sparc64: Fix perf_arch_get_caller_regs().
Subsystems (2)
David Miller
maintainerPATCH 1/1 sparc64: Fix perf_arch_get_caller_regs().
After b0f82b81fe6bbcf78d478071f33e44554726bc81 ("perf: Drop the skip argument from perf_arch_fetch_regs_caller") the build broke on sparc64 due to the lack of a module symbol export of __perf_arch_fetch_caller_regs.
But that assembler helper can actually be complete eliminated now that the semantics of this interface have been greatly simplified.
---
Frederic, I'll merge this to Linus via the sparc tree, this is just FYI...
arch/sparc/include/asm/perf_event.h | 25 +++++++++-- arch/sparc/kernel/helpers.S | 75 ----------------------------------- 2 files changed, 20 insertions(+), 80 deletions(-)
arch/sparc/include/asm/perf_event.h +20 -5
@@ -10,11 +10,26 @@ extern void set_perf_event_pending(void);extern void init_hw_perf_events (void ); -extern void -__perf_arch_fetch_caller_regs (struct pt_regs *regs,unsigned long ip,int skip); - -#define perf_arch_fetch_caller_regs (pt_regs, ip) \ - __perf_arch_fetch_caller_regs(pt_regs, ip, 1); +#define perf_arch_fetch_caller_regs (regs, ip) \ +do { \ + unsigned long _pstate, _asi, _pil, _i7, _fp; \ + __asm__ __volatile__("rdpr %%pstate, %0\n\t" \ + "rd %%asi, %1\n\t" \ + "rdpr %%pil, %2\n\t" \ + "mov %%i7, %3\n\t" \ + "mov %%i6, %4\n\t" \ + : "=r" (_pstate), \ + "=r" (_asi), \ + "=r" (_pil), \ + "=r" (_i7), \ + "=r" (_fp)); \ + (regs)->tstate = (_pstate << 8) | \ + (_asi << 24) | (_pil << 20); \ + (regs)->tpc = (ip); \ + (regs)->tnpc = (regs)->tpc + 4; \ + (regs)->u_regs[UREG_I6] = _fp; \ + (regs)->u_regs[UREG_I7] = _i7; \ +} while (0)#else static inline void init_hw_perf_events (void ) { }#endif
arch/sparc/kernel/helpers.S +0 -75
81 lines not on this page · raw message
Signature · 1 line
1.7.2.1
Quoted 5 lines · David Miller
After
b0f82b81fe6bbcf78d478071f33e44554726bc81("perf: Drop the skip argument fromperf_arch_fetch_regs_caller") the build broke on sparc64 due to the lack of a module symbol export of__perf_arch_fetch_caller_regs.
Quoted 8 lines
But that assembler helper can actually be complete eliminated now that the semantics of this interface have been greatly simplified.
Signed-off-by David S. Miller <[email protected]>---
Frederic, I'll merge this to Linus via the sparc tree, this is just FYI...
Ok thanks a lot! I have no more access to a Sparc box, so I'm sorry I can't anymore test perf and ftrace on this architecture :)
David Miller
maintainerFrom: Frederic Weisbecker <[email protected]>
Date: Mon, 9 Aug 2010 19:14:42 +0200
Quoted 2 lines
Ok thanks a lot! I have no more access to a Sparc box, so I'm sorry I can't anymore test perf and ftrace on this architecture :)
You don't need a sparc box to test the build.