[patch] mm: madvise avoid exclusive mmap_sem

[patch] mm: madvise avoid exclusive mmap_sem

[Nick Piggin]

Avoid down_write of the mmap_sem in madvise when we can help it.

Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>

Index: linux-2.6/mm/madvise.c
===================================================================
--- linux-2.6.orig/mm/madvise.c
+++ linux-2.6/mm/madvise.c
@@ -12,6 +12,24 @@
 #include <linux/hugetlb.h>
 
 /*
+ * Any behaviour which results in changes to the vma->vm_flags needs to
+ * take mmap_sem for writing. Others, which simply traverse vmas, need
+ * to only take it for reading.
+ */
+static int madvise_need_mmap_write(int behavior)
+{
+	switch (behavior) {
+	case MADV_REMOVE:
+	case MADV_WILLNEED:
+	case MADV_DONTNEED:
+		return 0;
+	default:
+		/* be safe, default to 1. list exceptions explicitly */
+		return 1;
+	}
+}
+
+/*
  * We can potentially split a vm area into separate
  * areas, each area with its own behavior.
  */
@@ -183,9 +201,9 @@ static long madvise_remove(struct vm_are
 			+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
 
 	/* vmtruncate_range needs to take i_mutex and i_alloc_sem */
-	up_write(&current->mm->mmap_sem);
+	up_read(&current->mm->mmap_sem);
 	error = vmtruncate_range(mapping->host, offset, endoff);
-	down_write(&current->mm->mmap_sem);
+	down_read(&current->mm->mmap_sem);
 	return error;
 }
 
@@ -270,7 +288,10 @@ asmlinkage long sys_madvise(unsigned lon
 	int error = -EINVAL;
 	size_t len;
 
-	down_write(&current->mm->mmap_sem);
+	if (madvise_need_mmap_write(behavior))
+		down_write(&current->mm->mmap_sem);
+	else
+		down_read(&current->mm->mmap_sem);
 
 	if (start & ~PAGE_MASK)
 		goto out;
@@ -332,6 +353,10 @@ asmlinkage long sys_madvise(unsigned lon
 			vma = find_vma(current->mm, start);
 	}
 out:
-	up_write(&current->mm->mmap_sem);
+	if (madvise_need_mmap_write(behavior))
+		up_write(&current->mm->mmap_sem);
+	else
+		up_read(&current->mm->mmap_sem);
+
 	return error;
 }

--
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>

Re: [patch] mm: madvise avoid exclusive mmap_sem

[Andrew Morton]

On Thu, 12 Apr 2007 02:56:38 +0200
Nick Piggin <npiggin@suse.de> wrote:

> Avoid down_write of the mmap_sem in madvise when we can help it.
> 
> Acked-by: Hugh Dickins <hugh@veritas.com>
> Signed-off-by: Nick Piggin <npiggin@suse.de>
> 
> Index: linux-2.6/mm/madvise.c
> ===================================================================
> --- linux-2.6.orig/mm/madvise.c
> +++ linux-2.6/mm/madvise.c
> @@ -12,6 +12,24 @@
>  #include <linux/hugetlb.h>
>  
>  /*
> + * Any behaviour which results in changes to the vma->vm_flags needs to
> + * take mmap_sem for writing. Others, which simply traverse vmas, need
> + * to only take it for reading.
> + */
> +static int madvise_need_mmap_write(int behavior)
> +{
> +	switch (behavior) {
> +	case MADV_REMOVE:
> +	case MADV_WILLNEED:
> +	case MADV_DONTNEED:
> +		return 0;
> +	default:
> +		/* be safe, default to 1. list exceptions explicitly */
> +		return 1;
> +	}
> +}

Are we sure that running zap_page_range() under down_read() is safe?
For hugepage regions too?

--
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>

Re: [patch] mm: madvise avoid exclusive mmap_sem

[Hugh Dickins]

On Fri, 13 Apr 2007, Andrew Morton wrote:
> On Thu, 12 Apr 2007 02:56:38 +0200
> Nick Piggin <npiggin@suse.de> wrote:
> 
> > Avoid down_write of the mmap_sem in madvise when we can help it.
> 
> Are we sure that running zap_page_range() under down_read() is safe?
> For hugepage regions too?

madvise_dontneed just says -EINVAL on a VM_HUGETLB vma, so I didn't
check the hugepage case further, it doesn't reach the zap_page_range.

We can indeed be sure that running zap_page_range on a normal vma
is safe under down_read (as opposed to the down_write there before),
because file truncation runs it without taking mmap_sem at all.

Hugh

--
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>