Re: [PATCH bpf-next v2 1/6] kallsyms: Add KALLSYMS_QUERY ioctl to /proc/kallsyms

[Samuel Wu]

On Fri, Oct 30, 2026 at 3:02 AM David Laight
<david.laight.linux@xxxxxxxxx> wrote:
>
> On Thu, 29 Oct 2026 15:05:45 -0700
> Samuel Wu <wusamuel@xxxxxxxxxx> wrote:
[...]
> > BY_NAME matches names exactly, like kallsyms_lookup_name(), so they
> > would be different symbols at different addresses. Please correct me
> > if I'm missing something here!
>
> Only that the user might really want one of the copies.
> Sometimes there is only the copy.

Good point. I'll have v3 add KALLSYMS_F_PREFIX + pos/next_pos
pagination to cover these. A prefix query for "foo." also finds the
copies when plain foo does not exist (expanded on this in my reply to
Andrii's Oct 30 mail in this thread).

[...]
> > > Have you tried measuring it with my patches that speed up the kernel
> > > 'search by name' code?
> > > That speeds the lookup by a factor of ~18 (testing the code in userspace).
> >
> > Not yet. Can you point me to the latest version? Happy to try it out.
>
> I've just forwarded you all the patches.
> (They include patches to the tests and scripts/kallsyms)
> I'll copy you (and the other people getting this email) for v3.

Thanks for forwarding the patches! I'll be looking out for your v3. I
tested them across baseline, baseline + my series, and baseline + my
series + your series, using the same two benchmarks and setups as v2
patchseries:

Loading a BPF program with 11 typeless ksyms via libbpf (8-CPU x86-64
Cuttlefish VM, 273k /proc/kallsyms lines; BY_NAME, unfiltered, so each
query also scans ~35.8k module symbols for ambiguity):
- Base: 156.1 ms
- Base + my v3 series: 2.05 ms
- Base + my v3 series + your v2 series: 1.95 ms

Perfetto symbolizer (8-CPU x86-64 VM, 175k /proc/kallsyms lines;
BY_ADDR; init / 50k cached symbolizations):
- Base: 129.1 ms / 28.0 ms
- Base + my v3 series: 0.04 ms / 3.4 ms
- Base + my v3 series + your v2 series: 0.04 ms / 3.1 ms

I'm seeing your series help most when the vmlinux name search
dominates: on the Cuttlefish VM, a vmlinux-filtered BY_NAME call is
5-7x faster (hit: 8.3 us -> 1.7 us, miss: 9.5 us -> 1.4 us).

[...]
> > > > +     __u64 query_flags;      /* in - query mode (enum kallsyms_query_flags) */
>
> There is no need for a command here - use the ioctl cmd.

Agreed, done in v3