[Devel] review last vz7 patches to port them to vz9

[Konstantin Khorenko]

On 31.12.2021 12:42, Vasily Averin wrote:
> Kirill, Kostja, Nikita, Pavel,
> 
> Please take look at attached vz7 patches to clarify:
> should them ported to vz9 or not?
> 

> db80e99e2409 mm: high order allocation detector

Surely we'll need HOAD in vz9, the review is still on me.
In my TODO list.

> f198023f31c6 fs/ext4: Make "lazytime" mount option by default

We need "lazytime" mount option for both ext4 and xfs,
but let's this time enforce it via userspace => filed a task
https://jira.sw.ru/browse/PSBM-137410

> 6dc84e083c94 pfcache: Fix unitialized s_csum_partial

i've put a note into jira task to port pfcache not to forget about this patch
https://jira.sw.ru/browse/PSBM-77581

> To Kostja: (configs, ub, fuse???)
> 
> affab14d5490 configs: change the default cpufreq governor to "performance"

[root at sb8 sys]# uname -a
Linux sb8.qa.sw.ru 5.14.0-4.vz9.11.32 #1 SMP Mon Dec 6 15:26:34 MSK 2021 x86_64 x86_64 x86_64 GNU/Linux

[root at sb8 sys]# cat /sys/module/cpufreq/parameters/default_governor
performance

[root at sb8 boot]# cat /boot/config-5.14.0-4.vz9.11.32 | grep CPU_FREQ | grep DEF
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

=> the patch is not needed, we already have performance governor by default.

> 2db0f3333215 ub: ub_ratelimit does not initialize ub_ratelimit->lock
> cd015897ebd4 ub: correct _ub_get_css() retvalue for dying/stopped containers
>      (VvS: it fixes /proc/bc/xxx output, is it still actual for vz9?)

beancounters have been dropped, the /proc/bc directory is left intact because systemd detects VZ 
virtualization by its presence.
=> the patches above are not needed.

> fs/fuse kio: share stucked requests count for prometheus stats
> fs/fuse kio: don't use kio with uninitialized statistic
> fs/fuse: a typo resulting in corruption of page cache and kernel crash
> fs/fuse: released handle could be used in fiemap
> fs/fuse: avoid triggering BUG_ON at file close
> fuse: illegal access to file in vstorage
> fuse/pcs: missing initialization of client role

=>
on me TODO:
* HOAD review and db80e99e2409 ("mm: high order allocation detector") in particular
* fuse commits - i'll handle them after merging commits from Acronis team.

--
Konstantin