On Mon, 15 Feb 2010, KAMEZAWA Hiroyuki wrote: > > I can't agree with that assessment, I don't think it's a desired result to > > ever panic the machine regardless of what /proc/sys/vm/panic_on_oom is set > > to because a lowmem page allocation fails especially considering, as > > mentioned in the changelog, these allocations are never __GFP_NOFAIL and > > returning NULL is acceptable. > > > please add > WARN_ON((high_zoneidx < ZONE_NORMAL) && (gfp_mask & __GFP_NOFAIL)) > somewhere. Then, it seems your patch makes sense. > high_zoneidx < ZONE_NORMAL is not the only case where this exists: it exists for __GFP_NOFAIL allocations that are not __GFP_FS as well and has for years, no special handling is now needed. There should be no cases of either (GFP_DMA | __GFP_NOFAIL, or GFP_NOFS | __GFP_NOFAIL) in my audit of the kernel code. And since __GFP_NOFAIL is not to be added anymore (see Andrew's dab48dab), there's no real reason to add a WARN_ON() here. > I don't like the "possibility" of inifinte loops. > The possibility of infinite loops has always existed in the page allocator for __GFP_NOFAIL allocations, that's precisely why it's deprecated and eventually we seek to remove it. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>