Re: [Potential Spoof] [PATCH] BUG/MAJOR: fd/threads, task/threads: ensure all spin locks are unlocked

2019-02-26 Thread Richard Russo
Olivier, I've been testing from master yesterday and today (with a reload every 60 second loop). As with any multithreaded testing, it's much easier to conclude something crashes than that it doesn't crash, but I didn't see any crashes or deadlocks in my test runs. So, I think it's good.

Re: [Potential Spoof] [PATCH] BUG/MAJOR: fd/threads, task/threads: ensure all spin locks are unlocked

2019-02-25 Thread Olivier Houchard
Hi Richard, On Wed, Feb 20, 2019 at 11:58:42PM +, Richard Russo wrote: > While continuing to test this, I ended up with a crash in > listener.c:listener_accept on a closed/closing listen socket where > fdtab[fd].owner is NULL by the time the thread gets there. This is possible > because

Re: [Potential Spoof] [PATCH] BUG/MAJOR: fd/threads, task/threads: ensure all spin locks are unlocked

2019-02-20 Thread Richard Russo
While continuing to test this, I ended up with a crash in listener.c:listener_accept on a closed/closing listen socket where fdtab[fd].owner is NULL by the time the thread gets there. This is possible because the fd.c: fdlist_process_cached_events unlocks the spinlock before calling