On Thu, Nov 18, 2021 at 5:03 PM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
>
> On Thu, Nov 18, 2021 at 03:21:02PM +0100, Arnd Bergmann wrote:
> > From: Arnd Bergmann <arnd@xxxxxxxx>
> >
> > gcc-11 points out a potential issue with integer overflow when
> > the iwl_dev_info_table[] array is empty:
> >
> > drivers/net/wireless/intel/iwlwifi/pcie/drv.c:1344:42: error: implicit conversion from 'unsigned long' to 'int' changes value from 18446744073709551615 to -1 [-Werror,-Wconstant-conversion]
> > for (i = ARRAY_SIZE(iwl_dev_info_table) - 1; i >= 0; i--) {
> > ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
>
> For what it's worth, I do see this warning with Clang when both
> CONFIG_IWLDVM and CONFIG_IWLMVM are disabled and looking through the GCC
> warning docs [1], I do not see a -Wconstant-conversion option? Maybe
> there is another warning that is similar but that warning right there
> appears to have come from clang, as it matches mine exactly.
>
> drivers/net/wireless/intel/iwlwifi/pcie/drv.c:1344:42: error: implicit conversion from 'unsigned long' to 'int' changes value from 18446744073709551615 to -1 [-Werror,-Wconstant-conversion]
> for (i = ARRAY_SIZE(iwl_dev_info_table) - 1; i >= 0; i--) {
> ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
> 1 error generated.
>
> [1]: https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
Ok, got it: it turns out this warning /also/ happens with gcc-11 and
the initial changelog was the one for matching the clang warning.
This is the gcc output, which is very similar but has a different
warning option.
drivers/net/wireless/intel/iwlwifi/pcie/drv.c: In function
'iwl_pci_find_dev_info':
include/linux/kernel.h:46:25: error: overflow in conversion from 'long
unsigned int' to 'int' changes value from '18446744073709551615' to
'-1' [-Werror=overflow]
46 | #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) +
__must_be_array(arr))
| ^
drivers/net/wireless/intel/iwlwifi/pcie/drv.c:1344:18: note: in
expansion of macro 'ARRAY_SIZE'
1344 | for (i = ARRAY_SIZE(iwl_dev_info_table) - 1; i >= 0; i--) {
My v2 patch only addressed the clang warning, while v3 works with both
gcc and clang. I can send a v4 if I should update the changelog again
to explain that, but I suppose it's still close enough.
Arnd
