The "L" AArch64 machine constraint, which we use for the "old" value in
an LL/SC cmpxchg(), generates an immediate that is suitable for a 64-bit
logical instruction. However, for cmpxchg() operations on types smaller
than 64 bits, this constraint can result in an invalid instruction which
is correctly rejected by GAS, such as EOR W1, W1, #0xffffffff.
Whilst we could special-case the constraint based on the cmpxchg size,
it's far easier to change the constraint to "K" and put up with using
a register for large 64-bit immediates. For out-of-line LL/SC atomics,
this is all moot anyway.
Reported-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The CAS instructions implicitly access only the relevant bits of the "old"
argument, so there is no need for explicit masking via type-casting as
there is in the LL/SC implementation.
Move the casting into the LL/SC code and remove it altogether for the LSE
implementation.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Our atomic instructions (either LSE atomics of LDXR/STXR sequences)
natively support byte, half-word, word and double-word memory accesses
so there is no need to mask the data register prior to being stored.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The cmpxchg implementation introduced by commit c342f78217 ("arm64:
cmpxchg: patch in lse instructions when supported by the CPU") performs
an apparently redundant register move of [old] to [oldval] in the
success case - it always uses the same register width as [oldval] was
originally loaded with, and is only executed when [old] and [oldval] are
known to be equal anyway.
The only effect it seemingly does have is to take up a surprising amount
of space in the kernel text, as removing it reveals:
text data bss dec hex filename
12426658 1348614 4499749 18275021 116dacd vmlinux.o.new
12429238 1348614 4499749 18277601 116e4e1 vmlinux.o.old
Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Implement FETCH-OP atomic primitives, these are very similar to the
existing OP-RETURN primitives we already have, except they return the
value of the atomic variable _before_ modification.
This is especially useful for irreversible operations -- such as
bitops (because it becomes impossible to reconstruct the state prior
to modification).
[wildea01: compile fixes for ll/sc]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steve Capper <steve.capper@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arch@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The current arm64 __cmpxchg_double{_mb} implementations carry out the
compare exchange by first comparing the old values passed in to the
values read from the pointer provided and by stashing the cumulative
bitwise difference in a 64-bit register.
By comparing the register content against 0, it is possible to detect if
the values read differ from the old values passed in, so that the compare
exchange detects whether it has to bail out or carry on completing the
operation with the exchange.
Given the current implementation, to detect the cmpxchg operation
status, the __cmpxchg_double{_mb} functions should return the 64-bit
stashed bitwise difference so that the caller can detect cmpxchg failure
by comparing the return value content against 0. The current implementation
declares the return value as an int, which means that the 64-bit
value stashing the bitwise difference is truncated before being
returned to the __cmpxchg_double{_mb} callers, which means that
any bitwise difference present in the top 32 bits goes undetected,
triggering false positives and subsequent kernel failures.
This patch fixes the issue by declaring the arm64 __cmpxchg_double{_mb}
return values as a long, so that the bitwise difference is
properly propagated on failure, restoring the expected behaviour.
Fixes: e9a4b79565 ("arm64: cmpxchg_dbl: patch in lse instructions when supported by the CPU")
Cc: <stable@vger.kernel.org> # 4.3+
Cc: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Commit 654672d4ba ("locking/atomics: Add _{acquire|release|relaxed}()
variants of some atomic operation") introduced a relaxed atomic API to
Linux that maps nicely onto the arm64 memory model, including the new
ARMv8.1 atomic instructions.
This patch hooks up the API to our relaxed atomic instructions, rather
than have them all expand to the full-barrier variants as they do
currently.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The ll/sc __cmpxchg_case_##name assembly mostly uses symbolic names for
operands, but in a single case uses %2 to refer to what is otherwise
known as %[v]. This makes the code more painful to read than is
necessary.
Use %[v] instead.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
If we attempt to atomic64_dec_if_positive on INT_MIN, we will underflow
and incorrectly decide that the original parameter was positive.
This patches fixes the broken condition code so that we handle this
corner case correctly.
Reviewed-by: Steve Capper <steve.capper@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
We don't need duplicate cmpxchg implementations, so use cmpxchg to
implement atomic{,64}_cmpxchg, like we do for xchg already.
Reviewed-by: Steve Capper <steve.capper@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The cost of changing a cacheline from shared to exclusive state can be
significant, especially when this is triggered by an exclusive store,
since it may result in having to retry the transaction.
This patch makes use of prfm to prefetch cachelines for write prior to
ldxr/stxr loops when using the ll/sc atomic routines.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
cmpxchg doesn't require memory barrier semantics when the value
comparison fails, so make the barrier conditional on success.
Reviewed-by: Steve Capper <steve.capper@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
We can perform the cmpxchg comparison using eor and cbnz which avoids
the "cc" clobber for the ll/sc case and consequently for the LSE case
where we may have to fall-back on the ll/sc code at runtime.
Reviewed-by: Steve Capper <steve.capper@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
On CPUs which support the LSE atomic instructions introduced in ARMv8.1,
it makes sense to use them in preference to ll/sc sequences.
This patch introduces runtime patching of our cmpxchg_double primitives
so that the LSE casp instruction is used instead.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
On CPUs which support the LSE atomic instructions introduced in ARMv8.1,
it makes sense to use them in preference to ll/sc sequences.
This patch introduces runtime patching of our cmpxchg primitives so that
the LSE cas instruction is used instead.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
On CPUs which support the LSE atomic instructions introduced in ARMv8.1,
it makes sense to use them in preference to ll/sc sequences.
This patch introduces runtime patching of atomic_t and atomic64_t
routines so that the call-site for the out-of-line ll/sc sequences is
patched with an LSE atomic instruction when we detect that
the CPU supports it.
If binutils is not recent enough to assemble the LSE instructions, then
the ll/sc sequences are inlined as though CONFIG_ARM64_LSE_ATOMICS=n.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In order to patch in the new atomic instructions at runtime, we need to
generate wrappers around the out-of-line exclusive load/store atomics.
This patch adds a new Kconfig option, CONFIG_ARM64_LSE_ATOMICS. which
causes our atomic functions to branch to the out-of-line ll/sc
implementations. To avoid the register spill overhead of the PCS, the
out-of-line functions are compiled with specific compiler flags to
force out-of-line save/restore of any registers that are usually
caller-saved.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In preparation for the Large System Extension (LSE) atomic instructions
introduced by ARM v8.1, move the current exclusive load/store (LL/SC)
atomics into their own header file.
Reviewed-by: Steve Capper <steve.capper@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
