[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <alpine.DEB.2.00.0906181212380.10979@chino.kir.corp.google.com>
Date: Thu, 18 Jun 2009 12:23:28 -0700 (PDT)
From: David Rientjes <rientjes@...gle.com>
To: David Miller <davem@...emloft.net>
cc: eric.dumazet@...il.com, jpiszcz@...idpixels.com,
linux-kernel@...r.kernel.org
Subject: Re: [patch] ipv4: don't warn about skb ack allocation failures
On Thu, 18 Jun 2009, David Miller wrote:
> > I disagree, page allocation failure messages show vital information about
> > the state of the VM so that we can find bugs and GFP_ATOMIC allocations
> > are the most common trigger for these diagnostic messages since
> > __GFP_WAIT allocations can trigger direct reclaim (and __GFP_FS
> > allocations can trigger the oom killer) to free memory and will retry the
> > allocation if ~__GFP_NORETRY.
>
> It's COMPLETELY and ABSOLUTELY normal for GFP_ATOMIC allocations to
> fail in the networking.
>
__GFP_NOWARN exists for that reason.
> If you warn it will just spam the logs, and on a router forwarding
> millions of packets per second are you sure that can ever be sane?
>
The spamming is ratelimited, but GFP_ATOMIC is really the only time we get
such diagnostic information since __GFP_WAIT allocations can reclaim,
__GFP_FS allocations can utilize the oom killer, and other order-0
allocations are implicitly ~__GFP_NORETRY.
As previously mentioned, GFP_ATOMIC allocations that are not __GFP_NOWARN
have been emitting these diagnostics since 2.5.53. This has been on your
TODO list for 6 1/2 years and now you insist all GFP_ATOMIC allocations
change their default behavior?
I understand what you're trying to avoid, but I disagree with the approach
of altering the default behavior of GFP_ATOMIC. I may suggest that
emitting the page allocation failures become a compile time option;
CONFIG_DEBUG_VM would be my suggestion.
> Use statistics and tracing if necessary, but log spam no way...
>
You need the meminfo that is emitted at the time of failure for it to be
useful.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists