Dear Tejun and cgroups mailing list users.
During validation of patches for cgoups I've noticed strange behaviour in cgroup v2
dd with bs=1 reads each 2nd line of cgroup.proc and cgroup.threads file
on Fedora 31 node
[test@localhost ~]$ dd if=/sys/fs/cgroup/cgroup.procs bs=1 | wc
195+0 records in
195+0 records out
195 bytes copied, 0,000149534 s, 1,3 MB/s
54 54 195
VvS: it reads 54 lines
[test@localhost ~]$ dd if=/sys/fs/cgroup/cgroup.procs bs=1000 | wc
0+1 records in
0+1 records out
389 bytes copied, 6,6956e-05 s, 5,8 MB/s
108 108 389
following one reads twice more, 108 lines.
[test@localhost ~]$ dd if=/sys/fs/cgroup/cgroup.procs bs=1 | wc
195+0 records in
195+0 records out
195 bytes copied, 0,000123775 s, 1,6 MB/s
54 54 195
following dd bs=1 reads 54 lines again
[test@localhost ~]$ uname -a
Linux localhost.localdomain 5.5.0-rc6-00150-g82efece #7 SMP Fri Jan 24 12:05:04 MSK 2020 x86_64 x86_64 x86_64 GNU/Linux
[test@localhost ~]$ mount | grep cgroup
cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,nsdelegate)
I prepared patches for .next seq_file funictions.
https://bugzilla.kernel.org/show_bug.cgi?id=206283
during its validation I've noticed described problem,
and did not found its reason yet.
Thank you,
Vasily Averin
