Skip to contentSkip to conversation

f2fs: add remount_fs callback support

PATCH In review Namjae Jeon 2 participants 5 messages Jun 1, 2013 – Jun 6, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 f2fs: add remount_fs callback support

From: Namjae Jeon <[email protected]>

Add the f2fs_remount function call which will be used during the filesystem remounting. This function will help us to change the mount options specific to f2fs.

Also modify the f2fs background_gc mount option, which will allow the user to dynamically trun on/off the garbage collection in f2fs based on the background_gc value. If background_gc=0, Garbage collection will be turned off & if background_gc=1, Garbage collection will be truned on.

By default the garbage collection is on in f2fs.

---
 Documentation/filesystems/f2fs.txt |    8 +-
 fs/f2fs/super.c                    |  223 +++++++++++++++++++++++-------------
 2 files changed, 148 insertions(+), 83 deletions(-)
Documentation/filesystems/f2fs.txt +6 -2
index bd3c56c..6a036ce 100644
@@ -98,8 +98,12 @@ Cleaning Overhead
 MOUNT OPTIONS
 ================================================================================
 
-background_gc_off      Turn off cleaning operations, namely garbage collection,
-		       triggered in background when I/O subsystem is idle.
+background_gc=%u       Turn on/off cleaning operations, namely garbage collection,
+                       triggered in background when I/O subsystem is idle. If
+                       background_gc=1, it will turn on the garbage collection &
+                       if background_gc=0, garbage collection will be truned off.
+                       Default value for this option is 1. So garbage collection
+                       is on by default.
 disable_roll_forward   Disable the roll-forward recovery routine
 discard                Issue discard/TRIM commands when a segment is cleaned.
 no_heap                Disable heap-style segment allocation which finds free
fs/f2fs/super.c +142 -81

272 lines not on this page · raw message

Signature · 9 lines
1.7.9.5


------------------------------------------------------------------------------
Get 100% visibility into Java/.NET code with AppDynamics Lite
It's a free troubleshooting tool designed for production
Get down to code-level detail for bottlenecks, with <2% overhead.
Download for free and get started troubleshooting in minutes.
http://p.sf.net/sfu/appdyn_d2d_ap2
#1 lore raw
Quoted 7 lines · Namjae Jeon

From: Namjae Jeon <[email protected]>

Add the f2fs_remount function call which will be used during the filesystem remounting. This function will help us to change the mount options specific to f2fs.

Quoted 6 lines

Also modify the f2fs background_gc mount option, which will allow the user to dynamically trun on/off the garbage collection in f2fs based on the background_gc value. If background_gc=0, Garbage collection will be turned off & if background_gc=1, Garbage collection will be truned on.

Hi Namjae,
I think splitting these two changes into single ones seems better.
Refer to the inline comments.

Thanks,
Gu

Quoted 140 lines

By default the garbage collection is on in f2fs.

---
 Documentation/filesystems/f2fs.txt |    8 +-
 fs/f2fs/super.c                    |  223 +++++++++++++++++++++++-------------
 2 files changed, 148 insertions(+), 83 deletions(-)
Documentation/filesystems/f2fs.txt +6 -2
index bd3c56c..6a036ce 100644
@@ -98,8 +98,12 @@ Cleaning Overhead
 MOUNT OPTIONS
 ================================================================================
 
-background_gc_off      Turn off cleaning operations, namely garbage collection,
-		       triggered in background when I/O subsystem is idle.
+background_gc=%u       Turn on/off cleaning operations, namely garbage collection,
+                       triggered in background when I/O subsystem is idle. If
+                       background_gc=1, it will turn on the garbage collection &
+                       if background_gc=0, garbage collection will be truned off.
+                       Default value for this option is 1. So garbage collection
+                       is on by default.
 disable_roll_forward   Disable the roll-forward recovery routine
 discard                Issue discard/TRIM commands when a segment is cleaned.
 no_heap                Disable heap-style segment allocation which finds free
fs/f2fs/super.c +82 -2
index 3ac305d..bcd68aa 100644
@@ -34,7 +34,7 @@
 static struct kmem_cache *f2fs_inode_cachep;
 
 enum {
-	Opt_gc_background_off,
+	Opt_gc_background,
 	Opt_disable_roll_forward,
 	Opt_discard,
 	Opt_noheap,
@@ -46,7 +46,7 @@ enum {
 };
 
 static match_table_t f2fs_tokens = {
-	{Opt_gc_background_off, "background_gc_off"},
+	{Opt_gc_background, "background_gc=%u"},
 	{Opt_disable_roll_forward, "disable_roll_forward"},
 	{Opt_discard, "discard"},
 	{Opt_noheap, "no_heap"},
@@ -76,6 +76,86 @@ static void init_once(void *foo)
 	inode_init_once(&fi->vfs_inode);
 }
 
+static int parse_options(struct super_block *sb, struct f2fs_sb_info *sbi,
+				char *options)
+{
+	substring_t args[MAX_OPT_ARGS];
+	char *p;
+	int arg = 0;
+
+	if (!options)
+		return 0;
+
+	while ((p = strsep(&options, ",")) != NULL) {
+		int token;
+		if (!*p)
+			continue;
+		/*
+		 * Initialize args struct so we know whether arg was
+		 * found; some options take optional arguments.
+		 */
+		args[0].to = args[0].from = NULL;
+		token = match_token(p, f2fs_tokens, args);
+
+		switch (token) {
+		case Opt_gc_background:
+			if (args->from && match_int(args, &arg))
+				return -EINVAL;
+			if (arg != 0 && arg != 1)
+				return -EINVAL;
+			if (arg == 0)
+				clear_opt(sbi, BG_GC);
+			else
+				set_opt(sbi, BG_GC);
+			break;
+		case Opt_disable_roll_forward:
+			set_opt(sbi, DISABLE_ROLL_FORWARD);
+			break;
+		case Opt_discard:
+			set_opt(sbi, DISCARD);
+			break;
+		case Opt_noheap:
+			set_opt(sbi, NOHEAP);
+			break;
+#ifdef CONFIG_F2FS_FS_XATTR
+		case Opt_nouser_xattr:
+			clear_opt(sbi, XATTR_USER);
+			break;
+#else
+		case Opt_nouser_xattr:
+			f2fs_msg(sb, KERN_INFO,
+				"nouser_xattr options not supported");
+			break;
+#endif
+#ifdef CONFIG_F2FS_FS_POSIX_ACL
+		case Opt_noacl:
+			clear_opt(sbi, POSIX_ACL);
+			break;
+#else
+		case Opt_noacl:
+			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
+			break;
+#endif
+		case Opt_active_logs:
+			if (args->from && match_int(args, &arg))
+				return -EINVAL;
+			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
+				return -EINVAL;
+			sbi->active_logs = arg;
+			break;
+		case Opt_disable_ext_identify:
+			set_opt(sbi, DISABLE_EXT_IDENTIFY);
+			break;
+		default:
+			f2fs_msg(sb, KERN_ERR,
+				"Unrecognized mount option \"%s\" or missing value",
+				p);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 {
 	struct f2fs_inode_info *fi;

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 300

Quoted 10 lines
fs/f2fs/super.c +3 -3
@@ -214,10 +294,10 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 
-	if (test_opt(sbi, BG_GC))
-		seq_puts(seq, ",background_gc_on");
+	if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
+		seq_printf(seq, ",background_gc=%u", 1);
 	else
-		seq_puts(seq, ",background_gc_off");
+		seq_printf(seq, ",background_gc=%u", 0);

Though simply option show is enough, but I think the "background_gc=on/off" is more friendly.

Quoted 23 lines
fs/f2fs/super.c +16 -0
 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 		seq_puts(seq, ",disable_roll_forward");
 	if (test_opt(sbi, DISCARD))
@@ -244,6 +324,52 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 	return 0;
 }
 
+static int f2fs_remount(struct super_block *sb, int *flags, char *data)
+{
+	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	unsigned long old_sb_flags;
+	struct f2fs_mount_info org_mount_opt;
+	int err, active_logs;
+
+	/**
+	 * Save the old mount options in case we
+	 * need to restore them.
+	 */
+	old_sb_flags = sb->s_flags;
+	org_mount_opt = sbi->mount_opt;
+	active_logs = sbi->active_logs;
+
+	/* parse mount options */

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 352

Quoted 10 lines
fs/f2fs/super.c +10 -0
+	err = parse_options(sb, sbi, data);
+	if (err)
+		goto restore_opts;
+
+	/**
+	 * We stop the GC thread if FS is mounted as RO
+	 * or if background_gc = 0 is passed in mount
+	 * option. Also sync the filesystem.
+	 */
+	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {

Another condition: The old mount is not RO.

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 363

Quoted 11 lines
fs/f2fs/super.c +11 -0
+		stop_gc_thread(sbi);
+		f2fs_sync_fs(sb, 1);
+	} else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
+		err = start_gc_thread(sbi);
+		if (err)
+			goto restore_opts;
+	}
+
+	/* Update the POSIXACL Flag */
+	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
+		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);

Maybe you forget to update the flags with MS_RDONLY or ~MS_RDONLY, if the flags changed.

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 367

Quoted 4 lines
fs/f2fs/super.c +4 -0
+	return 0;
+
+restore_opts:
+	sb->s_flags = old_sb_flags;

There is no need to restore sb->s_flags, parse_options() did not change it. no need to store the old sb->s_flags too.

Quoted 99 lines
fs/f2fs/super.c +6 -72
+	sbi->mount_opt = org_mount_opt;
+	sbi->active_logs = active_logs;
+	return err;
+}
+
 static struct super_operations f2fs_sops = {
 	.alloc_inode	= f2fs_alloc_inode,
 	.drop_inode	= f2fs_drop_inode,
@@ -256,6 +382,7 @@ static struct super_operations f2fs_sops = {
 	.freeze_fs	= f2fs_freeze,
 	.unfreeze_fs	= f2fs_unfreeze,
 	.statfs		= f2fs_statfs,
+	.remount_fs	= f2fs_remount,
 };
 
 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
@@ -303,78 +430,6 @@ static const struct export_operations f2fs_export_ops = {
 	.get_parent = f2fs_get_parent,
 };
 
-static int parse_options(struct super_block *sb, struct f2fs_sb_info *sbi,
-				char *options)
-{
-	substring_t args[MAX_OPT_ARGS];
-	char *p;
-	int arg = 0;
-
-	if (!options)
-		return 0;
-
-	while ((p = strsep(&options, ",")) != NULL) {
-		int token;
-		if (!*p)
-			continue;
-		/*
-		 * Initialize args struct so we know whether arg was
-		 * found; some options take optional arguments.
-		 */
-		args[0].to = args[0].from = NULL;
-		token = match_token(p, f2fs_tokens, args);
-
-		switch (token) {
-		case Opt_gc_background_off:
-			clear_opt(sbi, BG_GC);
-			break;
-		case Opt_disable_roll_forward:
-			set_opt(sbi, DISABLE_ROLL_FORWARD);
-			break;
-		case Opt_discard:
-			set_opt(sbi, DISCARD);
-			break;
-		case Opt_noheap:
-			set_opt(sbi, NOHEAP);
-			break;
-#ifdef CONFIG_F2FS_FS_XATTR
-		case Opt_nouser_xattr:
-			clear_opt(sbi, XATTR_USER);
-			break;
-#else
-		case Opt_nouser_xattr:
-			f2fs_msg(sb, KERN_INFO,
-				"nouser_xattr options not supported");
-			break;
-#endif
-#ifdef CONFIG_F2FS_FS_POSIX_ACL
-		case Opt_noacl:
-			clear_opt(sbi, POSIX_ACL);
-			break;
-#else
-		case Opt_noacl:
-			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
-			break;
-#endif
-		case Opt_active_logs:
-			if (args->from && match_int(args, &arg))
-				return -EINVAL;
-			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
-				return -EINVAL;
-			sbi->active_logs = arg;
-			break;
-		case Opt_disable_ext_identify:
-			set_opt(sbi, DISABLE_EXT_IDENTIFY);
-			break;
-		default:
-			f2fs_msg(sb, KERN_ERR,
-				"Unrecognized mount option \"%s\" or missing value",
-				p);
-			return -EINVAL;
-		}
-	}
-	return 0;
-}
 
 static loff_t max_file_size(unsigned bits)
 {
@@ -674,10 +729,16 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 				"Cannot recover all fsync data errno=%ld", err);
 	}
 

fs/f2fs/super.c static int f2fs_fill_super(struct super_block *sb, void *data, int silent) · line 737

Quoted 10 lines
fs/f2fs/super.c +6 -4
-	/* After POR, we can run background GC thread */
-	err = start_gc_thread(sbi);
-	if (err)
-		goto fail;
+	/* After POR, we can run background GC thread.*/
+	if (!(sb->s_flags & MS_RDONLY)) {
+		/**
+		 * If filesystem is mounted as read-only then
+		 * do not start the gc_thread.
+		 */

It seems that the comment here is against with the logic.

fs/f2fs/super.c static int f2fs_fill_super(struct super_block *sb, void *data, int silent) · line 744

Quoted 7 lines
fs/f2fs/super.c +4 -0
+		err = start_gc_thread(sbi);
+		if (err)
+			goto fail;
+	}
 
 	err = f2fs_build_stats(sbi);
 	if (err)

------------------------------------------------------------------------------
How ServiceNow helps IT people transform IT departments:
1. A cloud service to automate IT design, transition and operations
2. Dashboards that offer high-level views of enterprise services
3. A single system of record for all IT processes
http://p.sf.net/sfu/servicenow-d2d-j

#2 lore raw

2013/6/4 Gu Zheng <[email protected]>:

Quoted 17 lines · Namjae Jeon
Quoted 8 lines

From: Namjae Jeon <[email protected]>

Add the f2fs_remount function call which will be used during the filesystem remounting. This function will help us to change the mount options specific to f2fs.

Quoted 6 lines

Also modify the f2fs background_gc mount option, which will allow the user to dynamically trun on/off the garbage collection in f2fs based on the background_gc value. If background_gc=0, Garbage collection will be turned off & if background_gc=1, Garbage collection will be truned on.

Quoted 1 line

Hi Namjae,

Hi. Gu.

Quoted 2 lines

I think splitting these two changes into single ones seems better.
Refer to the inline comments.

I don't think so. Mount option background_gc is changed to make remount_fs working in the correct way.

Quoted 6 lines

Thanks,
Gu

Though simply option show is enough, but I think the "background_gc=on/off" is more friendly.

Yes, Agree. I will update.

Quoted 11 lines

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 352

Quoted 8 lines
fs/f2fs/super.c +7 -0
+
+     /**
+      * We stop the GC thread if FS is mounted as RO
+      * or if background_gc = 0 is passed in mount
+      * option. Also sync the filesystem.
+      */
+     if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {

Another condition: The old mount is not RO.

I don't think that it is needed. I think current condition check can be covered about all cases.
Am I missing something ?

Quoted 14 lines

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 363

Quoted 12 lines
fs/f2fs/super.c +11 -0
+             stop_gc_thread(sbi);
+             f2fs_sync_fs(sb, 1);
+     } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
+             err = start_gc_thread(sbi);
+             if (err)
+                     goto restore_opts;
+     }
+
+     /* Update the POSIXACL Flag */
+      sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
+             (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
Quoted 1 line

Maybe you forget to update the flags with MS_RDONLY or ~MS_RDONLY, if the flags changed.

No, we don't need to check this flags. sb-s_flags will be updated by MS_REMOUNT of vfs.(do_remount_sb)

Quoted 9 lines

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 367

Quoted 5 lines
fs/f2fs/super.c +4 -0
+     return 0;
+
+restore_opts:
+     sb->s_flags = old_sb_flags;

There is no need to restore sb->s_flags, parse_options() did not change it. no need to store the old sb->s_flags too.

Yes, right, I will update.

Quoted 13 lines

fs/f2fs/super.c static int f2fs_fill_super(struct super_block *sb, void *data, int silent) · line 737

Quoted 12 lines
fs/f2fs/super.c +6 -4
-     /* After POR, we can run background GC thread */
-     err = start_gc_thread(sbi);
-     if (err)
-             goto fail;
+     /* After POR, we can run background GC thread.*/
+     if (!(sb->s_flags & MS_RDONLY)) {
+             /**
+              * If filesystem is mounted as read-only then
+              * do not start the gc_thread.
+              */
Quoted 1 line

It seems that the comment here is against with the logic.

hum.. Okay, I will update comment to avoid some confusion.

Thanks for review :)
I will post v2 patch including your opinion soon.

------------------------------------------------------------------------------
How ServiceNow helps IT people transform IT departments:
1. A cloud service to automate IT design, transition and operations
2. Dashboards that offer high-level views of enterprise services
3. A single system of record for all IT processes
http://p.sf.net/sfu/servicenow-d2d-j

#3 lore raw

Hi Namjae,

Quoted 18 lines · Namjae Jeon

2013/6/4 Gu Zheng <[email protected]>:

Quoted 17 lines · Namjae Jeon
Quoted 14 lines

From: Namjae Jeon <[email protected]>

Add the f2fs_remount function call which will be used during the filesystem remounting. This function will help us to change the mount options specific to f2fs.

Also modify the f2fs background_gc mount option, which will allow the user to dynamically trun on/off the garbage collection in f2fs based on the background_gc value. If background_gc=0, Garbage collection will be turned off & if background_gc=1, Garbage collection will be truned on.

Quoted 7 lines
Quoted 1 line

Hi Namjae,

Hi. Gu.

Quoted 2 lines

I think splitting these two changes into single ones seems better.
Refer to the inline comments.

I don't think so. Mount option background_gc is changed to make remount_fs working in the correct way.

Yes, I know. Maybe you somewhat misread my words.
Though remount_fs is dependent on changing background_gc option, but the change of background_gc option and the adding remount_fs support are two different changes.
In order to make each patch simple and clear, maybe you need to split into single ones, such as:
[PATCH 1/3] f2fs: Modify the f2fs background_gc mount option
[PATCH 2/3] f2fs: add remount_fs callback support
[PATCH 3/3] f2fs: reorganise the function get_victim_by_default

Just a personal suggestion, if you think it is worthless, please ignore it.:)

Quoted 19 lines
Quoted 7 lines

Thanks,
Gu

Though simply option show is enough, but I think the "background_gc=on/off" is more friendly.

Yes, Agree. I will update.

Quoted 10 lines
Quoted 8 lines
fs/f2fs/super.c +7 -0
+
+     /**
+      * We stop the GC thread if FS is mounted as RO
+      * or if background_gc = 0 is passed in mount
+      * option. Also sync the filesystem.
+      */
+     if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
Quoted 4 lines
Quoted 1 line

Another condition: The old mount is not RO.

I don't think that it is needed. I think current condition check can be covered about all cases.
Am I missing something ?

Maybe. If the old mount is RO, so does the remount. It still can pass the judgement here, right? Though the following stop_gc_thread() and f2fs_sync_fs() can handle this case well, but this is unnecessary and needless. If we add additional judgement of whether old mount is not RO can avoid this.

Thanks,
Gu

Quoted 49 lines
Quoted 15 lines

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 363

Quoted 13 lines
fs/f2fs/super.c +11 -0
+             stop_gc_thread(sbi);
+             f2fs_sync_fs(sb, 1);
+     } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
+             err = start_gc_thread(sbi);
+             if (err)
+                     goto restore_opts;
+     }
+
+     /* Update the POSIXACL Flag */
+      sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
+             (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
Quoted 1 line

Maybe you forget to update the flags with MS_RDONLY or ~MS_RDONLY, if the flags changed.

No, we don't need to check this flags. sb-s_flags will be updated by MS_REMOUNT of vfs.(do_remount_sb)

Quoted 9 lines

fs/f2fs/super.c static int f2fs_show_options(struct seq_file *seq, struct dentry *root) · line 367

Quoted 5 lines
fs/f2fs/super.c +4 -0
+     return 0;
+
+restore_opts:
+     sb->s_flags = old_sb_flags;

There is no need to restore sb->s_flags, parse_options() did not change it. no need to store the old sb->s_flags too.

Yes, right, I will update.

Quoted 13 lines

fs/f2fs/super.c static int f2fs_fill_super(struct super_block *sb, void *data, int silent) · line 737

Quoted 12 lines
fs/f2fs/super.c +6 -4
-     /* After POR, we can run background GC thread */
-     err = start_gc_thread(sbi);
-     if (err)
-             goto fail;
+     /* After POR, we can run background GC thread.*/
+     if (!(sb->s_flags & MS_RDONLY)) {
+             /**
+              * If filesystem is mounted as read-only then
+              * do not start the gc_thread.
+              */
Quoted 1 line

It seems that the comment here is against with the logic.

hum.. Okay, I will update comment to avoid some confusion.

Thanks for review :)
I will post v2 patch including your opinion soon.

#4 lore raw

[snip]

Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 9 lines
fs/f2fs/super.c +7 -0
+
+     /**
+      * We stop the GC thread if FS is mounted as RO
+      * or if background_gc = 0 is passed in mount
+      * option. Also sync the filesystem.
+      */
+     if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
Quoted 5 lines
Quoted 4 lines
Quoted 1 line

Another condition: The old mount is not RO.

I don't think that it is needed. I think current condition check can be covered about all cases.
Am I missing something ?

Hi Gu.

Quoted 4 lines

Maybe. If the old mount is RO, so does the remount. It still can pass the judgement here, right? Though the following stop_gc_thread() and f2fs_sync_fs() can handle this case well, but this is unnecessary and needless. If we add additional judgement of whether old mount is not RO can avoid this.

Okay, I understand.
I will update it on v2 patch.

Thanks~.

Quoted 5 lines

Thanks,
Gu

Quoted 2 lines
Quoted 2 lines

------------------------------------------------------------------------------
How ServiceNow helps IT people transform IT departments:
1. A cloud service to automate IT design, transition and operations
2. Dashboards that offer high-level views of enterprise services
3. A single system of record for all IT processes
http://p.sf.net/sfu/servicenow-d2d-j