* Re: [PATCH v1] perf trace: Fix runtime error of index out of bounds
2025-01-22 19:11 ` Namhyung Kim
@ 2025-01-23 23:11 ` Howard Chu
2025-01-24 4:30 ` Howard Chu
1 sibling, 0 replies; 5+ messages in thread
From: Howard Chu @ 2025-01-23 23:11 UTC (permalink / raw)
To: Namhyung Kim
Cc: acme, jolsa, irogers, adrian.hunter, linux-perf-users,
linux-kernel
Hello Namhyung,
On Wed, Jan 22, 2025 at 11:11 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Tue, Jan 21, 2025 at 06:55:19PM -0800, Howard Chu wrote:
> > libtraceevent parses and returns an array of argument fields, sometimes
> > larger than RAW_SYSCALL_ARGS_NUM (6) because it includes "__syscall_nr",
> > idx will traverse to index 6 (7th element) whereas sc->fmt->arg holds 6
> > elements max, creating an out-of-bounds access. This runtime error is
> > found by UBsan. The error message:
>
> I'm curious how I miss this. Is this only for some specific syscalls or
> always happening? Do you which commit introduced this?
On my x86-64 machine, I found that this error can occur for syscalls
with 6 arguments, which results in nr_fields being 7. These syscalls
are:
copy_file_range
epoll_pwait
epoll_pwait2
futex
futex_wait
io_pgetevents
io_uring_enter
mbind
mmap
move_pages
preadv2
process_vm_readv
process_vm_writev
pselect6
pwritev2
recvfrom
sendto
splice
But the runtime error occurs only with the futex syscall, when
accessing sc->fmt->arg[idx]. This is an array of length 6.
>
> Thanks,
> Namhyung
>
> >
> > perf $ sudo UBSAN_OPTIONS=print_stacktrace=1 ./perf trace -a --max-events=1
> > builtin-trace.c:1966:35: runtime error: index 6 out of bounds for type 'syscall_arg_fmt [6]'
> > #0 0x5c04956be5fe in syscall__alloc_arg_fmts /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:1966
> > #1 0x5c04956c0510 in trace__read_syscall_info /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:2110
> > #2 0x5c04956c372b in trace__syscall_info /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:2436
> > #3 0x5c04956d2f39 in trace__init_syscalls_bpf_prog_array_maps /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:3897
> > #4 0x5c04956d6d25 in trace__run /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:4335
> > #5 0x5c04956e112e in cmd_trace /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:5502
> > #6 0x5c04956eda7d in run_builtin /home/howard/hw/linux-perf/tools/perf/perf.c:351
> > #7 0x5c04956ee0a8 in handle_internal_command /home/howard/hw/linux-perf/tools/perf/perf.c:404
> > #8 0x5c04956ee37f in run_argv /home/howard/hw/linux-perf/tools/perf/perf.c:448
> > #9 0x5c04956ee8e9 in main /home/howard/hw/linux-perf/tools/perf/perf.c:556
> > #10 0x79eb3622a3b7 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
> > #11 0x79eb3622a47a in __libc_start_main_impl ../csu/libc-start.c:360
> > #12 0x5c04955422d4 in _start (/home/howard/hw/linux-perf/tools/perf/perf+0x4e02d4) (BuildId: 5b6cab2d59e96a4341741765ad6914a4d784dbc6)
> >
> > 0.000 ( 0.014 ms): Chrome_ChildIO/117244 write(fd: 238, buf: !, count: 1) = 1
> >
> > Signed-off-by: Howard Chu <howardchu95@gmail.com>
> > ---
> > tools/perf/builtin-trace.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
> > index d7c7d29291fb..8d3260bad10a 100644
> > --- a/tools/perf/builtin-trace.c
> > +++ b/tools/perf/builtin-trace.c
> > @@ -2108,7 +2108,7 @@ static int trace__read_syscall_info(struct trace *trace, int id)
> > }
> >
> > if (syscall__alloc_arg_fmts(sc, IS_ERR(sc->tp_format) ?
> > - RAW_SYSCALL_ARGS_NUM : sc->tp_format->format.nr_fields))
> > + RAW_SYSCALL_ARGS_NUM : sc->tp_format->format.nr_fields - 1))
Looking back, unfortunately, this is incorrect, because, as Arnaldo
mentioned, the '__syscall_nr' in tracefs does not exist in older
kernels. You mentioned missing the runtime error; maybe check the
tracefs on your machine to see if there is a '__syscall_nr'.
$ cat /sys/kernel/tracing/events/syscalls/sys_enter_futex/format |
grep __syscall_nr
field:int __syscall_nr; offset:8; size:4; signed:1;
And I'll send v2 that will have the correct logic.
Thanks,
Howard
> > return -ENOMEM;
> >
> > sc->args = sc->tp_format->format.fields;
> > --
> > 2.45.2
> >
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH v1] perf trace: Fix runtime error of index out of bounds
2025-01-22 19:11 ` Namhyung Kim
2025-01-23 23:11 ` Howard Chu
@ 2025-01-24 4:30 ` Howard Chu
1 sibling, 0 replies; 5+ messages in thread
From: Howard Chu @ 2025-01-24 4:30 UTC (permalink / raw)
To: Namhyung Kim
Cc: acme, jolsa, irogers, adrian.hunter, linux-perf-users,
linux-kernel
Hello again,
On Wed, Jan 22, 2025 at 11:11 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Tue, Jan 21, 2025 at 06:55:19PM -0800, Howard Chu wrote:
> > libtraceevent parses and returns an array of argument fields, sometimes
> > larger than RAW_SYSCALL_ARGS_NUM (6) because it includes "__syscall_nr",
> > idx will traverse to index 6 (7th element) whereas sc->fmt->arg holds 6
> > elements max, creating an out-of-bounds access. This runtime error is
> > found by UBsan. The error message:
>
> I'm curious how I miss this. Is this only for some specific syscalls or
> always happening? Do you which commit introduced this?
Forgot to mention the commit: it is (5e58fcfaf4c6: 'perf trace: Allow
allocating sc->arg_fmt even without the syscall tracepoint') that
introduced the bug. However, I wasn't able to build the perf tool at
this commit because it dates back to 2017, so I couldn’t bisect and
test it.
As for the runtime error, it occurs with the first 6-argument syscall
that calls the syscall__alloc_arg_fmts() function and causes the
out-of-bounds access. That’s going to be the futex syscall. If I skip
the futex syscall, mmap is next.
static int syscall__alloc_arg_fmts(struct syscall *sc, int nr_args)
** the 'nr_args' passed in can be 7
{
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));
** arg_fmt has 7 elements
if (sc->arg_fmt == NULL)
return -1;
for (idx = 0; idx < nr_args; ++idx) {
** idx can be 6 (7th element), legal access for sc->arg_fmt
** but not for sc->fmt->arg, length of 6
if (sc->fmt)
sc->arg_fmt[idx] = sc->fmt->arg[idx];
}
sc->nr_args = nr_args;
return 0;
}
Thanks,
Howard
On Wed, Jan 22, 2025 at 11:11 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Tue, Jan 21, 2025 at 06:55:19PM -0800, Howard Chu wrote:
> > libtraceevent parses and returns an array of argument fields, sometimes
> > larger than RAW_SYSCALL_ARGS_NUM (6) because it includes "__syscall_nr",
> > idx will traverse to index 6 (7th element) whereas sc->fmt->arg holds 6
> > elements max, creating an out-of-bounds access. This runtime error is
> > found by UBsan. The error message:
>
> I'm curious how I miss this. Is this only for some specific syscalls or
> always happening? Do you which commit introduced this?
>
> Thanks,
> Namhyung
>
> >
> > perf $ sudo UBSAN_OPTIONS=print_stacktrace=1 ./perf trace -a --max-events=1
> > builtin-trace.c:1966:35: runtime error: index 6 out of bounds for type 'syscall_arg_fmt [6]'
> > #0 0x5c04956be5fe in syscall__alloc_arg_fmts /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:1966
> > #1 0x5c04956c0510 in trace__read_syscall_info /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:2110
> > #2 0x5c04956c372b in trace__syscall_info /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:2436
> > #3 0x5c04956d2f39 in trace__init_syscalls_bpf_prog_array_maps /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:3897
> > #4 0x5c04956d6d25 in trace__run /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:4335
> > #5 0x5c04956e112e in cmd_trace /home/howard/hw/linux-perf/tools/perf/builtin-trace.c:5502
> > #6 0x5c04956eda7d in run_builtin /home/howard/hw/linux-perf/tools/perf/perf.c:351
> > #7 0x5c04956ee0a8 in handle_internal_command /home/howard/hw/linux-perf/tools/perf/perf.c:404
> > #8 0x5c04956ee37f in run_argv /home/howard/hw/linux-perf/tools/perf/perf.c:448
> > #9 0x5c04956ee8e9 in main /home/howard/hw/linux-perf/tools/perf/perf.c:556
> > #10 0x79eb3622a3b7 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
> > #11 0x79eb3622a47a in __libc_start_main_impl ../csu/libc-start.c:360
> > #12 0x5c04955422d4 in _start (/home/howard/hw/linux-perf/tools/perf/perf+0x4e02d4) (BuildId: 5b6cab2d59e96a4341741765ad6914a4d784dbc6)
> >
> > 0.000 ( 0.014 ms): Chrome_ChildIO/117244 write(fd: 238, buf: !, count: 1) = 1
> >
> > Signed-off-by: Howard Chu <howardchu95@gmail.com>
> > ---
> > tools/perf/builtin-trace.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
> > index d7c7d29291fb..8d3260bad10a 100644
> > --- a/tools/perf/builtin-trace.c
> > +++ b/tools/perf/builtin-trace.c
> > @@ -2108,7 +2108,7 @@ static int trace__read_syscall_info(struct trace *trace, int id)
> > }
> >
> > if (syscall__alloc_arg_fmts(sc, IS_ERR(sc->tp_format) ?
> > - RAW_SYSCALL_ARGS_NUM : sc->tp_format->format.nr_fields))
> > + RAW_SYSCALL_ARGS_NUM : sc->tp_format->format.nr_fields - 1))
> > return -ENOMEM;
> >
> > sc->args = sc->tp_format->format.fields;
> > --
> > 2.45.2
> >
^ permalink raw reply [flat|nested] 5+ messages in thread