On 28/11/2018 05:15, Theodore Y. Ts'o wrote: > On Wed, Nov 28, 2018 at 03:16:33AM +0300, Andrey Jr. Melnikov wrote: >> Corrupted inodes - always directory, not touched at least year or >> more for writing. Something wrong when updating atime? I've just seen the errors come back despite having MQ off :( However, this time it took 5 days for them to come back, so MQ must play a role here. Also, indeed, they happened after fstrim ran and this time *only* on the SSD disks reported below, another clue? This time the errors were "just" orphaned inodes + invalid free inode counts, all repaired without issues by fsck. > > We're not sure. The frustrating thing is that it's not reproducing > for me. I run extensive regression tests, and I'm using 4.19 on my > development laptop without notcing any problems. If I could reproduce > it, I could debug it, but since I can't, I need to rely on those who > are seeing the problem to help pinpoint the problem. > > I'm trying to figure out common factors from those people who are > reporting problems. > > (a) What distribution are you running (it appears that many people > reporting problems are running Ubuntu, but this may be a sampling > issue; lots of people run Ubuntu)? (For the record, I'm using Debian > Testing.) Ubuntu 18.10 here > > (b) What hardware are you using? (SSD? SATA-attached? > NVMe-attached?) The errors occured on both SSD: - Samsung SSD 850 EVO 1TB, firmware rev EMT03B6Q - OCZ-AGILITY3, firmware rev 2.25 and spinning rust: - Seagate ST2000DX001-1CM164, firmware revision CC43 > > (c) Are you using LVM? LUKS (e.g., disk encrypted)? LUKS. Both the Samsung and the Seagate use DM for encryption. > (d) are you using discard? One theory is a recent discard change may > be in play. How do you use discard? (mount option, fstrim, etc.) fstrim runs weekly and the Samsung SSD is mounted with rw,nosuid,nodev,noatime,discard,helper=crypt marek > > - Ted >
