Re: [PATCH] Document that PF_KTHREAD _is_ ABI
From: Alexey Dobriyan
Date: Sat Mar 20 2021 - 13:43:08 EST
On Sat, Mar 20, 2021 at 10:23:12AM -0700, Andy Lutomirski wrote:
> > On Mar 20, 2021, at 9:31 AM, Alexey Dobriyan <adobriyan@xxxxxxxxx> wrote:
> > PF_KTHREAD value is visible via field number 9 of /proc/*/stat
> > $ sudo cat /proc/2/stat
> > 2 (kthreadd) S 0 0 0 0 -1 2129984 0 ...
> > ^^^^^^^
> > It is used by at least systemd to check for kernel-threadness:
> > https://github.com/systemd/systemd/blob/main/src/basic/process-util.c#L354
> > src/basic/process-util.c:is_kernel_thread()
> Could we fix it differently and more permanently by modifying the proc
> code to display the values systemd expects?
Right now there is no need to fix anything because 4 bits are available.
I put a comment so that PF_KTHREAD won't be moved accidently definitely
> > It means that the value can't be changed despite perceived notion that
> > task_struct flags are internal to kernel and can be shuffled at whim.
> > Formally, _all_ struct task_struct::flags PF_* values are kernel ABI
> > which is a disaster.
> > I hope we can mask everything but few flags and hope for the best :^)
> > +/*
> > + * PF_KTHREAD is part of kernel ABI, visible via value #9 in /proc/$pid/stat
> > + */
> > #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
I think everything should be masked except PF_KTHREAD and maybe few
flags for which known users exist and then don't touch them.
Some flags are clearly internal like PF_MEMALLOC and PF_IDLE.
Some aren't -- PF_FORKNOEXEC. However it is silly for userspace to query it
because programs knows if it forked but didn't exec without external help.