[v2,4/5] swapfile: use ->read_iter and ->write_iter

Commit Message

Omar Sandoval Dec. 20, 2014, 3:18 a.m. UTC

Using ->direct_IO and ->readpage for the generic swap file
infrastructure requires all sorts of nasty workarounds. ->readpage
implementations don't play nicely with swap cache pages, and ->direct_IO
implementations have different locking conventions for every filesystem.
Instead, use ->read_iter/->write_iter with an ITER_BVEC and let the
filesystem take care of it. This will also allow us to easily transition
to kernel AIO if that gets merged in the future.

Cc: Mel Gorman <[e-mail]>
Signed-off-by: Omar Sandoval <[e-mail]>
---
 mm/page_io.c  | 30 +++++++++++++++++++++++-------
 mm/swapfile.c | 11 ++++++++++-
 2 files changed, 33 insertions(+), 8 deletions(-)

Comments

Al Viro Dec. 20, 2014, 6:13 a.m. UTC | #1

On Fri, Dec 19, 2014 at 07:18:28PM -0800, Omar Sandoval wrote:

> +		ret = swap_file->f_op->read_iter(&kiocb, &to);
> +		if (ret == PAGE_SIZE) {
> +			SetPageUptodate(page);
>  			count_vm_event(PSWPIN);
> +			ret = 0;
> +		} else {
> +			ClearPageUptodate(page);
> +			SetPageError(page);
> +		}
> +		unlock_page(page);

Umm...  What's to guarantee that ->read_iter() won't try lock_page() on what
turns out to be equal to page?
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Omar Sandoval Dec. 22, 2014, 7:32 a.m. UTC | #2

On Sat, Dec 20, 2014 at 06:13:37AM +0000, Al Viro wrote:
> On Fri, Dec 19, 2014 at 07:18:28PM -0800, Omar Sandoval wrote:
> 
> > +		ret = swap_file->f_op->read_iter(&kiocb, &to);
> > +		if (ret == PAGE_SIZE) {
> > +			SetPageUptodate(page);
> >  			count_vm_event(PSWPIN);
> > +			ret = 0;
> > +		} else {
> > +			ClearPageUptodate(page);
> > +			SetPageError(page);
> > +		}
> > +		unlock_page(page);
> 
> Umm...  What's to guarantee that ->read_iter() won't try lock_page() on what
> turns out to be equal to page?

Ergh. I don't see why ->read_iter would be screwing around in the swap
cache or with the pages in the iterator, anything in particular you can
see happening?

Patch

diff --git a/mm/page_io.c b/mm/page_io.c
index 532a39b..61165b0 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -263,7 +263,6 @@  int __swap_writepage(struct page *page, struct writeback_control *wbc,
 	if (sis->flags & SWP_FILE) {
 		struct kiocb kiocb;
 		struct file *swap_file = sis->swap_file;
-		struct address_space *mapping = swap_file->f_mapping;
 		struct iov_iter from;
 		struct bio_vec bv = {
 			.bv_page = page,
@@ -279,9 +278,7 @@  int __swap_writepage(struct page *page, struct writeback_control *wbc,
 
 		set_page_writeback(page);
 		unlock_page(page);
-		ret = mapping->a_ops->direct_IO(ITER_BVEC | WRITE,
-						&kiocb, &from,
-						kiocb.ki_pos);
+		ret = swap_file->f_op->write_iter(&kiocb, &from);
 		if (ret == PAGE_SIZE) {
 			count_vm_event(PSWPOUT);
 			ret = 0;
@@ -344,12 +341,31 @@  int swap_readpage(struct page *page)
 	}
 
 	if (sis->flags & SWP_FILE) {
+		struct kiocb kiocb;
 		struct file *swap_file = sis->swap_file;
-		struct address_space *mapping = swap_file->f_mapping;
+		struct iov_iter to;
+		struct bio_vec bv = {
+			.bv_page = page,
+			.bv_len = PAGE_SIZE,
+			.bv_offset = 0,
+		};
+
+		iov_iter_bvec(&to, ITER_BVEC | READ, &bv, 1, PAGE_SIZE);
+
+		init_sync_kiocb(&kiocb, swap_file);
+		kiocb.ki_pos = page_file_offset(page);
+		kiocb.ki_nbytes = PAGE_SIZE;
 
-		ret = mapping->a_ops->readpage(swap_file, page);
-		if (!ret)
+		ret = swap_file->f_op->read_iter(&kiocb, &to);
+		if (ret == PAGE_SIZE) {
+			SetPageUptodate(page);
 			count_vm_event(PSWPIN);
+			ret = 0;
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+		unlock_page(page);
 		return ret;
 	}
 
diff --git a/mm/swapfile.c b/mm/swapfile.c
index 63f55cc..4e14122 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -2379,7 +2379,16 @@  SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
 		name = NULL;
 		goto bad_swap;
 	}
-	swap_file = file_open_name(name, O_RDWR|O_LARGEFILE, 0);
+	swap_file = file_open_name(name, O_RDWR | O_LARGEFILE | O_DIRECT, 0);
+	if (swap_file == ERR_PTR(-EINVAL)) {
+		/*
+		 * XXX: there are several filesystems that implement ->bmap but
+		 * not ->direct_IO. It's unlikely that anyone is using a
+		 * swapfile on, e.g., the MINIX fs, but this kludge will keep us
+		 * from getting a complaint from the one person who does.
+		 */
+		swap_file = file_open_name(name, O_RDWR | O_LARGEFILE, 0);
+	}
 	if (IS_ERR(swap_file)) {
 		error = PTR_ERR(swap_file);
 		swap_file = NULL;