On Wed, Mar 09, 2022 at 09:10:29PM +1100, Stephen Rothwell wrote:
> Hi all,
>
> After merging the kspp tree, today's linux-next build (x86_64
> allmodconfig) failed like this:
>
> In file included from include/linux/bitops.h:33,
> from include/linux/kernel.h:22,
> from drivers/platform/x86/thinkpad_acpi.c:37:
> drivers/platform/x86/thinkpad_acpi.c: In function 'tpacpi_dytc_profile_init':
> arch/x86/include/asm/bitops.h:207:22: error: array subscript 'long unsigned int[0]' is partly outside array bounds of 'int[1]' [-Werror=array-bounds]
> 207 | (addr[nr >> _BITOPS_LONG_SHIFT])) != 0;
> | ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~
> drivers/platform/x86/thinkpad_acpi.c:10385:18: note: while referencing 'output'
> 10385 | int err, output;
> | ^~~~~~
>
> Actually the test_bit() in line 10409.
>
> Caused by commit
>
> e1c21608e3cf ("platform/x86: thinkpad_acpi: Add PSC mode support")
>
> from the drivers-x86 tree interacting with the enablement of
> -Werror=array-bounds.
>
> I have reverted that commit for today.
Looking at https://lore.kernel.org/all/20220225182505.7234-1-markpearson@xxxxxxxxxx/
The problem is with the cast of "int output" to "unsigned long". Past
fixes have either made the variable unsigned long, or adding a new
variable and removal of casts. e.g. this pseudo-patch:
+ unsigned long bits;
...
+ bits = output;
- if (test_bit(DYTC_FC_MMC, (void *)&output)) { /* MMC MODE */
+ if (test_bit(DYTC_FC_MMC, &bits)) { /* MMC MODE */
...
- } else if (test_bit(DYTC_FC_PSC, (void *)&output)) { /*PSC MODE */
+ } else if (test_bit(DYTC_FC_PSC, &bits)) { /*PSC MODE */
--
Kees Cook
