I can't spend more time on it, but it appears that the bug is stable (i.e., for a particular revision, I can repeatably trigger, or not trigger the bug). However, when I bisect it using kvm, it doesn't make any sense. This is what I am doing to repro the bug.
/dev/vdd is a 220G ext4 file system, created using LVM. I compile a 64-bit kernel, and then to test for the bug, I do the following commands: e2fsck -fy /dev/vdd mount -t ext4 /dev/vdd /vdd cd /vdd rm -f file1 fallocate -l 215g file1 mke2fs -t ext4 -F file1 If you see lots of ext4 error messages in the dmesg (I'm logged into the KVM environment via a serial console, so it's really obvious when there are areas), then that's the bug. A sample of the ext4 error message is: [ 109.811751] EXT4-fs error (device vdd): ext4_ext_map_blocks:3770: inode #12: comm mke2fs.static: bad extent address lblock: 1023, depth: 2 pblock 0 Since I am doing a reverse bisection to try to find a bug, "bisect good" means that the ext4 error messages are detected. "bisect bad" means that the mke2fs succeeded. (Note that in order for this test to be valid, you need to make sure that you see the Discarding blocks: done message in mke2fs. If instead you see some message about the discard ioctl failing --- which will happen if you use a 32-bit kernel or a 32-bit userspace binary of mke2fs --- then the mke2fs will always succeed w/o errors.) So here's my bisection log. Unfortunately, the place where this ends up makes no sense (inside some ARM arch code, which wouldn't be applicable since this was being tested on an x86_64 kernel running under KVM). I can't spend any more time on this, and since I can't reproduce the problem at all on anything newer than 3.3, it's not particularly interesting to me as an upstream developer.... Hopefully some Canonical kernel developer or Ubuntu user can take this further.... git bisect start # git bisect good 805a6af8dba5dfdd35ec35dc52ec0122400b2610 git bisect good 22cdfca5641817060dd724a9c30442f5c0675fcd # conf git bisect good 5872ddaaf05bf25e3ab90580295ebc946405928c # conf git bisect good 34c80b1d93e6e20ca9dea0baf583a5b5510d92d4 # conf git bisect good 98793265b429a3f0b3f1750e74d67cd4d740d162 # conf git bisect good 12e55508921865aefdd80fd17afe70c191afbd1b # conf git bisect good 6d889d03ab1417645e76e129834f76204bae37c0 # conf git bisect good 2ac9d7aaccbd598b5bd19ac40761b723bb675442 # conf git bisect good 6b6fa98c3e6d29eb7b515d807139590a27e6b9b0 # conf git bisect good e8cbce976050a9f874a8b07012ddeb9b9eb59603 # conf git bisect good e735489976c356a189f7332e6ff4b0dee2ede434 # conf git bisect good 5c77f5608a5081632b9530e5f3260f8b63247bd8 # conf git bisect good ca4c22d391b8a5ac2b79135213ca68abc7aeb984 # conf git bisect good a481377013c1c07dac0421886db7a6b4c3081c49 # conf # git bisect bad c16fa4f2ad19908a47c63d8fa436a1178438c7e7 git bisect bad e4e11180dfa545233e5145919b75b7fac88638df # conf git bisect bad 979ecef5b89a8003902299566d9cdc08de34a3ee # conf git bisect bad e1482a1708e48add8d2f4ecc949885e0e552d9e8 # conf (#conf means I double checked the KVM console logs to make sure I didn't confuse myself with the reverse bisection and screwed up the sense of git bisect good/bad while doing the reverse bisection....) -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1031518 Title: fallocate followed by mke2fs of file image causes ext4_error To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1031518/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
