Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1445031 > unrolled thread

[PATCH 1/3] f2fs: avoid memory allocation failure due to a long length

Back to article view | Back to linux.kernel

Contents

  [PATCH 1/3] f2fs: avoid memory allocation failure due to a long length Jaegeuk Kim <jaegeuk@kernel.org> - 2016-07-17 08:30 +0200
    [PATCH 3/3] f2fs: support clone_file_range Jaegeuk Kim <jaegeuk@kernel.org> - 2016-07-17 08:30 +0200
      Re: [PATCH 3/3] f2fs: support clone_file_range Christoph Hellwig <hch@infradead.org> - 2016-07-19 05:50 +0200
        Re: [PATCH 3/3] f2fs: support clone_file_range Jaegeuk Kim <jaegeuk@kernel.org> - 2016-07-19 06:50 +0200

#1445031 — [PATCH 1/3] f2fs: avoid memory allocation failure due to a long length

We need to avoid ENOMEM due to unexpected long length.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
---
 fs/f2fs/file.c | 46 ++++++++++++++++++++++++++++------------------
 1 file changed, 28 insertions(+), 18 deletions(-)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 17b3059..3573b07 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -1039,33 +1039,43 @@ static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
 
 static int __exchange_data_block(struct inode *src_inode,
 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
-			int len, bool full)
+			pgoff_t len, bool full)
 {
 	block_t *src_blkaddr;
 	int *do_replace;
+	pgoff_t olen;
 	int ret;
 
-	src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * len, GFP_KERNEL);
-	if (!src_blkaddr)
-		return -ENOMEM;
+	while (len) {
+		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK, len);
 
-	do_replace = f2fs_kvzalloc(sizeof(int) * len, GFP_KERNEL);
-	if (!do_replace) {
-		kvfree(src_blkaddr);
-		return -ENOMEM;
-	}
+		src_blkaddr = f2fs_kvzalloc(sizeof(block_t) * olen, GFP_KERNEL);
+		if (!src_blkaddr)
+			return -ENOMEM;
 
-	ret = __read_out_blkaddrs(src_inode, src_blkaddr, do_replace, src, len);
-	if (ret)
-		goto roll_back;
+		do_replace = f2fs_kvzalloc(sizeof(int) * olen, GFP_KERNEL);
+		if (!do_replace) {
+			kvfree(src_blkaddr);
+			return -ENOMEM;
+		}
 
-	ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
-					do_replace, src, dst, len, full);
-	if (ret)
-		goto roll_back;
+		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
+					do_replace, src, olen);
+		if (ret)
+			goto roll_back;
 
-	kvfree(src_blkaddr);
-	kvfree(do_replace);
+		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
+					do_replace, src, dst, olen, full);
+		if (ret)
+			goto roll_back;
+
+		src += olen;
+		dst += olen;
+		len -= olen;
+
+		kvfree(src_blkaddr);
+		kvfree(do_replace);
+	}
 	return 0;
 
 roll_back:
-- 
2.8.3

[toc] | [next] | [standalone]

#1445032 — [PATCH 3/3] f2fs: support clone_file_range

This patch implements clone_file_range in f2fs.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
---
 fs/f2fs/file.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index c2b7e35..37480f3 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -2312,6 +2312,12 @@ ssize_t f2fs_copy_file_range(struct file *file_in, loff_t pos_in,
 	return ret;
 }
 
+int f2fs_clone_file_range(struct file *file_in, loff_t pos_in,
+		struct file *file_out, loff_t pos_out, u64 len)
+{
+	return f2fs_clone_files(file_in, pos_in, file_out, pos_out, len);
+}
+
 #ifdef CONFIG_COMPAT
 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
@@ -2359,6 +2365,7 @@ const struct file_operations f2fs_file_operations = {
 	.compat_ioctl	= f2fs_compat_ioctl,
 #endif
 	.copy_file_range = f2fs_copy_file_range,
+	.clone_file_range = f2fs_clone_file_range,
 	.splice_read	= generic_file_splice_read,
 	.splice_write	= iter_file_splice_write,
 };
-- 
2.8.3

[toc] | [prev] | [next] | [standalone]

#1446005 — Re: [PATCH 3/3] f2fs: support clone_file_range

On Sat, Jul 16, 2016 at 11:24:27PM -0700, Jaegeuk Kim wrote:
> This patch implements clone_file_range in f2fs.

[...]

> +int f2fs_clone_file_range(struct file *file_in, loff_t pos_in,
> +		struct file *file_out, loff_t pos_out, u64 len)
> +{
> +	return f2fs_clone_files(file_in, pos_in, file_out, pos_out, len);
> +}
> +

Falling back from copy to clone should be done in the VFS.  I look into
implementing the fallback, and the code is trivial, but we don't seem
to have any useful coverage for copy_file_range yet, so I didn't dare
to add it yet.  How did you test copy and clone in f2fs?  And how did
you handle the difference in corner cases (e.g. the lacking 0 special
case in copy)?

[toc] | [prev] | [next] | [standalone]

#1446019 — Re: [PATCH 3/3] f2fs: support clone_file_range

On Mon, Jul 18, 2016 at 08:47:36PM -0700, Christoph Hellwig wrote:
> On Sat, Jul 16, 2016 at 11:24:27PM -0700, Jaegeuk Kim wrote:
> > This patch implements clone_file_range in f2fs.
> 
> [...]
> 
> > +int f2fs_clone_file_range(struct file *file_in, loff_t pos_in,
> > +		struct file *file_out, loff_t pos_out, u64 len)
> > +{
> > +	return f2fs_clone_files(file_in, pos_in, file_out, pos_out, len);
> > +}
> > +
> 
> Falling back from copy to clone should be done in the VFS.  I look into
> implementing the fallback, and the code is trivial, but we don't seem
> to have any useful coverage for copy_file_range yet, so I didn't dare
> to add it yet.  How did you test copy and clone in f2fs?  And how did
> you handle the difference in corner cases (e.g. the lacking 0 special
> case in copy)?

Frankly speaking, I confused the behaviors without being aware of corner cases.
And thus, I dropped the patches already. Instead, I realized that what I did
was moving a range of blocks from one file to another file. So, I've been
testing a new ioctl, F2FS_IOC_MOVE_RANGE.
In terms of coverage, I just did some unit tests only.

Thanks,

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web