[lkml]   [1999]   [Oct]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: locking question: do_mmap(), do_munmap()
Alexander Viro wrote:
> On Sun, 10 Oct 1999, Alexander Viro wrote:
> >
> > [Cc'd to mingo]
> >
> > On Sun, 10 Oct 1999, Manfred Spraul wrote:
> >
> > > I've started adding "assert_down()" and "assert_kernellocked()" macros,
> > > and now I don't see the login prompt any more...
> > >
> > > eg. sys_mprotect calls merge_segments without lock_kernel().
> >
> > Manfred, Andrea - please stop it. Yes, it does and yes, it should.

Yes, it should cause oops?

> > Plonking the big lock around every access to VM is _not_ a solution

I never did that, I'll never do that, I only notice that the current
code is filled with races.

> >. If
> > swapper doesn't use mmap_sem - _swapper_ should be fixed. How the hell
> > does lock_kernel() have smaller deadlock potential than
> > down(&mm->mmap_sem)?

lock_kernel() is dropped on thread switch, the semaphore is not dropped.

> OK, folks. Code in swapper (unuse_process(), right?) is called only from
> sys_swapoff(). It's a syscall. Andrea, could you show a scenario for
> deadlock here? OK, some process (but not the process doing swapoff()) may
> have the map locked So? it is not going to release the thing - we are
> seriously screwed anyway (read: we already are in deadlock). We don't hold
> the semaphore ourselves.

AFAIK the problem is OOM:
* a process accesses a not-present, ie page fault:
handle_mm_fault(): this process own mm->mmap_sem.
-> (eg.) do_wp_page().
-> get_free_page().
now get_free_page() notices that there is no free memory.
--> wakeup kswapd.

* the swapper runs, and it tries to swap out data from that process.
mm->mmap_sem is already acquired --> lock-up.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:54    [W:0.080 / U:0.168 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site