Re: [PATCH 01/10] perf, trace: Remove IRQ-disable fromperf/tracepoint interaction

From: Peter Zijlstra
Date: Sun May 23 2010 - 14:16:36 EST


On Sun, 2010-05-23 at 22:11 +1000, Paul Mackerras wrote:
> On Fri, May 21, 2010 at 11:02:02AM +0200, Peter Zijlstra wrote:
>
> > entry = (struct trace_entry *)raw_data;
> > - tracing_generic_entry_update(entry, *irq_flags, pc);
> > + tracing_generic_entry_update(entry, regs->flags, pc);
> > entry->type = type;
>
> This is breaking on powerpc -- we don't have a "flags" element in
> struct pt_regs. What is it trying to do? Get the interrupt enable
> state, as local_irq_save(flags) would provide?

Yes, I fouled that up.

On x86 perf_arch_fetch_caller_regs() stores pt_regs::flags, which is the
same as used in local_save_flags().

So to avoid another local_save_flags() for the ftrace code, I wanted to
reuse it, and totally forgot the !x86 side of things.

Possibly we could remove the local_save_flags() from the
perf_arch_fetch_caller_regs() as I'm not sure we actually use
pt_regs::flags (Frederic?) and leave the ftrace thing to use
local_save_flags().

Something like:


---
arch/x86/kernel/cpu/perf_event.c | 1 -
kernel/trace/trace_event_perf.c | 4 +++-
2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index fd4db0d..856aeeb 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1717,7 +1717,6 @@ void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int ski
*/
regs->bp = rewind_frame_pointer(skip + 1);
regs->cs = __KERNEL_CS;
- local_save_flags(regs->flags);
}

unsigned long perf_instruction_pointer(struct pt_regs *regs)
diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c
index 26b8607..cb6f365 100644
--- a/kernel/trace/trace_event_perf.c
+++ b/kernel/trace/trace_event_perf.c
@@ -157,6 +157,7 @@ __kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
struct pt_regs *regs, int *rctxp)
{
struct trace_entry *entry;
+ unsigned long flags;
char *raw_data;
int pc;

@@ -174,7 +175,8 @@ __kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
memset(&raw_data[size - sizeof(u64)], 0, sizeof(u64));

entry = (struct trace_entry *)raw_data;
- tracing_generic_entry_update(entry, regs->flags, pc);
+ local_save_flags(flags);
+ tracing_generic_entry_update(entry, flags, pc);
entry->type = type;

return raw_data;


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/