copy_page() on a kmalloc-ed page with DEBUG_SLAB enabled (was "zram: do not use copy_page with non-page alinged address")

[Sergey Senozhatsky]

Hello,

I'll fork it into a separate thread and Cc more MM people.
sorry for top-posting.

Minchan reported that doing copy_page() on a kmalloc(PAGE_SIZE) page
with DEBUG_SLAB enabled can cause a memory corruption (See below or
lkml.kernel.org/r/1492042622-12074-2-git-send-email-minchan@xxxxxxxxxx )

that's an interesting problem. arm64 copy_page(), for instance, wants src
and dst to be page aligned, which is reasonable, while generic copy_page(),
on the contrary, simply does memcpy(). there are, probably, other callpaths
that do copy_page() on kmalloc-ed pages and I'm wondering if there is some
sort of a generic fix to the problem.

> > On (04/13/17 09:17), Minchan Kim wrote:
> > > The copy_page is optimized memcpy for page-alinged address.
> > > If it is used with non-page aligned address, it can corrupt memory which
> > > means system corruption. With zram, it can happen with
> > > 
> > > 1. 64K architecture
> > > 2. partial IO
> > > 3. slub debug
> > > 
> > > Partial IO need to allocate a page and zram allocates it via kmalloc.
> > > With slub debug, kmalloc(PAGE_SIZE) doesn't return page-size aligned
> > > address. And finally, copy_page(mem, cmem) corrupts memory.
> > 
> > which would be the case for many other copy_page() calls in the kernel.
> > right? if so - should the fix be in copy_page() then?
> 
> I thought about it but was not sure it's good idea by several reasons
> (but don't want to discuss it in this thread).
> 
> Anyway, it's stable stuff so I don't want to make the patch bloat.
> If you believe it is right direction and valuable, you could be
> a volunteer. :)

	-ss