Re: [PATCH v5 2/2] lib: checksum: Use aligned accesses for ip_fast_csum and csum_ipv6_magic tests

[Guenter Roeck]

Hi,

On Tue, Jan 30, 2024 at 11:10:04AM -0800, Charlie Jenkins wrote:
> The test cases for ip_fast_csum and csum_ipv6_magic were using arbitrary
> alignment of data to iterate through random inputs. ip_fast_csum should
> have the data aligned along (14 + NET_IP_ALIGN) bytes and
> csum_ipv6_magic should have data aligned along 32-bit boundaries.
> 
> While this is being changed, fix up the awkward offset code in
> test_csum_ipv6_magic and use a struct instead.
> 
> Signed-off-by: Charlie Jenkins <charlie@xxxxxxxxxxxx>
> Tested-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> Fixes: 6f4c45cbcb00 ("kunit: Add tests for csum_ipv6_magic and ip_fast_csum")

So this works on little endian systems. Unfortunately, I still get

[   18.447037]     # test_ip_fast_csum: ASSERTION FAILED at lib/checksum_kunit.c:463
[   18.447037]     Expected ( u64)csum_result == ( u64)expected, but
[   18.447037]         ( u64)csum_result == 33754 (0x83da)
[   18.447037]         ( u64)expected == 55939 (0xda83)
[   18.455565]     not ok 4 test_ip_fast_csum
[   18.463570]     # test_csum_ipv6_magic: ASSERTION FAILED at lib/checksum_kunit.c:487
[   18.463570]     Expected ( u64)csum_result == ( u64)expected, but
[   18.463570]         ( u64)csum_result == 12221 (0x2fbd)
[   18.463570]         ( u64)expected == 37721 (0x9359)
[   18.470679]     not ok 5 test_csum_ipv6_magic

when running the test on big endian systems such as hppa/parisc or sparc.

Guenter