perf/x86/amd/ibs: Fix reading of the IBS OpData register and thus pre…

…cise RIP validity

[ Upstream commit 317b96b ]

The loop that reads all the IBS MSRs into *buf stopped one MSR short of
reading the IbsOpData register, which contains the RipInvalid status bit.

Fix the offset_max assignment so the MSR gets read, so the RIP invalid
evaluation is based on what the IBS h/w output, instead of what was
left in memory.

Signed-off-by: Kim Phillips <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Vince Weaver <[email protected]>
Fixes: d47e823 ("perf/x86-ibs: Take instruction pointer from ibs sample")
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>

arch/x86/events/amd/ibs.c

@@ -625,7 +625,7 @@ static int perf_ibs_handle_irq(struct perf_ibs *perf_ibs, struct pt_regs *iregs)
 	if (event->attr.sample_type & PERF_SAMPLE_RAW)
 		offset_max = perf_ibs->offset_max;
 	else if (check_rip)
-		offset_max = 2;
+		offset_max = 3;
 	else
 		offset_max = 1;
 	do {