Re: [ANOMALIES]: 2.4.2 - __alloc_pages: failed - Causes more then just msgs

[Shawn Starr <spstarr@sh0n.net>]

It may not be an important message but what does happen is /dev/dsp becomes
hung and no sound works after the fault. So something is definately wrong.

Shawn.

Marcelo Tosatti wrote:

> On Mon, 26 Feb 2001, Alan Cox wrote:
>
> > > We can add an allocation flag (__GFP_NO_CRITICAL?) which can be used by
> > > sg_low_malloc() (and other non critical allocations) to fail previously
> > > and not print the message.
> >
> > It is just for debugging. The message can go. If anytbing it would be more
> > useful to tack Failed alloc data on the end of /proc/slabinfo
>
> The issue is not the warn message.
>
> Non critical allocations (such as this case of sg_low_malloc()) are trying
> to get additional memory to optimize things -- we want the allocator to be
> lazy and fail previously instead doing hard work. If kswapd cannot keep up
> with the memory pressure, we're surely in a memory shortage state.
>
> Its better to get out of the memory shortage instead running into OOM
> because of some optimization, I guess.
>
> Another example of such a flag is swapin readahead.

--
Hugged a Tux today? (tm)