[RFC PATCH] fat: truncate timestamps returned from fat_getattr

[RFC PATCH] fat: truncate timestamps returned from fat_getattr

[Frank Sorenson]

Author: Frank Sorenson <sorenson@redhat.com>
Date:   2018-03-23 23:28:32 -0500

    fat: add function to truncate timestamps returned from fat_getattr
    
    vfat/msdos timestamps are stored on-disk with several different
    granularities, some of them lower resolution than timespec_trunc()
    can provide.  However, they are only truncated as they are written
    to disk, so the timestamps in-memory for new or modified
    files/directories may be different from the same timestamps after
    a remount, as the now-truncated times are re-read from the on-disk
    format.
    
    This patch adds a function to adjust the timestamps returned from
    fat_getattr() so that userspace sees consistent times across a
    remount.  The existing functions are used to convert the timespec
    to the on-disk representation, then back to timespec again.
    
    Signed-off-by: Frank Sorenson <sorenson@redhat.com>

diff --git a/fs/fat/fat.h b/fs/fat/fat.h
index 8fc1093da47d..0367be021b23 100644
--- a/fs/fat/fat.h
+++ b/fs/fat/fat.h
@@ -410,6 +410,7 @@ extern void fat_time_fat2unix(struct msdos_sb_info *sbi, struct timespec *ts,
 			      __le16 __time, __le16 __date, u8 time_cs);
 extern void fat_time_unix2fat(struct msdos_sb_info *sbi, struct timespec *ts,
 			      __le16 *time, __le16 *date, u8 *time_cs);
+extern void fat_trunc_kstat_times(struct msdos_sb_info *sbi, struct kstat *stat);
 extern int fat_sync_bhs(struct buffer_head **bhs, int nr_bhs);
 
 int fat_cache_init(void);
diff --git a/fs/fat/file.c b/fs/fat/file.c
index 4724cc9ad650..11639e393ef7 100644
--- a/fs/fat/file.c
+++ b/fs/fat/file.c
@@ -370,6 +370,7 @@ int fat_getattr(const struct path *path, struct kstat *stat,
 {
 	struct inode *inode = d_inode(path->dentry);
 	generic_fillattr(inode, stat);
+	fat_trunc_kstat_times(MSDOS_SB(inode->i_sb), stat);
 	stat->blksize = MSDOS_SB(inode->i_sb)->cluster_size;
 
 	if (MSDOS_SB(inode->i_sb)->options.nfs == FAT_NFS_NOSTALE_RO) {
diff --git a/fs/fat/misc.c b/fs/fat/misc.c
index f9bdc1e01c98..52f40edefd12 100644
--- a/fs/fat/misc.c
+++ b/fs/fat/misc.c
@@ -262,6 +262,29 @@ void fat_time_unix2fat(struct msdos_sb_info *sbi, struct timespec *ts,
 }
 EXPORT_SYMBOL_GPL(fat_time_unix2fat);
 
+void fat_trunc_kstat_times(struct msdos_sb_info *sbi, struct kstat *stat)
+{
+	if (stat->ino == MSDOS_ROOT_INO)
+		stat->mtime = stat->ctime = stat->atime =
+			(struct timespec){0, 0};
+	else {
+		__le16 time;
+		__le16 date;
+		u8 ctime_cs;
+
+		fat_time_unix2fat(sbi, &stat->mtime, &time, &date, NULL);
+		fat_time_fat2unix(sbi, &stat->mtime, time, date, 0);
+
+		fat_time_unix2fat(sbi, &stat->ctime, &time, &date,
+			sbi->options.isvfat ? &ctime_cs :  NULL);
+		fat_time_fat2unix(sbi, &stat->ctime, time, date,
+			sbi->options.isvfat ? ctime_cs : 0);
+
+		fat_time_unix2fat(sbi, &stat->atime, &time, &date, NULL);
+		fat_time_fat2unix(sbi, &stat->atime, 0, date, 0);
+	}
+}
+
 int fat_sync_bhs(struct buffer_head **bhs, int nr_bhs)
 {
 	int i, err = 0;

Re: [RFC PATCH] fat: truncate timestamps returned from fat_getattr

[OGAWA Hirofumi]

Frank Sorenson <sorenson@redhat.com> writes:

Hi,

> Author: Frank Sorenson <sorenson@redhat.com>
> Date:   2018-03-23 23:28:32 -0500
>
>     fat: add function to truncate timestamps returned from fat_getattr
>     
>     vfat/msdos timestamps are stored on-disk with several different
>     granularities, some of them lower resolution than timespec_trunc()
>     can provide.  However, they are only truncated as they are written
>     to disk, so the timestamps in-memory for new or modified
>     files/directories may be different from the same timestamps after
>     a remount, as the now-truncated times are re-read from the on-disk
>     format.
>     
>     This patch adds a function to adjust the timestamps returned from
>     fat_getattr() so that userspace sees consistent times across a
>     remount.  The existing functions are used to convert the timespec
>     to the on-disk representation, then back to timespec again.
>     
>     Signed-off-by: Frank Sorenson <sorenson@redhat.com>

Already tried to use ->update_time() callback? With quick look though,
it seems to be possible to control timestamp granularity with
->update_time(), and add fat_current_time() or such to get proper
timestamp for FAT.

I think we should give consistent in-core inode timestamp too. For
example, overlayfs may be copy timestamps from in-core inode, etc., then
this stat workaround will not work.

Thanks.

> diff --git a/fs/fat/fat.h b/fs/fat/fat.h
> index 8fc1093da47d..0367be021b23 100644
> --- a/fs/fat/fat.h
> +++ b/fs/fat/fat.h
> @@ -410,6 +410,7 @@ extern void fat_time_fat2unix(struct msdos_sb_info *sbi, struct timespec *ts,
>  			      __le16 __time, __le16 __date, u8 time_cs);
>  extern void fat_time_unix2fat(struct msdos_sb_info *sbi, struct timespec *ts,
>  			      __le16 *time, __le16 *date, u8 *time_cs);
> +extern void fat_trunc_kstat_times(struct msdos_sb_info *sbi, struct kstat *stat);
>  extern int fat_sync_bhs(struct buffer_head **bhs, int nr_bhs);
>  
>  int fat_cache_init(void);
> diff --git a/fs/fat/file.c b/fs/fat/file.c
> index 4724cc9ad650..11639e393ef7 100644
> --- a/fs/fat/file.c
> +++ b/fs/fat/file.c
> @@ -370,6 +370,7 @@ int fat_getattr(const struct path *path, struct kstat *stat,
>  {
>  	struct inode *inode = d_inode(path->dentry);
>  	generic_fillattr(inode, stat);
> +	fat_trunc_kstat_times(MSDOS_SB(inode->i_sb), stat);
>  	stat->blksize = MSDOS_SB(inode->i_sb)->cluster_size;
>  
>  	if (MSDOS_SB(inode->i_sb)->options.nfs == FAT_NFS_NOSTALE_RO) {
> diff --git a/fs/fat/misc.c b/fs/fat/misc.c
> index f9bdc1e01c98..52f40edefd12 100644
> --- a/fs/fat/misc.c
> +++ b/fs/fat/misc.c
> @@ -262,6 +262,29 @@ void fat_time_unix2fat(struct msdos_sb_info *sbi, struct timespec *ts,
>  }
>  EXPORT_SYMBOL_GPL(fat_time_unix2fat);
>  
> +void fat_trunc_kstat_times(struct msdos_sb_info *sbi, struct kstat *stat)
> +{
> +	if (stat->ino == MSDOS_ROOT_INO)
> +		stat->mtime = stat->ctime = stat->atime =
> +			(struct timespec){0, 0};
> +	else {
> +		__le16 time;
> +		__le16 date;
> +		u8 ctime_cs;
> +
> +		fat_time_unix2fat(sbi, &stat->mtime, &time, &date, NULL);
> +		fat_time_fat2unix(sbi, &stat->mtime, time, date, 0);
> +
> +		fat_time_unix2fat(sbi, &stat->ctime, &time, &date,
> +			sbi->options.isvfat ? &ctime_cs :  NULL);
> +		fat_time_fat2unix(sbi, &stat->ctime, time, date,
> +			sbi->options.isvfat ? ctime_cs : 0);
> +
> +		fat_time_unix2fat(sbi, &stat->atime, &time, &date, NULL);
> +		fat_time_fat2unix(sbi, &stat->atime, 0, date, 0);
> +	}
> +}
> +
>  int fat_sync_bhs(struct buffer_head **bhs, int nr_bhs)
>  {
>  	int i, err = 0;
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [RFC PATCH] fat: truncate timestamps returned from fat_getattr

[Frank Sorenson]

On 04/01/2018 01:02 AM, OGAWA Hirofumi wrote:
> Frank Sorenson <sorenson@redhat.com> writes:
> 
> Already tried to use ->update_time() callback? With quick look though,
> it seems to be possible to control timestamp granularity with
> ->update_time(), and add fat_current_time() or such to get proper
> timestamp for FAT.
> 
> I think we should give consistent in-core inode timestamp too. For
> example, overlayfs may be copy timestamps from in-core inode, etc., then
> this stat workaround will not work.


I'll see what I can do with update_time.  Thanks.

Frank