lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  PHC 
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Wed, 14 Aug 2019 00:13:59 +0000
Subject: [Bug 203317] WARNING: CPU: 2 PID: 925 at fs/ext4/inode.c:3897

howaboutsynergy is abandoned account everywhere ( changed:

           What    |Removed                     |Added
                 CC|                            |

--- Comment #4 from howaboutsynergy is abandoned account everywhere ( ---
The commit that attempts to fix this(and thus links to this issue) is causing a
system freeze (where sysctl+s,u,b still works though) when attempting to cause
a memory pressure via script `memfreeze`(ran as variant 1) which I used in
order to test
that prevents disk thrashing, but this system freeze happens without that patch

commit aa56a292ce623734ddd30f52d73f527d1f3529b5
    drm/i915/userptr: Acquire the page lock around set_page_dirty()

The freeze seems to be happening right before OOM-killer is about to kill
something, apparently.

I double checked that v5.3-rc4 with this commit reverted does indeed get rid of
the system freeze.

`memfreeze` script is this:



#run this multiple times consecutively, but not at once.

#this should freeze i87k for about 1-2 seconds(actually 5 seconds - try running
top of watch -n1 -d cat /proc/meminfo) while running out of memory temporarily
just before OOM triggers and kills one of the threads:
#^ old comments

#Ensure watchers are running:
ensure_this_runs_already() {
  local cmdtorun="$*"
  if ! pgrep --full "^${cmdtorun}$" >/dev/null; then
    #shellcheck disable=SC2086
    xfce4-terminal -x $cmdtorun  #XXX: unquoted! else it will fail to launch!
ensure_this_runs_already "watch -n.1 -d cat /proc/meminfo"
#ensure_this_runs_already watch -n.1 -d cat /proc/meminfo #yes this works too!
I even tested it.
ensure_this_runs_already "top -d 0.5"
ensure_this_runs_already "sudo iotop -d 5 -o -b"
#exit 1

#XXX tested on i87k host
systemctl --user stop psd #otherwise have to manually rename 2+1 profile dirs
from *-backup to * after the system crashes(and it does with
5.3.0-rc4-gd45331b00ddb kernel, not with 5.2.0 (git) though, or 5.2.4 (stable)
if test "${0##*/}" == "memfreeze2"; then
elif test "${0##*/}" == "memfreeze3"; then
echo "!! Using memfreeze variant '$variant'"

if test "$1" != "keepswap"; then
  echo "! swapoff first"
  sudo swapoff -a
sync #because possibly crash! in fact guaranteed crash in
5.3.0-rc4-gd45331b00ddb even without any le*.patch-es

if test "$variant" == "1"; then
  #this is a remnant from when I've tested this on QubesOS
  alloc="$(awk '/MemAvailable/{printf "%d\n", $2 + 4000;}' < /proc/meminfo)"
  echo "Will alloc: $alloc kB for each of the $threads concurrent threads."
  echo "MemTotal before: $(awk '/MemTotal/{printf "%d kB\n", $2;}' <
  time stress --vm-bytes "${alloc}k" --vm-keep -m $threads --timeout $timeout
  echo "exit code: $?"
  awk '/MemTotal/{printf "MemTotal afterwards: %d kB\n", $2;}' < /proc/meminfo
elif test "$variant" == "2"; then
  time stress -m 220 --vm-bytes 10000000000 --timeout 20
elif test "$variant" == "3"; then
  #XXX say bye bye to Xorg, courtesy of kernel's OOM-killer FIXME: somehow
disable oom-killer or what?!
  time stress -m 3000 --vm-bytes 10M --timeout 50
  echo "!! memfreeze variant not implemented"

You are receiving this mail because:
You are watching the assignee of the bug.

Powered by blists - more mailing lists