Re: [Patch bpf-next v5 1/3] bpf: introduce timeout hash map

On Fri, Jan 29, 2021 at 6:14 AM Jamal Hadi Salim [off-list ref] wrote:

On 2021-01-29 9:06 a.m., Jamal Hadi Salim wrote:

quoted

Which leads to:
Why not extend the general feature so one can register for optional
callbacks not just for expire but also add/del/update on specific
entries or table?
add/del/update could be sourced from other kernel programs or user space
and the callback would be invoked before an entry is added/deleted etc.
(just like it is here for expiry).

Sorry - shouldve read the rest of the thread:
Agree with Cong that you want per-map but there are use cases where you
want it per entry (eg the add/del/update case).

That was my point as well.
bpf_timer api should be generic, so that users can do both.
The program could use bpf_timer one for each flow and bpf_timer for each map.
And timers without maps.

On 2021-01-29 10:14 p.m., Alexei Starovoitov wrote:

On Fri, Jan 29, 2021 at 6:14 AM Jamal Hadi Salim [off-list ref] wrote:

quoted

On 2021-01-29 9:06 a.m., Jamal Hadi Salim wrote:

quoted

Which leads to:
Why not extend the general feature so one can register for optional
callbacks not just for expire but also add/del/update on specific
entries or table?
add/del/update could be sourced from other kernel programs or user space
and the callback would be invoked before an entry is added/deleted etc.
(just like it is here for expiry).

Sorry - shouldve read the rest of the thread:
Agree with Cong that you want per-map but there are use cases where you
want it per entry (eg the add/del/update case).

That was my point as well.
bpf_timer api should be generic, so that users can do both.
The program could use bpf_timer one for each flow and bpf_timer for each map.
And timers without maps.

I like it. Sensible to also have callback invocations for map
changes i.e entry create/update/delete (maybe map create/destroy).

cheers,
jamal