On 2000-06-23 09:41 -0700, Matthew Dillon <[EMAIL PROTECTED]> wrote:
> Slight problem: We've run out of mount option flags.
But there already ist MNT_SOFTDEP in <sys/mount.h> ...
#define MNT_SUIDDIR 0x00100000 /* special handling of SUID on dirs */
#define MNT_SOFTDEP 0x00200000 /* soft updates being done */
#define MNT_NOSYMFOLLOW 0x00400000 /* do not follow symlinks */
Hmmm, just checked in <ufs/ffs/fs.h> and found that the soft-updates state
is already kept in the fs_flags element of struct fs (the super-block).
/*
* Super block for an FFS file system.
*/
struct fs {
...
int8_t fs_flags; /* see FS_ flags below */
...
};
/*
* Filesystem flags.
*/
#define FS_UNCLEAN 0x01 /* filesystem not clean at mount */
#define FS_DOSOFTDEP 0x02 /* filesystem using soft dependencies */
And in fsck/setup.c, I found:
bufinit();
if (sblock.fs_flags & FS_DOSOFTDEP)
usedsoftdep = 1;
else
usedsoftdep = 0;
return (1);
So its obvious, that we could make "-o softdep" a mount option, do away
with the tunefs option, and have the *kernel* record the use of soft-updates
on a R/W mounted file system when it marks the file system FS_UNCLEAN in
the fs_flags files in the super-block. If a file system is unmounted
cleanly, the FS_DOSOFTDEP flag can be cleared along with the FS_UNCLEAN
bit, since the next mount is free to decide whether to use soft-updates
or not.
This is more safe than the tunefs method, since FS_DOSOFTDEP will only be
set if the last R/W mount was on a kernel that supported soft-updates ...
Only problem I see (but that is not different from the current situation)
is that "mount -u" can't change the soft-updates state, if a file-system
is mounted R/W. This should probably be inforced in the kernel (not the
mount command) and there should be *no* way to "force" mount to override
this lock.
Regards, STefan
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
