* [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats @ 2025-03-24 23:52 Howard Chu 2025-03-24 23:58 ` Howard Chu 2025-03-26 18:21 ` Namhyung Kim 0 siblings, 2 replies; 6+ messages in thread From: Howard Chu @ 2025-03-24 23:52 UTC (permalink / raw) To: acme Cc: mingo, namhyung, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel, Howard Chu In my previous fix of runtime error(Link: https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), I made a mistake of decrementing one unconditionally, regardless of whether an extra 'syscall_nr' or 'nr' field was present in libtraceevent's tp_format. This may cause perf trace to allocate one fewer arg_fmt entry than needed for the accurate representation of syscall arguments. This patch corrects the mistake by checking the presence of'syscall_nr' or 'nr', and adjusting the length of arg_fmt[] accordingly. Signed-off-by: Howard Chu <howardchu95@gmail.com> --- tools/perf/builtin-trace.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index a102748bd0c9..ad6dad8e2453 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -2139,6 +2139,7 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) char tp_name[128]; const char *name; int err; + bool extra_nr = false; if (sc->nonexistent) return -EEXIST; @@ -2176,20 +2177,22 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) return err; } + sc->args = sc->tp_format->format.fields; + if (!strcmp(sc->args->name, "__syscall_nr") || !strcmp(sc->args->name, "nr")) + extra_nr = true; /* * The tracepoint format contains __syscall_nr field, so it's one more * than the actual number of syscall arguments. */ - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) + if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - (extra_nr ? 1 : 0))) return -ENOMEM; - sc->args = sc->tp_format->format.fields; /* * We need to check and discard the first variable '__syscall_nr' * or 'nr' that mean the syscall number. It is needless here. * So drop '__syscall_nr' or 'nr' field but does not exist on older kernels. */ - if (sc->args && (!strcmp(sc->args->name, "__syscall_nr") || !strcmp(sc->args->name, "nr"))) { + if (sc->args && extra_nr) { sc->args = sc->args->next; --sc->nr_args; } -- 2.45.2 ^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats 2025-03-24 23:52 [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats Howard Chu @ 2025-03-24 23:58 ` Howard Chu 2025-03-26 18:21 ` Namhyung Kim 1 sibling, 0 replies; 6+ messages in thread From: Howard Chu @ 2025-03-24 23:58 UTC (permalink / raw) To: acme Cc: mingo, namhyung, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel Hello, Forgot to add, there should be: Fixes: c7b87ce0dd10 ("perf trace: Fix runtime error of index out of bounds") Thanks, Howard On Mon, Mar 24, 2025 at 04:52:45PM -0700, Howard Chu wrote: > In my previous fix of runtime error(Link: > https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), > I made a mistake of decrementing one unconditionally, regardless of > whether an extra 'syscall_nr' or 'nr' field was present in > libtraceevent's tp_format. This may cause perf trace to allocate one > fewer arg_fmt entry than needed for the accurate representation of syscall > arguments. > > This patch corrects the mistake by checking the presence of'syscall_nr' or > 'nr', and adjusting the length of arg_fmt[] accordingly. > > Signed-off-by: Howard Chu <howardchu95@gmail.com> > --- > tools/perf/builtin-trace.c | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c > index a102748bd0c9..ad6dad8e2453 100644 > --- a/tools/perf/builtin-trace.c > +++ b/tools/perf/builtin-trace.c > @@ -2139,6 +2139,7 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > char tp_name[128]; > const char *name; > int err; > + bool extra_nr = false; > > if (sc->nonexistent) > return -EEXIST; > @@ -2176,20 +2177,22 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > return err; > } > > + sc->args = sc->tp_format->format.fields; > + if (!strcmp(sc->args->name, "__syscall_nr") || !strcmp(sc->args->name, "nr")) > + extra_nr = true; > /* > * The tracepoint format contains __syscall_nr field, so it's one more > * than the actual number of syscall arguments. > */ > - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) > + if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - (extra_nr ? 1 : 0))) > return -ENOMEM; > > - sc->args = sc->tp_format->format.fields; > /* > * We need to check and discard the first variable '__syscall_nr' > * or 'nr' that mean the syscall number. It is needless here. > * So drop '__syscall_nr' or 'nr' field but does not exist on older kernels. > */ > - if (sc->args && (!strcmp(sc->args->name, "__syscall_nr") || !strcmp(sc->args->name, "nr"))) { > + if (sc->args && extra_nr) { > sc->args = sc->args->next; > --sc->nr_args; > } > -- > 2.45.2 > ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats 2025-03-24 23:52 [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats Howard Chu 2025-03-24 23:58 ` Howard Chu @ 2025-03-26 18:21 ` Namhyung Kim 2025-03-26 18:30 ` Howard Chu 1 sibling, 1 reply; 6+ messages in thread From: Namhyung Kim @ 2025-03-26 18:21 UTC (permalink / raw) To: Howard Chu Cc: acme, mingo, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel On Mon, Mar 24, 2025 at 04:52:45PM -0700, Howard Chu wrote: > In my previous fix of runtime error(Link: > https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), > I made a mistake of decrementing one unconditionally, regardless of > whether an extra 'syscall_nr' or 'nr' field was present in > libtraceevent's tp_format. This may cause perf trace to allocate one > fewer arg_fmt entry than needed for the accurate representation of syscall > arguments. > > This patch corrects the mistake by checking the presence of'syscall_nr' or > 'nr', and adjusting the length of arg_fmt[] accordingly. Thanks for the fix. I've noticed this too but I feel like we can make syscall__alloc_arg_fmts() a bit simpler. How about this? Thanks, Namhyung ---8<--- diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 6ac51925ea4249c2..b9bdab52f5801c3a 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -2022,9 +2022,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) { int idx; - if (nr_args == RAW_SYSCALL_ARGS_NUM && sc->fmt && sc->fmt->nr_args != 0) - nr_args = sc->fmt->nr_args; - sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt)); if (sc->arg_fmt == NULL) return -1; @@ -2034,7 +2031,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) sc->arg_fmt[idx] = sc->fmt->arg[idx]; } - sc->nr_args = nr_args; return 0; } @@ -2176,14 +2172,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) return err; } - /* - * The tracepoint format contains __syscall_nr field, so it's one more - * than the actual number of syscall arguments. - */ - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) - return -ENOMEM; - sc->args = sc->tp_format->format.fields; + sc->nr_args = sc->tp_format->format.nr_fields; + /* * We need to check and discard the first variable '__syscall_nr' * or 'nr' that mean the syscall number. It is needless here. @@ -2194,6 +2185,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) --sc->nr_args; } + if (syscall__alloc_arg_fmts(sc, sc->nr_args)) + return -ENOMEM; + sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit"); sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat"); -- 2.49.0.395.g12beb8f557-goog ^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats 2025-03-26 18:21 ` Namhyung Kim @ 2025-03-26 18:30 ` Howard Chu 2025-03-26 19:44 ` Namhyung Kim 0 siblings, 1 reply; 6+ messages in thread From: Howard Chu @ 2025-03-26 18:30 UTC (permalink / raw) To: Namhyung Kim Cc: acme, mingo, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel Hello Namhyung, LGTM. This is better. :) Thanks, Howard On Wed, Mar 26, 2025 at 11:21 AM Namhyung Kim <namhyung@kernel.org> wrote: > > On Mon, Mar 24, 2025 at 04:52:45PM -0700, Howard Chu wrote: > > In my previous fix of runtime error(Link: > > https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), > > I made a mistake of decrementing one unconditionally, regardless of > > whether an extra 'syscall_nr' or 'nr' field was present in > > libtraceevent's tp_format. This may cause perf trace to allocate one > > fewer arg_fmt entry than needed for the accurate representation of syscall > > arguments. > > > > This patch corrects the mistake by checking the presence of'syscall_nr' or > > 'nr', and adjusting the length of arg_fmt[] accordingly. > > Thanks for the fix. I've noticed this too but I feel like we can make > syscall__alloc_arg_fmts() a bit simpler. > > How about this? > > Thanks, > Namhyung > > > ---8<--- > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c > index 6ac51925ea4249c2..b9bdab52f5801c3a 100644 > --- a/tools/perf/builtin-trace.c > +++ b/tools/perf/builtin-trace.c > @@ -2022,9 +2022,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > { > int idx; > > - if (nr_args == RAW_SYSCALL_ARGS_NUM && sc->fmt && sc->fmt->nr_args != 0) > - nr_args = sc->fmt->nr_args; > - > sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt)); > if (sc->arg_fmt == NULL) > return -1; > @@ -2034,7 +2031,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > sc->arg_fmt[idx] = sc->fmt->arg[idx]; > } > > - sc->nr_args = nr_args; > return 0; > } > > @@ -2176,14 +2172,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > return err; > } > > - /* > - * The tracepoint format contains __syscall_nr field, so it's one more > - * than the actual number of syscall arguments. > - */ > - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) > - return -ENOMEM; > - > sc->args = sc->tp_format->format.fields; > + sc->nr_args = sc->tp_format->format.nr_fields; > + > /* > * We need to check and discard the first variable '__syscall_nr' > * or 'nr' that mean the syscall number. It is needless here. > @@ -2194,6 +2185,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > --sc->nr_args; > } > > + if (syscall__alloc_arg_fmts(sc, sc->nr_args)) > + return -ENOMEM; > + > sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit"); > sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat"); > > -- > 2.49.0.395.g12beb8f557-goog > ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats 2025-03-26 18:30 ` Howard Chu @ 2025-03-26 19:44 ` Namhyung Kim 2025-03-26 19:52 ` Howard Chu 0 siblings, 1 reply; 6+ messages in thread From: Namhyung Kim @ 2025-03-26 19:44 UTC (permalink / raw) To: Howard Chu Cc: acme, mingo, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel On Wed, Mar 26, 2025 at 11:30:11AM -0700, Howard Chu wrote: > Hello Namhyung, > > LGTM. This is better. :) Great, can you please resend it with my Suggested-by? Thanks, Namhyung > > On Wed, Mar 26, 2025 at 11:21 AM Namhyung Kim <namhyung@kernel.org> wrote: > > > > On Mon, Mar 24, 2025 at 04:52:45PM -0700, Howard Chu wrote: > > > In my previous fix of runtime error(Link: > > > https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), > > > I made a mistake of decrementing one unconditionally, regardless of > > > whether an extra 'syscall_nr' or 'nr' field was present in > > > libtraceevent's tp_format. This may cause perf trace to allocate one > > > fewer arg_fmt entry than needed for the accurate representation of syscall > > > arguments. > > > > > > This patch corrects the mistake by checking the presence of'syscall_nr' or > > > 'nr', and adjusting the length of arg_fmt[] accordingly. > > > > Thanks for the fix. I've noticed this too but I feel like we can make > > syscall__alloc_arg_fmts() a bit simpler. > > > > How about this? > > > > Thanks, > > Namhyung > > > > > > ---8<--- > > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c > > index 6ac51925ea4249c2..b9bdab52f5801c3a 100644 > > --- a/tools/perf/builtin-trace.c > > +++ b/tools/perf/builtin-trace.c > > @@ -2022,9 +2022,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > > { > > int idx; > > > > - if (nr_args == RAW_SYSCALL_ARGS_NUM && sc->fmt && sc->fmt->nr_args != 0) > > - nr_args = sc->fmt->nr_args; > > - > > sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt)); > > if (sc->arg_fmt == NULL) > > return -1; > > @@ -2034,7 +2031,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > > sc->arg_fmt[idx] = sc->fmt->arg[idx]; > > } > > > > - sc->nr_args = nr_args; > > return 0; > > } > > > > @@ -2176,14 +2172,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > > return err; > > } > > > > - /* > > - * The tracepoint format contains __syscall_nr field, so it's one more > > - * than the actual number of syscall arguments. > > - */ > > - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) > > - return -ENOMEM; > > - > > sc->args = sc->tp_format->format.fields; > > + sc->nr_args = sc->tp_format->format.nr_fields; > > + > > /* > > * We need to check and discard the first variable '__syscall_nr' > > * or 'nr' that mean the syscall number. It is needless here. > > @@ -2194,6 +2185,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > > --sc->nr_args; > > } > > > > + if (syscall__alloc_arg_fmts(sc, sc->nr_args)) > > + return -ENOMEM; > > + > > sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit"); > > sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat"); > > > > -- > > 2.49.0.395.g12beb8f557-goog > > ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats 2025-03-26 19:44 ` Namhyung Kim @ 2025-03-26 19:52 ` Howard Chu 0 siblings, 0 replies; 6+ messages in thread From: Howard Chu @ 2025-03-26 19:52 UTC (permalink / raw) To: Namhyung Kim Cc: acme, mingo, mark.rutland, alexander.shishkin, jolsa, irogers, adrian.hunter, peterz, kan.liang, linux-perf-users, linux-kernel Hello Namhyung, On Wed, Mar 26, 2025 at 12:44 PM Namhyung Kim <namhyung@kernel.org> wrote: > > On Wed, Mar 26, 2025 at 11:30:11AM -0700, Howard Chu wrote: > > Hello Namhyung, > > > > LGTM. This is better. :) > > Great, can you please resend it with my Suggested-by? Sure. Thanks, Howard > > Thanks, > Namhyung > > > > > On Wed, Mar 26, 2025 at 11:21 AM Namhyung Kim <namhyung@kernel.org> wrote: > > > > > > On Mon, Mar 24, 2025 at 04:52:45PM -0700, Howard Chu wrote: > > > > In my previous fix of runtime error(Link: > > > > https://lore.kernel.org/linux-perf-users/20250122025519.361873-1-howardchu95@gmail.com/), > > > > I made a mistake of decrementing one unconditionally, regardless of > > > > whether an extra 'syscall_nr' or 'nr' field was present in > > > > libtraceevent's tp_format. This may cause perf trace to allocate one > > > > fewer arg_fmt entry than needed for the accurate representation of syscall > > > > arguments. > > > > > > > > This patch corrects the mistake by checking the presence of'syscall_nr' or > > > > 'nr', and adjusting the length of arg_fmt[] accordingly. > > > > > > Thanks for the fix. I've noticed this too but I feel like we can make > > > syscall__alloc_arg_fmts() a bit simpler. > > > > > > How about this? > > > > > > Thanks, > > > Namhyung > > > > > > > > > ---8<--- > > > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c > > > index 6ac51925ea4249c2..b9bdab52f5801c3a 100644 > > > --- a/tools/perf/builtin-trace.c > > > +++ b/tools/perf/builtin-trace.c > > > @@ -2022,9 +2022,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > > > { > > > int idx; > > > > > > - if (nr_args == RAW_SYSCALL_ARGS_NUM && sc->fmt && sc->fmt->nr_args != 0) > > > - nr_args = sc->fmt->nr_args; > > > - > > > sc->arg_fmt = calloc(nr_args, sizeof(*sc->arg_fmt)); > > > if (sc->arg_fmt == NULL) > > > return -1; > > > @@ -2034,7 +2031,6 @@ static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args) > > > sc->arg_fmt[idx] = sc->fmt->arg[idx]; > > > } > > > > > > - sc->nr_args = nr_args; > > > return 0; > > > } > > > > > > @@ -2176,14 +2172,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > > > return err; > > > } > > > > > > - /* > > > - * The tracepoint format contains __syscall_nr field, so it's one more > > > - * than the actual number of syscall arguments. > > > - */ > > > - if (syscall__alloc_arg_fmts(sc, sc->tp_format->format.nr_fields - 1)) > > > - return -ENOMEM; > > > - > > > sc->args = sc->tp_format->format.fields; > > > + sc->nr_args = sc->tp_format->format.nr_fields; > > > + > > > /* > > > * We need to check and discard the first variable '__syscall_nr' > > > * or 'nr' that mean the syscall number. It is needless here. > > > @@ -2194,6 +2185,9 @@ static int syscall__read_info(struct syscall *sc, struct trace *trace) > > > --sc->nr_args; > > > } > > > > > > + if (syscall__alloc_arg_fmts(sc, sc->nr_args)) > > > + return -ENOMEM; > > > + > > > sc->is_exit = !strcmp(name, "exit_group") || !strcmp(name, "exit"); > > > sc->is_open = !strcmp(name, "open") || !strcmp(name, "openat"); > > > > > > -- > > > 2.49.0.395.g12beb8f557-goog > > > ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2025-03-26 19:52 UTC | newest] Thread overview: 6+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2025-03-24 23:52 [PATCH v1] perf trace: Fix possible insufficient allocation of argument formats Howard Chu 2025-03-24 23:58 ` Howard Chu 2025-03-26 18:21 ` Namhyung Kim 2025-03-26 18:30 ` Howard Chu 2025-03-26 19:44 ` Namhyung Kim 2025-03-26 19:52 ` Howard Chu
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).