[PATCHv2 00/10] ACCESS_ONCE and non-scalar accesses

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



As discussed on LKML http://marc.info/?i=54611D86.4040306%40de.ibm.com
ACCESS_ONCE might fail with specific compiler for non-scalar accesses.

Here is a set of patches to tackle that problem.

The first patch is already in kvm/next and this series is against rc3 (as
kvm/next so that we can avoid a merge conflict as soon as this series
has stabilized).
The 2nd patch introduces READ_ONCE/ASSIGN_ONCE as suggested by Linus.
The 2nd to last patch will force ACCESS_ONCE to error-out if it is used
on non-scalar accesses.

I have cross-compiled the resulting kernel with defconfig for
microblaze, m68k, alpha, s390,x86_64, i686, sparc, sparc64, mips,
ia64, arm and arm64.
Will Deacon pointed me to the right defconfig for arm32 to also trigger
a finding here.
Runtime testing was only done on s390x.

There is a small problem left with sparc (32bit) and m68k:

mm/rmap.c: In function 'mm_find_pmd':
include/linux/compiler.h:220:72: warning: '__val' may be used uninitialized in this function [-Wmaybe-uninitialized]
       ({ typeof(p) __val; __read_once_size(&p, &__val, sizeof(__val)); __val; })
                                                                        ^
include/linux/compiler.h:220:20: note: '__val' was declared here
       ({ typeof(p) __val; __read_once_size(&p, &__val, sizeof(__val)); __val; })
                    ^
mm/rmap.c:584:9: note: in expansion of macro 'READ_ONCE'
  pmde = READ_ONCE(*pmd);

Reason is that for both architectures pmd_t is long[16]. WTF?

So the next spin will either fix m68k/sparc or use a barrier + ACCESS_ONCE.

Comments?

Christian Borntraeger (10):
  KVM: s390: Fix ipte locking
  kernel: Provide READ_ONCE and ASSIGN_ONCE
  mm: replace ACCESS_ONCE with READ_ONCE
  x86/spinlock: Replace ACCESS_ONCE with READ_ONCE/ASSIGN_ONCE
  x86: Replace ACCESS_ONCE in gup with READ_ONCE
  mips: Replace ACCESS_ONCE in gup with READ_ONCE
  arm64: Replace ACCESS_ONCE for spinlock code with READ_ONCE
  arm: Replace ACCESS_ONCE for spinlock code with READ_ONCE
  tighten rules for ACCESS ONCE
  KVM: s390: change ipte lock from barrier to READ_ONCE

 arch/arm/include/asm/spinlock.h   |  4 ++--
 arch/arm64/include/asm/spinlock.h |  4 ++--
 arch/mips/mm/gup.c                |  2 +-
 arch/s390/kvm/gaccess.c           | 14 +++++++------
 arch/x86/include/asm/spinlock.h   |  8 +++----
 arch/x86/mm/gup.c                 |  2 +-
 include/linux/compiler.h          | 44 ++++++++++++++++++++++++++++++++++++++-
 mm/gup.c                          |  2 +-
 mm/memory.c                       |  2 +-
 mm/rmap.c                         |  2 +-
 10 files changed, 64 insertions(+), 20 deletions(-)


--
To unsubscribe from this list: send the line "unsubscribe linux-x86_64" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux ia64]     [Linux Kernel]     [DCCP]     [Linux ARM]     [Yosemite News]     [Linux SCSI]     [Linux Hams]
  Powered by Linux