Re: [PATCH 1/1] mm: introduce process_reap system call

* Suren Baghdasaryan:

On Wed, Jul 7, 2021 at 2:47 AM Florian Weimer [off-list ref] wrote:

quoted

* Suren Baghdasaryan:

quoted

The API is as follows,

          int process_reap(int pidfd, unsigned int flags);

        DESCRIPTION
          The process_reap() system call is used to free the memory of a
          dying process.

          The pidfd selects the process referred to by the PID file
          descriptor.
          (See pidofd_open(2) for further information)

          The flags argument is reserved for future use; currently, this
          argument must be specified as 0.

        RETURN VALUE
          On success, process_reap() returns 0. On error, -1 is returned
          and errno is set to indicate the error.

I think the manual page should mention what it means for a process to be
“dying”, and how to move a process to this state.

Thanks for the suggestion, Florian! Would replacing "dying process"
with "process which was sent a SIGKILL signal" be sufficient?

That explains very clearly the requirement, but it raises the question
why this isn't an si_code flag for rt_sigqueueinfo, reusing the existing
system call.

Thanks,
Florian

On Wed, Jul 7, 2021 at 10:41 PM Florian Weimer [off-list ref] wrote:

* Suren Baghdasaryan:

quoted

On Wed, Jul 7, 2021 at 2:47 AM Florian Weimer [off-list ref] wrote:

quoted

* Suren Baghdasaryan:

quoted

The API is as follows,

          int process_reap(int pidfd, unsigned int flags);

        DESCRIPTION
          The process_reap() system call is used to free the memory of a
          dying process.

          The pidfd selects the process referred to by the PID file
          descriptor.
          (See pidofd_open(2) for further information)

          The flags argument is reserved for future use; currently, this
          argument must be specified as 0.

        RETURN VALUE
          On success, process_reap() returns 0. On error, -1 is returned
          and errno is set to indicate the error.

I think the manual page should mention what it means for a process to be
“dying”, and how to move a process to this state.

Thanks for the suggestion, Florian! Would replacing "dying process"
with "process which was sent a SIGKILL signal" be sufficient?

That explains very clearly the requirement, but it raises the question
why this isn't an si_code flag for rt_sigqueueinfo, reusing the existing
system call.

I think you are suggesting to use sigqueue() to deliver the signal and
perform the reaping when a special value accompanies it. This would be
somewhat similar to my early suggestion to use a flag in
pidfd_send_signal() (see:
https://lore.kernel.org/patchwork/patch/1060407) to implement memory
reaping which has another advantage of operation on PIDFDs instead of
PIDs which can be recycled.
kill()/pidfd_send_signal()/sigqueue() are supposed to deliver the
signal and return without blocking. Changing that behavior was
considered unacceptable in these discussions. On the other hand using
some kthread to do the reaping asynchronously has its disadvantages:
userspace can't control the priority and cpu affinity of the thread
doing the reaping and this work is not charged towards the caller. In
the end a separate blocking syscall was deemed appropriate for this
operation. More details can be found in the links I posted in the
description of the patch.

Thanks,
Florian

On Thursday 2021-07-08 08:05, Suren Baghdasaryan wrote:

quoted

That explains very clearly the requirement, but it raises the question
why this isn't an si_code flag for rt_sigqueueinfo, reusing the existing
system call.

I think you are suggesting to use sigqueue() to deliver the signal and
perform the reaping when a special value accompanies it. This would be
somewhat similar to my early suggestion to use a flag in
pidfd_send_signal() (see:
https://lore.kernel.org/patchwork/patch/1060407) to implement memory
reaping which has another advantage of operation on PIDFDs instead of
PIDs which can be recycled.
kill()/pidfd_send_signal()/sigqueue() are supposed to deliver the
signal and return without blocking. Changing that behavior was
considered unacceptable in these discussions.

The way I understood the request is that a userspace program (or perhaps two,
if so desired) should issue _two_ calls, one to deliver the signal,
one to perform the reap portion:

	uinfo.si_code = SI_QUEUE;
	sigqueue(pid, SIGKILL, &uinfo);
	uinfo.si_code = SI_REAP;
	sigqueue(pid, SIGKILL, &uinfo);

On Mon, Jul 12, 2021 at 5:51 AM Jan Engelhardt [off-list ref] wrote:

On Thursday 2021-07-08 08:05, Suren Baghdasaryan wrote:

quoted

quoted

That explains very clearly the requirement, but it raises the question
why this isn't an si_code flag for rt_sigqueueinfo, reusing the existing
system call.

I think you are suggesting to use sigqueue() to deliver the signal and
perform the reaping when a special value accompanies it. This would be
somewhat similar to my early suggestion to use a flag in
pidfd_send_signal() (see:
https://lore.kernel.org/patchwork/patch/1060407) to implement memory
reaping which has another advantage of operation on PIDFDs instead of
PIDs which can be recycled.
kill()/pidfd_send_signal()/sigqueue() are supposed to deliver the
signal and return without blocking. Changing that behavior was
considered unacceptable in these discussions.

The way I understood the request is that a userspace program (or perhaps two,
if so desired) should issue _two_ calls, one to deliver the signal,
one to perform the reap portion:

        uinfo.si_code = SI_QUEUE;
        sigqueue(pid, SIGKILL, &uinfo);
        uinfo.si_code = SI_REAP;
        sigqueue(pid, SIGKILL, &uinfo);

This approach would still lead to the same discussion: by design,
sigqueue/kill/pidfd_send_signal deliver the signal but do not wait for
the signal to be processed by the recipient. Changing that would be a
behavior change. Therefore we would have to follow this pattern and
implement memory reaping in an asynchronous manner using a
kthread/workqueue and it won't be done in the context of the calling
process. This is undesirable because we lose the ability to control
priority and cpu affinity for this operation and work won't be charged
to the caller.
That's why the proposed syscall performs memory reaping in the
caller's context and blocks until the operation is done. In this
proposal, your sequence looks like this:

pidfd_send_signal(pidfd, SIGKILL, NULL, 0);
process_reap(pidfd, 0);

except we decided to rename process_reap() to process_mrelease() in
the next revision.

On Monday 2021-07-12 20:39, Suren Baghdasaryan wrote:

quoted

The way I understood the request is that a userspace program (or perhaps two,
if so desired) should issue _two_ calls, one to deliver the signal,
one to perform the reap portion:

        uinfo.si_code = SI_QUEUE;
        sigqueue(pid, SIGKILL, &uinfo);
        uinfo.si_code = SI_REAP;
        sigqueue(pid, SIGKILL, &uinfo);

This approach would still lead to the same discussion: by design,
sigqueue/kill/pidfd_send_signal deliver the signal but do not wait for
the signal to be processed by the recipient.

Oh, so the only reason not to do that is because there is some POSIX
specification that says the sigqueue API should be non-waiting for all
possible parameter values (with an implied "present and future
values!"), not because there's some hurdle to actually add a wait
inside within rt_sigqueueinfo if the REAP flag is set.
Gotcha.