On Wed, Apr 16, 2025 at 02:44:19PM +0200, Thomas Weißschuh wrote:
> usercopy of 64 bit values does not work on 32-bit SPARC:
>
> # usercopy_test_valid: EXPECTATION FAILED at
> lib/tests/usercopy_kunit.c:209
> Expected val_u64 == 0x5a5b5c5d6a6b6c6d, but
> val_u64 == 1515936861 (0x5a5b5c5d)
> 0x5a5b5c5d6a6b6c6d == 6510899242581322861 (0x5a5b5c5d6a6b6c6d)
>
> Disable the test.
>
> Fixes: 4c5d7bc63775 ("usercopy: Add tests for all get_user() sizes")
> Signed-off-by: Thomas Weißschuh <[email protected]>
> ---
> To be honest I think sparc32 wants to support usercopy of 64 bit values.
> But it does seem to be broken.
>
> (+Cc SPARC maintainers)
I wonder if these archs need to explicitly fail to build if they trigger
a 64-bit wide put_user/get_user call? It seems weird to me that it
should be possible at all to provide access to the function if we know
it's going to 100% fail.
-Kees
> ---
> lib/tests/usercopy_kunit.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/lib/tests/usercopy_kunit.c b/lib/tests/usercopy_kunit.c
> index
> 77fa00a13df775074a19772bfbc9120ef33634bb..80f8abe10968c1d91c61006f1eaa63858c633872
> 100644
> --- a/lib/tests/usercopy_kunit.c
> +++ b/lib/tests/usercopy_kunit.c
> @@ -27,6 +27,7 @@
> !defined(CONFIG_MICROBLAZE) && \
> !defined(CONFIG_NIOS2) && \
> !defined(CONFIG_PPC32) && \
> + !defined(CONFIG_SPARC32) && \
> !defined(CONFIG_SUPERH))
> # define TEST_U64
> #endif
>
> ---
> base-commit: 1a1d569a75f3ab2923cb62daf356d102e4df2b86
> change-id: 20250416-kunit-sparc-usercopy-c36de3ca4ef8
>
> Best regards,
> --
> Thomas Weißschuh <[email protected]>
>
--
Kees Cook
