by Christian Brauner

[permalink] [raw]

Subject: Re: [PATCH 1/2] pid: add pidfd_open()

On Wed, May 15, 2019 at 04:38:58PM +0200, Oleg Nesterov wrote:
> On 05/15, Christian Brauner wrote:
> >
> > +SYSCALL_DEFINE2(pidfd_open, pid_t, pid, unsigned int, flags)
> > +{
> > + int fd, ret;
> > + struct pid *p;
> > + struct task_struct *tsk;
> > +
> > + if (flags)
> > + return -EINVAL;
> > +
> > + if (pid <= 0)
> > + return -EINVAL;
> > +
> > + p = find_get_pid(pid);
> > + if (!p)
> > + return -ESRCH;
> > +
> > + rcu_read_lock();
> > + tsk = pid_task(p, PIDTYPE_PID);
>
> You do not need find_get_pid() before rcu_lock and put_pid() at the end.
> You can just do find_vpid() under rcu_read_lock().

Will do.

>
> > + if (!tsk)
> > + ret = -ESRCH;
> > + else if (unlikely(!thread_group_leader(tsk)))
> > + ret = -EINVAL;
>
> it seems that you can do a single check
>
> tsk = pid_task(p, PIDTYPE_TGID);
> if (!tsk)
> ret = -ESRCH;
>
> this even looks more correct if we race with exec changing the leader.

The logic here being that you can only reach the thread_group leader
from struct pid if PIDTYPE_PID == PIDTYPE_TGID for this struct pid?

Thanks, Oleg.
Christian

2019-05-15 14:53:18

On Thu, May 16, 2019 at 04:03:27PM +0200, Jann Horn wrote:
> On Thu, May 16, 2019 at 3:08 PM Christian Brauner <[email protected]> wrote:
> > On Wed, May 15, 2019 at 10:45:06AM -0700, Daniel Colascione wrote:
> > > On Wed, May 15, 2019 at 3:04 AM Christian Brauner <[email protected]> wrote:
> > > >
> > > > This adds the pidfd_open() syscall. It allows a caller to retrieve pollable
> > > > pidfds for a process which did not get created via CLONE_PIDFD, i.e. for a
> > > > process that is created via traditional fork()/clone() calls that is only
> > > > referenced by a PID:
> [...]
> > > > +/**
> > > > + * pidfd_open() - Open new pid file descriptor.
> > > > + *
> > > > + * @pid: pid for which to retrieve a pidfd
> > > > + * @flags: flags to pass
> > > > + *
> > > > + * This creates a new pid file descriptor with the O_CLOEXEC flag set for
> > > > + * the process identified by @pid. Currently, the process identified by
> > > > + * @pid must be a thread-group leader. This restriction currently exists
> > > > + * for all aspects of pidfds including pidfd creation (CLONE_PIDFD cannot
> > > > + * be used with CLONE_THREAD) and pidfd polling (only supports thread group
> > > > + * leaders).
> > > > + *
> > > > + * Return: On success, a cloexec pidfd is returned.
> > > > + * On error, a negative errno number will be returned.
> > > > + */
> > > > +SYSCALL_DEFINE2(pidfd_open, pid_t, pid, unsigned int, flags)
> > > > +{
> [...]
> > > > + if (pid <= 0)
> > > > + return -EINVAL;
> > >
> > > WDYT of defining pid == 0 to mean "open myself"?
> >
> > I'm torn. It be a nice shortcut of course but pid being 0 is usually an
> > indicator for child processes. So unless the getpid() before
> > pidfd_open() is an issue I'd say let's leave it as is. If you really
> > want the shortcut might -1 be better?
>
> Joining the bikeshed painting club: Please don't allow either 0 or -1
> as shortcut for "self". James Forshaw found an Android security bug a
> while back (https://bugs.chromium.org/p/project-zero/issues/detail?id=727)
> that passed a PID to getpidcon(), except that the PID was 0
> (placeholder for oneway binder transactions), and then the service
> thought it was talking to itself. You could pick some other number and
> provide a #define for that, but I think pidfd_open(getpid(), ...)
> makes more sense.

Yes, I agree. I left it as is for v1, i.e. no shortcut; getpid() should
do.

Christian

2019-05-16 14:55:44

by Aleksa Sarai

[permalink] [raw]

Subject: Re: [PATCH 1/2] pid: add pidfd_open()

On 2019-05-16, Christian Brauner <[email protected]> wrote:
> On Wed, May 15, 2019 at 10:45:06AM -0700, Daniel Colascione wrote:
> > On Wed, May 15, 2019 at 3:04 AM Christian Brauner <[email protected]> wrote:
> > > + if (pid <= 0)
> > > + return -EINVAL;
> >
> > WDYT of defining pid == 0 to mean "open myself"?
>
> I'm torn. It be a nice shortcut of course but pid being 0 is usually an
> indicator for child processes. So unless the getpid() before
> pidfd_open() is an issue I'd say let's leave it as is. If you really
> want the shortcut might -1 be better?

I'd suggest not using negative numbers, and instead reserving them for
PIDTYPE_TGID if we ever want to have that in the future. IMHO, doing

pfd = pidfd_open(getpid(), 0);

is not the end of the world.

--
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Attachments:

(No filename) (921.00 B)
signature.asc (849.00 B)
Download all attachments