xfs
[Top] [All Lists]

RE: open sleeps

To: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Subject: RE: open sleeps
From: Sebastian Brings <sbrings@xxxxxxxxxxxxxx>
Date: Thu, 19 Jun 2008 05:48:28 -0400
Accept-language: en-US
Acceptlanguage: en-US
In-reply-to: <20080619084311.GA16736@xxxxxxxxxxxxx>
References: <4859EE54.6050801@xxxxxxxx> <20080619062118.GY3700@disturbed> <4859FF40.8010206@xxxxxxxx> <20080619084311.GA16736@xxxxxxxxxxxxx>
Sender: xfs-bounce@xxxxxxxxxxx
Thread-index: AcjR6OaTzUVNwjE2QXyam8WFCInnggACGDQQ
Thread-topic: open sleeps

> -----Original Message-----
> From: xfs-bounce@xxxxxxxxxxx [mailto:xfs-bounce@xxxxxxxxxxx]
> On Behalf Of Christoph Hellwig
> Sent: Donnerstag, 19. Juni 2008 10:43
> To: Brian May
> Cc: xfs@xxxxxxxxxxx
> Subject: Re: open sleeps
>
> On Thu, Jun 19, 2008 at 04:40:00PM +1000, Brian May wrote:
> > Does the following help? I still have the logs of the other
> processes, if
> > required (just in case it is some weird interaction between multiple
> > processes?)
> >
> > It seems to be pretty consistent with lock_timer_base,
> every time I look
> > (assuming I haven't read the stack trace upside down...).
> >
> > Jun 19 16:33:30 hq kernel: grep          S 00000000     0
> 12793  12567                     (NOTLB)
> >
> > Jun 19 16:33:30 hq kernel:        f0c23e7c 00200082
> 000a1089 00000000
> > 00000010 00000008 cd0db550 dfa97550
> > Jun 19 16:33:30 hq kernel:        34f84262 00273db2
> 0008a1dc 00000001
> > cd0db660 c20140a0 dfe1cbe8 00200286
> > Jun 19 16:33:30 hq kernel:        c0125380 a4dbf26b
> dfa6a000 00200286
> > 000000ff 00000000 00000000 a4dbf26b
> > Jun 19 16:33:30 hq kernel: Call Trace:
> >
> > Jun 19 16:33:30 hq kernel:  [<c0125380>] lock_timer_base+0x15/0x2f
> >
> > Jun 19 16:33:30 hq kernel:  [<c027f960>] schedule_timeout+0x71/0x8c
> >
> > Jun 19 16:33:30 hq kernel:  [<c0124a81>] process_timeout+0x0/0x5
> >
> > Jun 19 16:33:30 hq kernel:  [<c016c801>] __break_lease+0x2a8/0x2b9
>
> That's the lease breaking code in the VFS, long before we call
> into XFS.  Looks like someone (samba?) has a least on this file and
> we're having trouble having it broken.  Try sending a report about
> this to linux-fsdevel@xxxxxxxxxxxxxxx
>
>

For curiosity: Is the lease breaking you mention breaking a kernel oplock samba 
may hold?



<Prev in Thread] Current Thread [Next in Thread>