Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2: nilfs2: add reader's lock for cno in nilfs_ioctl_sync nilfs2: delete unnecessary condition in load_segment_summary nilfs2: move iterator to write log into segment buffer nilfs2: get rid of s_dirt flag use nilfs2: get rid of nilfs_segctor_req struct nilfs2: delete unnecessary condition in nilfs_dat_translate nilfs2: fix potential hang in nilfs_error on errors=remount-ro nilfs2: use mnt_want_write in ioctls where write access is needed nilfs2: issue discard request after cleaning segments
2010-03-03 08:53:49 -08:00 · 2010-03-03 08:53:49 -08:00 · feaf77d51a
--- a/Documentation/filesystems/nilfs2.txt
+++ b/Documentation/filesystems/nilfs2.txt
@ -74,6 +74,9 @@ norecovery		Disable recovery of the filesystem on mount.
 			This disables every write access on the device for
 			read-only mounts or snapshots.  This option will fail
 			for r/w mounts on an unclean volume.
+discard			Issue discard/TRIM commands to the underlying block
+			device when blocks are freed.  This is useful for SSD
+			devices and sparse/thinly-provisioned LUNs.

 NILFS2 usage
 ============
--- a/fs/nilfs2/dat.c
+++ b/fs/nilfs2/dat.c
@ -388,8 +388,7 @@ int nilfs_dat_translate(struct inode *dat, __u64 vblocknr, sector_t *blocknrp)
 		ret = -ENOENT;
 		goto out;
 	}
-	if (blocknrp != NULL)
-		*blocknrp = blocknr;
+	*blocknrp = blocknr;

 out:
 	kunmap_atomic(kaddr, KM_USER0);
--- a/fs/nilfs2/ioctl.c
+++ b/fs/nilfs2/ioctl.c
@ -26,6 +26,7 @@
 #include <linux/capability.h>	/* capable() */
 #include <linux/uaccess.h>	/* copy_from_user(), copy_to_user() */
 #include <linux/vmalloc.h>
+#include <linux/mount.h>	/* mnt_want_write(), mnt_drop_write() */
 #include <linux/nilfs2_fs.h>
 #include "nilfs.h"
 #include "segment.h"
@ -107,20 +108,28 @@ static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,

 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
+
+	ret = mnt_want_write(filp->f_path.mnt);
+	if (ret)
+		return ret;
+
+	ret = -EFAULT;
 	if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
-		return -EFAULT;
+		goto out;

 	mutex_lock(&nilfs->ns_mount_mutex);
+
 	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 	ret = nilfs_cpfile_change_cpmode(
 		cpfile, cpmode.cm_cno, cpmode.cm_mode);
-	if (unlikely(ret < 0)) {
+	if (unlikely(ret < 0))
 		nilfs_transaction_abort(inode->i_sb);
-		mutex_unlock(&nilfs->ns_mount_mutex);
-		return ret;
-	}
-	nilfs_transaction_commit(inode->i_sb); /* never fails */
+	else
+		nilfs_transaction_commit(inode->i_sb); /* never fails */
+
 	mutex_unlock(&nilfs->ns_mount_mutex);
+out:
+	mnt_drop_write(filp->f_path.mnt);
 	return ret;
 }

@ -135,16 +144,23 @@ nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp,

 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
+
+	ret = mnt_want_write(filp->f_path.mnt);
+	if (ret)
+		return ret;
+
+	ret = -EFAULT;
 	if (copy_from_user(&cno, argp, sizeof(cno)))
-		return -EFAULT;
+		goto out;

 	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 	ret = nilfs_cpfile_delete_checkpoint(cpfile, cno);
-	if (unlikely(ret < 0)) {
+	if (unlikely(ret < 0))
 		nilfs_transaction_abort(inode->i_sb);
-		return ret;
-	}
-	nilfs_transaction_commit(inode->i_sb); /* never fails */
+	else
+		nilfs_transaction_commit(inode->i_sb); /* never fails */
+out:
+	mnt_drop_write(filp->f_path.mnt);
 	return ret;
 }

@ -496,12 +512,19 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;

-	if (copy_from_user(argv, argp, sizeof(argv)))
-		return -EFAULT;
+	ret = mnt_want_write(filp->f_path.mnt);
+	if (ret)
+		return ret;

+	ret = -EFAULT;
+	if (copy_from_user(argv, argp, sizeof(argv)))
+		goto out;
+
+	ret = -EINVAL;
 	nsegs = argv[4].v_nmembs;
 	if (argv[4].v_size != argsz[4])
-		return -EINVAL;
+		goto out;
+
 	/*
 	 * argv[4] points to segment numbers this ioctl cleans.  We
 	 * use kmalloc() for its buffer because memory used for the
@ -509,9 +532,10 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	 */
 	kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base,
 			       nsegs * sizeof(__u64));
-	if (IS_ERR(kbufs[4]))
-		return PTR_ERR(kbufs[4]);
-
+	if (IS_ERR(kbufs[4])) {
+		ret = PTR_ERR(kbufs[4]);
+		goto out;
+	}
 	nilfs = NILFS_SB(inode->i_sb)->s_nilfs;

 	for (n = 0; n < 4; n++) {
@ -563,10 +587,12 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 		nilfs_remove_all_gcinode(nilfs);
 	clear_nilfs_gc_running(nilfs);

- out_free:
+out_free:
 	while (--n >= 0)
 		vfree(kbufs[n]);
 	kfree(kbufs[4]);
+out:
+	mnt_drop_write(filp->f_path.mnt);
 	return ret;
 }

@ -575,13 +601,17 @@ static int nilfs_ioctl_sync(struct inode *inode, struct file *filp,
 {
 	__u64 cno;
 	int ret;
+	struct the_nilfs *nilfs;

 	ret = nilfs_construct_segment(inode->i_sb);
 	if (ret < 0)
 		return ret;

 	if (argp != NULL) {
-		cno = NILFS_SB(inode->i_sb)->s_nilfs->ns_cno - 1;
+		nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
+		down_read(&nilfs->ns_segctor_sem);
+		cno = nilfs->ns_cno - 1;
+		up_read(&nilfs->ns_segctor_sem);
 		if (copy_to_user(argp, &cno, sizeof(cno)))
 			return -EFAULT;
 	}
--- a/fs/nilfs2/recovery.c
+++ b/fs/nilfs2/recovery.c
@ -39,7 +39,6 @@ enum {
 	NILFS_SEG_FAIL_IO,
 	NILFS_SEG_FAIL_MAGIC,
 	NILFS_SEG_FAIL_SEQ,
-	NILFS_SEG_FAIL_CHECKSUM_SEGSUM,
 	NILFS_SEG_FAIL_CHECKSUM_SUPER_ROOT,
 	NILFS_SEG_FAIL_CHECKSUM_FULL,
 	NILFS_SEG_FAIL_CONSISTENCY,
@ -71,10 +70,6 @@ static int nilfs_warn_segment_error(int err)
 		printk(KERN_WARNING
 		       "NILFS warning: Sequence number mismatch\n");
 		break;
-	case NILFS_SEG_FAIL_CHECKSUM_SEGSUM:
-		printk(KERN_WARNING
-		       "NILFS warning: Checksum error in segment summary\n");
-		break;
 	case NILFS_SEG_FAIL_CHECKSUM_SUPER_ROOT:
 		printk(KERN_WARNING
 		       "NILFS warning: Checksum error in super root\n");
@ -206,19 +201,15 @@ int nilfs_read_super_root_block(struct super_block *sb, sector_t sr_block,
 * @pseg_start: start disk block number of partial segment
 * @seg_seq: sequence number requested
 * @ssi: pointer to nilfs_segsum_info struct to store information
- * @full_check: full check flag
- *              (0: only checks segment summary CRC, 1: data CRC)
 */
 static int
 load_segment_summary(struct nilfs_sb_info *sbi, sector_t pseg_start,
-		     u64 seg_seq, struct nilfs_segsum_info *ssi,
-		     int full_check)
+		     u64 seg_seq, struct nilfs_segsum_info *ssi)
 {
 	struct buffer_head *bh_sum;
 	struct nilfs_segment_summary *sum;
-	unsigned long offset, nblock;
-	u64 check_bytes;
-	u32 crc, crc_sum;
+	unsigned long nblock;
+	u32 crc;
 	int ret = NILFS_SEG_FAIL_IO;

 	bh_sum = sb_bread(sbi->s_super, pseg_start);
@ -237,34 +228,24 @@ load_segment_summary(struct nilfs_sb_info *sbi, sector_t pseg_start,
 		ret = NILFS_SEG_FAIL_SEQ;
 		goto failed;
 	}
-	if (full_check) {
-		offset = sizeof(sum->ss_datasum);
-		check_bytes =
-			((u64)ssi->nblocks << sbi->s_super->s_blocksize_bits);
-		nblock = ssi->nblocks;
-		crc_sum = le32_to_cpu(sum->ss_datasum);
-		ret = NILFS_SEG_FAIL_CHECKSUM_FULL;
-	} else { /* only checks segment summary */
-		offset = sizeof(sum->ss_datasum) + sizeof(sum->ss_sumsum);
-		check_bytes = ssi->sumbytes;
-		nblock = ssi->nsumblk;
-		crc_sum = le32_to_cpu(sum->ss_sumsum);
-		ret = NILFS_SEG_FAIL_CHECKSUM_SEGSUM;
-	}

+	nblock = ssi->nblocks;
 	if (unlikely(nblock == 0 ||
 		     nblock > sbi->s_nilfs->ns_blocks_per_segment)) {
 		/* This limits the number of blocks read in the CRC check */
 		ret = NILFS_SEG_FAIL_CONSISTENCY;
 		goto failed;
 	}
-	if (calc_crc_cont(sbi, bh_sum, &crc, offset, check_bytes,
+	if (calc_crc_cont(sbi, bh_sum, &crc, sizeof(sum->ss_datasum),
+			  ((u64)nblock << sbi->s_super->s_blocksize_bits),
 			  pseg_start, nblock)) {
 		ret = NILFS_SEG_FAIL_IO;
 		goto failed;
 	}
-	if (crc == crc_sum)
+	if (crc == le32_to_cpu(sum->ss_datasum))
 		ret = 0;
+	else
+		ret = NILFS_SEG_FAIL_CHECKSUM_FULL;
 failed:
 	brelse(bh_sum);
 out:
@ -598,7 +579,7 @@ static int nilfs_do_roll_forward(struct the_nilfs *nilfs,

 	while (segnum != ri->ri_segnum || pseg_start <= ri->ri_pseg_start) {

-		ret = load_segment_summary(sbi, pseg_start, seg_seq, &ssi, 1);
+		ret = load_segment_summary(sbi, pseg_start, seg_seq, &ssi);
 		if (ret) {
 			if (ret == NILFS_SEG_FAIL_IO) {
 				err = -EIO;
@ -821,7 +802,7 @@ int nilfs_search_super_root(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi,

 	for (;;) {
 		/* Load segment summary */
-		ret = load_segment_summary(sbi, pseg_start, seg_seq, &ssi, 1);
+		ret = load_segment_summary(sbi, pseg_start, seg_seq, &ssi);
 		if (ret) {
 			if (ret == NILFS_SEG_FAIL_IO)
 				goto failed;
--- a/fs/nilfs2/segbuf.c
+++ b/fs/nilfs2/segbuf.c
@ -40,6 +40,11 @@ struct nilfs_write_info {
 };


+static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
+			      struct the_nilfs *nilfs);
+static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf);
+
+
 static struct kmem_cache *nilfs_segbuf_cachep;

 static void nilfs_segbuf_init_once(void *obj)
@ -302,6 +307,19 @@ void nilfs_truncate_logs(struct list_head *logs,
 	}
 }

+int nilfs_write_logs(struct list_head *logs, struct the_nilfs *nilfs)
+{
+	struct nilfs_segment_buffer *segbuf;
+	int ret = 0;
+
+	list_for_each_entry(segbuf, logs, sb_list) {
+		ret = nilfs_segbuf_write(segbuf, nilfs);
+		if (ret)
+			break;
+	}
+	return ret;
+}
+
 int nilfs_wait_on_logs(struct list_head *logs)
 {
 	struct nilfs_segment_buffer *segbuf;
--- a/fs/nilfs2/segbuf.h
+++ b/fs/nilfs2/segbuf.h
@ -166,13 +166,10 @@ nilfs_segbuf_add_file_buffer(struct nilfs_segment_buffer *segbuf,
 	segbuf->sb_sum.nfileblk++;
 }

-int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
-		       struct the_nilfs *nilfs);
-int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf);
-
 void nilfs_clear_logs(struct list_head *logs);
 void nilfs_truncate_logs(struct list_head *logs,
 			 struct nilfs_segment_buffer *last);
+int nilfs_write_logs(struct list_head *logs, struct the_nilfs *nilfs);
 int nilfs_wait_on_logs(struct list_head *logs);

 static inline void nilfs_destroy_logs(struct list_head *logs)
--- a/fs/nilfs2/segment.c
+++ b/fs/nilfs2/segment.c
@ -1764,14 +1764,9 @@ static int nilfs_segctor_prepare_write(struct nilfs_sc_info *sci,
 static int nilfs_segctor_write(struct nilfs_sc_info *sci,
 			       struct the_nilfs *nilfs)
 {
-	struct nilfs_segment_buffer *segbuf;
-	int ret = 0;
+	int ret;

-	list_for_each_entry(segbuf, &sci->sc_segbufs, sb_list) {
-		ret = nilfs_segbuf_write(segbuf, nilfs);
-		if (ret)
-			break;
-	}
+	ret = nilfs_write_logs(&sci->sc_segbufs, nilfs);
 	list_splice_tail_init(&sci->sc_segbufs, &sci->sc_write_logs);
 	return ret;
 }
@ -1937,8 +1932,7 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
 {
 	struct nilfs_segment_buffer *segbuf;
 	struct page *bd_page = NULL, *fs_page = NULL;
-	struct nilfs_sb_info *sbi = sci->sc_sbi;
-	struct the_nilfs *nilfs = sbi->s_nilfs;
+	struct the_nilfs *nilfs = sci->sc_sbi->s_nilfs;
 	int update_sr = (sci->sc_super_root != NULL);

 	list_for_each_entry(segbuf, &sci->sc_write_logs, sb_list) {
@ -2020,7 +2014,7 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
 	if (update_sr) {
 		nilfs_set_last_segment(nilfs, segbuf->sb_pseg_start,
 				       segbuf->sb_sum.seg_seq, nilfs->ns_cno++);
-		sbi->s_super->s_dirt = 1;
+		set_nilfs_sb_dirty(nilfs);

 		clear_bit(NILFS_SC_HAVE_DELTA, &sci->sc_flags);
 		clear_bit(NILFS_SC_DIRTY, &sci->sc_flags);
@ -2425,43 +2419,43 @@ int nilfs_construct_dsync_segment(struct super_block *sb, struct inode *inode,
 	return err;
 }

-struct nilfs_segctor_req {
-	int mode;
-	__u32 seq_accepted;
-	int sc_err;  /* construction failure */
-	int sb_err;  /* super block writeback failure */
-};
-
 #define FLUSH_FILE_BIT	(0x1) /* data file only */
 #define FLUSH_DAT_BIT	(1 << NILFS_DAT_INO) /* DAT only */

-static void nilfs_segctor_accept(struct nilfs_sc_info *sci,
-				 struct nilfs_segctor_req *req)
+/**
+ * nilfs_segctor_accept - record accepted sequence count of log-write requests
+ * @sci: segment constructor object
+ */
+static void nilfs_segctor_accept(struct nilfs_sc_info *sci)
 {
-	req->sc_err = req->sb_err = 0;
 	spin_lock(&sci->sc_state_lock);
-	req->seq_accepted = sci->sc_seq_request;
+	sci->sc_seq_accepted = sci->sc_seq_request;
 	spin_unlock(&sci->sc_state_lock);

 	if (sci->sc_timer)
 		del_timer_sync(sci->sc_timer);
 }

-static void nilfs_segctor_notify(struct nilfs_sc_info *sci,
-				 struct nilfs_segctor_req *req)
+/**
+ * nilfs_segctor_notify - notify the result of request to caller threads
+ * @sci: segment constructor object
+ * @mode: mode of log forming
+ * @err: error code to be notified
+ */
+static void nilfs_segctor_notify(struct nilfs_sc_info *sci, int mode, int err)
 {
 	/* Clear requests (even when the construction failed) */
 	spin_lock(&sci->sc_state_lock);

-	if (req->mode == SC_LSEG_SR) {
+	if (mode == SC_LSEG_SR) {
 		sci->sc_state &= ~NILFS_SEGCTOR_COMMIT;
-		sci->sc_seq_done = req->seq_accepted;
-		nilfs_segctor_wakeup(sci, req->sc_err ? : req->sb_err);
+		sci->sc_seq_done = sci->sc_seq_accepted;
+		nilfs_segctor_wakeup(sci, err);
 		sci->sc_flush_request = 0;
 	} else {
-		if (req->mode == SC_FLUSH_FILE)
+		if (mode == SC_FLUSH_FILE)
 			sci->sc_flush_request &= ~FLUSH_FILE_BIT;
-		else if (req->mode == SC_FLUSH_DAT)
+		else if (mode == SC_FLUSH_DAT)
 			sci->sc_flush_request &= ~FLUSH_DAT_BIT;

 		/* re-enable timer if checkpoint creation was not done */
@ -2472,30 +2466,37 @@ static void nilfs_segctor_notify(struct nilfs_sc_info *sci,
 	spin_unlock(&sci->sc_state_lock);
 }

-static int nilfs_segctor_construct(struct nilfs_sc_info *sci,
-				   struct nilfs_segctor_req *req)
+/**
+ * nilfs_segctor_construct - form logs and write them to disk
+ * @sci: segment constructor object
+ * @mode: mode of log forming
+ */
+static int nilfs_segctor_construct(struct nilfs_sc_info *sci, int mode)
 {
 	struct nilfs_sb_info *sbi = sci->sc_sbi;
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 	int err = 0;

+	nilfs_segctor_accept(sci);
+
 	if (nilfs_discontinued(nilfs))
-		req->mode = SC_LSEG_SR;
-	if (!nilfs_segctor_confirm(sci)) {
-		err = nilfs_segctor_do_construct(sci, req->mode);
-		req->sc_err = err;
-	}
+		mode = SC_LSEG_SR;
+	if (!nilfs_segctor_confirm(sci))
+		err = nilfs_segctor_do_construct(sci, mode);
+
 	if (likely(!err)) {
-		if (req->mode != SC_FLUSH_DAT)
+		if (mode != SC_FLUSH_DAT)
 			atomic_set(&nilfs->ns_ndirtyblks, 0);
 		if (test_bit(NILFS_SC_SUPER_ROOT, &sci->sc_flags) &&
 		    nilfs_discontinued(nilfs)) {
 			down_write(&nilfs->ns_sem);
-			req->sb_err = nilfs_commit_super(sbi,
-					nilfs_altsb_need_update(nilfs));
+			err = nilfs_commit_super(
+				sbi, nilfs_altsb_need_update(nilfs));
 			up_write(&nilfs->ns_sem);
 		}
 	}
+
+	nilfs_segctor_notify(sci, mode, err);
 	return err;
 }

@ -2526,7 +2527,6 @@ int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
 	struct nilfs_sc_info *sci = NILFS_SC(sbi);
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 	struct nilfs_transaction_info ti;
-	struct nilfs_segctor_req req = { .mode = SC_LSEG_SR };
 	int err;

 	if (unlikely(!sci))
@ -2547,10 +2547,8 @@ int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
 	list_splice_tail_init(&nilfs->ns_gc_inodes, &sci->sc_gc_inodes);

 	for (;;) {
-		nilfs_segctor_accept(sci, &req);
-		err = nilfs_segctor_construct(sci, &req);
+		err = nilfs_segctor_construct(sci, SC_LSEG_SR);
 		nilfs_remove_written_gcinodes(nilfs, &sci->sc_gc_inodes);
-		nilfs_segctor_notify(sci, &req);

 		if (likely(!err))
 			break;
@ -2560,6 +2558,16 @@ int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
 		set_current_state(TASK_INTERRUPTIBLE);
 		schedule_timeout(sci->sc_interval);
 	}
+	if (nilfs_test_opt(sbi, DISCARD)) {
+		int ret = nilfs_discard_segments(nilfs, sci->sc_freesegs,
+						 sci->sc_nfreesegs);
+		if (ret) {
+			printk(KERN_WARNING
+			       "NILFS warning: error %d on discard request, "
+			       "turning discards off for the device\n", ret);
+			nilfs_clear_opt(sbi, DISCARD);
+		}
+	}

 out_unlock:
 	sci->sc_freesegs = NULL;
@ -2573,13 +2581,9 @@ static void nilfs_segctor_thread_construct(struct nilfs_sc_info *sci, int mode)
 {
 	struct nilfs_sb_info *sbi = sci->sc_sbi;
 	struct nilfs_transaction_info ti;
-	struct nilfs_segctor_req req = { .mode = mode };

 	nilfs_transaction_lock(sbi, &ti, 0);
-
-	nilfs_segctor_accept(sci, &req);
-	nilfs_segctor_construct(sci, &req);
-	nilfs_segctor_notify(sci, &req);
+	nilfs_segctor_construct(sci, mode);

 	/*
 	 * Unclosed segment should be retried.  We do this using sc_timer.
@ -2635,6 +2639,7 @@ static int nilfs_segctor_flush_mode(struct nilfs_sc_info *sci)
 static int nilfs_segctor_thread(void *arg)
 {
 	struct nilfs_sc_info *sci = (struct nilfs_sc_info *)arg;
+	struct the_nilfs *nilfs = sci->sc_sbi->s_nilfs;
 	struct timer_list timer;
 	int timeout = 0;

@ -2680,7 +2685,6 @@ static int nilfs_segctor_thread(void *arg)
 	} else {
 		DEFINE_WAIT(wait);
 		int should_sleep = 1;
-		struct the_nilfs *nilfs;

 		prepare_to_wait(&sci->sc_wait_daemon, &wait,
 				TASK_INTERRUPTIBLE);
@ -2701,8 +2705,8 @@ static int nilfs_segctor_thread(void *arg)
 		finish_wait(&sci->sc_wait_daemon, &wait);
 		timeout = ((sci->sc_state & NILFS_SEGCTOR_COMMIT) &&
 			   time_after_eq(jiffies, sci->sc_timer->expires));
-		nilfs = sci->sc_sbi->s_nilfs;
-		if (sci->sc_super->s_dirt && nilfs_sb_need_update(nilfs))
+
+		if (nilfs_sb_dirty(nilfs) && nilfs_sb_need_update(nilfs))
 			set_nilfs_discontinued(nilfs);
 	}
 	goto loop;
@ -2797,12 +2801,9 @@ static void nilfs_segctor_write_out(struct nilfs_sc_info *sci)
 	do {
 		struct nilfs_sb_info *sbi = sci->sc_sbi;
 		struct nilfs_transaction_info ti;
-		struct nilfs_segctor_req req = { .mode = SC_LSEG_SR };

 		nilfs_transaction_lock(sbi, &ti, 0);
-		nilfs_segctor_accept(sci, &req);
-		ret = nilfs_segctor_construct(sci, &req);
-		nilfs_segctor_notify(sci, &req);
+		ret = nilfs_segctor_construct(sci, SC_LSEG_SR);
 		nilfs_transaction_unlock(sbi);

 	} while (ret && retrycount-- > 0);
@ -2865,8 +2866,15 @@ int nilfs_attach_segment_constructor(struct nilfs_sb_info *sbi)
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 	int err;

-	/* Each field of nilfs_segctor is cleared through the initialization
-	   of super-block info */
+	if (NILFS_SC(sbi)) {
+		/*
+		 * This happens if the filesystem was remounted
+		 * read/write after nilfs_error degenerated it into a
+		 * read-only mount.
+		 */
+		nilfs_detach_segment_constructor(sbi);
+	}
+
 	sbi->s_sc_info = nilfs_segctor_new(sbi);
 	if (!sbi->s_sc_info)
 		return -ENOMEM;
--- a/fs/nilfs2/segment.h
+++ b/fs/nilfs2/segment.h
@ -116,6 +116,7 @@ struct nilfs_segsum_pointer {
 * @sc_wait_daemon: Daemon wait queue
 * @sc_wait_task: Start/end wait queue to control segctord task
 * @sc_seq_request: Request counter
+ * @sc_seq_accept: Accepted request count
 * @sc_seq_done: Completion counter
 * @sc_sync: Request of explicit sync operation
 * @sc_interval: Timeout value of background construction
@ -169,6 +170,7 @@ struct nilfs_sc_info {
 	wait_queue_head_t	sc_wait_task;

 	__u32			sc_seq_request;
+	__u32			sc_seq_accepted;
 	__u32			sc_seq_done;

 	int			sc_sync;
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@ -96,9 +96,6 @@ void nilfs_error(struct super_block *sb, const char *function,
 	if (!(sb->s_flags & MS_RDONLY)) {
 		struct the_nilfs *nilfs = sbi->s_nilfs;

-		if (!nilfs_test_opt(sbi, ERRORS_CONT))
-			nilfs_detach_segment_constructor(sbi);
-
 		down_write(&nilfs->ns_sem);
 		if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
 			nilfs->ns_mount_state |= NILFS_ERROR_FS;
@ -301,7 +298,7 @@ int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
 		memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
 		nilfs->ns_sbwtime[1] = t;
 	}
-	sbi->s_super->s_dirt = 0;
+	clear_nilfs_sb_dirty(nilfs);
 	return nilfs_sync_super(sbi, dupsb);
 }

@ -345,7 +342,7 @@ static int nilfs_sync_fs(struct super_block *sb, int wait)
 		err = nilfs_construct_segment(sb);

 	down_write(&nilfs->ns_sem);
-	if (sb->s_dirt)
+	if (nilfs_sb_dirty(nilfs))
 		nilfs_commit_super(sbi, 1);
 	up_write(&nilfs->ns_sem);

@ -481,6 +478,8 @@ static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_printf(seq, ",order=strict");
 	if (nilfs_test_opt(sbi, NORECOVERY))
 		seq_printf(seq, ",norecovery");
+	if (nilfs_test_opt(sbi, DISCARD))
+		seq_printf(seq, ",discard");

 	return 0;
 }
@ -550,7 +549,7 @@ static const struct export_operations nilfs_export_ops = {
 enum {
 	Opt_err_cont, Opt_err_panic, Opt_err_ro,
 	Opt_nobarrier, Opt_snapshot, Opt_order, Opt_norecovery,
-	Opt_err,
+	Opt_discard, Opt_err,
 };

 static match_table_t tokens = {
@ -561,6 +560,7 @@ static match_table_t tokens = {
 	{Opt_snapshot, "cp=%u"},
 	{Opt_order, "order=%s"},
 	{Opt_norecovery, "norecovery"},
+	{Opt_discard, "discard"},
 	{Opt_err, NULL}
 };

@ -614,6 +614,9 @@ static int parse_options(char *options, struct super_block *sb)
 		case Opt_norecovery:
 			nilfs_set_opt(sbi, NORECOVERY);
 			break;
+		case Opt_discard:
+			nilfs_set_opt(sbi, DISCARD);
+			break;
 		default:
 			printk(KERN_ERR
 			       "NILFS: Unrecognized mount option \"%s\"\n", p);
--- a/fs/nilfs2/the_nilfs.c
+++ b/fs/nilfs2/the_nilfs.c
@ -646,6 +646,44 @@ int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
 	goto out;
 }

+int nilfs_discard_segments(struct the_nilfs *nilfs, __u64 *segnump,
+			    size_t nsegs)
+{
+	sector_t seg_start, seg_end;
+	sector_t start = 0, nblocks = 0;
+	unsigned int sects_per_block;
+	__u64 *sn;
+	int ret = 0;
+
+	sects_per_block = (1 << nilfs->ns_blocksize_bits) /
+		bdev_logical_block_size(nilfs->ns_bdev);
+	for (sn = segnump; sn < segnump + nsegs; sn++) {
+		nilfs_get_segment_range(nilfs, *sn, &seg_start, &seg_end);
+
+		if (!nblocks) {
+			start = seg_start;
+			nblocks = seg_end - seg_start + 1;
+		} else if (start + nblocks == seg_start) {
+			nblocks += seg_end - seg_start + 1;
+		} else {
+			ret = blkdev_issue_discard(nilfs->ns_bdev,
+						   start * sects_per_block,
+						   nblocks * sects_per_block,
+						   GFP_NOFS,
+						   DISCARD_FL_BARRIER);
+			if (ret < 0)
+				return ret;
+			nblocks = 0;
+		}
+	}
+	if (nblocks)
+		ret = blkdev_issue_discard(nilfs->ns_bdev,
+					   start * sects_per_block,
+					   nblocks * sects_per_block,
+					   GFP_NOFS, DISCARD_FL_BARRIER);
+	return ret;
+}
+
 int nilfs_count_free_blocks(struct the_nilfs *nilfs, sector_t *nblocks)
 {
 	struct inode *dat = nilfs_dat_inode(nilfs);
--- a/fs/nilfs2/the_nilfs.h
+++ b/fs/nilfs2/the_nilfs.h
@ -38,6 +38,7 @@ enum {
 				   the latest checkpoint was loaded */
 	THE_NILFS_DISCONTINUED,	/* 'next' pointer chain has broken */
 	THE_NILFS_GC_RUNNING,	/* gc process is running */
+	THE_NILFS_SB_DIRTY,	/* super block is dirty */
 };

 /**
@ -197,6 +198,7 @@ THE_NILFS_FNS(INIT, init)
 THE_NILFS_FNS(LOADED, loaded)
 THE_NILFS_FNS(DISCONTINUED, discontinued)
 THE_NILFS_FNS(GC_RUNNING, gc_running)
+THE_NILFS_FNS(SB_DIRTY, sb_dirty)

 /* Minimum interval of periodical update of superblocks (in seconds) */
 #define NILFS_SB_FREQ		10
@ -221,6 +223,7 @@ struct the_nilfs *find_or_create_nilfs(struct block_device *);
 void put_nilfs(struct the_nilfs *);
 int init_nilfs(struct the_nilfs *, struct nilfs_sb_info *, char *);
 int load_nilfs(struct the_nilfs *, struct nilfs_sb_info *);
+int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
 struct nilfs_sb_info *nilfs_find_sbinfo(struct the_nilfs *, int, __u64);
 int nilfs_checkpoint_is_mounted(struct the_nilfs *, __u64, int);
--- a/include/linux/nilfs2_fs.h
+++ b/include/linux/nilfs2_fs.h
@ -153,6 +153,7 @@ struct nilfs_super_root {
 						   semantics also for data */
 #define NILFS_MOUNT_NORECOVERY		0x4000  /* Disable write access during
 						   mount-time recovery */
+#define NILFS_MOUNT_DISCARD		0x8000  /* Issue DISCARD requests */


 /**