On 10/27/21 4:06 PM, Gabe Al-Ghalith wrote:
I'm running a 28TB volume on USB3 with xfs. I have many other volumes with NTFS formatting that haven't had issues, but both xfs volumes have had trouble. When I run xfs_repair on a drive that hangs and crashes, I get a segfault. I'm running Fedora 33 with Linux 5.14.11-100.fc33.x86_64. My machine has 1.5TB of RAM. I seem to reach the segmentation fault on what looks like recursive reconstruction of ".." directory inodes: I'm including the run log as of Phase 6 (which is where the failure happens) inline below: Thanks, Gabe
Can you try creating an xfs_metadump of the filesystem, and see if the segfault is reproducible? Then, if you are willing to share it (in private) we could do some debugging. To create the xfs_metadump: # xfs_metadump /dev/$WHATEVER metadump.meta To see if the image reproduces the problem, # xfs_mdrestore metadump.meta metadump.img (Note that both metadump.meta and metadump.img are metadata-only, contain no file data, and will take FAR less than 28T of disk space.) If so, and if you're OK with sharing the obfuscated metadump (it contains no file data, but may contain unobfuscated short file names, xattrs etc), then compress it and send it my way, offline. Thanks, -Eric
