Re: [PATCH 08/20] mm: Optimize fullmm TLB flushing

[Catalin Marinas <catalin.marinas@arm.com>]

On Thu, Jun 28, 2012 at 10:57:21PM +0100, Benjamin Herrenschmidt wrote:
> On Thu, 2012-06-28 at 18:52 +0200, Peter Zijlstra wrote:
> > No I think you're right (as always).. also an IPI will not force
> > schedule the thread that might be running on the receiving cpu, also
> > we'd have to wait for any such schedule to complete in order to
> > guarantee the mm isn't lazily used anymore.
> > 
> > Bugger.. 
> 
> You can still do it if the mm count is 1 no ? Ie, current is the last
> holder of a reference to the mm struct... which will probably be the
> common case for short lived programs.

BTW, can we not move the free_pgtables() call in exit_mmap() to
__mmdrop()? Something like below but I'm not entirely sure about its
implications:


diff --git a/include/linux/mm.h b/include/linux/mm.h
index b36d08c..507ee9f 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1372,6 +1372,7 @@ extern void unlink_file_vma(struct vm_area_struct *);
 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
 	unsigned long addr, unsigned long len, pgoff_t pgoff);
 extern void exit_mmap(struct mm_struct *);
+extern void exit_pgtables(struct mm_struct *mm);
 
 extern int mm_take_all_locks(struct mm_struct *mm);
 extern void mm_drop_all_locks(struct mm_struct *mm);
diff --git a/kernel/fork.c b/kernel/fork.c
index ab5211b..3412b1a 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -588,6 +588,7 @@ struct mm_struct *mm_alloc(void)
 void __mmdrop(struct mm_struct *mm)
 {
 	BUG_ON(mm == &init_mm);
+	exit_pgtables(mm);
 	mm_free_pgd(mm);
 	destroy_context(mm);
 	mmu_notifier_mm_destroy(mm);
diff --git a/mm/mmap.c b/mm/mmap.c
index 074b487..d9ebfdb 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2269,7 +2269,6 @@ void exit_mmap(struct mm_struct *mm)
 {
 	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
-	unsigned long nr_accounted = 0;
 
 	/* mm's last user has gone, and its about to be pulled down */
 	mmu_notifier_release(mm);
@@ -2291,11 +2290,23 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb_gather_mmu(&tlb, mm, 1);
+	tlb_gather_mmu(&tlb, mm, 0);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	unmap_vmas(&tlb, vma, 0, -1);
+	tlb_finish_mmu(&tlb, 0, -1);
+}
+
+void exit_pgtables(struct mm_struct *mm)
+{
+	struct mmu_gather tlb;
+	struct vm_area_struct *vma;
+	unsigned long nr_accounted = 0;
 
+	vma = mm->mmap;
+	if (!vma)
+		return;
+	tlb_gather_mmu(&tlb, mm, 1);
 	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, TASK_SIZE);
 	tlb_finish_mmu(&tlb, 0, -1);
 

-- 
Catalin

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>