Re: [PATCH bpf-next v12 1/5] bpf: Move constants blinding out of arch-specific JITs

[Xu Kuohai]

On 4/4/2026 12:04 AM, Emil Tsalapatis wrote:
> >   	/* Pass 3: Adjust jump offset and write final image */
> >   	if (build_body(&ctx, extra_pass) ||
> > -		WARN_ON_ONCE(ctx.idx != ctx.epilogue_offset)) {
> > -		prog = orig_prog;
> > +		WARN_ON_ONCE(ctx.idx != ctx.epilogue_offset))
> This thunk is slightly different now, the WARN_ON_ONCE() won't be checked
> if build_body() succeeds.

Sorry, I do not see the difference here. The WARN_ON_ONCE() is still checked
when build_body() returns 0 (succeeds) due to the '||' evaluation.

> Do we even need it? AFAICT the only case it
> wouldn't trigger if build_body() fails is if it did so at the very last
> instruction. Alternatively, should we check it if build_body() succeeds
> instead to retain the old behavior?
>
> >   		goto out_free_hdr;
> > -	}
> >