Running libgpiod applications under valgrind results in the following warning: ==3006== Invalid read of size 8 ==3006== at 0x10C867: line_request_values (core.c:711) ==3006== by 0x10CDA6: gpiod_line_request_bulk (core.c:849) ==3006== by 0x10AE27: main (gpioset.c:323) ==3006== Address 0x4a4d370 is 0 bytes after a block of size 16 alloc'd ==3006== at 0x483F790: malloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so) ==3006== by 0x10B884: gpiod_line_bulk_new (core.c:109) ==3006== by 0x10DBB0: gpiod_chip_get_lines (helpers.c:24) ==3006== by 0x10ADC3: main (gpioset.c:313) This is because the foreach loop reads the next value before checking that index is still in bounds. Add a test to avoid reading past the end of the allocation. This bug is not present a released version of libgpiod. Fixes: 2b02d7ae1aa6 ("treewide: rework struct gpiod_line_bulk") Signed-off-by: Joel Stanley <joel@xxxxxxxxx> --- lib/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/lib/core.c b/lib/core.c index 6ef09baec0f5..4463a7014776 100644 --- a/lib/core.c +++ b/lib/core.c @@ -178,7 +178,8 @@ GPIOD_API void gpiod_line_bulk_foreach_line(struct gpiod_line_bulk *bulk, #define line_bulk_foreach_line(bulk, line, index) \ for ((index) = 0, (line) = (bulk)->lines[0]; \ (index) < (bulk)->num_lines; \ - (index)++, (line) = (bulk)->lines[(index)]) + (index)++, \ + (line) = (index) < (bulk)->num_lines ? (bulk)->lines[(index)] : NULL) GPIOD_API bool gpiod_is_gpiochip_device(const char *path) { -- 2.34.1