rickl...@linux.vnet.ibm.com writes: > From: Rick Lindsley <rickl...@linux.vnet.ibm.com> >
> However, the non-radix path falls through to the old, hashed slice code > (slice_get_unmapped_area, etc.) and these code paths still inspect > task_size. The same attention to addr_limit made in (for example) > radix__arch_get_unmapped_area() should also be applied to (correspondingly) > slice_get_unmapped_area(). I missed this part earlier. I guess that should be fixed in radix code. This came in via fbfef9027c2a7ad9277755509fdb849dbccfe8c1 (powerpc/mm: Switch some TASK_SIZE checks to use mm_context addr_limit). That patch needs update. When we switched from mm->task_size to mm->context.addr_limit in latest version of the patch, we missed updating the above correctly. I have now send a version which should update this correctly. https://lists.ozlabs.org/pipermail/linuxppc-dev/2017-April/156781.html With this we use addr_limit only for slice mask optimization and addr serach limit. All the boundary check is now based on mm->task_size. We will later consolidate TASK_SIZE/mm->task_size/mm->context.addr_limit -aneesh
