Jean-Yves Migeon <jeanyves.mig...@free.fr> wrote: > On 18.04.2011 10:35, Mindaugas Rasiukevicius wrote: > > We used to check the return of big size allocations, when kmem(9) could > > fail even with KM_SLEEP due to KVA starvation. However, pretty much all > > kernel does not perform checks for smaller allocations and since the bug > > was fixed - we are no longer checking for big ones as well. > > > > This applies to all allocators. > > So, any thread sleeping for an allocation cannot be interrupted? >
Nope. However, for nearly all cases there is no need for that. For the previous reserve_pages() case, pointed out by Cherry, code should rather be restructured to pre-allocate memory before acquiring the locks. That would have avoided the problem. -- Mindaugas
