Commit major SMP cleanups and move the BGL (big giant lock) in the

    syscall path inward.  A system call may select whether it needs the MP
    lock or not (the default being that it does need it).

    A great deal of conditional SMP code for various deadended experiments
    has been removed.  'cil' and 'cml' have been removed entirely, and the
    locking around the cpl has been removed.  The conditional
    separately-locked fast-interrupt code has been removed, meaning that
    interrupts must hold the CPL now (but they pretty much had to anyway).
    Another reason for doing this is that the original separate-lock for
    interrupts just doesn't apply to the interrupt thread mechanism being
    contemplated.

    Modifications to the cpl may now ONLY occur while holding the MP
    lock.  For example, if an otherwise MP safe syscall needs to mess with
    the cpl, it must hold the MP lock for the duration and must (as usual)
    save/restore the cpl in a nested fashion.

    This is precursor work for the real meat coming later: avoiding having
    to hold the MP lock for common syscalls and I/O's and interrupt threads.
    It is expected that the spl mechanisms and new interrupt threading
    mechanisms will be able to run in tandem, allowing a slow piecemeal
    transition to occur.

    This patch should result in a moderate performance improvement due to
    the considerable amount of code that has been removed from the critical
    path, especially the simplification of the spl*() calls.  The real
    performance gains will come later.

Approved by: jkh
Reviewed by: current, bde (exception.s)
Some work taken from: luoqi's patch

1 files changed, 12 insertions, 6 deletions

diff --git a/sys/amd64/include/cpu.h b/sys/amd64/include/cpu.h
index 18049d0..c6aa46f 100644
--- a/sys/amd64/include/cpu.h
+++ b/sys/amd64/include/cpu.h
@@ -82,10 +82,13 @@
 /*
  * Preempt the current process if in interrupt from user mode,
  * or after the current trap/syscall if in system mode.
+ *
+ * XXX: if astpending is later changed to an |= here due to more flags being
+ * added, we will have an atomicy problem.  The type of atomicy we need is
+ * a non-locked orl.
  */
-#define	need_resched()		do { want_resched = 1; aston(); } while (0)
-
-#define	resched_wanted()	want_resched
+#define	need_resched()		do { astpending = AST_RESCHED|AST_PENDING; } while (0)
+#define	resched_wanted()	(astpending & AST_RESCHED)
 
 /*
  * Arrange to handle pending profiling ticks before returning to user mode.
@@ -100,10 +103,15 @@
 /*
  * Notify the current process (p) that it has a signal pending,
  * process as soon as possible.
+ *
+ * XXX: aston() really needs to be an atomic (not locked, but an orl),
+ * in case need_resched() is set by an interrupt.  But with astpending a
+ * per-cpu variable this is not trivial to do efficiently.  For now we blow
+ * it off (asynchronous need_resched() conflicts are not critical).
  */
 #define	signotify(p)	aston()
 
-#define	aston()			do { astpending = 1; } while (0)
+#define	aston()			do { astpending |= AST_PENDING; } while (0)
 #define astoff()
 
 /*
@@ -126,11 +134,9 @@
 }
 
 #ifdef _KERNEL
-extern int	astpending;
 extern char	btext[];
 extern char	etext[];
 extern u_char	intr_nesting_level;
-extern int	want_resched;	/* resched was called */
 
 void	fork_trampoline __P((void));
 void	fork_return __P((struct proc *, struct trapframe));