Re: [syzbot] [cgroups?] WARNING in css_rstat_exit

[JP Kobryn]

Thanks Michal and Shakeel for the input on this. I will be sending out a
series to harden css_create(). I was able to open a small window for the
placement of css_rstat_init() that satisfies existing constraints and
allows for error handling before refcounts come into play.

On 7/17/25 11:46 AM, Shakeel Butt wrote:
> On Thu, Jul 17, 2025 at 03:28:27PM +0200, Michal Koutný wrote:
> > Thanks for looking into this JP.
> > You seem to tracked down the cause with uncleaned rstat, beware that the
> > approach in the patch would leave reference imbalance after
> > init_and_link_css() though.
>
> Yeah I discussed the same with JP and I think JP is planning to move the
> css_rstat_init() before init_and_link_css() and a second param to
> css_rstat_init() to differentiate between css_is_self() or not.