[Bug 219166] ext4 hang when setting echo noop > /sys/block/sda/queue/scheduler

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=219166

--- Comment #6 from Richard W.M. Jones (rjones@redhat.com) ---
Yes I can reproduce this inside a software emulated VM with another 6.11.0
Fedora kernel.  I will bisect this later, but for now reproduction instructions
are given below.

(1) Install a Fedora 40 virtual machine.  I used the command below but other
ways are available:

virt-builder fedora-40 --size=10G --root-password=password:123456

(2) Run the VM in qemu with software emulation (TCG):

qemu-system-x86_64 -machine accel=tcg -cpu qemu64 -m 4096 -drive
file=/var/tmp/fedora-40.qcow2,format=qcow2,if=virtio

(3) Inside the VM, log in as root/123456, install fio, and update the kernel:

dnf install fedora-repos-rawhide
dnf install fio
dnf update kernel
reboot

(should upgrade to 6.11.0 and boot into that kernel).

(4) Inside the VM, in one terminal run:

while true; do echo noop > /sys/block/sda/queue/scheduler 2>/dev/null ; done

(5) Inside the VM, in another terminal run fio with the following config or
similar:

[global]
name=fio-rand-write
filename=/root/fio-rand-write
rw=randwrite
bs=4K
numjobs=4
time_based
runtime=1h
group_reporting

[file1]
size=1G
ioengine=libaio
iodepth=64

(6) After a while the fio process ETA will start counting up (since one or more
threads have got stuck and are not making progress).  Also logging in is
problematic and many common commands like 'dmesg' or 'ps' hang.  I could only
recover by rebooting.

-- 
You may reply to this email to add a comment.

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