On Tue, 11 Jun 2013 18:16:08 +0300 Pekka Enberg <penberg@xxxxxxxxxx> wrote: > On Tue, 11 Jun 2013, Sasha Levin wrote: > >> I think that leaving the warning makes sense to catch similar > >> things which are actually bugs - we had a similar issue with > >> /dev/kmsg (if I remember correctly) which actually pointed to > >> a bug. > > On 6/11/13 6:14 PM, Christoph Lameter wrote: > > Right. Requesting an allocation larger than even supported by the page > > allocator from the slab allocators that are specializing in allocations of > > small objects is usually an indication of a problem in the code. > > So you're OK with going forward with Sasha's patch? Yes please. slab should honour __GFP_NOWARN. __GFP_NOWARN is frequently used by kernel code to probe for "how big an allocation can I get". That's a bit lame, but it's used on slow paths and is pretty simple. In the case of pipe_set_size(), it's userspace who is doing the probing: an application can request a huge pipe buffer and if that fails, try again with a smaller one. It's just wrong to emit a kernel warning in this case. Plus, we've already reported the failure anyway, by returning -ENOMEM from pipe_fcntl(). -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>