Re: bpf_errno. Was: [PATCH RFC bpf-next 1/3] bpf: report probe fault to BPF stderr

On 2025/10/11 02:55, Eduard Zingerman wrote:

On Fri, 2025-10-10 at 20:05 +0800, Menglong Dong wrote:

[...]

quoted

save errno to r0(Eduard)
-----------------------------------
Save the errno to r0 in the exception handler of BPF_PROBE_MEM,
and read r0 with a __kfun in BPF program. (Not sure if I understand
it correctly).

This sounds effective, but won't this break the usage of r0? I mean,
the r0 can be used by the BPF program somewhere.

What I meant is that for cases when someone wants to check for memory
access error, there is already bpf_probe_read_kernel(). It's return
value in r0 and is defined for both success and failure cases.

The problem with it, is that it has a function call overhead.
But we can workaround that for 1,2,4,8 byte accesses, by replacing
helper call by some `BPF_LDX | BPF_PROBE_MEM1 | <size>`,
where BPF_PROBE_MEM1 is different from BPF_PROBE_MEM and tells
jit that exception handler for this memory access needs to set
r0 to -EFAULT if it is executed.

The inconvenient part here is that one can't do chaining,
like a->b->c, using bpf_probe_read_kernel().
One needs to insert bpf_probe_read_kernel() call at each step of a
chain, which is a bit of a pain.  Maybe it can be alleviated using
some vararg macro.

Thanks for the explication, and I see now.

Interesting idea, and I think this is something that we can do
despite this problem, which can optimize the performance of
bpf_probe_read_kernel().

Thanks!
Menglong Dong

[...]