On Fri, Jun 09, 2023 at 09:24:10AM +0100, Mark Rutland wrote:

On Thu, Jun 08, 2023 at 04:55:40PM -0700, Andrii Nakryiko wrote:

quoted

On Thu, Jun 8, 2023 at 4:27 PM Steven Rostedt [off-list ref] wrote:

quoted

On Thu, 8 Jun 2023 15:43:03 -0700 Andrii Nakryiko [off-list ref] wrote:

quoted

On Thu, Jun 8, 2023 at 2:26 PM Jiri Olsa [off-list ref] wrote:

quoted

There are BPF tools that allow user to specify regex/glob of kernel
functions to attach to. This regex/glob is checked against
available_filter_functions to check which functions are traceable. All
good. But then also it's important to have corresponding memory
addresses for selected functions (for many reasons, e.g., to have
non-ambiguous and fast attachment by address instead of by name, or
for some post-processing based on captured IP addresses, etc). And
that means that now we need to also parse /proc/kallsyms and
cross-join it with data fetched from available_filter_functions.

All this is unnecessary if avalable_filter_functions would just
provide function address in the first place. It's a huge
simplification. And saves memory and CPU.

Do you need the address of the function entry-point or the address of the
patch-site within the function? Those can differ, and the rec->ip address won't
necessarily equal the address in /proc/kallsyms, so the pointer in
/proc/kallsyms won't (always) match the address we could print for the ftrace site.

On arm64, today we can have offsets of +0, +4, and +8, and within a single
kernel image different functions can have different offsets. I suspect in
future that we may have more potential offsets (e.g. due to changes for HW/SW
CFI).

so we need that for kprobe_multi bpf link, which is based on fprobe,
and that uses ftrace_set_filter_ips to setup the ftrace_ops filter

and ftrace_set_filter_ips works fine with ip address being the address
of the patched instruction (it's matched in ftrace_location)

but right, I did not realize this.. it might cause confusion if people
don't know it's patch-side addresses..  not sure if there's easy way to
get real function address out of rec->ip, but it will also get more
complicated on x86 when IBT is enabled, will check

or we could just use patch-side addresses and reflect that in the file's
name like 'available_filter_functions_patch_addrs' .. it's already long
name ;-)

jirka