On Mon, 2024-07-01 at 17:42 -0700, Andrii Nakryiko wrote:
[...]
> You don't seem to have a case where offset is not a multiple of 8
> (though it would have to be a sub-register spill which would be
> "rejected" anyway, so not sure if there is anything to add here)
My bad, will add such a test-case.
>
> > diff --git a/tools/testing/selftests/bpf/prog_tests/verifier.c b/tools/testing/selftests/bpf/prog_tests/verifier.c
[...]
> > +void test_verifier_nocsr(void)
> > +{
> > +#if defined(__x86_64__)
> > + RUN(verifier_nocsr);
> > +#endif /* __x86_64__ */
>
> maybe #else <mark-as-skipped> ?
Right, makes sense.
[...]
> > +++ b/tools/testing/selftests/bpf/progs/verifier_nocsr.c
> > @@ -0,0 +1,437 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +
> > +#include <linux/bpf.h>
> > +#include <bpf/bpf_helpers.h>
> > +#include "bpf_misc.h"
> > +
> > +#define __xlated_bpf_get_smp_processor_id \
> > + __xlated(": w0 = ") \
>
> how will this work for no_alu32 mode?
The patch is applied by kernel, and it does not care about alu32
compiler flags:
insn_buf[0] = BPF_MOV32_IMM(BPF_REG_0, (u32)(unsigned long)&pcpu_hot.cpu_number);
insn_buf[1] = BPF_MOV64_PERCPU_REG(BPF_REG_0, BPF_REG_0);
insn_buf[2] = BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, 0);
[...]
