > btree space waste bytes: 325265468 > file data blocks allocated: 6169163599872 > referenced 575636733952
At least your fs is good to go. > > Could somebody please answer my initial questions about this obscure > warning? Only two places will output this warning, and they are both used by original mode. So at least if you're using lowmem mode, you won't encounter this warning line. And for your line at 4144, git blames shows it's from the ancient day of btrfs-progs. According to a quick glance of it, it seems that during the check, it encountered a shared node, which is quite common for fs with snapshots, but it doesn't do the cleanup correctly. I think it's caused by some extent tree problem, which lacks some backref or has some mismatch backref, to cause the shared node checking code behaved unexpectedly. (runtime glitch) Thanks, Qu > > Thanks, >
signature.asc
Description: OpenPGP digital signature
