2007-02-27 20:38:30

by Kalpak Shah

[permalink] [raw]
Subject: [PATCH take3 1/1] nanosecond timestamps

Hi,

I have incorporated all the changes that have been suggested. Dave, thanks for simplifying the macros. I have also moved the EXT4_INODE_GET_XTIME macros ahead, so that they are called after ei->i_extra_isize is initialized, as pointed out by Johann Lombardi.

Signed-off-by: Andreas Dilger <[email protected]>
Signed-off-by: Kalpak Shah <[email protected]>

Index: linux-2.6.20/fs/ext4/ialloc.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/ialloc.c
+++ linux-2.6.20/fs/ext4/ialloc.c
@@ -563,7 +563,8 @@ got:
inode->i_ino = ino;
/* This is the optimal IO size (for stat), not the fs block size */
inode->i_blocks = 0;
- inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_mtime = inode->i_atime = inode->i_ctime = ei->i_crtime =
+ ext4_current_time(inode);

memset(ei->i_data, 0, sizeof(ei->i_data));
ei->i_dir_start_lookup = 0;
@@ -595,9 +596,8 @@ got:
spin_unlock(&sbi->s_next_gen_lock);

ei->i_state = EXT4_STATE_NEW;
- ei->i_extra_isize =
- (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) ?
- sizeof(struct ext4_inode) - EXT4_GOOD_OLD_INODE_SIZE : 0;
+
+ ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize;

ret = inode;
if(DQUOT_ALLOC_INODE(inode)) {
Index: linux-2.6.20/fs/ext4/inode.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/inode.c
+++ linux-2.6.20/fs/ext4/inode.c
@@ -727,7 +727,7 @@ static int ext4_splice_branch(handle_t *

/* We are done with atomic stuff, now do the rest of housekeeping */

- inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_ctime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);

/* had we spliced it onto indirect block? */
@@ -2441,7 +2441,7 @@ do_indirects:
ext4_discard_reservation(inode);

mutex_unlock(&ei->truncate_mutex);
- inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);

/*
@@ -2676,10 +2676,6 @@ void ext4_read_inode(struct inode * inod
}
inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
inode->i_size = le32_to_cpu(raw_inode->i_size);
- inode->i_atime.tv_sec = le32_to_cpu(raw_inode->i_atime);
- inode->i_ctime.tv_sec = le32_to_cpu(raw_inode->i_ctime);
- inode->i_mtime.tv_sec = le32_to_cpu(raw_inode->i_mtime);
- inode->i_atime.tv_nsec = inode->i_ctime.tv_nsec = inode->i_mtime.tv_nsec = 0;

ei->i_state = 0;
ei->i_dir_start_lookup = 0;
@@ -2755,6 +2751,11 @@ void ext4_read_inode(struct inode * inod
} else
ei->i_extra_isize = 0;

+ EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
+ EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
+ EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
+ EXT4_INODE_GET_XTIME(i_crtime, ei, raw_inode);
+
if (S_ISREG(inode->i_mode)) {
inode->i_op = &ext4_file_inode_operations;
inode->i_fop = &ext4_file_operations;
@@ -2835,9 +2836,12 @@ static int ext4_do_update_inode(handle_t
}
raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
raw_inode->i_size = cpu_to_le32(ei->i_disksize);
- raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
- raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
- raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
+
+ EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
+ EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
+ EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
+ EXT4_INODE_SET_XTIME(i_crtime, ei, raw_inode);
+
raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
raw_inode->i_flags = cpu_to_le32(ei->i_flags);
Index: linux-2.6.20/fs/ext4/ioctl.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/ioctl.c
+++ linux-2.6.20/fs/ext4/ioctl.c
@@ -96,7 +96,7 @@ int ext4_ioctl (struct inode * inode, st
ei->i_flags = flags;

ext4_set_inode_flags(inode);
- inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_ctime = ext4_current_time(inode);

err = ext4_mark_iloc_dirty(handle, inode, &iloc);
flags_err:
@@ -133,7 +133,7 @@ flags_err:
return PTR_ERR(handle);
err = ext4_reserve_inode_write(handle, inode, &iloc);
if (err == 0) {
- inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_ctime = ext4_current_time(inode);
inode->i_generation = generation;
err = ext4_mark_iloc_dirty(handle, inode, &iloc);
}
Index: linux-2.6.20/fs/ext4/namei.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/namei.c
+++ linux-2.6.20/fs/ext4/namei.c
@@ -1282,7 +1282,7 @@ static int add_dirent_to_buf(handle_t *h
* happen is that the times are slightly out of date
* and/or different from the directory change time.
*/
- dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC;
+ dir->i_mtime = dir->i_ctime = ext4_current_time(dir);
ext4_update_dx_flag(dir);
dir->i_version++;
ext4_mark_inode_dirty(handle, dir);
@@ -2058,7 +2058,7 @@ static int ext4_rmdir (struct inode * di
* recovery. */
inode->i_size = 0;
ext4_orphan_add(handle, inode);
- inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
+ inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);
drop_nlink(dir);
ext4_update_dx_flag(dir);
@@ -2108,13 +2108,13 @@ static int ext4_unlink(struct inode * di
retval = ext4_delete_entry(handle, dir, de, bh);
if (retval)
goto end_unlink;
- dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC;
+ dir->i_ctime = dir->i_mtime = ext4_current_time(dir);
ext4_update_dx_flag(dir);
ext4_mark_inode_dirty(handle, dir);
drop_nlink(inode);
if (!inode->i_nlink)
ext4_orphan_add(handle, inode);
- inode->i_ctime = dir->i_ctime;
+ inode->i_ctime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);
retval = 0;

@@ -2199,7 +2199,7 @@ retry:
if (IS_DIRSYNC(dir))
handle->h_sync = 1;

- inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_ctime = ext4_current_time(inode);
ext4_inc_count(handle, inode);
atomic_inc(&inode->i_count);

@@ -2301,7 +2301,7 @@ static int ext4_rename (struct inode * o
* Like most other Unix systems, set the ctime for inodes on a
* rename.
*/
- old_inode->i_ctime = CURRENT_TIME_SEC;
+ old_inode->i_ctime = ext4_current_time(old_inode);
ext4_mark_inode_dirty(handle, old_inode);

/*
@@ -2334,9 +2334,9 @@ static int ext4_rename (struct inode * o

if (new_inode) {
drop_nlink(new_inode);
- new_inode->i_ctime = CURRENT_TIME_SEC;
+ new_inode->i_ctime = ext4_current_time(new_inode);
}
- old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME_SEC;
+ old_dir->i_ctime = old_dir->i_mtime = ext4_current_time(old_dir);
ext4_update_dx_flag(old_dir);
if (dir_bh) {
BUFFER_TRACE(dir_bh, "get_write_access");
Index: linux-2.6.20/fs/ext4/super.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/super.c
+++ linux-2.6.20/fs/ext4/super.c
@@ -1631,6 +1631,8 @@ static int ext4_fill_super (struct super
sbi->s_inode_size);
goto failed_mount;
}
+ if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
+ sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
}
sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
le32_to_cpu(es->s_log_frag_size);
@@ -1847,6 +1849,32 @@ static int ext4_fill_super (struct super
}

ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
+
+ /* determine the minimum size of new large inodes, if present */
+ if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
+ sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
+ EXT4_GOOD_OLD_INODE_SIZE;
+ if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
+ EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
+ if (sbi->s_want_extra_isize <
+ le16_to_cpu(es->s_want_extra_isize))
+ sbi->s_want_extra_isize =
+ le16_to_cpu(es->s_want_extra_isize);
+ if (sbi->s_want_extra_isize <
+ le16_to_cpu(es->s_min_extra_isize))
+ sbi->s_want_extra_isize =
+ le16_to_cpu(es->s_min_extra_isize);
+ }
+ }
+ /* Check if enough inode space is available */
+ if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
+ sbi->s_inode_size) {
+ sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
+ EXT4_GOOD_OLD_INODE_SIZE;
+ printk(KERN_INFO "EXT4-fs: required extra inode space not"
+ "available.\n");
+ }
+
/*
* akpm: core read_super() calls in here with the superblock locked.
* That deadlocks, because orphan cleanup needs to lock the superblock
Index: linux-2.6.20/fs/ext4/xattr.c
===================================================================
--- linux-2.6.20.orig/fs/ext4/xattr.c
+++ linux-2.6.20/fs/ext4/xattr.c
@@ -1004,7 +1004,7 @@ ext4_xattr_set_handle(handle_t *handle,
}
if (!error) {
ext4_xattr_update_super_block(handle, inode->i_sb);
- inode->i_ctime = CURRENT_TIME_SEC;
+ inode->i_ctime = ext4_current_time(inode);
error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
/*
* The bh is consumed by ext4_mark_iloc_dirty, even with
Index: linux-2.6.20/include/linux/ext4_fs.h
===================================================================
--- linux-2.6.20.orig/include/linux/ext4_fs.h
+++ linux-2.6.20/include/linux/ext4_fs.h
@@ -282,7 +282,7 @@ struct ext4_inode {
__le16 i_uid; /* Low 16 bits of Owner Uid */
__le32 i_size; /* Size in bytes */
__le32 i_atime; /* Access time */
- __le32 i_ctime; /* Creation time */
+ __le32 i_ctime; /* Inode Change time */
__le32 i_mtime; /* Modification time */
__le32 i_dtime; /* Deletion Time */
__le16 i_gid; /* Low 16 bits of Group Id */
@@ -331,10 +331,57 @@ struct ext4_inode {
} osd2; /* OS dependent 2 */
__le16 i_extra_isize;
__le16 i_pad1;
+ __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
+ __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
+ __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
+ __le32 i_crtime; /* File Creation time */
+ __le32 i_crtime_extra; /* extra File Creation time (nsec << 2 | epoch) */
};

#define i_size_high i_dir_acl

+#define EXT4_EPOCH_BITS 2
+#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
+#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
+
+#define EXT4_FITS_IN_INODE(ext4_inode, field) \
+ ((offsetof(typeof(*ext4_inode), field) + \
+ sizeof((ext4_inode)->field)) \
+ <= (EXT4_GOOD_OLD_INODE_SIZE + \
+ le16_to_cpu((ext4_inode)->i_extra_isize))) \
+
+static inline __le32 ext4_encode_extra_time(struct timespec *time)
+{
+ return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
+ time->tv_sec >> 32 : 0) |
+ ((time->tv_nsec << 2) & EXT4_NSEC_MASK));
+}
+
+static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra) {
+ if (sizeof(time->tv_sec) > 4)
+ time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
+ << 32;
+ time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;
+}
+
+#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
+do { \
+ if (EXT4_FITS_IN_INODE(raw_inode, xtime)) \
+ (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
+ if (EXT4_FITS_IN_INODE(raw_inode, xtime ## _extra)) \
+ (raw_inode)->xtime ## _extra = \
+ ext4_encode_extra_time(&(inode)->xtime); \
+} while (0)
+
+#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
+do { \
+ if (EXT4_FITS_IN_INODE(raw_inode, xtime)) \
+ (inode)->xtime.tv_sec = le32_to_cpu((raw_inode)->xtime); \
+ if (EXT4_FITS_IN_INODE(raw_inode, xtime ## _extra)) \
+ ext4_decode_extra_time(&(inode)->xtime, \
+ raw_inode->xtime ## _extra); \
+} while (0)
+
#if defined(__KERNEL__) || defined(__linux__)
#define i_reserved1 osd1.linux1.l_i_reserved1
#define i_frag osd2.linux2.l_i_frag
@@ -513,7 +560,9 @@ struct ext4_super_block {
/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
__le32 s_r_blocks_count_hi; /* Reserved blocks count */
__le32 s_free_blocks_count_hi; /* Free blocks count */
- __u32 s_reserved[169]; /* Padding to the end of the block */
+ __le16 s_min_extra_isize; /* All inodes have at least # bytes */
+ __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
+ __u32 s_reserved[168]; /* Padding to the end of the block */
};

#ifdef __KERNEL__
@@ -526,6 +575,13 @@ static inline struct ext4_inode_info *EX
return container_of(inode, struct ext4_inode_info, vfs_inode);
}

+static inline struct timespec ext4_current_time(struct inode *inode)
+{
+ return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
+ current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
+}
+
+
static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
{
return ino == EXT4_ROOT_INO ||
@@ -596,6 +652,7 @@ static inline int ext4_valid_inum(struct
#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
+#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040

#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
@@ -613,6 +670,7 @@ static inline int ext4_valid_inum(struct
EXT4_FEATURE_INCOMPAT_64BIT)
#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
+ EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE| \
EXT4_FEATURE_RO_COMPAT_BTREE_DIR)

/*
Index: linux-2.6.20/include/linux/ext4_fs_i.h
===================================================================
--- linux-2.6.20.orig/include/linux/ext4_fs_i.h
+++ linux-2.6.20/include/linux/ext4_fs_i.h
@@ -153,6 +153,7 @@ struct ext4_inode_info {

unsigned long i_ext_generation;
struct ext4_ext_cache i_cached_extent;
+ struct timespec i_crtime;
};

#endif /* _LINUX_EXT4_FS_I */
Index: linux-2.6.20/include/linux/ext4_fs_sb.h
===================================================================
--- linux-2.6.20.orig/include/linux/ext4_fs_sb.h
+++ linux-2.6.20/include/linux/ext4_fs_sb.h
@@ -89,6 +89,7 @@ struct ext4_sb_info {
unsigned long s_ext_blocks;
unsigned long s_ext_extents;
#endif
+ unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
};

#endif /* _LINUX_EXT4_FS_SB */


Thanks,
Kalpak Shah.


2007-02-27 22:45:27

by Andreas Dilger

[permalink] [raw]
Subject: Re: [PATCH take3 1/1] nanosecond timestamps

On Feb 27, 2007 15:14 +0530, Kalpak Shah wrote:
> +#define EXT4_EPOCH_BITS 2
> +
> +static inline __le32 ext4_encode_extra_time(struct timespec *time)
> +{
> + return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
> + time->tv_sec >> 32 : 0) |
> + ((time->tv_nsec << 2) & EXT4_NSEC_MASK));

This should be "(time->tv_nsec << EXT4_EPOCH_BITS)". We don't strictly
need the EXT4_NSEC_MASK because cpu_to_le32() will truncate the field to
32 bits anyways.

> +static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra) {
> + if (sizeof(time->tv_sec) > 4)
> + time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
> + << 32;
> + time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;

This should be ">> EXT4_EPOCH_BITS". Similarly, le32_to_cpu() will truncate
extra to 32 bits and we shift away the "epoch" part of the field, so we don't
need the EXT4_NSEC_MASK at all.

> +#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
> +do { \
> + if (EXT4_FITS_IN_INODE(raw_inode, xtime)) \
> + (inode)->xtime.tv_sec = le32_to_cpu((raw_inode)->xtime); \

We might consider doing this in the caller for the crtime field. That is
only read/written in a single place, and it is the only one that needs the
extra check to determine if the seconds field is in the body of the inode.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.