[lkml]   [2007]   [Nov]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: msync(2) bug(?), returns AOP_WRITEPAGE_ACTIVATE to userland
On Mon, 5 Nov 2007, Dave Hansen wrote:
> Actually, I think your s/while/if/ change is probably a decent fix.

Any resemblance to a decent fix is purely coincidental.

> Barring any other races, that loop should always have made progress on
> mnt->__mnt_writers the way it is written. If we get to:
> > lock_and_coalesce_cpu_mnt_writer_counts();
> ----------------->HERE
> > mnt_unlock_cpus();
> and don't have a positive mnt->__mnt_writers, we know something is going
> badly. We WARN_ON() there, which should at least give an earlier
> warning that the system is not doing well. But it doesn't fix the
> inevitable. Could you try the attached patch and see if it at least
> warns you earlier?

Thanks, Dave, yes, that gives me a nice warning:

leak detected on mount(c25ebd80) writers count: -65537
WARNING: at fs/namespace.c:249 handle_write_count_underflow()
[<c0103486>] show_trace_log_lvl+0x1b/0x2e
[<c01034b6>] show_trace+0x16/0x1b
[<c0103589>] dump_stack+0x19/0x1e
[<c0171906>] handle_write_count_underflow+0x4c/0x60
[<c0171983>] mnt_drop_write+0x69/0x8e
[<c0160211>] __fput+0xff/0x162
[<c016010d>] fput+0x2e/0x33
[<c01b8f63>] unionfs_file_release+0xc2/0x1c5
[<c01601a1>] __fput+0x8f/0x162
[<c016010d>] fput+0x2e/0x33
[<c015ec9d>] filp_close+0x50/0x5d
[<c015ed1e>] sys_close+0x74/0xb4
[<c01026ce>] sysenter_past_esp+0x5f/0x85

and the test then goes quietly on its way instead of hanging. Though
I imagine, with your patch or mine, that it's then making an unfortunate
frequency of calls to lock_and_coalesce_longer_name_than_I_care_to_type
thereafter. But it's hardly your responsibility to optimize for bugs

The 2.6.23-mm1 tree has MNT_USER at 0x200, so I adjusted your flag to
#define MNT_IMBALANCED_WRITE_COUNT 0x400 /* just for debugging */

> I have a decent guess what the bug is, too. In the unionfs code:

I'll let Erez take it from there...

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

 \ /
  Last update: 2007-11-05 20:07    [W:0.077 / U:8.756 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site