Messages in this thread | | | Date | Thu, 1 Sep 2022 07:31:58 -0700 | Subject | Re: [PATCH 5/5] perf intel-pt: Support itrace option flag d+e to log on error | From | Andi Kleen <> |
| |
On 9/1/2022 4:00 AM, Adrian Hunter wrote: > Pass d+e option and log size via intel_pt_log_enable(). Allocate a buffer > for log messages and provide intel_pt_log_dump_buf() to dump and reset the > buffer upon decoder errors. > > Example: > > $ sudo perf record -e intel_pt// sleep 1 > [ perf record: Woken up 1 times to write data ] > [ perf record: Captured and wrote 0.094 MB perf.data ] > $ sudo perf config itrace.debug-log-buffer-size=300 > $ sudo perf script --itrace=ed+e+o | head -20 > Dumping debug log buffer (first line may be sliced) > Other > ffffffff96ca22f6: 48 89 e5 Other > ffffffff96ca22f9: 65 48 8b 05 ff e0 38 69 Other > ffffffff96ca2301: 48 3d c0 a5 c1 98 Other > ffffffff96ca2307: 74 08 Jcc +8 > ffffffff96ca2311: 5d Other > ffffffff96ca2312: c3 Ret > ERROR: Bad RET compression (TNT=N) at 0xffffffff96ca2312 > End of debug log buffer dump > instruction trace error type 1 time 15913.537143482 cpu 5 pid 36292 tid 36292 ip 0xffffffff96ca2312 code 6: Trace doesn't match instruction > Dumping debug log buffer (first line may be sliced) > Other > ffffffff96ce7fe9: f6 47 2e 20 Other > ffffffff96ce7fed: 74 11 Jcc +17 > ffffffff96ce7fef: 48 8b 87 28 0a 00 00 Other > ffffffff96ce7ff6: 5d Other > ffffffff96ce7ff7: 48 8b 40 18 Other > ffffffff96ce7ffb: c3 Ret > ERROR: Bad RET compression (TNT=N) at 0xffffffff96ce7ffb > Warning: > 8 instruction trace errors > > Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> > --- > tools/perf/Documentation/perf-intel-pt.txt | 5 +- > .../perf/util/intel-pt-decoder/intel-pt-log.c | 94 ++++++++++++++++++- > .../perf/util/intel-pt-decoder/intel-pt-log.h | 3 +- > tools/perf/util/intel-pt.c | 20 +++- > 4 files changed, 117 insertions(+), 5 deletions(-) > > diff --git a/tools/perf/Documentation/perf-intel-pt.txt b/tools/perf/Documentation/perf-intel-pt.txt > index d5ddb968bcf4..92464a5d7eaf 100644 > --- a/tools/perf/Documentation/perf-intel-pt.txt > +++ b/tools/perf/Documentation/perf-intel-pt.txt > @@ -989,10 +989,13 @@ must be preceded by either '+' or '-'. The flags support by Intel PT are: > > -a Suppress logging of perf events > +a Log all perf events > + +e Output only on decoding errors (size configurable) > +o Output to stdout instead of "intel_pt.log" > > By default, logged perf events are filtered by any specified time ranges, but > -flag +a overrides that. > +flag +a overrides that. The +e flag can be useful for analyzing errors. By > +default, the log size in that case is 16384 bytes, but can be altered by > +linkperf:perf-config[1] e.g. perf config itrace.debug-log-buffer-size=30000 > > In addition, the period of the "instructions" event can be specified. e.g. > > diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-log.c b/tools/perf/util/intel-pt-decoder/intel-pt-log.c > index 5f5dfc8753f3..ea96dcae187a 100644 > --- a/tools/perf/util/intel-pt-decoder/intel-pt-log.c > +++ b/tools/perf/util/intel-pt-decoder/intel-pt-log.c > @@ -5,12 +5,16 @@ > */ > > #include <stdio.h> > +#include <stdlib.h> > #include <stdint.h> > #include <inttypes.h> > #include <stdarg.h> > #include <stdbool.h> > #include <string.h> > > +#include <linux/zalloc.h> > +#include <linux/kernel.h> > + > #include "intel-pt-log.h" > #include "intel-pt-insn-decoder.h" > > @@ -21,15 +25,20 @@ > static FILE *f; > static char log_name[MAX_LOG_NAME]; > bool intel_pt_enable_logging; > +static bool intel_pt_dump_log_on_error; > +static unsigned int intel_pt_log_on_error_size; > +static struct log_buf log_buf; > > void *intel_pt_log_fp(void) > { > return f; > } > > -void intel_pt_log_enable(void) > +void intel_pt_log_enable(bool dump_log_on_error, unsigned int log_on_error_size) > { > intel_pt_enable_logging = true; > + intel_pt_dump_log_on_error = dump_log_on_error; > + intel_pt_log_on_error_size = log_on_error_size; > } > > void intel_pt_log_disable(void) > @@ -74,6 +83,87 @@ static void intel_pt_print_no_data(uint64_t pos, int indent) > fprintf(f, " "); > } > > +#define DFLT_BUF_SZ (16 * 1024) > + > +struct log_buf { > + char *buf; > + size_t buf_sz; > + size_t head; > + bool wrapped; > + FILE *backend; > +}; > + > +static ssize_t log_buf__write(void *cookie, const char *buf, size_t size) > +{ > + struct log_buf *b = cookie; > + size_t sz = size; > + > + if (!b->buf) > + return size; > + > + while (sz) { > + size_t space = b->buf_sz - b->head; > + size_t n = min(space, sz); > + > + memcpy(b->buf + b->head, buf, n); > + sz -= n; > + buf += n; > + b->head += n; > + if (sz && b->head >= b->buf_sz) { > + b->head = 0; > + b->wrapped = true; > + } > + }
I would add an assert or somesuch for the case where sz is larger than the buffer size, which would otherwise loop forever.
> + > +static void log_buf__dump(struct log_buf *b) > +{ > + if (!b->buf) > + return; > + > + fflush(f); > + fprintf(b->backend, "Dumping debug log buffer (first line may be sliced)\n");
Should be easy to skip the first line, no?
Rest looks good.
Reviewed-by: Andi Kleen <ak@linux.intel.com>
-Andi
| |