On 06/14/2018 02:28 AM, Ram Pai wrote:
Assortment of fixes to pkey.Patch 1 makes pkey consumable in multithreaded applications. Patch 2 fixes fork behavior to inherit the key attributes. Patch 3 A off-by-one bug made one key unusable. Fixes it. Patch 4 Execute-only key is preallocated. Patch 5 Makes pkey-0 less special. Patch 6 Deny by default permissions on all unallocated keys. Passes all selftests on powerpc. Also behavior verified to be correct by Florian. Changelog: v2: . fixed merge conflict with upstream code. . Add patch 6. Makes the behavior consistent with that on x86.
(Except signal handling, but I agree with Ram that the POWER behavior is the correct one.)
Ram Pai (6): powerpc/pkeys: Enable all user-allocatable pkeys at init. powerpc/pkeys: Save the pkey registers before fork powerpc/pkeys: fix calculation of total pkeys. powerpc/pkeys: Preallocate execute-only key powerpc/pkeys: make protection key 0 less special powerpc/pkeys: Deny read/write/execute by default
I tested the whole series with the new selftests, with the printamr.c program I posted earlier, and the glibc test for pkey_alloc &c. The latter required some test fixes, but now passes as well. As far as I can tell, everything looks good now.
Tested-By: Florian Weimer <fwei...@redhat.com> Thanks, Florian
