Merge tag 'pidfd.v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner...

Pull pidfd updates from Christian Brauner:
 "Various places in the kernel have picked up pidfds.

  The two most recent additions have probably been the ability to use
  pidfds in bpf maps and the usage of pidfds in mm-based syscalls such
  as process_mrelease() and process_madvise().

  The same pattern to turn a pidfd into a struct task exists in two
  places. One of those places used PIDTYPE_TGID while the other one used
  PIDTYPE_PID even though it is clearly documented in all pidfd-helpers
  that pidfds __currently__ only refer to thread-group leaders (subject
  to change in the future if need be).

  This isn't a bug per se but has the potential to be one if we allow
  pidfds to refer to individual threads. If that happens we want to
  audit all codepaths that make use of them to ensure they can deal with
  pidfds refering to individual threads.

  This adds a simple helper to turn a pidfd into a struct task making it
  easy to grep for such places. Plus, it gets rid of code-duplication"

* tag 'pidfd.v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
  mm: use pidfd_get_task()
  pid: add pidfd_get_task() helper

diff --cc mm/oom_kill.c
index 195b366,70d399d..1ddabef
--- 1/mm/oom_kill.c
--- 2/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@@ -1149,8 -1150,7 +1149,7 @@@ SYSCALL_DEFINE2(process_mrelease, int, 
        struct task_struct *task;
        struct task_struct *p;
        unsigned int f_flags;
 -      bool reap = true;
 +      bool reap = false;
-       struct pid *pid;
        long ret = 0;
  
        if (flags)
@@@ -1200,12 -1194,9 +1193,10 @@@
        mmap_read_unlock(mm);
  
  drop_mm:
 -      mmdrop(mm);
 +      if (mm)
 +              mmput(mm);
  put_task:
        put_task_struct(task);
- put_pid:
-       put_pid(pid);
        return ret;
  #else
        return -ENOSYS;