On Tue, Feb 24, 2026 at 9:59 PM Mahdi Faramarzpour <[email protected]> wrote: > > On Wed, Feb 25, 2026 at 7:35 AM Alexei Starovoitov > <[email protected]> wrote: > > > > On Wed, Feb 18, 2026 at 1:20 AM Mahdi Faramarzpour <[email protected]> > > wrote: > > > > > > From: Mahdi Faramarzpour <[email protected]> > > > > > > This commit fixes the integer parsing of -t option. The cli parser > > > only relies on errno to detect parsing errors. The manpage for > > > strtol (https://man7.org/linux/man-pages/man3/strtol.3.html) > > > states that the said function "MAY" set errno to EINVAL in case the > > > conversion fails. Currently on some systems, this leads to a silent > > > failure with return value not being exactly documented in the > > > manpages (probably zero). The reliable way to validate the input is > > > to check whether the endptr has been bumped all the way to the end > > > of the string or not. > > > > > > Fixes: 146e30554a53 ("selftests/xsk: add option to run single test") > > > Signed-off-by: Mahdi Faramarzpour <[email protected]> > > > --- > > > v2: > > > - fix style issues > > > v1: > > > https://lore.kernel.org/all/[email protected]/ > > > --- > > > tools/testing/selftests/bpf/xskxceiver.c | 5 +++-- > > > 1 file changed, 3 insertions(+), 2 deletions(-) > > > > > > diff --git a/tools/testing/selftests/bpf/xskxceiver.c > > > b/tools/testing/selftests/bpf/xskxceiver.c > > > index 05b3cebc5..5e4095fd1 100644 > > > --- a/tools/testing/selftests/bpf/xskxceiver.c > > > +++ b/tools/testing/selftests/bpf/xskxceiver.c > > > @@ -200,6 +200,7 @@ static void parse_command_line(struct ifobject > > > *ifobj_tx, struct ifobject *ifobj > > > struct ifobject *ifobj; > > > u32 interface_nb = 0; > > > int option_index, c; > > > + char *eptr; > > > > > > opterr = 0; > > > > > > @@ -248,8 +249,8 @@ static void parse_command_line(struct ifobject > > > *ifobj_tx, struct ifobject *ifobj > > > break; > > > case 't': > > > errno = 0; > > > - opt_run_test = strtol(optarg, NULL, 0); > > > - if (errno) > > > + opt_run_test = strtol(optarg, &eptr, 0); > > > + if (errno || *eptr) > > > > This is unnecessary. It was required we would have hit segfaults by now. > This patch does not prevent segfaults; it prevents silent misparsing. > On systems where strtol() does not set > errno for non-numeric input, passing -t abc results in opt_run_test > being 0 without triggering usage(). > Checking endptr is the documented way to validate full-string conversion.
Prove me wrong with a selftest.
