RE: [PATCH 19/24] sunrpc: Remove unnecessary OOM logging messages

[Myklebust, Trond]

> -----Original Message-----
> From: Boaz Harrosh [mailto:bharrosh@xxxxxxxxxxx]
> Sent: Monday, August 29, 2011 5:55 PM
> To: David Miller
> Cc: Myklebust, Trond; joe@xxxxxxxxxxx; bfields@xxxxxxxxxxxx;
> neilb@xxxxxxx; linux-nfs@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 19/24] sunrpc: Remove unnecessary OOM logging
> messages
> 
> On 08/29/2011 02:37 PM, David Miller wrote:
> > From: "Myklebust, Trond" <Trond.Myklebust@xxxxxxxxxx>
> > Date: Mon, 29 Aug 2011 14:36:17 -0700
> >
> >> Big NACK...
> >>
> >> By whose standard are those "not useful"?
> >
> > By mine, that's for sure.  It's duplicating something that the
> allocation
> > layers are already going to print.
> 
> I have a question about that. Are the dprints going to show the stack
> backtrace?
> Otherwise how can I see which exact allocation failed and was not
> properly handled?
> 
> If yes above? then I'm not sure I like it either, because am I'll be
> getting a full
> stack backtrace for every failed allocation?
> 
> But I might like it if I try. How do I turn on allocation failures
> prints?
> Can I filter out to print only GFP_KERNEL failures and or other GFP
> combinations?

Right. If every GFP_ATOMIC or GFP_NOWAIT is going to print out stack traces, then we're heading for absolute insanity. If not, then the existing dprintk()s make a lot more sense, 'cos they are turned on only when the administrator notices a problem, and is trying to debug it.

Trond

N‹§²æìr¸›yúèšØb²X¬¶ÇvØ^–)Þ{.nÇ+‰·¥Š{±‘êçzX§¶›¡Ü}©ž²ÆzÚ&j:+v‰¨¾«‘êçzZ+€Ê+zf£¢·hšˆ§~†­†Ûiÿûàz¹®w¥¢¸?™¨è­Ú&¢)ßf”ù^jÇy§m…á@A«a¶Úÿ0¶ìh®å’i