diff options
author | Vineet Gupta <vgupta@synopsys.com> | 2014-11-13 10:24:01 (GMT) |
---|---|---|
committer | Vineet Gupta <vgupta@synopsys.com> | 2015-06-25 00:29:23 (GMT) |
commit | d57f727264f1425a94689bafc7e99e502cb135b5 (patch) | |
tree | c1a73c49583de1880c14be29b09a27344f695a45 /arch/arc/include | |
parent | eaf0ecc33f82b9c46528d1646575dd8caf586a3d (diff) | |
download | linux-d57f727264f1425a94689bafc7e99e502cb135b5.tar.xz |
ARC: add compiler barrier to LLSC based cmpxchg
When auditing cmpxchg call sites, Chuck noted that gcc was optimizing
away some of the desired LDs.
| do {
| new = old = *ipi_data_ptr;
| new |= 1U << msg;
| } while (cmpxchg(ipi_data_ptr, old, new) != old);
was generating to below
| 8015cef8: ld r2,[r4,0] <-- First LD
| 8015cefc: bset r1,r2,r1
|
| 8015cf00: llock r3,[r4] <-- atomic op
| 8015cf04: brne r3,r2,8015cf10
| 8015cf08: scond r1,[r4]
| 8015cf0c: bnz 8015cf00
|
| 8015cf10: brne r3,r2,8015cf00 <-- Branch doesn't go to orig LD
Although this was fixed by adding a ACCESS_ONCE in this call site, it
seems safer (for now at least) to add compiler barrier to LLSC based
cmpxchg
Reported-by: Chuck Jordan <cjordan@synopsys,com>
Cc: <stable@vger.kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Diffstat (limited to 'arch/arc/include')
-rw-r--r-- | arch/arc/include/asm/cmpxchg.h | 9 |
1 files changed, 5 insertions, 4 deletions
diff --git a/arch/arc/include/asm/cmpxchg.h b/arch/arc/include/asm/cmpxchg.h index 03cd689..90de5c5 100644 --- a/arch/arc/include/asm/cmpxchg.h +++ b/arch/arc/include/asm/cmpxchg.h @@ -25,10 +25,11 @@ __cmpxchg(volatile void *ptr, unsigned long expected, unsigned long new) " scond %3, [%1] \n" " bnz 1b \n" "2: \n" - : "=&r"(prev) - : "r"(ptr), "ir"(expected), - "r"(new) /* can't be "ir". scond can't take limm for "b" */ - : "cc"); + : "=&r"(prev) /* Early clobber, to prevent reg reuse */ + : "r"(ptr), /* Not "m": llock only supports reg direct addr mode */ + "ir"(expected), + "r"(new) /* can't be "ir". scond can't take LIMM for "b" */ + : "cc", "memory"); /* so that gcc knows memory is being written here */ return prev; } |