login
Search
Hi, here is the stuff we have scheduled for 2.6.28. The patches contains various small fixes and cleanups, but there is something interesting as well. The first interesting thing is new "bulk read" functionality. The idea is that many NAND flashes support "bulk read" in some form. For example, OneNAND has "read while load" feature, which allows reading consecutive NAND pages faster than reading them one-by-one. So we've made UBIFS benefit from this feature and introduced new "bulk_read" mount option. With this option enabled UBIFS reads files a little ahead if the file data sits in consecutive physical addresses. For example, if user-space asks to read page zero of a file, and page 0-4 are in consecutive flash addressed, UBIFS reads pages 0-4 and populates them to the Page Cache. Note, this is disabled by default and UBIFS has to be explicitly mounted mounted with "bulk_read" option. The reason for this is that we consider this feature as experimental so far. Note, UBIFS does not use VFS read-ahead and actually explicitly disables it. This is because MTD is synchronous and all I/O is done synchronously, so read-ahead actually slows things down for UBIFS, instead of improving them. So the "bulk read" feature is basically internal UBIFS read-ahead implementation. We are able to gain 4-5MiB/s of read speed on OneNAND with bulk read enabled. The second interesting thing is new "no_chk_data_crc" mount option which disables data CRC32 checking. UBIFS always checks CRC of everything it reads from flash by default. On ARM platform this results in ~30% of total CPU usage in profiles, which is quite high. But many modern flashes are very reliable (e.g., OneNAND), and one does not need that level of protection. So now it is possible to disable CRC checking for _data_. However: * internal indexing information CRC is always checked; * when replaying the journal, data CRC is always checked; * on write, CRC is always calculated. With this mount option we are able to gain ...
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
When inserting into a full znode it is split into two
znodes. Because data node keys are usually consecutive,
it is better to try to keep them together. This patch
does a better job of that.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/tnc.c | 54 +++++++++++++++++++++++++++++++++---------------------
1 files changed, 33 insertions(+), 21 deletions(-)
diff --git a/fs/ubifs/tnc.c b/fs/ubifs/tnc.c
index 66dc571..e0878a4 100644
--- a/fs/ubifs/tnc.c
+++ b/fs/ubifs/tnc.c
@@ -1962,7 +1962,7 @@ static int tnc_insert(struct ubifs_info *c, struct ubifs_znode *znode,
{
struct ubifs_znode *zn, *zi, *zp;
int i, keep, move, appending = 0;
- union ubifs_key *key = &zbr->key;
+ union ubifs_key *key = &zbr->key, *key1;
ubifs_assert(n >= 0 && n <= c->fanout);
@@ -2003,20 +2003,33 @@ again:
zn->level = znode->level;
/* Decide where to split */
- if (znode->level == 0 && n == c->fanout &&
- key_type(c, key) == UBIFS_DATA_KEY) {
- union ubifs_key *key1;
-
- /*
- * If this is an inode which is being appended - do not split
- * it because no other zbranches can be inserted between
- * zbranches of consecutive data nodes anyway.
- */
- key1 = &znode->zbranch[n - 1].key;
- if (key_inum(c, key1) == key_inum(c, key) &&
- key_type(c, key1) == UBIFS_DATA_KEY &&
- key_block(c, key1) == key_block(c, key) - 1)
- appending = 1;
+ if (znode->level == 0 && key_type(c, key) == UBIFS_DATA_KEY) {
+ /* Try not to split consecutive data keys */
+ if (n == c->fanout) {
+ key1 = &znode->zbranch[n - 1].key;
+ if (key_inum(c, key1) == key_inum(c, key) &&
+ key_type(c, key1) == UBIFS_DATA_KEY)
+ appending = 1;
+ } else
+ goto check_split;
+ } else if (appending && n != c->fanout) {
+ /* Try not to split consecutive data keys */
+ appending = 0;
+check_split:
+ if (n >= (c->fanout + 1) / 2) {
+ key1 = &znode->zbranch[0].key;
+ if (key_inum(c, ...From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Znode may refer both data nodes and indexing nodes
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/ubifs.h | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index c3ac5a8..49b06c9 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -707,8 +707,8 @@ struct ubifs_jhead {
* struct ubifs_zbranch - key/coordinate/length branch stored in znodes.
* @key: key
* @znode: znode address in memory
- * @lnum: LEB number of the indexing node
- * @offs: offset of the indexing node within @lnum
+ * @lnum: LEB number of the target node (indexing node or data node)
+ * @offs: target node offset within @lnum
* @len: target node length
*/
struct ubifs_zbranch {
--
1.5.4.1
--
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
Also add debugging checks for LPT size and separate
out c->check_lpt_free from unrelated bitfields.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/debug.c | 37 ++++++++++
fs/ubifs/debug.h | 6 ++
fs/ubifs/lpt.c | 3 +-
fs/ubifs/lpt_commit.c | 185 ++++++++++++++++++++++++++++++++++++++++++++++---
fs/ubifs/ubifs.h | 10 ++-
5 files changed, 228 insertions(+), 13 deletions(-)
diff --git a/fs/ubifs/debug.c b/fs/ubifs/debug.c
index 32071ec..7186400 100644
--- a/fs/ubifs/debug.c
+++ b/fs/ubifs/debug.c
@@ -655,6 +655,43 @@ void dbg_dump_lprops(struct ubifs_info *c)
}
}
+void dbg_dump_lpt_info(struct ubifs_info *c)
+{
+ int i;
+
+ spin_lock(&dbg_lock);
+ printk(KERN_DEBUG "\tlpt_sz: %lld\n", c->lpt_sz);
+ printk(KERN_DEBUG "\tpnode_sz: %d\n", c->pnode_sz);
+ printk(KERN_DEBUG "\tnnode_sz: %d\n", c->nnode_sz);
+ printk(KERN_DEBUG "\tltab_sz: %d\n", c->ltab_sz);
+ printk(KERN_DEBUG "\tlsave_sz: %d\n", c->lsave_sz);
+ printk(KERN_DEBUG "\tbig_lpt: %d\n", c->big_lpt);
+ printk(KERN_DEBUG "\tlpt_hght: %d\n", c->lpt_hght);
+ printk(KERN_DEBUG "\tpnode_cnt: %d\n", c->pnode_cnt);
+ printk(KERN_DEBUG "\tnnode_cnt: %d\n", c->nnode_cnt);
+ printk(KERN_DEBUG "\tdirty_pn_cnt: %d\n", c->dirty_pn_cnt);
+ printk(KERN_DEBUG "\tdirty_nn_cnt: %d\n", c->dirty_nn_cnt);
+ printk(KERN_DEBUG "\tlsave_cnt: %d\n", c->lsave_cnt);
+ printk(KERN_DEBUG "\tspace_bits: %d\n", c->space_bits);
+ printk(KERN_DEBUG "\tlpt_lnum_bits: %d\n", c->lpt_lnum_bits);
+ printk(KERN_DEBUG "\tlpt_offs_bits: %d\n", c->lpt_offs_bits);
+ printk(KERN_DEBUG "\tlpt_spc_bits: %d\n", c->lpt_spc_bits);
+ printk(KERN_DEBUG "\tpcnt_bits: %d\n", c->pcnt_bits);
+ printk(KERN_DEBUG "\tlnum_bits: %d\n", c->lnum_bits);
+ printk(KERN_DEBUG "\tLPT root is at %d:%d\n", c->lpt_lnum, c->lpt_offs);
+ printk(KERN_DEBUG "\tLPT head is at %d:%d\n",
+ ...From: Adrian Hunter <ext-adrian.hunter@nokia.com>
'is_a_node()' function was reading from a buffer before
checking the buffer length, resulting in an OOPS as
follows:
BUG: unable to handle kernel paging request at f8f74002
IP: [<f8f9783f>] :ubifs:ubifs_unpack_bits+0xca/0x233
*pde = 19e95067 *pte = 00000000
Oops: 0000 [#1] PREEMPT SMP
Modules linked in: ubifs ubi mtdchar bio2mtd mtd brd video output
[last unloaded: mtd]
Pid: 6414, comm: integck Not tainted (2.6.27-rc6ubifs34 #23)
EIP: 0060:[<f8f9783f>] EFLAGS: 00010246 CPU: 0
EIP is at ubifs_unpack_bits+0xca/0x233 [ubifs]
EAX: 00000000 EBX: f6090630 ECX: d9badcfc EDX: 00000000
ESI: 00000004 EDI: f8f74002 EBP: d9badcec ESP: d9badcc0
DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
Process integck (pid: 6414, ti=d9bac000 task=f727dae0 task.ti=d9bac000)
Stack: 00000006 f7306240 00000002 00000000 d9badcfc d9badd00 0000001c 00000000
f6090630 f6090630 f8f74000 d9badd10 f8fa1cc9 00000000 f8f74002 00000000
f8f74002 f60fe128 f6090630 f8f74000 d9badd68 f8fa1e46 00000000 0001e000
Call Trace:
[<f8fa1cc9>] ? is_a_node+0x30/0x90 [ubifs]
[<f8fa1e46>] ? dbg_check_ltab+0x11d/0x5bd [ubifs]
[<f8fa388f>] ? ubifs_lpt_start_commit+0x42/0xed3 [ubifs]
[<c038e76a>] ? mutex_unlock+0x8/0xa
[<f8f9625d>] ? ubifs_tnc_start_commit+0x1c8/0xedb [ubifs]
[<f8f8d90b>] ? do_commit+0x187/0x523 [ubifs]
[<c038e76a>] ? mutex_unlock+0x8/0xa
[<f8f7ca17>] ? bud_wbuf_callback+0x22/0x28 [ubifs]
[<f8f8dd1d>] ? ubifs_run_commit+0x76/0xc0 [ubifs]
[<f8f8032c>] ? ubifs_sync_fs+0xd2/0xe6 [ubifs]
[<c01a2e97>] ? vfs_quota_sync+0x0/0x17e
[<c01a5ba6>] ? quota_sync_sb+0x26/0xbb
[<c01a2e97>] ? vfs_quota_sync+0x0/0x17e
[<c01a5c5d>] ? sync_dquots+0x22/0x12c
[<c0173d1b>] ? __fsync_super+0x19/0x68
[<c0173d75>] ? fsync_super+0xb/0x19
[<c0174065>] ? generic_shutdown_super+0x22/0xe7
[<c01a31fc>] ? vfs_quota_off+0x0/0x5fd
[<f8f7cf4d>] ? ubifs_kill_sb+0x31/0x35 [ubifs]
[<c01741f9>] ? deactivate_super+0x5e/0x71
[<c0187610>] ? ...From: Julien Brunel <brunel@diku.dk> In case of error, the function kthread_create returns an ERR pointer, but never returns a NULL pointer. So a NULL test that comes before an IS_ERR test should be deleted. The semantic match that finds this problem is as follows: (http://www.emn.fr/x-info/coccinelle/) // <smpl> @match_bad_null_test@ expression x, E; statement S1,S2; @@ x = kthread_create(...) ... when != x = E * if (x == NULL) S1 else S2 // </smpl> Signed-off-by: Julien Brunel <brunel@diku.dk> Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com> --- fs/ubifs/super.c | 4 ---- 1 files changed, 0 insertions(+), 4 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 667c72d..d87b0cf 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -1032,8 +1032,6 @@ static int mount_ubifs(struct ubifs_info *c) /* Create background thread */ c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); - if (!c->bgt) - c->bgt = ERR_PTR(-EINVAL); if (IS_ERR(c->bgt)) { err = PTR_ERR(c->bgt); c->bgt = NULL; @@ -1347,8 +1345,6 @@ static int ubifs_remount_rw(struct ubifs_info *c) /* Create background thread */ c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); - if (!c->bgt) - c->bgt = ERR_PTR(-EINVAL); if (IS_ERR(c->bgt)) { err = PTR_ERR(c->bgt); c->bgt = NULL; -- 1.5.4.1 --
From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
The "bulk_read" and "no_chk_data_crc" have only 2 values -
0 and 1. We already have bit-fields in corresponding data
structers, so make "bulk_read" and "no_chk_data_crc"
bit-fields as well.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/ubifs.h | 24 +++++++++++-------------
1 files changed, 11 insertions(+), 13 deletions(-)
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index d6ae3f7..542cbaf 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -331,6 +331,7 @@ struct ubifs_gced_idx_leb {
* this inode
* @dirty: non-zero if the inode is dirty
* @xattr: non-zero if this is an extended attribute inode
+ * @bulk_read: non-zero if bulk-read should be used
* @ui_mutex: serializes inode write-back with the rest of VFS operations,
* serializes "clean <-> dirty" state changes, serializes bulk-read,
* protects @dirty, @ui_size, and @xattr_size
@@ -343,7 +344,6 @@ struct ubifs_gced_idx_leb {
* @compr_type: default compression type used for this inode
* @last_page_read: page number of last page read (for bulk read)
* @read_in_a_row: number of consecutive pages read in a row (for bulk read)
- * @bulk_read: indicates whether bulk-read should be used
* @data_len: length of the data attached to the inode
* @data: inode's data
*
@@ -385,6 +385,7 @@ struct ubifs_inode {
unsigned int xattr_names;
unsigned int dirty:1;
unsigned int xattr:1;
+ unsigned int bulk_read:1;
struct mutex ui_mutex;
spinlock_t ui_lock;
loff_t synced_i_size;
@@ -393,7 +394,6 @@ struct ubifs_inode {
int compr_type;
pgoff_t last_page_read;
pgoff_t read_in_a_row;
- int bulk_read;
int data_len;
void *data;
};
@@ -940,6 +940,7 @@ struct ubifs_mount_opts {
* @cmt_state: commit state
* @cs_lock: commit state lock
* @cmt_wq: wait queue to sleep on if the log is full and a commit is running
+ *
* @fast_unmount: do not run journal ...From: Adrian Hunter <ext-adrian.hunter@nokia.com> Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com> --- fs/ubifs/tnc.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/fs/ubifs/tnc.c b/fs/ubifs/tnc.c index e0878a4..d27fd91 100644 --- a/fs/ubifs/tnc.c +++ b/fs/ubifs/tnc.c @@ -1600,7 +1600,7 @@ out: * An enormous hole could cause bulk-read to encompass too many * page cache pages, so limit the number here. */ - if (bu->blk_cnt >= UBIFS_MAX_BULK_READ) + if (bu->blk_cnt > UBIFS_MAX_BULK_READ) bu->blk_cnt = UBIFS_MAX_BULK_READ; /* * Ensure that bulk-read covers a whole number of page cache -- 1.5.4.1 --
From: Adrian Hunter <ext-adrian.hunter@nokia.com> UBIFS read performance can be improved by skipping the CRC check when data nodes are read. This option can be used if the underlying media is considered to be highly reliable. Note that CRCs are always checked for metadata. Read speed on Arm platform with OneNAND goes from 19 MiB/s to 27 MiB/s with data CRC checking disabled. Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com> --- Documentation/filesystems/ubifs.txt | 6 ++++++ fs/ubifs/io.c | 11 ++++++++--- fs/ubifs/scan.c | 2 +- fs/ubifs/super.c | 34 +++++++++++++++++++++++++++++++--- fs/ubifs/tnc.c | 6 +++++- fs/ubifs/ubifs.h | 11 ++++++++++- 6 files changed, 61 insertions(+), 9 deletions(-) diff --git a/Documentation/filesystems/ubifs.txt b/Documentation/filesystems/ubifs.txt index 340512c..dd84ea3 100644 --- a/Documentation/filesystems/ubifs.txt +++ b/Documentation/filesystems/ubifs.txt @@ -89,6 +89,12 @@ fast_unmount do not commit on unmount; this option makes bulk_read read more in one go to take advantage of flash media that read faster sequentially no_bulk_read (*) do not bulk-read +no_chk_data_crc skip checking of CRCs on data nodes in order to + improve read performance. Use this option only + if the flash media is highly reliable. The effect + of this option is that corruption of the contents + of a file can go unnoticed. +chk_data_crc (*) do not skip checking CRCs on data nodes Quick usage instructions diff --git a/fs/ubifs/io.c b/fs/ubifs/io.c index 054363f..40e2790 100644 --- a/fs/ubifs/io.c +++ b/fs/ubifs/io.c @@ -74,6 +74,7 @@ void ubifs_ro_mode(struct ubifs_info *c, int err) * @lnum: logical eraseblock number * @offs: offset within the logical eraseblock * @quiet: print no messages + * @chk_crc: indicates whether to always check the CRC * * This function checks node magic number and ...
From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
We cannot store bit-fields together if the processes which
change them may race, unless we serialize them.
Thus, move the nospc and nospc_rp bit-fields eway from
the mount option/constant bit-fields, to avoid races.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/ubifs.h | 17 +++++++++--------
1 files changed, 9 insertions(+), 8 deletions(-)
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index 542cbaf..c3ac5a8 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -334,7 +334,7 @@ struct ubifs_gced_idx_leb {
* @bulk_read: non-zero if bulk-read should be used
* @ui_mutex: serializes inode write-back with the rest of VFS operations,
* serializes "clean <-> dirty" state changes, serializes bulk-read,
- * protects @dirty, @ui_size, and @xattr_size
+ * protects @dirty, @bulk_read, @ui_size, and @xattr_size
* @ui_lock: protects @synced_i_size
* @synced_i_size: synchronized size of inode, i.e. the value of inode size
* currently stored on the flash; used only for regular file
@@ -944,10 +944,6 @@ struct ubifs_mount_opts {
* @fast_unmount: do not run journal commit before un-mounting
* @big_lpt: flag that LPT is too big to write whole during commit
* @check_lpt_free: flag that indicates LPT GC may be needed
- * @nospace: non-zero if the file-system does not have flash space (used as
- * optimization)
- * @nospace_rp: the same as @nospace, but additionally means that even reserved
- * pool is full
* @no_chk_data_crc: do not check CRCs when reading data nodes (except during
* recovery)
* @bulk_read: enable bulk-reads
@@ -1017,12 +1013,17 @@ struct ubifs_mount_opts {
* but which still have to be taken into account because
* the index has not been committed so far
* @space_lock: protects @budg_idx_growth, @budg_data_growth, ...From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
'ubifs_get_lprops()' and 'ubifs_release_lprops()' basically wrap
mutex lock and unlock. We have them because we want lprops subsystem
be separate and as independent as possible. And we planned better
locking rules for lprops.
Anyway, because they are short, it is better to inline them.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/lprops.c | 28 ----------------------------
fs/ubifs/misc.h | 27 +++++++++++++++++++++++++++
fs/ubifs/ubifs.h | 2 --
3 files changed, 27 insertions(+), 30 deletions(-)
diff --git a/fs/ubifs/lprops.c b/fs/ubifs/lprops.c
index 3659b88..f27176e 100644
--- a/fs/ubifs/lprops.c
+++ b/fs/ubifs/lprops.c
@@ -461,18 +461,6 @@ static void change_category(struct ubifs_info *c, struct ubifs_lprops *lprops)
}
/**
- * ubifs_get_lprops - get reference to LEB properties.
- * @c: the UBIFS file-system description object
- *
- * This function locks lprops. Lprops have to be unlocked by
- * 'ubifs_release_lprops()'.
- */
-void ubifs_get_lprops(struct ubifs_info *c)
-{
- mutex_lock(&c->lp_mutex);
-}
-
-/**
* calc_dark - calculate LEB dark space size.
* @c: the UBIFS file-system description object
* @spc: amount of free and dirty space in the LEB
@@ -643,22 +631,6 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
}
/**
- * ubifs_release_lprops - release lprops lock.
- * @c: the UBIFS file-system description object
- *
- * This function has to be called after each 'ubifs_get_lprops()' call to
- * unlock lprops.
- */
-void ubifs_release_lprops(struct ubifs_info *c)
-{
- ubifs_assert(mutex_is_locked(&c->lp_mutex));
- ubifs_assert(c->lst.empty_lebs >= 0 &&
- c->lst.empty_lebs <= c->main_lebs);
-
- mutex_unlock(&c->lp_mutex);
-}
-
-/**
* ubifs_get_lp_stats - get lprops statistics.
* @c: UBIFS file-system description object
* @st: return statistics
diff --git a/fs/ubifs/misc.h b/fs/ubifs/misc.h
index ...From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Commit the journal when the FS is sync'ed. This will make
statfs provide better free space report. And we anyway
advice our users to sync the FS if they want better statfs
report.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/super.c | 12 ++++++++++++
1 files changed, 12 insertions(+), 0 deletions(-)
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index dae1c62..7e1f3ef 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -418,6 +418,7 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
{
struct ubifs_info *c = sb->s_fs_info;
int i, ret = 0, err;
+ long long bud_bytes;
if (c->jheads)
for (i = 0; i < c->jhead_cnt; i++) {
@@ -425,6 +426,17 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
if (err && !ret)
ret = err;
}
+
+ /* Commit the journal unless it has too few data */
+ spin_lock(&c->buds_lock);
+ bud_bytes = c->bud_bytes;
+ spin_unlock(&c->buds_lock);
+ if (bud_bytes > c->leb_size) {
+ err = ubifs_run_commit(c);
+ if (err)
+ return err;
+ }
+
/*
* We ought to call sync for c->ubi but it does not have one. If it had
* it would in turn call mtd->sync, however mtd operations are
--
1.5.4.1
--
From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/super.c | 9 +++------
1 files changed, 3 insertions(+), 6 deletions(-)
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index cf078b5..dae1c62 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -1465,12 +1465,9 @@ out:
* commit_on_unmount - commit the journal when un-mounting.
* @c: UBIFS file-system description object
*
- * This function is called during un-mounting and it commits the journal unless
- * the "fast unmount" mode is enabled. It also avoids committing the journal if
- * it contains too few data.
- *
- * Sometimes recovery requires the journal to be committed at least once, and
- * this function takes care about this.
+ * This function is called during un-mounting and re-mounting, and it commits
+ * the journal unless the "fast unmount" mode is enabled. It also avoids
+ * committing the journal if it contains too few data.
*/
static void commit_on_unmount(struct ubifs_info *c)
{
--
1.5.4.1
--
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
Bulk-read skips uptodate pages but this was putting its
array index out and causing it to treat subsequent pages
as holes.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/file.c | 16 +++++++++++-----
1 files changed, 11 insertions(+), 5 deletions(-)
diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
index 2f20a49..51cf511 100644
--- a/fs/ubifs/file.c
+++ b/fs/ubifs/file.c
@@ -595,7 +595,7 @@ out:
static int populate_page(struct ubifs_info *c, struct page *page,
struct bu_info *bu, int *n)
{
- int i = 0, nn = *n, offs = bu->zbranch[0].offs, hole = 1, read = 0;
+ int i = 0, nn = *n, offs = bu->zbranch[0].offs, hole = 0, read = 0;
struct inode *inode = page->mapping->host;
loff_t i_size = i_size_read(inode);
unsigned int page_block;
@@ -609,6 +609,7 @@ static int populate_page(struct ubifs_info *c, struct page *page,
end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
if (!i_size || page->index > end_index) {
+ hole = 1;
memset(addr, 0, PAGE_CACHE_SIZE);
goto out_hole;
}
@@ -617,10 +618,10 @@ static int populate_page(struct ubifs_info *c, struct page *page,
while (1) {
int err, len, out_len, dlen;
- if (nn >= bu->cnt ||
- key_block(c, &bu->zbranch[nn].key) != page_block)
+ if (nn >= bu->cnt) {
+ hole = 1;
memset(addr, 0, UBIFS_BLOCK_SIZE);
- else {
+ } else if (key_block(c, &bu->zbranch[nn].key) == page_block) {
struct ubifs_data_node *dn;
dn = bu->buf + (bu->zbranch[nn].offs - offs);
@@ -643,8 +644,13 @@ static int populate_page(struct ubifs_info *c, struct page *page,
memset(addr + len, 0, UBIFS_BLOCK_SIZE - len);
nn += 1;
- hole = 0;
read = (i << UBIFS_BLOCK_SHIFT) + len;
+ } else if (key_block(c, &bu->zbranch[nn].key) < page_block) {
+ nn += 1;
+ continue;
+ } else {
+ hole = 1;
+ memset(addr, 0, UBIFS_BLOCK_SIZE);
}
if (++i >= UBIFS_BLOCKS_PER_PAGE)
break;
--
1.5.4.1
--
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
Make garbage collection try to keep data nodes from the same
inode together and in ascending order. This improves
performance when reading those nodes especially when bulk-read
is used.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/gc.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++-------
1 files changed, 72 insertions(+), 10 deletions(-)
diff --git a/fs/ubifs/gc.c b/fs/ubifs/gc.c
index a6b633a..0bef650 100644
--- a/fs/ubifs/gc.c
+++ b/fs/ubifs/gc.c
@@ -96,6 +96,48 @@ static int switch_gc_head(struct ubifs_info *c)
}
/**
+ * joinup - bring data nodes for an inode together.
+ * @c: UBIFS file-system description object
+ * @sleb: describes scanned LEB
+ * @inum: inode number
+ * @blk: block number
+ * @data: list to which to add data nodes
+ *
+ * This function looks at the first few nodes in the scanned LEB @sleb and adds
+ * them to @data if they are data nodes from @inum and have a larger block
+ * number than @blk. This function returns %0 on success and a negative error
+ * code on failure.
+ */
+static int joinup(struct ubifs_info *c, struct ubifs_scan_leb *sleb, ino_t inum,
+ unsigned int blk, struct list_head *data)
+{
+ int err, cnt = 6, lnum = sleb->lnum, offs;
+ struct ubifs_scan_node *snod, *tmp;
+ union ubifs_key *key;
+
+ list_for_each_entry_safe(snod, tmp, &sleb->nodes, list) {
+ key = &snod->key;
+ if (key_inum(c, key) == inum &&
+ key_type(c, key) == UBIFS_DATA_KEY &&
+ key_block(c, key) > blk) {
+ offs = snod->offs;
+ err = ubifs_tnc_has_node(c, key, 0, lnum, offs, 0);
+ if (err < 0)
+ return err;
+ list_del(&snod->list);
+ if (err) {
+ list_add_tail(&snod->list, data);
+ blk = key_block(c, key);
+ } else
+ kfree(snod);
+ cnt = 6;
+ } else if (--cnt == 0)
+ break;
+ }
+ return 0;
+}
+
+/**
* move_nodes - move nodes.
* @c: UBIFS file-system description object
* @sleb: describes nodes to ...From: Adrian Hunter <ext-adrian.hunter@nokia.com>
The comparison was working, but more by accident than design.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/tnc_misc.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/ubifs/tnc_misc.c b/fs/ubifs/tnc_misc.c
index a25c1cc..b48db99 100644
--- a/fs/ubifs/tnc_misc.c
+++ b/fs/ubifs/tnc_misc.c
@@ -480,8 +480,8 @@ int ubifs_tnc_read_node(struct ubifs_info *c, struct ubifs_zbranch *zbr,
}
/* Make sure the key of the read node is correct */
- key_read(c, key, &key1);
- if (memcmp(node + UBIFS_KEY_OFFSET, &key1, c->key_len)) {
+ key_read(c, node + UBIFS_KEY_OFFSET, &key1);
+ if (!keys_eq(c, key, &key1)) {
ubifs_err("bad key in node at LEB %d:%d",
zbr->lnum, zbr->offs);
dbg_tnc("looked for key %s found node's key %s",
--
1.5.4.1
--
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/file.c | 10 ++++++++--
fs/ubifs/ubifs-media.h | 1 -
2 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
index cdcfe95..2f20a49 100644
--- a/fs/ubifs/file.c
+++ b/fs/ubifs/file.c
@@ -147,6 +147,12 @@ static int do_readpage(struct page *page)
err = ret;
if (err != -ENOENT)
break;
+ } else if (block + 1 == beyond) {
+ int dlen = le32_to_cpu(dn->size);
+ int ilen = i_size & (UBIFS_BLOCK_SIZE - 1);
+
+ if (ilen && ilen < dlen)
+ memset(addr + ilen, 0, dlen - ilen);
}
}
if (++i >= UBIFS_BLOCKS_PER_PAGE)
@@ -601,7 +607,7 @@ static int populate_page(struct ubifs_info *c, struct page *page,
addr = zaddr = kmap(page);
- end_index = (i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+ end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;
if (!i_size || page->index > end_index) {
memset(addr, 0, PAGE_CACHE_SIZE);
goto out_hole;
@@ -649,7 +655,7 @@ static int populate_page(struct ubifs_info *c, struct page *page,
if (end_index == page->index) {
int len = i_size & (PAGE_CACHE_SIZE - 1);
- if (len < read)
+ if (len && len < read)
memset(zaddr + len, 0, read - len);
}
diff --git a/fs/ubifs/ubifs-media.h b/fs/ubifs/ubifs-media.h
index a9ecbd9..0b37804 100644
--- a/fs/ubifs/ubifs-media.h
+++ b/fs/ubifs/ubifs-media.h
@@ -75,7 +75,6 @@
*/
#define UBIFS_BLOCK_SIZE 4096
#define UBIFS_BLOCK_SHIFT 12
-#define UBIFS_BLOCK_MASK 0x00000FFF
/* UBIFS padding byte pattern (must not be first or last byte of node magic) */
#define UBIFS_PADDING_BYTE 0xCE
--
1.5.4.1
--
From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com> This commit adds a reserved pool size print and tweaks the prints to make them look nicer. It also fixes and cleans-up some comments. Additionally, it deletes some blank lines to make the code look a little nicer. In other words, nothing essential. Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com> --- fs/ubifs/budget.c | 26 ++++++++++++++------------ fs/ubifs/lprops.c | 6 +----- fs/ubifs/super.c | 16 +++++++++------- 3 files changed, 24 insertions(+), 24 deletions(-) diff --git a/fs/ubifs/budget.c b/fs/ubifs/budget.c index 73db464..1a4973e 100644 --- a/fs/ubifs/budget.c +++ b/fs/ubifs/budget.c @@ -414,19 +414,21 @@ static int do_budget_space(struct ubifs_info *c) * @c->lst.empty_lebs + @c->freeable_cnt + @c->idx_gc_cnt - * @c->lst.taken_empty_lebs * - * @empty_lebs are available because they are empty. @freeable_cnt are - * available because they contain only free and dirty space and the - * index allocation always occurs after wbufs are synch'ed. - * @idx_gc_cnt are available because they are index LEBs that have been - * garbage collected (including trivial GC) and are awaiting the commit - * before they can be unmapped - note that the in-the-gaps method will - * grab these if it needs them. @taken_empty_lebs are empty_lebs that - * have already been allocated for some purpose (also includes those - * LEBs on the @idx_gc list). + * @c->lst.empty_lebs are available because they are empty. + * @c->freeable_cnt are available because they contain only free and + * dirty space, @c->idx_gc_cnt are available because they are index + * LEBs that have been garbage collected and are awaiting the commit + * before they can be used. And the in-the-gaps method will grab these + * if it needs them. @c->lst.taken_empty_lebs are empty LEBs that have + * already been allocated for some purpose. * - * Note, @taken_empty_lebs may temporarily be higher by one because ...
From: Hirofumi Nakagawa <hnakagawa@miraclelinux.com>
IS_ERR() macro already has unlikely(), so do not use constructions
like 'if (unlikely(IS_ERR())'.
Signed-off-by: Hirofumi Nakagawa <hnakagawa@miraclelinux.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/find.c | 4 ++--
fs/ubifs/gc.c | 8 ++++----
fs/ubifs/tnc.c | 4 ++--
fs/ubifs/xattr.c | 2 +-
4 files changed, 9 insertions(+), 9 deletions(-)
diff --git a/fs/ubifs/find.c b/fs/ubifs/find.c
index 47814cd..717d79c 100644
--- a/fs/ubifs/find.c
+++ b/fs/ubifs/find.c
@@ -901,11 +901,11 @@ static int get_idx_gc_leb(struct ubifs_info *c)
* it is needed now for this commit.
*/
lp = ubifs_lpt_lookup_dirty(c, lnum);
- if (unlikely(IS_ERR(lp)))
+ if (IS_ERR(lp))
return PTR_ERR(lp);
lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
lp->flags | LPROPS_INDEX, -1);
- if (unlikely(IS_ERR(lp)))
+ if (IS_ERR(lp))
return PTR_ERR(lp);
dbg_find("LEB %d, dirty %d and free %d flags %#x",
lp->lnum, lp->dirty, lp->free, lp->flags);
diff --git a/fs/ubifs/gc.c b/fs/ubifs/gc.c
index 02aba36..a6b633a 100644
--- a/fs/ubifs/gc.c
+++ b/fs/ubifs/gc.c
@@ -653,7 +653,7 @@ int ubifs_gc_start_commit(struct ubifs_info *c)
*/
while (1) {
lp = ubifs_fast_find_freeable(c);
- if (unlikely(IS_ERR(lp))) {
+ if (IS_ERR(lp)) {
err = PTR_ERR(lp);
goto out;
}
@@ -665,7 +665,7 @@ int ubifs_gc_start_commit(struct ubifs_info *c)
if (err)
goto out;
lp = ubifs_change_lp(c, lp, c->leb_size, 0, lp->flags, 0);
- if (unlikely(IS_ERR(lp))) {
+ if (IS_ERR(lp)) {
err = PTR_ERR(lp);
goto out;
}
@@ -680,7 +680,7 @@ int ubifs_gc_start_commit(struct ubifs_info *c)
/* Record index freeable LEBs for unmapping after commit */
while (1) {
lp = ubifs_fast_find_frdi_idx(c);
- if (unlikely(IS_ERR(lp))) {
+ if (IS_ERR(lp)) {
err = PTR_ERR(lp);
goto out;
}
@@ -696,7 +696,7 @@ int ubifs_gc_start_commit(struct ubifs_info ...From: Adrian Hunter <ext-adrian.hunter@nokia.com>
Some flash media are capable of reading sequentially at faster rates.
UBIFS bulk-read facility is designed to take advantage of that, by
reading in one go consecutive data nodes that are also located
consecutively in the same LEB.
Read speed on Arm platform with OneNAND goes from 17 MiB/s to
19 MiB/s.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
Documentation/filesystems/ubifs.txt | 3 +
fs/ubifs/file.c | 248 ++++++++++++++++++++++++++++++
fs/ubifs/key.h | 22 +++-
fs/ubifs/super.c | 31 ++++
fs/ubifs/tnc.c | 283 +++++++++++++++++++++++++++++++++++
fs/ubifs/ubifs.h | 45 ++++++-
6 files changed, 629 insertions(+), 3 deletions(-)
diff --git a/Documentation/filesystems/ubifs.txt b/Documentation/filesystems/ubifs.txt
index 6a0d70a..340512c 100644
--- a/Documentation/filesystems/ubifs.txt
+++ b/Documentation/filesystems/ubifs.txt
@@ -86,6 +86,9 @@ norm_unmount (*) commit on unmount; the journal is committed
fast_unmount do not commit on unmount; this option makes
unmount faster, but the next mount slower
because of the need to replay the journal.
+bulk_read read more in one go to take advantage of flash
+ media that read faster sequentially
+no_bulk_read (*) do not bulk-read
Quick usage instructions
diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
index 3d698e2..cdcfe95 100644
--- a/fs/ubifs/file.c
+++ b/fs/ubifs/file.c
@@ -577,8 +577,256 @@ out:
return copied;
}
+/**
+ * populate_page - copy data nodes into a page for bulk-read.
+ * @c: UBIFS file-system description object
+ * @page: page
+ * @bu: bulk-read information
+ * @n: next zbranch slot
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int populate_page(struct ubifs_info *c, struct page *page,
+ struct bu_info *bu, int *n)
+{
+ int i = 0, nn = *n, ...From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
'dbg_dump_inode()' is quite outdated and does not print all
the fileds.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/debug.c | 42 +++++++++++++++++++++++++-----------------
1 files changed, 25 insertions(+), 17 deletions(-)
diff --git a/fs/ubifs/debug.c b/fs/ubifs/debug.c
index d7f7645..32071ec 100644
--- a/fs/ubifs/debug.c
+++ b/fs/ubifs/debug.c
@@ -222,30 +222,38 @@ void dbg_dump_inode(const struct ubifs_info *c, const struct inode *inode)
{
const struct ubifs_inode *ui = ubifs_inode(inode);
- printk(KERN_DEBUG "inode %lu\n", inode->i_ino);
- printk(KERN_DEBUG "size %llu\n",
+ printk(KERN_DEBUG "Dump in-memory inode:");
+ printk(KERN_DEBUG "\tinode %lu\n", inode->i_ino);
+ printk(KERN_DEBUG "\tsize %llu\n",
(unsigned long long)i_size_read(inode));
- printk(KERN_DEBUG "nlink %u\n", inode->i_nlink);
- printk(KERN_DEBUG "uid %u\n", (unsigned int)inode->i_uid);
- printk(KERN_DEBUG "gid %u\n", (unsigned int)inode->i_gid);
- printk(KERN_DEBUG "atime %u.%u\n",
+ printk(KERN_DEBUG "\tnlink %u\n", inode->i_nlink);
+ printk(KERN_DEBUG "\tuid %u\n", (unsigned int)inode->i_uid);
+ printk(KERN_DEBUG "\tgid %u\n", (unsigned int)inode->i_gid);
+ printk(KERN_DEBUG "\tatime %u.%u\n",
(unsigned int)inode->i_atime.tv_sec,
(unsigned int)inode->i_atime.tv_nsec);
- printk(KERN_DEBUG "mtime %u.%u\n",
+ printk(KERN_DEBUG "\tmtime %u.%u\n",
(unsigned int)inode->i_mtime.tv_sec,
(unsigned int)inode->i_mtime.tv_nsec);
- printk(KERN_DEBUG "ctime %u.%u\n",
+ printk(KERN_DEBUG "\tctime %u.%u\n",
(unsigned int)inode->i_ctime.tv_sec,
(unsigned int)inode->i_ctime.tv_nsec);
- printk(KERN_DEBUG "creat_sqnum %llu\n", ui->creat_sqnum);
- printk(KERN_DEBUG "xattr_size %u\n", ui->xattr_size);
- printk(KERN_DEBUG "xattr_cnt ...From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
When UBIFS switches to R/O mode because of an error,
it is reasonable to enable data CRC checking.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
---
fs/ubifs/io.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/fs/ubifs/io.c b/fs/ubifs/io.c
index 40e2790..0168271 100644
--- a/fs/ubifs/io.c
+++ b/fs/ubifs/io.c
@@ -62,6 +62,7 @@ void ubifs_ro_mode(struct ubifs_info *c, int err)
{
if (!c->ro_media) {
c->ro_media = 1;
+ c->no_chk_data_crc = 0;
ubifs_warn("switched to read-only mode, error %d", err);
dbg_dump_stack();
}
--
1.5.4.1
--
From: Adrian Hunter <ext-adrian.hunter@nokia.com>
sync_fs can be called even if the file system is mounted
read-only. Ensure the commit is not run in that case.
Reported-by: Zoltan Sogor <weth@inf.u-szeged.hu>
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
---
fs/ubifs/super.c | 19 ++++++++++---------
1 files changed, 10 insertions(+), 9 deletions(-)
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 7e1f3ef..7fd759d 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -420,21 +420,22 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
int i, ret = 0, err;
long long bud_bytes;
- if (c->jheads)
+ if (c->jheads) {
for (i = 0; i < c->jhead_cnt; i++) {
err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
if (err && !ret)
ret = err;
}
- /* Commit the journal unless it has too few data */
- spin_lock(&c->buds_lock);
- bud_bytes = c->bud_bytes;
- spin_unlock(&c->buds_lock);
- if (bud_bytes > c->leb_size) {
- err = ubifs_run_commit(c);
- if (err)
- return err;
+ /* Commit the journal unless it has too little data */
+ spin_lock(&c->buds_lock);
+ bud_bytes = c->bud_bytes;
+ spin_unlock(&c->buds_lock);
+ if (bud_bytes > c->leb_size) {
+ err = ubifs_run_commit(c);
+ if (err)
+ return err;
+ }
}
/*
--
1.5.4.1
--
Actually this example is a little wrong, because UBIFS starts doing bulk-read only if few pages were read sequentially before. So, user would have to read pages 0-3 sequentially, then UBIFS would start doing bulk-read for this inode. -- Best regards, Artem Bityutskiy (Битюцкий Артём) --
| Greg KH | Og dreams of kernels |
| Jens Axboe | [PATCH 31/33] Fusion: sg chaining support |
| Arnd Bergmann | Re: finding your own dead "CONFIG_" variables |
| Mark Brown | [PATCH 2/2] Subject: natsemi: Allow users to disable workaround for DspCfg reset |
| Tony Breeds | [LGUEST] Look in object dir for .config |
git: | |
| Brian Downing | Re: Git in a Nutshell guide |
| John Benes | Re: master has some toys |