linux.git/arch, branch v4.10-rc2

mm: optimize PageWaiters bit use for unlock_page()

2016-12-29T19:03:15+00:00

In commit 62906027091f ("mm: add PageWaiters indicating tasks are
waiting for a page bit") Nick Piggin made our page locking no longer
unconditionally touch the hashed page waitqueue, which not only helps
performance in general, but is particularly helpful on NUMA machines
where the hashed wait queues can bounce around a lot.

However, the "clear lock bit atomically and then test the waiters bit"
sequence turns out to be much more expensive than it needs to be,
because you get a nasty stall when trying to access the same word that
just got updated atomically.

On architectures where locking is done with LL/SC, this would be trivial
to fix with a new primitive that clears one bit and tests another
atomically, but that ends up not working on x86, where the only atomic
operations that return the result end up being cmpxchg and xadd.  The
atomic bit operations return the old value of the same bit we changed,
not the value of an unrelated bit.

On x86, we could put the lock bit in the high bit of the byte, and use
"xadd" with that bit (where the overflow ends up not touching other
bits), and look at the other bits of the result.  However, an even
simpler model is to just use a regular atomic "and" to clear the lock
bit, and then the sign bit in eflags will indicate the resulting state
of the unrelated bit #7.

So by moving the PageWaiters bit up to bit #7, we can atomically clear
the lock bit and test the waiters bit on x86 too.  And architectures
with LL/SC (which is all the usual RISC suspects), the particular bit
doesn't matter, so they are fine with this approach too.

This avoids the extra access to the same atomic word, and thus avoids
the costly stall at page unlock time.

The only downside is that the interface ends up being a bit odd and
specialized: clear a bit in a byte, and test the sign bit.  Nick doesn't
love the resulting name of the new primitive, but I'd rather make the
name be descriptive and very clear about the limitation imposed by
trying to work across all relevant architectures than make it be some
generic thing that doesn't make the odd semantics explicit.

So this introduces the new architecture primitive

    clear_bit_unlock_is_negative_byte();

and adds the trivial implementation for x86.  We have a generic
non-optimized fallback (that just does a "clear_bit()"+"test_bit(7)"
combination) which can be overridden by any architecture that can do
better.  According to Nick, Power has the same hickup x86 has, for
example, but some other architectures may not even care.

All these optimizations mean that my page locking stress-test (which is
just executing a lot of small short-lived shell scripts: "make test" in
the git source tree) no longer makes our page locking look horribly bad.
Before all these optimizations, just the unlock_page() costs were just
over 3% of all CPU overhead on "make test".  After this, it's down to
0.66%, so just a quarter of the cost it used to be.

(The difference on NUMA is bigger, but there this micro-optimization is
likely less noticeable, since the big issue on NUMA was not the accesses
to 'struct page', but the waitqueue accesses that were already removed
by Nick's earlier commit).

Acked-by: Nick Piggin 
Cc: Dave Hansen 
Cc: Bob Peterson 
Cc: Steven Whitehouse 
Cc: Andrew Lutomirski 
Cc: Andreas Gruenbacher 
Cc: Peter Zijlstra 
Cc: Mel Gorman 
Signed-off-by: Linus Torvalds

x86/mce/AMD: Make the init code more robust

2016-12-27T01:30:24+00:00

If mce_device_init() fails then the mce device pointer is NULL and the
AMD mce code happily dereferences it.

Add a sanity check.

Reported-by: Markus Trippelsdorf 
Reported-by: Boris Ostrovsky 
Signed-off-by: Thomas Gleixner 
Signed-off-by: Linus Torvalds

arm64: don't pull uaccess.h into *.S

2016-12-26T18:05:17+00:00

Split asm-only parts of arm64 uaccess.h into a new header and use that
from *.S.

Signed-off-by: Al Viro

powerpc: Fix build warning on 32-bit PPC

2016-12-26T00:12:20+00:00

I am getting the following warning when I build kernel 4.9-git on my
PowerBook G4 with a 32-bit PPC processor:

    AS      arch/powerpc/kernel/misc_32.o
  arch/powerpc/kernel/misc_32.S:299:7: warning: "CONFIG_FSL_BOOKE" is not defined [-Wundef]

This problem is evident after commit 989cea5c14be ("kbuild: prevent
lib-ksyms.o rebuilds"); however, this change in kbuild only exposes an
error that has been in the code since 2005 when this source file was
created.  That was with commit 9994a33865f4 ("powerpc: Introduce
entry_{32,64}.S, misc_{32,64}.S, systbl.S").

The offending line does not make a lot of sense.  This error does not
seem to cause any errors in the executable, thus I am not recommending
that it be applied to any stable versions.

Thanks to Nicholas Piggin for suggesting this solution.

Fixes: 9994a33865f4 ("powerpc: Introduce entry_{32,64}.S, misc_{32,64}.S, systbl.S")
Signed-off-by: Larry Finger 
Cc: Nicholas Piggin 
Cc: Benjamin Herrenschmidt 
Cc: Paul Mackerras 
Cc: Michael Ellerman 
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Linus Torvalds

Merge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2016-12-25T22:30:04+00:00

Pull timer type cleanups from Thomas Gleixner:
 "This series does a tree wide cleanup of types related to
  timers/timekeeping.

   - Get rid of cycles_t and use a plain u64. The type is not really
     helpful and caused more confusion than clarity

   - Get rid of the ktime union. The union has become useless as we use
     the scalar nanoseconds storage unconditionally now. The 32bit
     timespec alike storage got removed due to the Y2038 limitations
     some time ago.

     That leaves the odd union access around for no reason. Clean it up.

  Both changes have been done with coccinelle and a small amount of
  manual mopping up"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ktime: Get rid of ktime_equal()
  ktime: Cleanup ktime_set() usage
  ktime: Get rid of the union
  clocksource: Use a plain u64 instead of cycle_t

Merge branch 'smp-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2016-12-25T22:05:56+00:00

Pull SMP hotplug notifier removal from Thomas Gleixner:
 "This is the final cleanup of the hotplug notifier infrastructure. The
  series has been reintgrated in the last two days because there came a
  new driver using the old infrastructure via the SCSI tree.

  Summary:

   - convert the last leftover drivers utilizing notifiers

   - fixup for a completely broken hotplug user

   - prevent setup of already used states

   - removal of the notifiers

   - treewide cleanup of hotplug state names

   - consolidation of state space

  There is a sphinx based documentation pending, but that needs review
  from the documentation folks"

* 'smp-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/armada-xp: Consolidate hotplug state space
  irqchip/gic: Consolidate hotplug state space
  coresight/etm3/4x: Consolidate hotplug state space
  cpu/hotplug: Cleanup state names
  cpu/hotplug: Remove obsolete cpu hotplug register/unregister functions
  staging/lustre/libcfs: Convert to hotplug state machine
  scsi/bnx2i: Convert to hotplug state machine
  scsi/bnx2fc: Convert to hotplug state machine
  cpu/hotplug: Prevent overwriting of callbacks
  x86/msr: Remove bogus cleanup from the error path
  bus: arm-ccn: Prevent hotplug callback leak
  perf/x86/intel/cstate: Prevent hotplug callback leak
  ARM/imx/mmcd: Fix broken cpu hotplug handling
  scsi: qedi: Convert to hotplug state machine

ktime: Cleanup ktime_set() usage

2016-12-25T16:21:22+00:00

ktime_set(S,N) was required for the timespec storage type and is still
useful for situations where a Seconds and Nanoseconds part of a time value
needs to be converted. For anything where the Seconds argument is 0, this
is pointless and can be replaced with a simple assignment.

Signed-off-by: Thomas Gleixner 
Cc: Peter Zijlstra

clocksource: Use a plain u64 instead of cycle_t

2016-12-25T10:04:12+00:00

There is no point in having an extra type for extra confusion. u64 is
unambiguous.

Conversion was done with the following coccinelle script:

@rem@
@@
-typedef u64 cycle_t;

@fix@
typedef cycle_t;
@@
-cycle_t
+u64

Signed-off-by: Thomas Gleixner 
Cc: Peter Zijlstra 
Cc: John Stultz

cpu/hotplug: Cleanup state names

2016-12-25T09:47:44+00:00

When the state names got added a script was used to add the extra argument
to the calls. The script basically converted the state constant to a
string, but the cleanup to convert these strings into meaningful ones did
not happen.

Replace all the useless strings with 'subsys/xxx/yyy:state' strings which
are used in all the other places already.

Signed-off-by: Thomas Gleixner 
Cc: Peter Zijlstra 
Cc: Sebastian Siewior 
Link: http://lkml.kernel.org/r/20161221192112.085444152@linutronix.de
Signed-off-by: Thomas Gleixner

x86/msr: Remove bogus cleanup from the error path

2016-12-25T09:47:41+00:00

The error cleanup which is invoked when the hotplug state setup failed
tries to remove the failed state, which is broken.

Fixes: 8fba38c937cd ("x86/msr: Convert to hotplug state machine")
Reported-by: kernel test robot 
Signed-off-by: Thomas Gleixner 
Cc: Sebastian Siewior