[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netbsd-port-alpha
Subject:    Re: Panic on -current with MULTIPROCESSOR
From:       Jason Thorpe <thorpej () me ! com>
Date:       2021-04-03 15:04:52
Message-ID: 3C020AE7-0495-4680-8F0F-BDD941B7A1B8 () me ! com
[Download RAW message or body]


> On Apr 1, 2021, at 9:22 AM, Jason Thorpe <thorpej@me.com> wrote:
> 
> 
> > On Apr 1, 2021, at 9:18 AM, Jason Thorpe <thorpej@me.com> wrote:
> > 
> > #define SPINLOCK_BACKOFF(count)                                 \
> > do {                                                            \
> > int __i;                                                \
> > for (__i = (count); __i != 0; __i--) {                  \
> > SPINLOCK_BACKOFF_HOOK;                          \
> > }                                                       \
> > if ((count) < SPINLOCK_BACKOFF_MAX)                     \
> > (count) += (count);                             \
> > } while (/* CONSTCOND */ 0);
> 
> Sorry, that's the backoff hook … the spin hook is called in _kernel_lock() (i.e. \
> acquiring the big lock)…. Ah, and I think this is the problem… the spin hook \
> does not appear to be called in the spin mutex case.

To close the loop on this, I sent John a patch off-list, and it appears to have fixed \
the problem for him. I've gone ahead and checked it in:

	http://mail-index.netbsd.org/source-changes/2021/04/03/msg128464.html

Thanks for testing it, John!

-- thorpej


[prev in list] [next in list] [prev in thread] [next in thread]