Page MenuHomeFreeBSD

Allow the BPF to be select for EVFILT_WRITE with kqueue.
ClosedPublic

Authored by harti on Sep 15 2021, 8:35 AM.
Tags
None
Referenced Files
F132507478: D31967.id.diff
Fri, Oct 17, 12:18 PM
F132507472: D31967.id96788.diff
Fri, Oct 17, 12:18 PM
F132507470: D31967.id95958.diff
Fri, Oct 17, 12:18 PM
F132507465: D31967.id95191.diff
Fri, Oct 17, 12:18 PM
F132460259: D31967.diff
Fri, Oct 17, 3:07 AM
Unknown Object (File)
Sun, Oct 5, 11:01 PM
Unknown Object (File)
Sep 11 2025, 5:35 PM
Unknown Object (File)
Sep 11 2025, 5:22 PM
Subscribers

Details

Summary

Allow the BPF to be select for EVFILT_WRITE with kqueue. This is required for Boost asio to be able to use a BPF file descriptor even if it is used only for reading. Asio insists on installing a write filter.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

harti requested review of this revision.Sep 15 2021, 8:35 AM
sys/net/bpf.c
2193

I'd prefer a bit less code duplication:

struct filterops *ops;

...

switch (kn->kn_filter) {
case EVFILT_READ:
   ops = &bpfread_filtops;
   break;
case EVFILT_WRITE:
   ops = &bpwrite_filtops;
   break;
default:
   return (EINVAL);
}

...

Also it seems you need to modify kqueue(2).

Reduce code duplication.
Update kqueue.2.

This revision is now accepted and ready to land.Oct 1 2021, 8:03 AM