[PATCH 1/2] ext4: Close race between direct IO and ext4_break_layouts()

[PATCH 1/2] ext4: Close race between direct IO and ext4_break_layouts()

[Dave Jiang]

From: Ross Zwisler <zwisler-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>

If the refcount of a page is lowered between the time that it is returned
by dax_busy_page() and when the refcount is again checked in
ext4_break_layouts() => ___wait_var_event(), the waiting function
ext4_wait_dax_page() will never be called.  This means that
ext4_break_layouts() will still have 'retry' set to false, so we'll stop
looping and never check the refcount of other pages in this inode.

Instead, always continue looping as long as dax_layout_busy_page() gives us
a page which it found with an elevated refcount.

Note that this works around the race exposed by my unit test, but I think
that there is another race that needs to be addressed, probably with
additional synchronization added between direct I/O and
{ext4,xfs}_break_layouts().

Signed-off-by: Ross Zwisler <ross.zwisler-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
Reviewed-by: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>
---
 fs/ext4/inode.c |    9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 8f6ad7667974..d2663a1e3ec2 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -4191,9 +4191,8 @@ int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
 	return 0;
 }
 
-static void ext4_wait_dax_page(struct ext4_inode_info *ei, bool *did_unlock)
+static void ext4_wait_dax_page(struct ext4_inode_info *ei)
 {
-	*did_unlock = true;
 	up_write(&ei->i_mmap_sem);
 	schedule();
 	down_write(&ei->i_mmap_sem);
@@ -4203,14 +4202,12 @@ int ext4_break_layouts(struct inode *inode)
 {
 	struct ext4_inode_info *ei = EXT4_I(inode);
 	struct page *page;
-	bool retry;
 	int error;
 
 	if (WARN_ON_ONCE(!rwsem_is_locked(&ei->i_mmap_sem)))
 		return -EINVAL;
 
 	do {
-		retry = false;
 		page = dax_layout_busy_page(inode->i_mapping);
 		if (!page)
 			return 0;
@@ -4218,8 +4215,8 @@ int ext4_break_layouts(struct inode *inode)
 		error = ___wait_var_event(&page->_refcount,
 				atomic_read(&page->_refcount) == 1,
 				TASK_INTERRUPTIBLE, 0, 0,
-				ext4_wait_dax_page(ei, &retry));
-	} while (error == 0 && retry);
+				ext4_wait_dax_page(ei));
+	} while (error == 0);
 
 	return error;
 }

[PATCH 2/2] [PATCH] xfs: Close race between direct IO and xfs_break_layouts()

[Dave Jiang]

This patch is the duplicate of ross's fix for ext4 for xfs.

If the refcount of a page is lowered between the time that it is returned
by dax_busy_page() and when the refcount is again checked in
xfs_break_layouts() => ___wait_var_event(), the waiting function
xfs_wait_dax_page() will never be called.  This means that
xfs_break_layouts() will still have 'retry' set to false, so we'll stop
looping and never check the refcount of other pages in this inode.

Instead, always continue looping as long as dax_layout_busy_page() gives us
a page which it found with an elevated refcount.

Signed-off-by: Dave Jiang <dave.jiang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 fs/xfs/xfs_file.c |    7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index a3e7767a5715..666c93fe5284 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -721,12 +721,10 @@ xfs_file_write_iter(
 
 static void
 xfs_wait_dax_page(
-	struct inode		*inode,
-	bool			*did_unlock)
+	struct inode		*inode)
 {
 	struct xfs_inode        *ip = XFS_I(inode);
 
-	*did_unlock = true;
 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL);
 	schedule();
 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
@@ -746,9 +744,10 @@ xfs_break_dax_layouts(
 	if (!page)
 		return 0;
 
+	*did_unlock = true;
 	return ___wait_var_event(&page->_refcount,
 			atomic_read(&page->_refcount) == 1, TASK_INTERRUPTIBLE,
-			0, 0, xfs_wait_dax_page(inode, did_unlock));
+			0, 0, xfs_wait_dax_page(inode));
 }
 
 int

Re: [PATCH 1/2] ext4: Close race between direct IO and ext4_break_layouts()

[Jan Kara]

On Tue 07-08-18 15:11:37, Dave Jiang wrote:
> From: Ross Zwisler <zwisler-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
> 
> If the refcount of a page is lowered between the time that it is returned
> by dax_busy_page() and when the refcount is again checked in
> ext4_break_layouts() => ___wait_var_event(), the waiting function
> ext4_wait_dax_page() will never be called.  This means that
> ext4_break_layouts() will still have 'retry' set to false, so we'll stop
> looping and never check the refcount of other pages in this inode.
> 
> Instead, always continue looping as long as dax_layout_busy_page() gives us
> a page which it found with an elevated refcount.
> 
> Note that this works around the race exposed by my unit test, but I think
> that there is another race that needs to be addressed, probably with
> additional synchronization added between direct I/O and
> {ext4,xfs}_break_layouts().

I'd just note that the race Ross suspected should be properly handled by
dax_layout_busy_page() so I think this last paragraph from the changelog
can go.

Also Ted, this fixes a problem in the DAX truncate patches you currently
carry in your tree so you can consider just pushing it with them during the
merge window. It's not necessary though - the patches already make the
problematic behavior much less likely, this patch just hopefully completely
closes the race window.

> Signed-off-by: Ross Zwisler <ross.zwisler-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
> Reviewed-by: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>

								Honza

> ---
>  fs/ext4/inode.c |    9 +++------
>  1 file changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 8f6ad7667974..d2663a1e3ec2 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -4191,9 +4191,8 @@ int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
>  	return 0;
>  }
>  
> -static void ext4_wait_dax_page(struct ext4_inode_info *ei, bool *did_unlock)
> +static void ext4_wait_dax_page(struct ext4_inode_info *ei)
>  {
> -	*did_unlock = true;
>  	up_write(&ei->i_mmap_sem);
>  	schedule();
>  	down_write(&ei->i_mmap_sem);
> @@ -4203,14 +4202,12 @@ int ext4_break_layouts(struct inode *inode)
>  {
>  	struct ext4_inode_info *ei = EXT4_I(inode);
>  	struct page *page;
> -	bool retry;
>  	int error;
>  
>  	if (WARN_ON_ONCE(!rwsem_is_locked(&ei->i_mmap_sem)))
>  		return -EINVAL;
>  
>  	do {
> -		retry = false;
>  		page = dax_layout_busy_page(inode->i_mapping);
>  		if (!page)
>  			return 0;
> @@ -4218,8 +4215,8 @@ int ext4_break_layouts(struct inode *inode)
>  		error = ___wait_var_event(&page->_refcount,
>  				atomic_read(&page->_refcount) == 1,
>  				TASK_INTERRUPTIBLE, 0, 0,
> -				ext4_wait_dax_page(ei, &retry));
> -	} while (error == 0 && retry);
> +				ext4_wait_dax_page(ei));
> +	} while (error == 0);
>  
>  	return error;
>  }
> 
-- 
Jan Kara <jack-IBi9RG/b67k@public.gmane.org>
SUSE Labs, CR

Re: [PATCH 2/2] [PATCH] xfs: Close race between direct IO and xfs_break_layouts()

[Jan Kara]

On Tue 07-08-18 15:11:43, Dave Jiang wrote:
> This patch is the duplicate of ross's fix for ext4 for xfs.
> 
> If the refcount of a page is lowered between the time that it is returned
> by dax_busy_page() and when the refcount is again checked in
> xfs_break_layouts() => ___wait_var_event(), the waiting function
> xfs_wait_dax_page() will never be called.  This means that
> xfs_break_layouts() will still have 'retry' set to false, so we'll stop
> looping and never check the refcount of other pages in this inode.
> 
> Instead, always continue looping as long as dax_layout_busy_page() gives us
> a page which it found with an elevated refcount.
> 
> Signed-off-by: Dave Jiang <dave.jiang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The patch looks good to me. You can add:

Reviewed-by: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>

Just one minor nit below:

> @@ -746,9 +744,10 @@ xfs_break_dax_layouts(
>  	if (!page)
>  		return 0;
>  
> +	*did_unlock = true;

I think it would be more understandable to name the argument of
xfs_break_dax_layouts() as 'retry' instead of 'did_unlock' as it's not
about unlocking anymore.

							Honza
-- 
Jan Kara <jack-IBi9RG/b67k@public.gmane.org>
SUSE Labs, CR

Re: [PATCH 2/2] [PATCH] xfs: Close race between direct IO and xfs_break_layouts()

[Dave Jiang]

On 08/08/2018 01:53 AM, Jan Kara wrote:
> On Tue 07-08-18 15:11:43, Dave Jiang wrote:
>> This patch is the duplicate of ross's fix for ext4 for xfs.
>>
>> If the refcount of a page is lowered between the time that it is returned
>> by dax_busy_page() and when the refcount is again checked in
>> xfs_break_layouts() => ___wait_var_event(), the waiting function
>> xfs_wait_dax_page() will never be called.  This means that
>> xfs_break_layouts() will still have 'retry' set to false, so we'll stop
>> looping and never check the refcount of other pages in this inode.
>>
>> Instead, always continue looping as long as dax_layout_busy_page() gives us
>> a page which it found with an elevated refcount.
>>
>> Signed-off-by: Dave Jiang <dave.jiang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> 
> The patch looks good to me. You can add:
> 
> Reviewed-by: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>
> 
> Just one minor nit below:
> 
>> @@ -746,9 +744,10 @@ xfs_break_dax_layouts(
>>  	if (!page)
>>  		return 0;
>>  
>> +	*did_unlock = true;
> 
> I think it would be more understandable to name the argument of
> xfs_break_dax_layouts() as 'retry' instead of 'did_unlock' as it's not
> about unlocking anymore.

Thanks for the review Jan! I will change. I was trying to decide between
less code change vs more clear definition. :)

> 
> 							Honza
>