[Bug 1194268] New: disable btrfs asserts in default kernels
https://bugzilla.suse.com/show_bug.cgi?id=1194268 Bug ID: 1194268 Summary: disable btrfs asserts in default kernels Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Kernel Assignee: kernel-bugs@opensuse.org Reporter: dmueller@suse.com QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- while chasing for I/O performance issues in virtualized environments (usecase Open Build Service package builds for example), I noticed that btrfs in our default kernel has asserts enabled. according to the Kconfig help setting, this setting is purely for development purposes. I suggested to move the setting CONFIG_BTRFS_ASSERT only into the debug flavors, however there was concerns that this affects finding memory corruptions and other hardware related issues. On the other side, in my brief testing it appeared disabling reduced code size noticeably (in dual-digit percentage range) and also improved IOPS on a FIO randrw test (on a cow file) by over 20%. I will replicate the tests to be sure. This bug is to track how to address the issue and making asserts purely developer only and disable them in production builds of the openSUSE distribution. -- You are receiving this mail because: You are on the CC list for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1194268 Takashi Iwai <tiwai@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |rgoldwyn@suse.com, | |tiwai@suse.com -- You are receiving this mail because: You are on the CC list for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1194268 Goldwyn Rodrigues <rgoldwyn@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |dsterba@suse.com -- You are receiving this mail because: You are on the CC list for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1194268 https://bugzilla.suse.com/show_bug.cgi?id=1194268#c1 --- Comment #1 from David Sterba <dsterba@suse.com> --- The reason for keeping asserts on is to have checks for memory access to metadata, as the checking function is inside the ASSERT. This can be separated so the asserts can be disabled and the memory checks still in place. Now implemented in the patchset https://lore.kernel.org/linux-btrfs/cover.1643904960.git.dsterba@suse.com/ The first patch is an optimization that should bring back the performance, so that's something to backport regardless of the rest of the series. -- You are receiving this mail because: You are on the CC list for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1194268 https://bugzilla.suse.com/show_bug.cgi?id=1194268#c2 Jiri Slaby <jslaby@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED CC| |jslaby@suse.com Resolution|--- |FIXED --- Comment #2 from Jiri Slaby <jslaby@suse.com> --- commit 234fdd2815ce8fe4da6782109580f3b166aeb97b Author: David Sterba <dsterba@suse.com> Date: Mon Jul 27 20:59:20 2020 +0200 btrfs: remove redundant check in up check_setget_bounds is in v6.0-rc1. SO I think we are done here? Or did you want to backport this to SLE? -- You are receiving this mail because: You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply@suse.com