public inbox for fstests@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH 0/1] cloner: add support for clone_file_range for cifs
@ 2018-09-11  5:28 Ronnie Sahlberg
  2018-09-11  5:28 ` [PATCH] " Ronnie Sahlberg
  0 siblings, 1 reply; 4+ messages in thread
From: Ronnie Sahlberg @ 2018-09-11  5:28 UTC (permalink / raw)
  To: fstests

This is a trivial patch that adds support for copy_file_range for cifs.
This uses the copy_file_range interface which cifs.ko supports.

Note that cloner.c uses an ioctl() for copying an entire file. This is
an older interface from before cifs.ko added proper copy_file_range
support but both interfaces eventually end up in the same codepaths
and use the same SMB2 commands on the wire.
Could argue that we could/should use copy_file_range() for the whole file
case as well but there is no harm leaving it as-is using the older interface.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH] cloner: add support for clone_file_range for cifs
  2018-09-11  5:28 [PATCH 0/1] cloner: add support for clone_file_range for cifs Ronnie Sahlberg
@ 2018-09-11  5:28 ` Ronnie Sahlberg
  2018-09-11  7:02   ` Christoph Hellwig
  0 siblings, 1 reply; 4+ messages in thread
From: Ronnie Sahlberg @ 2018-09-11  5:28 UTC (permalink / raw)
  To: fstests

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
---
 src/cloner.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/src/cloner.c b/src/cloner.c
index ffad82f0..9f9156e5 100644
--- a/src/cloner.c
+++ b/src/cloner.c
@@ -120,6 +120,18 @@ clone_file_range_btrfs(int src_fd, int dst_fd, uint64_t src_off,
 }
 
 static int
+clone_file_range_cifs(int fd_in, int fd_out, loff_t off_in,
+		      loff_t off_out, size_t len)
+{
+	int ret;
+
+	ret = copy_file_range(fd_in, &off_in, fd_out, &off_out, len, 0);
+	if (ret < 0)
+		ret = errno;
+	return 0;
+}
+
+static int
 clone_file_range(unsigned int fs_type, int src_fd, int dst_fd, uint64_t src_off,
 		 uint64_t dst_off, uint64_t len)
 {
@@ -128,7 +140,9 @@ clone_file_range(unsigned int fs_type, int src_fd, int dst_fd, uint64_t src_off,
 		return clone_file_range_btrfs(src_fd, dst_fd, src_off, dst_off,
 					      len);
 		break;
-	case CIFS_MAGIC_NUMBER:	/* only supports full file server-side copies */
+	case CIFS_MAGIC_NUMBER:
+		return clone_file_range_cifs(src_fd, dst_fd, src_off, dst_off,
+					     len);
 	default:
 		return ENOTSUP;
 		break;
-- 
2.13.3

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] cloner: add support for clone_file_range for cifs
  2018-09-11  5:28 ` [PATCH] " Ronnie Sahlberg
@ 2018-09-11  7:02   ` Christoph Hellwig
  2018-09-11  7:26     ` ronnie sahlberg
  0 siblings, 1 reply; 4+ messages in thread
From: Christoph Hellwig @ 2018-09-11  7:02 UTC (permalink / raw)
  To: Ronnie Sahlberg; +Cc: fstests

On Tue, Sep 11, 2018 at 03:28:06PM +1000, Ronnie Sahlberg wrote:
> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
> ---
>  src/cloner.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/src/cloner.c b/src/cloner.c
> index ffad82f0..9f9156e5 100644
> --- a/src/cloner.c
> +++ b/src/cloner.c
> @@ -120,6 +120,18 @@ clone_file_range_btrfs(int src_fd, int dst_fd, uint64_t src_off,
>  }
>  
>  static int
> +clone_file_range_cifs(int fd_in, int fd_out, loff_t off_in,
> +		      loff_t off_out, size_t len)
> +{
> +	int ret;
> +
> +	ret = copy_file_range(fd_in, &off_in, fd_out, &off_out, len, 0);
> +	if (ret < 0)
> +		ret = errno;
> +	return 0;
> +}

This is weird.  cloner seems to test clone, in which case it should
test FICLONE (BTRFS_IOC_CLONE) / FICLONERANGE(BTRFS_IOC_CLONE_RANGE).

But it seems to check file system magic numbers which is rather bogus
to start with.  I think the right fix here is to remove all the magic
number checks, always try FICLONE, and also remove support for the
odd legacy cifs ioctl.

>  {
> @@ -128,7 +140,9 @@ clone_file_range(unsigned int fs_type, int src_fd, int dst_fd, uint64_t src_off,
>  		return clone_file_range_btrfs(src_fd, dst_fd, src_off, dst_off,
>  					      len);
>  		break;
> -	case CIFS_MAGIC_NUMBER:	/* only supports full file server-side copies */
> +	case CIFS_MAGIC_NUMBER:
> +		return clone_file_range_cifs(src_fd, dst_fd, src_off, dst_off,
> +					     len);
>  	default:
>  		return ENOTSUP;
>  		break;

And independent break after a return is just insane.  Someone needs
to fix this program to stop the eye bleeding..

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] cloner: add support for clone_file_range for cifs
  2018-09-11  7:02   ` Christoph Hellwig
@ 2018-09-11  7:26     ` ronnie sahlberg
  0 siblings, 0 replies; 4+ messages in thread
From: ronnie sahlberg @ 2018-09-11  7:26 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: Ronnie Sahlberg, fstests

On Tue, Sep 11, 2018 at 5:02 PM, Christoph Hellwig <hch@infradead.org> wrote:
> On Tue, Sep 11, 2018 at 03:28:06PM +1000, Ronnie Sahlberg wrote:
>> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
>> ---
>>  src/cloner.c | 16 +++++++++++++++-
>>  1 file changed, 15 insertions(+), 1 deletion(-)
>>
>> diff --git a/src/cloner.c b/src/cloner.c
>> index ffad82f0..9f9156e5 100644
>> --- a/src/cloner.c
>> +++ b/src/cloner.c
>> @@ -120,6 +120,18 @@ clone_file_range_btrfs(int src_fd, int dst_fd, uint64_t src_off,
>>  }
>>
>>  static int
>> +clone_file_range_cifs(int fd_in, int fd_out, loff_t off_in,
>> +                   loff_t off_out, size_t len)
>> +{
>> +     int ret;
>> +
>> +     ret = copy_file_range(fd_in, &off_in, fd_out, &off_out, len, 0);
>> +     if (ret < 0)
>> +             ret = errno;
>> +     return 0;
>> +}
>
> This is weird.  cloner seems to test clone, in which case it should
> test FICLONE (BTRFS_IOC_CLONE) / FICLONERANGE(BTRFS_IOC_CLONE_RANGE).
>
> But it seems to check file system magic numbers which is rather bogus
> to start with.  I think the right fix here is to remove all the magic
> number checks, always try FICLONE, and also remove support for the
> odd legacy cifs ioctl.

Thanks.  That is a lot bigger change but it is the right thing to do.

Please disregard this patch and I will make a new patch with your suggestions.

>
>>  {
>> @@ -128,7 +140,9 @@ clone_file_range(unsigned int fs_type, int src_fd, int dst_fd, uint64_t src_off,
>>               return clone_file_range_btrfs(src_fd, dst_fd, src_off, dst_off,
>>                                             len);
>>               break;
>> -     case CIFS_MAGIC_NUMBER: /* only supports full file server-side copies */
>> +     case CIFS_MAGIC_NUMBER:
>> +             return clone_file_range_cifs(src_fd, dst_fd, src_off, dst_off,
>> +                                          len);
>>       default:
>>               return ENOTSUP;
>>               break;
>
> And independent break after a return is just insane.  Someone needs
> to fix this program to stop the eye bleeding..

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2018-09-11 12:24 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-09-11  5:28 [PATCH 0/1] cloner: add support for clone_file_range for cifs Ronnie Sahlberg
2018-09-11  5:28 ` [PATCH] " Ronnie Sahlberg
2018-09-11  7:02   ` Christoph Hellwig
2018-09-11  7:26     ` ronnie sahlberg

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox