On Mon, Jun 29, 2015 at 2:33 PM, Steve French
<steve.french@xxxxxxxxxxxxxxx> wrote:
> On Mon, Jun 29, 2015 at 12:27 PM, Darrick J. Wong
> <darrick.wong@xxxxxxxxxx> wrote:
>>
>> On Sun, Jun 28, 2015 at 09:21:05PM -0500, Steve French wrote:
>> > Update the patch to correct target file size.
>> >
>> > Getting fantastic copy performance with cp --reflink over SMB3.11
>> > using the new FSCTL_DUPLICATE_EXTENTS.
>> >
>> > This FSCTL was added in the SMB3.11 dialect (testing was
>> > against REFS file system) so have put it as a 3.11 protocol
>> > specific operation ("vers=3.1.1" on the mount). Tested at
>> > the SMB3 plugfest in Redmond.
>> >
>> > It depends on the new FS Attribute (BLOCK_REFCOUNTING) which
>> > is used to advertise support for the ability to do this ioctl
>> > (if you can support multiple files pointing to the same block
>> > than this refcounting ability or equivalent is needed to
>> > support the new reflink-like duplicate extent SMB3 ioctl.
>> >
>> > Signed-off-by: Steve French <steve.french@xxxxxxxxxxxxxxx>
>> > ---
>> > fs/cifs/cifsglob.h | 3 +++
>> > fs/cifs/cifspdu.h | 2 ++
>> > fs/cifs/ioctl.c | 16 +++++++++++++---
>> > fs/cifs/smb2ops.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
>> > fs/cifs/smb2pdu.h | 8 ++++++++
>> > fs/cifs/smbfsctl.h | 1 +
>> > 6 files changed, 75 insertions(+), 3 deletions(-)
>> >
>> > diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
>> > index a0212ec..81194e6 100644
>> > --- a/fs/cifs/cifsglob.h
>> > +++ b/fs/cifs/cifsglob.h
>> > @@ -390,6 +390,9 @@ struct smb_version_operations {
>> > int (*clone_range)(const unsigned int, struct cifsFileInfo *src_file,
>> > struct cifsFileInfo *target_file, u64 src_off, u64 len,
>> > u64 dest_off);
>> > + int (*duplicate_extents)(const unsigned int, struct cifsFileInfo *src,
>> > + struct cifsFileInfo *target_file, u64 src_off, u64 len,
>> > + u64 dest_off);
>> > int (*validate_negotiate)(const unsigned int, struct cifs_tcon *);
>> > ssize_t (*query_all_EAs)(const unsigned int, struct cifs_tcon *,
>> > const unsigned char *, const unsigned char *, char *,
>> > diff --git a/fs/cifs/cifspdu.h b/fs/cifs/cifspdu.h
>> > index 998a66f..47b030d 100644
>> > --- a/fs/cifs/cifspdu.h
>> > +++ b/fs/cifs/cifspdu.h
>> > @@ -2255,6 +2255,8 @@ typedef struct {
>> >
>> >
>> > /* List of FileSystemAttributes - see 2.5.1 of MS-FSCC */
>> > +#define FILE_SUPPORTS_SPARSE_VDL 0x10000000 /* faster nonsparse extend */
>> > +#define FILE_SUPPORTS_BLOCK_REFCOUNTING 0x08000000 /* allow ioctl dup extents */
>> > #define FILE_SUPPORT_INTEGRITY_STREAMS 0x04000000
>> > #define FILE_SUPPORTS_USN_JOURNAL 0x02000000
>> > #define FILE_SUPPORTS_OPEN_BY_FILE_ID 0x01000000
>> > diff --git a/fs/cifs/ioctl.c b/fs/cifs/ioctl.c
>> > index 8b7898b..7843b19 100644
>> > --- a/fs/cifs/ioctl.c
>> > +++ b/fs/cifs/ioctl.c
>> > @@ -31,12 +31,14 @@
>> > #include "cifsproto.h"
>> > #include "cifs_debug.h"
>> > #include "cifsfs.h"
>> > +#include <linux/btrfs.h>
>> >
>> > #define CIFS_IOCTL_MAGIC 0xCF
>> > #define CIFS_IOC_COPYCHUNK_FILE _IOW(CIFS_IOCTL_MAGIC, 3, int)
>> >
>> > static long cifs_ioctl_clone(unsigned int xid, struct file *dst_file,
>> > - unsigned long srcfd, u64 off, u64 len, u64 destoff)
>> > + unsigned long srcfd, u64 off, u64 len, u64 destoff,
>> > + bool dup_extents)
>> > {
>> > int rc;
>> > struct cifsFileInfo *smb_file_target = dst_file->private_data;
>> > @@ -109,9 +111,14 @@ static long cifs_ioctl_clone(unsigned int xid, struct file *dst_file,
>> > truncate_inode_pages_range(&target_inode->i_data, destoff,
>> > PAGE_CACHE_ALIGN(destoff + len)-1);
>> >
>> > - if (target_tcon->ses->server->ops->clone_range)
>> > + if (dup_extents && target_tcon->ses->server->ops->duplicate_extents)
>> > + rc = target_tcon->ses->server->ops->duplicate_extents(xid,
>> > + smb_file_src, smb_file_target, off, len, destoff);
>> > + else if (!dup_extents && target_tcon->ses->server->ops->clone_range)
>> > rc = target_tcon->ses->server->ops->clone_range(xid,
>> > smb_file_src, smb_file_target, off, len, destoff);
>> > + else
>> > + rc = -EOPNOTSUPP;
>> >
>> > /* force revalidate of size and timestamps of target file now
>> > that target is updated on the server */
>> > @@ -205,7 +212,10 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
>> > }
>> > break;
>> > case CIFS_IOC_COPYCHUNK_FILE:
>> > - rc = cifs_ioctl_clone(xid, filep, arg, 0, 0, 0);
>> > + rc = cifs_ioctl_clone(xid, filep, arg, 0, 0, 0, false);
>> > + break;
>> > + case BTRFS_IOC_CLONE:
>> > + rc = cifs_ioctl_clone(xid, filep, arg, 0, 0, 0, true);
>>
>> Any interest in supporting BTRFS_IOC_CLONE_RANGE or BTRFS_IOC_EXTENT_SAME?
>> It looks like you could easily support the former, and the latter would
>> enable things like duperemove. I've been working on a pile of xfstests to
>> exercise these three ioctls, will post them later today, I hope.
>>
>> --D
>>
>
> So far Samba (server) only uses three btrfs ioctls, but that includes
> clone range:
>
> ./source3/modules/vfs_btrfs.c:#define BTRFS_IOC_CLONE_RANGE ...
> ./source3/modules/vfs_btrfs.c:#define BTRFS_IOC_SNAP_DESTROY ...
> ./source3/modules/vfs_btrfs.c:#define BTRFS_IOC_SNAP_CREATE_V2 ...
>
> But btrfs clone range is only (optionally if present) used to handle
> the "CopyChunk" API (SMB3 and CIFS network protocol requests), and
> Samba server doesn't have code for the FSCTL_DUPLICATE_EXTENTS fsctl
> (SMB3 network protocol request) yet. It looks trivial to map that to
> BTRFS_IOC_CLONE_RANGE in the Samba server. Would need to add trivial
> parsing in smbd/smb2_ioctl_network_fs.c and a new VFS entry point in
> the Samba VFS and a small helper routine in
> source3/modules/vfs_btrfs.c
I should also mention that cifs.ko can already do fast(er) server side
copy today by using CopyChunk on SMB2/SMB3 but it is emulated in some
cases on the Samba server (e.g. when not running on btrfs or with
vfs_btrfs enabled) but "cp --reflink" semantics required that it
actually do the clone_range/duplicate_extents and Samba can't
guarantee that with CopyChunk (FSCTL_SRV_COPYCHUNK_WRITE) - on the
other hand the new smb3 network protocol request (SMB3
FSCTL_DUPLICATE_EXTENTS) does guarantee that it is a reflink-like
copy.
We need an "smb3cp" utility (perhaps added to cifs-utils) that can try
the various types of server side copy and accept parms which allow the
user to specify the behavior they want (and the offsets they want) -
since few people know about the existing ioctl in cifs.ko which calls
CopyChunk (and FSCTL_SRV_COPYCHUNK_WRITE is not as fast as
FSCTL_DUPLICATE_EXTENTS)
I was also looking into using DUPLICATE_EXTENTS to enable the two new
fallocate features (insert range and collapse range) because it can be
called with source and target as the same file.
--
Steve French
Principal Systems Engineer: Protocols
W: 512-918-9276
C: 512-501-9669
www.primarydata.com
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
