On Wed, Mar 21, 2018 at 03:58:31PM -0700, Dan Williams wrote:
> xfs_break_dax_layouts(), similar to xfs_break_leased_layouts(), scans
> for busy / pinned dax pages and waits for those pages to go idle before
> any potential extent unmap operation.
>
> dax_layout_busy_page() handles synchronizing against new page-busy
> events (get_user_pages). It invalidates all mappings to trigger the
> get_user_pages slow path which will eventually block on the xfs inode
> lock held in XFS_MMAPLOCK_EXCL mode. If dax_layout_busy_page() finds a
> busy page it returns it for xfs to wait for the page-idle event that
> will fire when the page reference count reaches 1 (recall ZONE_DEVICE
> pages are idle at count 1, see generic_dax_pagefree()).
>
> While waiting, the XFS_MMAPLOCK_EXCL lock is dropped in order to not
> deadlock the process that might be trying to elevate the page count of
> more pages before arranging for any of them to go idle. I.e. the typical
> case of submitting I/O is that iov_iter_get_pages() elevates the
> reference count of all pages in the I/O before starting I/O on the first
> page. The process of elevating the reference count of all pages involved
> in an I/O may cause faults that need to take XFS_MMAPLOCK_EXCL.
>
> Cc: Jan Kara <[email protected]>
> Cc: Dave Chinner <[email protected]>
> Cc: "Darrick J. Wong" <[email protected]>
> Cc: Ross Zwisler <[email protected]>
> Reviewed-by: Christoph Hellwig <[email protected]>
> Signed-off-by: Dan Williams <[email protected]>
> ---
> fs/xfs/xfs_file.c | 59
> ++++++++++++++++++++++++++++++++++++++++++++---------
> 1 file changed, 49 insertions(+), 10 deletions(-)
>
> diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
> index 7f37fadf007e..d4573f93fddb 100644
> --- a/fs/xfs/xfs_file.c
> +++ b/fs/xfs/xfs_file.c
> @@ -752,6 +752,38 @@ xfs_file_write_iter(
> return ret;
> }
>
> +static void
> +xfs_wait_var_event(
> + struct inode *inode,
> + uint iolock,
> + bool *did_unlock)
> +{
> + struct xfs_inode *ip = XFS_I(inode);
> +
> + *did_unlock = true;
> + xfs_iunlock(ip, iolock);
> + schedule();
> + xfs_ilock(ip, iolock);
> +}
> +
> +static int
> +xfs_break_dax_layouts(
> + struct inode *inode,
> + uint iolock,
> + bool *did_unlock)
> +{
> + struct page *page;
> +
> + *did_unlock = false;
> + page = dax_layout_busy_page(inode->i_mapping);
> + if (!page)
> + return 0;
> +
> + return ___wait_var_event(&page->_refcount,
> + atomic_read(&page->_refcount) == 1, TASK_INTERRUPTIBLE,
> + 0, 0, xfs_wait_var_event(inode, iolock, did_unlock));
> +}
> +
> int
> xfs_break_layouts(
> struct inode *inode,
> @@ -766,16 +798,23 @@ xfs_break_layouts(
> | (reason == BREAK_UNMAPI
> ? XFS_MMAPLOCK_EXCL : 0)));
>
> - switch (reason) {
> - case BREAK_UNMAPI:
> - /* fall through */
> - case BREAK_WRITE:
> - error = xfs_break_leased_layouts(inode, iolock, &did_unlock);
> - break;
> - default:
> - error = -EINVAL;
> - break;
> - }
> + do {
> + switch (reason) {
> + case BREAK_UNMAPI:
> + error = xfs_break_dax_layouts(inode, *iolock,
> + &did_unlock);
> + /* fall through */
> + case BREAK_WRITE:
> + if (error || did_unlock)
> + break;
> + error = xfs_break_leased_layouts(inode, iolock,
> + &did_unlock);
> + break;
> + default:
> + error = -EINVAL;
> + break;
> + }
> + } while (error == 0 && did_unlock);
Nitpick: I think did_unlock should probably be called done in this context,
and how about something like this:
for (;;) {
switch (reason) {
case BREAK_UNMAPI:
error = xfs_break_dax_layouts(inode, *iolock, &done);
if (error || done)
return error;
/*fallthrough*/
case BREAK_WRITE:
error = xfs_break_leased_layouts(inode, iolock, &done);
if (error || done)
return error;
break;
default:
WARN_ON_ONCE(1);
return -EINVAL;
}
}
