Spent some time debugging this with a GDB script to print the number of
watches and timeouts, from that determined a way to trigger both, and
made a reliable reproducer - combining many parallel launches of avahi-
browse -a -t with a .local DNS lookup.

Don't really need the GDB script at this point because you can clearly
see the CPU usage growing, but attaching it out of interest. However the
cleanup_watches gets called so many times and GDB is so slow that ends
up being the bottleneck when the bug exists - 100% CPU on GDB processing
all of the breakpoints. Perhaps I should try bpftrace next time :)


** Attachment added: "gdbinit"
   
https://bugs.launchpad.net/ubuntu/+source/avahi/+bug/1799265/+attachment/5919883/+files/gdbinit

** Patch removed: "avahi-lp1799265-jammy.debdiff"
   
https://bugs.launchpad.net/ubuntu/+source/avahi/+bug/1799265/+attachment/5919602/+files/avahi-lp1799265-jammy.debdiff

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1799265

Title:
  avahi-daemon high cpu, unusable networking

To manage notifications about this bug go to:
https://bugs.launchpad.net/avahi/+bug/1799265/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to