diff mbox

[1/6] Btrfs: add mount option for dax

Message ID 1481147110-20048-2-git-send-email-bo.li.liu@oracle.com (mailing list archive)
State New, archived
Headers show

Commit Message

Liu Bo Dec. 7, 2016, 9:45 p.m. UTC 
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
---
 fs/btrfs/ctree.h |  1 +
 fs/btrfs/super.c | 40 +++++++++++++++++++++++++++++++++++++++-
 2 files changed, 40 insertions(+), 1 deletion(-)

Comments

kernel test robot Dec. 8, 2016, 2:44 a.m. UTC | #1 
Hi Liu,

[auto build test WARNING on tip/perf/core]
[also build test WARNING on v4.9-rc8 next-20161207]
[cannot apply to btrfs/next]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Liu-Bo/btrfs-dax-IO/20161208-082651
config: x86_64-randconfig-s2-12081004 (attached as .config)
compiler: gcc-4.4 (Debian 4.4.7-8) 4.4.7
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All warnings (new ones prefixed by >>):

   fs/btrfs/super.c: In function 'btrfs_parse_options':
>> fs/btrfs/super.c:414: warning: unused variable 'set_bdev'
   fs/btrfs/super.o: warning: objtool: btrfs_statfs()+0x2af: function has unreachable instruction

vim +/set_bdev +414 fs/btrfs/super.c

   398	 * XXX JDM: This needs to be cleaned up for remount.
   399	 */
   400	int btrfs_parse_options(struct btrfs_root *root, char *options,
   401				unsigned long new_flags)
   402	{
   403		struct btrfs_fs_info *info = root->fs_info;
   404		substring_t args[MAX_OPT_ARGS];
   405		char *p, *num, *orig = NULL;
   406		u64 cache_gen;
   407		int intarg;
   408		int ret = 0;
   409		char *compress_type;
   410		bool compress_force = false;
   411		enum btrfs_compression_type saved_compress_type;
   412		bool saved_compress_force;
   413		int no_compress = 0;
 > 414		int set_bdev = 0;
   415	
   416		cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
   417		if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
   418			btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
   419		else if (cache_gen)
   420			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
   421	
   422		/*

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation
Dave Chinner Dec. 9, 2016, 4:47 a.m. UTC | #2 
On Wed, Dec 07, 2016 at 01:45:05PM -0800, Liu Bo wrote:
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> ---
>  fs/btrfs/ctree.h |  1 +
>  fs/btrfs/super.c | 40 +++++++++++++++++++++++++++++++++++++++-
>  2 files changed, 40 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
> index 0b8ce2b..e54c6e6 100644
> --- a/fs/btrfs/ctree.h
> +++ b/fs/btrfs/ctree.h
> @@ -1317,6 +1317,7 @@ static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_root *root)
>  #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
>  #define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
>  #define BTRFS_MOUNT_NOLOGREPLAY		(1 << 27)
> +#define BTRFS_MOUNT_DAX		(1 << 28)
>  
>  #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
>  #define BTRFS_DEFAULT_MAX_INLINE	(2048)
> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> index 74ed5aa..9b18f3d 100644
> --- a/fs/btrfs/super.c
> +++ b/fs/btrfs/super.c
> @@ -323,7 +323,7 @@ enum {
>  	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
>  	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
>  	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
> -	Opt_nologreplay, Opt_norecovery,
> +	Opt_nologreplay, Opt_norecovery, Opt_dax,

Can we please not create more filesystems with a DAX mount option?
This was only even an enabler, and not meant to be a permanent
thing. The permanent functionality for DAX is supposed to be
per-inode inheritable DAX flags - not mount options - so that
applications can choose on a per file basis to enable/disable DAX
access as they see fit.

This also enables the filesystem to reject the attempt to turn on
DAX if the set of contexts for the file are not DAX compatible....

Cheers,

Dave.
Liu Bo Dec. 9, 2016, 6:41 p.m. UTC | #3 
On Fri, Dec 09, 2016 at 03:47:20PM +1100, Dave Chinner wrote:
> On Wed, Dec 07, 2016 at 01:45:05PM -0800, Liu Bo wrote:
> > Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> > ---
> >  fs/btrfs/ctree.h |  1 +
> >  fs/btrfs/super.c | 40 +++++++++++++++++++++++++++++++++++++++-
> >  2 files changed, 40 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
> > index 0b8ce2b..e54c6e6 100644
> > --- a/fs/btrfs/ctree.h
> > +++ b/fs/btrfs/ctree.h
> > @@ -1317,6 +1317,7 @@ static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_root *root)
> >  #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
> >  #define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
> >  #define BTRFS_MOUNT_NOLOGREPLAY		(1 << 27)
> > +#define BTRFS_MOUNT_DAX		(1 << 28)
> >  
> >  #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
> >  #define BTRFS_DEFAULT_MAX_INLINE	(2048)
> > diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> > index 74ed5aa..9b18f3d 100644
> > --- a/fs/btrfs/super.c
> > +++ b/fs/btrfs/super.c
> > @@ -323,7 +323,7 @@ enum {
> >  	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
> >  	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
> >  	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
> > -	Opt_nologreplay, Opt_norecovery,
> > +	Opt_nologreplay, Opt_norecovery, Opt_dax,
> 
> Can we please not create more filesystems with a DAX mount option?
> This was only even an enabler, and not meant to be a permanent
> thing. The permanent functionality for DAX is supposed to be
> per-inode inheritable DAX flags - not mount options - so that
> applications can choose on a per file basis to enable/disable DAX
> access as they see fit.
> 
> This also enables the filesystem to reject the attempt to turn on
> DAX if the set of contexts for the file are not DAX compatible....

Sounds good, I'll try to update it to use inode DAX flag directly.

Thanks,

-liubo
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Dave Chinner Dec. 9, 2016, 9:58 p.m. UTC | #4 
On Fri, Dec 09, 2016 at 10:41:34AM -0800, Liu Bo wrote:
> On Fri, Dec 09, 2016 at 03:47:20PM +1100, Dave Chinner wrote:
> > On Wed, Dec 07, 2016 at 01:45:05PM -0800, Liu Bo wrote:
> > > Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> > > ---
> > >  fs/btrfs/ctree.h |  1 +
> > >  fs/btrfs/super.c | 40 +++++++++++++++++++++++++++++++++++++++-
> > >  2 files changed, 40 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
> > > index 0b8ce2b..e54c6e6 100644
> > > --- a/fs/btrfs/ctree.h
> > > +++ b/fs/btrfs/ctree.h
> > > @@ -1317,6 +1317,7 @@ static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_root *root)
> > >  #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
> > >  #define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
> > >  #define BTRFS_MOUNT_NOLOGREPLAY		(1 << 27)
> > > +#define BTRFS_MOUNT_DAX		(1 << 28)
> > >  
> > >  #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
> > >  #define BTRFS_DEFAULT_MAX_INLINE	(2048)
> > > diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> > > index 74ed5aa..9b18f3d 100644
> > > --- a/fs/btrfs/super.c
> > > +++ b/fs/btrfs/super.c
> > > @@ -323,7 +323,7 @@ enum {
> > >  	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
> > >  	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
> > >  	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
> > > -	Opt_nologreplay, Opt_norecovery,
> > > +	Opt_nologreplay, Opt_norecovery, Opt_dax,
> > 
> > Can we please not create more filesystems with a DAX mount option?
> > This was only even an enabler, and not meant to be a permanent
> > thing. The permanent functionality for DAX is supposed to be
> > per-inode inheritable DAX flags - not mount options - so that
> > applications can choose on a per file basis to enable/disable DAX
> > access as they see fit.
> > 
> > This also enables the filesystem to reject the attempt to turn on
> > DAX if the set of contexts for the file are not DAX compatible....
> 
> Sounds good, I'll try to update it to use inode DAX flag directly.

xfs_io already has chattr/lsattr support (+/-x) for the FS_XFLAG_DAX
flag in FS_IOC_FS{GS}ETXATTR, and you can have a look at the XFS
code in xfs_ioctl.c for the operations that are needed to
dynamically change the S_DAX flag on an inode (e.g
xfs_ioctl_setattr_dax_invalidate())

Cheers,

Dave.
diff mbox

Patch

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 0b8ce2b..e54c6e6 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1317,6 +1317,7 @@  static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_root *root)
 #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
 #define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
 #define BTRFS_MOUNT_NOLOGREPLAY		(1 << 27)
+#define BTRFS_MOUNT_DAX		(1 << 28)
 
 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 74ed5aa..9b18f3d 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -323,7 +323,7 @@  enum {
 	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
-	Opt_nologreplay, Opt_norecovery,
+	Opt_nologreplay, Opt_norecovery, Opt_dax,
 #ifdef CONFIG_BTRFS_DEBUG
 	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 #endif
@@ -383,6 +383,7 @@  static const match_table_t tokens = {
 	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 	{Opt_fatal_errors, "fatal_errors=%s"},
 	{Opt_commit_interval, "commit=%d"},
+	{Opt_dax, "dax"},
 #ifdef CONFIG_BTRFS_DEBUG
 	{Opt_fragment_data, "fragment=data"},
 	{Opt_fragment_metadata, "fragment=metadata"},
@@ -410,6 +411,7 @@  int btrfs_parse_options(struct btrfs_root *root, char *options,
 	enum btrfs_compression_type saved_compress_type;
 	bool saved_compress_force;
 	int no_compress = 0;
+	int set_bdev = 0;
 
 	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 	if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
@@ -470,6 +472,40 @@  int btrfs_parse_options(struct btrfs_root *root, char *options,
 			btrfs_clear_opt(info->mount_opt, NODATACOW);
 			btrfs_clear_opt(info->mount_opt, NODATASUM);
 			break;
+#ifdef CONFIG_FS_DAX
+		case Opt_dax:
+			btrfs_set_and_info(info, DAX, "setting dax");
+			/*
+			 * sb->s_blocksize is set to root->sectorsize
+			 * sb->s_bdev is required, but btrfs doesn't set it
+			 * because of multi-device, so here we set it
+			 * temporarily.
+			 * We allows only one device in dax case.
+			 */
+			if (!info->sb->s_bdev) {
+				info->sb->s_bdev = info->fs_devices->latest_bdev;
+				set_bdev = 1;
+			}
+			ret = bdev_dax_supported(info->sb, info->sb->s_blocksize);
+			if (set_bdev)
+				info->sb->s_bdev = NULL;
+			if (ret)
+				goto out;
+
+			/* dax inode doesn't need inline. */
+			info->max_inline = 0;
+			btrfs_info(info, "max_inline at %llu", info->max_inline);
+
+			btrfs_clear_opt(info->mount_opt, COMPRESS);
+			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
+			btrfs_set_opt(info->mount_opt, NODATACOW);
+			btrfs_set_opt(info->mount_opt, NODATASUM);
+			btrfs_info(info,
+				   "setting nodatacow, compression disabled");
+
+			/* dax doesn't expect other fancy options. */
+			goto out;
+#endif
 		case Opt_nodatacow:
 			if (!btrfs_test_opt(info, NODATACOW)) {
 				if (!btrfs_test_opt(info, COMPRESS) ||
@@ -1232,6 +1268,8 @@  static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 		seq_puts(seq, ",nodatasum");
 	if (btrfs_test_opt(info, NODATACOW))
 		seq_puts(seq, ",nodatacow");
+	if (btrfs_test_opt(info, DAX))
+		seq_puts(seq, ",dax");
 	if (btrfs_test_opt(info, NOBARRIER))
 		seq_puts(seq, ",nobarrier");
 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)