On 02/02/18 20:43, Christopher Lameter wrote: > On Thu, 1 Feb 2018, Igor Stoppa wrote: > >>> Would it not be better to use compound page allocations here? [...] > Ok its compound_head(). See also the use in the SLAB and SLUB allocator. > >> During hardened user copy permission check, I need to confirm if the >> memory range that would be exposed to userspace is a legitimate >> sub-range of a pmalloc allocation. > > If you save the size in the head page struct then you could do that pretty > fast. Ok, now I get what you mean. But it doesn't seem to fit the intended use case, for other reasons (maybe the same, from 2 different POV): - compound pages are aggregates of regular pages, in numbers that are powers of 2, while the amount of pages to allocate is not known upfront. One *could* give a hint to pmalloc about how many pages to allocate every time there is a need to grow the pool. Iow it would be the size of a chunk. But I'm afraid the granularity would still be pretty low, so maybe it would be 2-4 times less. - the property of the compound page will affect the property of all the pages in the compound, so when one is write protected, it can generate a lot of wasted memory, if there is too much slack (because of the order) With vmalloc, I can allocate any number of pages, minimizing the waste. Finally, there was a discussion about optimization: http://www.openwall.com/lists/kernel-hardening/2017/08/07/2 The patch I sent does indeed take advantage of the new information, not just for pmalloc use. I have not measured if/where/what there is gain, but it does look like the extra info can be exploited also elsewhere. -- igor -- 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>
