Re: [PATCH 4/7] x86,tlb: make lazy TLB mode lazier

[Peter Zijlstra]

On Thu, Jul 19, 2018 at 09:45:47AM -0700, Andy Lutomirski wrote:
> After some grepping, there are very few users.  The
> only nontrivial ones are the ones in kernel/ and mm/mmu_context.c that
> are involved in the rather complicated dance of refcounting active_mm.

Something like so should work I suppose.

It keeps the field, but ensure that for all tasks tsk->active_mm ==
tsk->mm. The implication is that switch_mm(.prev == NULL) should work
(it does for x86, we don't care about @prev).


diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index d84eafcf5e29..7bfd850c6bf7 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2780,12 +2780,8 @@ static __always_inline struct rq *
 context_switch(struct rq *rq, struct task_struct *prev,
               struct task_struct *next, struct rq_flags *rf)
 {
-       struct mm_struct *mm, *oldmm;
-
        prepare_task_switch(rq, prev, next);
 
-       mm = next->mm;
-       oldmm = prev->active_mm;
        /*
         * For paravirt, this is coupled with an exit in switch_to to
         * combine the page table reload and the switch backend into
@@ -2800,16 +2796,32 @@ context_switch(struct rq *rq, struct task_struct *prev,
         * membarrier after storing to rq->curr, before returning to
         * user-space.
         */
-       if (!mm) {
-               next->active_mm = oldmm;
-               mmgrab(oldmm);
-               enter_lazy_tlb(oldmm, next);
-       } else
-               switch_mm_irqs_off(oldmm, mm, next);
-
-       if (!prev->mm) {
-               prev->active_mm = NULL;
-               rq->prev_mm = oldmm;
+
+       /*
+        * kernel -> kernel   lazy + transfer active
+        *   user -> kernel   lazy + mmgrab() active
+        *
+        * kernel ->   user   switch + mmdrop() active
+        *   user ->   user   switch
+        */
+       if (!next->mm) {                                // to kernel
+               enter_lazy_tlb(prev->active_mm, next);
+
+#ifdef ARCH_NO_ACTIVE_MM
+               next->active_mm = prev->active_mm;
+               if (prev->mm)                           // from user
+                       mmgrab(prev->active_mm);
+#endif
+       } else {                                        // to user
+               switch_mm_irqs_off(prev->active_mm, next->mm, next);
+
+#ifdef ARCH_NO_ACTIVE_MM
+               if (!prev->mm) {                        // from kernel
+                       /* will mmdrop() in finish_task_switch(). */
+                       rq->prev_mm = prev->active_mm;
+                       prev->active_mm = NULL;
+               }
+#endif
        }
 
        rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
diff --git a/mm/mmu_context.c b/mm/mmu_context.c
index 3e612ae748e9..770bb615d115 100644
--- a/mm/mmu_context.c
+++ b/mm/mmu_context.c
@@ -12,6 +12,10 @@
 
 #include <asm/mmu_context.h>
 
+#ifndef finish_arch_post_lock_switch
+# define finish_arch_post_lock_switch()        do { } while (0)
+#endif
+
 /*
  * use_mm
  *     Makes the calling kernel thread take on the specified
@@ -33,12 +37,14 @@ void use_mm(struct mm_struct *mm)
        tsk->mm = mm;
        switch_mm(active_mm, mm, tsk);
        task_unlock(tsk);
-#ifdef finish_arch_post_lock_switch
        finish_arch_post_lock_switch();
-#endif
 
+#ifdef ARCH_NO_ACTIVE_MM
+       WARN_ON_ONCE(active_mm != NULL);
+#else
        if (active_mm != mm)
                mmdrop(active_mm);
+#endif
 }
 EXPORT_SYMBOL_GPL(use_mm);
 
@@ -57,8 +63,22 @@ void unuse_mm(struct mm_struct *mm)
        task_lock(tsk);
        sync_mm_rss(mm);
        tsk->mm = NULL;
+
+       WARN_ON_ONCE(tsk->active_mm != mm);
+
+#ifdef ARCH_NO_ACTIVE_MM
+       switch_mm(tsk->active_mm, &init_mm, tsk);
+       mmdrop(tsk->active_mm);
+       tsk->active_mm = NULL;
+#else
        /* active_mm is still 'mm' */
        enter_lazy_tlb(mm, tsk);
+#endif
+
        task_unlock(tsk);
+
+#ifdef ARCH_NO_ACTIVE_MM
+       finish_arch_post_lock_switch();
+#endif
 }
 EXPORT_SYMBOL_GPL(unuse_mm);