linux/tools/arch/sparc/include/asm/barrier_64.h
Arnaldo Carvalho de Melo 94cdda6b98 perf tools: Move sparc barrier.h stuff to tools/arch/sparc/include/asm/barrier.h
We will need it for atomic.h, so move it from the ad-hoc tools/perf/
place to a tools/ subset of the kernel arch/ hierarchy.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-f0d04b9x63grt30nahpw9ei0@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-05-08 16:05:05 -03:00

43 lines
1.6 KiB
C

#ifndef __TOOLS_LINUX_SPARC64_BARRIER_H
#define __TOOLS_LINUX_SPARC64_BARRIER_H
/* Copied from the kernel sources to tools/:
*
* These are here in an effort to more fully work around Spitfire Errata
* #51. Essentially, if a memory barrier occurs soon after a mispredicted
* branch, the chip can stop executing instructions until a trap occurs.
* Therefore, if interrupts are disabled, the chip can hang forever.
*
* It used to be believed that the memory barrier had to be right in the
* delay slot, but a case has been traced recently wherein the memory barrier
* was one instruction after the branch delay slot and the chip still hung.
* The offending sequence was the following in sym_wakeup_done() of the
* sym53c8xx_2 driver:
*
* call sym_ccb_from_dsa, 0
* movge %icc, 0, %l0
* brz,pn %o0, .LL1303
* mov %o0, %l2
* membar #LoadLoad
*
* The branch has to be mispredicted for the bug to occur. Therefore, we put
* the memory barrier explicitly into a "branch always, predicted taken"
* delay slot to avoid the problem case.
*/
#define membar_safe(type) \
do { __asm__ __volatile__("ba,pt %%xcc, 1f\n\t" \
" membar " type "\n" \
"1:\n" \
: : : "memory"); \
} while (0)
/* The kernel always executes in TSO memory model these days,
* and furthermore most sparc64 chips implement more stringent
* memory ordering than required by the specifications.
*/
#define mb() membar_safe("#StoreLoad")
#define rmb() __asm__ __volatile__("":::"memory")
#define wmb() __asm__ __volatile__("":::"memory")
#endif /* !(__TOOLS_LINUX_SPARC64_BARRIER_H) */