Please sign in to comment.
x86-32, mm: Rip out x86_32 NUMA remapping code
This code was an optimization for 32-bit NUMA systems. It has probably been the cause of a number of subtle bugs over the years, although the conditions to excite them would have been hard to trigger. Essentially, we remap part of the kernel linear mapping area, and then sometimes part of that area gets freed back in to the bootmem allocator. If those pages get used by kernel data structures (say mem_map or a dentry), there's no big deal. But, if anyone ever tried to use the linear mapping for these pages _and_ cared about their physical address, bad things happen. For instance, say you passed __GFP_ZERO to the page allocator and then happened to get handed one of these pages, it zero the remapped page, but it would make a pte to the _old_ page. There are probably a hundred other ways that it could screw with things. We don't need to hang on to performance optimizations for these old boxes any more. All my 32-bit NUMA systems are long dead and buried, and I probably had access to more than most people. This code is causing real things to break today: https://lkml.org/lkml/2013/1/9/376 I looked in to actually fixing this, but it requires surgery to way too much brittle code, as well as stuff like per_cpu_ptr_to_phys(). [ hpa: Cc: this for -stable, since it is a memory corruption issue. However, an alternative is to simply mark NUMA as depends BROKEN rather than EXPERIMENTAL in the X86_32 subclause... ] Link: http://lkml.kernel.org/r/20130131005616.1C79F411@kernel.stglabs.ibm.com Signed-off-by: H. Peter Anvin <firstname.lastname@example.org> Cc: <email@example.com>
- Loading branch information...
Showing with 0 additions and 174 deletions.