Re: [PATCH RFC bpf-next 3/4] bpf/selftests: add tests to validate proper arguments alignment on ARM64

On Mon Apr 28, 2025 at 10:41 PM CEST, Alexis Lothoré wrote:

On Mon Apr 28, 2025 at 6:52 PM CEST, Eduard Zingerman wrote:

...

Alexis Lothoré alexis.lothore@bootlin.com writes:

If things really behaved correctly, f would not have the correct value but would still be handled as a 16 bytes value, so the test would not fail with "actual 35 != 43", but something like "actual 27254487904906932132179118915584 != 43" (43 << 64 | 35) I guess. I still need to sort this out.

And so indeed, the broken value is a big one:

(gdb) p skel->bss->t11_f $4 = 793209995169510719523 (gdb) p/x skel->bss->t11_f $5 = 0x2b0000000000000023 (gdb)

But we see the 35 (0x23) value in the error log because the formatters used in ASSERT_EQ truncate the actual value.

Alexis