On Mon, Jul 29, 2019 at 9:05 AM Swâmi Petaramesh <sw...@petaramesh.org> wrote: > > On 7/29/19 4:55 PM, Swâmi Petaramesh wrote: > > Well All the errors I detailed today happen on the SAME FS, and this fs > > is a BTRFS that was created on a new HD with a recent kernel (surely >= > > 4.19) only a few months ago. > > > > And the errors I have one this one, As far as I can tell, look exactly > > like what happened on the same machines SSD as soons as I installer a > > 5.2 kernel... > > Plus I just decided to “btrfs check” the SSD FS from my machine (not yet > showing errors), which I completely reformatted using 5.2 3 days ago > (after having fully tested the SSD error-free itself)... > > And btrfs check tells me that this FS is now completely corrupt as well > :-((( > > The list of files in error has been scrolling for five minutes now :-(((
Without both dmesg and btrfs check output it's not very useful. I've got a case where a file system scrubs fine and btrfs check complains, but turns out it's because of nocow files that are compressed via defrag path. The files are fine, there is no corruption, it's just noise. But the only way to know it is to always include the full dmesg and check output - I personally think snippets and trimmed logs are annoying. In this case we don't actually have anything to go on, therefore the problem could be anything, therefore we need all the information available. It seems unlikely drive related, as so many drives are involved. Same for logic board or RAM. I think these days with ext4 and XFS using checksumming for metadata, that if it were device-mapper or blk-mq related, they too would see errors. And yet of course many people are using kernel 5.2 with Btrfs and aren't having problems. So, it's just inherently tedious work to narrow down what's causing the problem. I think it's totally reasonable to go back to 5.1 for a while and make certain the problems aren't happening there. If they are, then I start to wonder about noisy power since you have so many different drives and setups affected. Some of the strangest problems I have ever seen in computing were directly attributed to noise on the power line. -- Chris Murphy
