Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#46796 - [linux-grsec] journald and other daemons killed with «PAX: size overflow detected»
Attached to Project:
Community Packages
Opened by Hermann Zahnweh (eigengrau) - Monday, 19 October 2015, 08:03 GMT
Last edited by Daniel Micay (thestinger) - Friday, 13 November 2015, 00:41 GMT
Opened by Hermann Zahnweh (eigengrau) - Monday, 19 October 2015, 08:03 GMT
Last edited by Daniel Micay (thestinger) - Friday, 13 November 2015, 00:41 GMT
|
DetailsDescription:
After updating linux-grsec from 4.2.3.201510072230-1 to 4.2.3.201510171833-1, I’ve noticed that most of the system daemons (inter alia journald) fail to start. From the dmesg output, it looks as if they are all being killed by PAX arising from a btrfs kernel function: [ 4.507532] PAX: size overflow detected in function btrfs_sync_file fs/btrfs/file.c:1871 cicus.676_106 max, count: 295, decl: btrfs_wait_ordered_range; num: 3; context: fndecl; [ 4.509936] CPU: 0 PID: 169 Comm: systemd-journal Not tainted 4.2.3.201510171833-1-grsec #1 [ 4.509939] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007 [ 4.509941] ffffffffbba0a847 430c9d1ad0298947 0000000000000000 ffffffffc0217cae [ 4.509945] ffffc900007ebd68 ffffffffbb5b53db 00000000000000a4 ffffffffc0217d77 [ 4.509948] ffffc900007ebd98 ffffffffbb1a1dd4 ffff88007ceecd68 0000000000000000 [ 4.509950] Call Trace: [ 4.509969] [<ffffffffc0217cae>] ? exit_btrfs_fs+0x16d07/0x2e599 [btrfs] [ 4.509974] [<ffffffffbb5b53db>] dump_stack+0x4c/0x7f [ 4.509980] [<ffffffffc0217d77>] ? exit_btrfs_fs+0x16dd0/0x2e599 [btrfs] [ 4.509984] [<ffffffffbb1a1dd4>] report_size_overflow+0x34/0x40 [ 4.509996] [<ffffffffc01914e2>] btrfs_sync_file+0xa2/0x4d0 [btrfs] [ 4.510000] [<ffffffffbb1d2d6e>] vfs_fsync_range+0x4e/0xd0 [ 4.510007] [<ffffffffbb1d2e5c>] do_fsync+0x3c/0x70 [ 4.510035] [<ffffffffbb1d3155>] SyS_fsync+0x15/0x30 [ 4.510038] [<ffffffffbb5bb5f0>] entry_SYSCALL_64_fastpath+0x12/0x8a [ 4.515412] systemd[1]: systemd-journald.service: Main process exited, code=killed, status=9/KILL [ 4.516041] systemd[1]: Failed to start Journal Service. I’ve also observed that the system was extremely sluggish, even when just typing on the tty, with 100% CPU load even at times when kernel and userland threads only sum up to ~5%. However, based on some previous observations, it’s likely that the load problem was already present in earlier versions. This is a KVM system booted with «pax_nouderef». Additional info: * linux-grsec 4.2.3.201510171833-1 * paxd 30-1 * complete dmesg found in the attachment |
This task depends upon
Comment by PaX Team (paxteam) -
Monday, 19 October 2015, 21:01 GMT
the kvm guest performance issue was fixed already however we're still trying to resolve the btrfs issues as it's not obvious yet what's going on (as in, it's not the usual false positive size overflow reports). for cross-linking: https://forums.grsecurity.net/viewtopic.php?t=4284 and there should be a report on the btrfs list too from another user.
Comment by Emese Revfy (ephox) -
Monday, 19 October 2015, 22:32 GMT
it is same as this bug: https://forums.grsecurity.net/viewtopic.php?f=1&t=4284&p=15617#p15617
Comment by Daniel Micay (thestinger) -
Wednesday, 21 October 2015, 06:18 GMT
This will no longer crash with the latest packages, since the detected overflow is a soft failure for now.
dmesg_full