"Scheduling while atomic"

walt wa1ter at myrealbox.com
Mon Nov 20 15:11:28 PST 2006


Matthew Dillon wrote:
>     ...it is not ok to schedule while holding a spinlock.  A panic
>     *will* occur if you tried to block while holding a spinlock.

Just today in the commits list you posted:

  Use spinlocks instead of tokens to interlock the objcache depot.
  The cpu-localized critical path is still lockless.

The problem I face is that I've not spent a lifetime dissecting
these problems the way you have, and yet I would like to understand
the problem space in at least a general sense even though I will
(sadly) never be a contributing kernel hacker.

What I really need/want, I guess, is a kernel-hacking-for-dummies
handbook so that I can understand the problems, at least, even if
I can't contribute any solutions.

[Heh!  After reading the paragraph above, I am reminded of the many
questions my wife asks me, which all basically translate to this one
question:  'Explain to me why men think the weird way that men think!']

I hope you can answer my question better than I've been able to
answer her question ;o)





More information about the Users mailing list