Skip to contentSkip to conversation

/sbin/mount and /proc/mounts difference

PATCH New Sven Vermeulen 5 participants 7 messages Nov 18, 2001 – Nov 19, 2001

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi,

When running "mount" I see:
~$ mount
/dev/hda6 on / type reiserfs (rw)
proc on /proc type proc (rw)
/dev/hda1 on /boot type ext2 (rw)
/dev/hda7 on /usr type reiserfs (rw,noatime,notail) /dev/hda8 on /home type reiserfs (rw,nosuid,nodev,noatime,notail) /dev/hda9 on /var type reiserfs (rw,noexec,nodev) /dev/hda10 on /tmp type reiserfs (rw,noexec,nosuid,nodev)

while /proc/mounts says
~$ cat /proc/mounts
/dev/root / reiserfs rw 0 0
/proc /proc proc rw 0 0
/dev/hda1 /boot ext2 rw 0 0
/dev/hda7 /usr reiserfs rw,noatime 0 0 /dev/hda8 /home reiserfs rw,noatime,nosuid,nodev 0 0 /dev/hda9 /var reiserfs rw,nodev,noexec 0 0 /dev/hda10 /tmp reiserfs rw,nosuid,nodev,noexec 0 0

As you can see the notail-option of reiserfs isn't listed on /proc/mounts, but it is on "mount".

Does this have any particular reason?

Wkr,
Sven Vermeulen

Signature · 2 lines
Unix, MS-DOS and Windows NT (also known as the Good, the Bad and the
Ugly). ~(Matt Welsh)
#1 lore raw
Quoted 5 lines · * Sven Vermeulen ([email protected]) wrote:

As you can see the notail-option of reiserfs isn't listed on /proc/mounts, but it is on "mount".

Does this have any particular reason?

Sure, mount reads /etc/mtab.

Stephen

#2 lore raw
In reply to Sven Vermeulen
Quoted 4 lines · Sven Vermeulen

As you can see the notail-option of reiserfs isn't listed on /proc/mounts, but it is on "mount".

Does this have any particular reason?

mount writes everything to /etc/mtab and displays that when asked.

The kernel doesn't ask the individual filesystem "drivers" if they have some options they would like to list there. So for /proc it simply lists the options the VFS knows about. Except for NFS that for some reson has code there.

I don't know if there are any plans to change that. Wouldn't be difficult to add something to super_operations.

fs/namespace.c:
show_vfsmnt
show_nfs_mount

If it is important to you, you could add a "show_reiserfs_mount" by copying how things are done for nfs.

/Urban

#3 lore raw
Quoted 8 lines · Urban Widmark

On Sun, 18 Nov 2001, Sven Vermeulen wrote:

I don't know if there are any plans to change that. Wouldn't be difficult to add something to super_operations.

fs/namespace.c:
show_vfsmnt
show_nfs_mount

Well, now that's easy to do. With the old code (FVO "old" equal to 2.4.15-pre2) you would end up with mind-boggling amount of buffer overruns _and_ ugly code in filesystems.

I would rather start messing with that area with cleaning up mount options' parsers - I've got a patch that does it, it just needs to be ported to recent trees.

#4 lore raw

PATCH 1/1 fs-specific options in /proc/mounts

Quoted 9 lines · Alexander Viro
Quoted 6 lines · Urban Widmark

I don't know if there are any plans to change that. Wouldn't be difficult to add something to super_operations.

fs/namespace.c:
show_vfsmnt
show_nfs_mount

Quoted 7 lines

Well, now that's easy to do. With the old code (FVO "old" equal to 2.4.15-pre2) you would end up with mind-boggling amount of buffer overruns _and_ ugly code in filesystems.

I would rather start messing with that area with cleaning up mount options' parsers - I've got a patch that does it, it just needs to be ported to recent trees.

OTOH... This stuff is really trivial now, so that can easily go into 2.4.
* new superblock method: ->show_options(). It's optional and if
present - used by show_vfsmnt() (i.e. /proc/mounts) to show fs-specific options.
* Prototype: int (*show_options)(struct seq_file *, struct vfsmount *);
Use seq_{putc,puts,printf,escape} for output (see fs/nfs/inode.c for example of use). Pointer to seq_file describes the output buffer (first argument of seq_...().
* Return 0 in case of success, negative - in case of fatal error
(e.g. your function needs to allocate memory and fails). read() from /proc/mounts will fail at that point, so if you see that e.g. server is all fucked up - say seq_puts(m, "server_all_fucked_up"); and return 0 instead of returning -EINVAL.
* The only lock held by caller is mount_sem. Method may block,
allocate memory, cause IO, yodda, yodda. It must not call path_walk() (e.g. filp_open() is off-limits) since that could deadlock if lookup triggers automount.

Patch below moves handling of NFS-specific options into such
method; if you want to show some fs-specific options in your fs - add ->show_options() in your super_operations.

fs/namespace.c +4 -47
diff -urN S15-pre6/fs/namespace.c S15-pre6-mounts/fs/namespace.c
@@ -19,9 +19,6 @@
 
 #include <asm/uaccess.h>
 
-#include <linux/nfs_fs.h>
-#include <linux/nfs_fs_sb.h>
-#include <linux/nfs_mount.h>
 #include <linux/seq_file.h>
 
 struct vfsmount *do_kern_mount(char *type, int flags, char *name, void *data);
@@ -198,50 +195,10 @@
 	seq_escape(m, s, " \t\n\\");
 }
 
-static void show_nfs_mount(struct seq_file *m, struct vfsmount *mnt)
-{
-	static struct proc_nfs_info {
-		int flag;
-		char *str;
-		char *nostr;
-	} nfs_info[] = {
-		{ NFS_MOUNT_SOFT, ",soft", ",hard" },
-		{ NFS_MOUNT_INTR, ",intr", "" },
-		{ NFS_MOUNT_POSIX, ",posix", "" },
-		{ NFS_MOUNT_TCP, ",tcp", ",udp" },
-		{ NFS_MOUNT_NOCTO, ",nocto", "" },
-		{ NFS_MOUNT_NOAC, ",noac", "" },
-		{ NFS_MOUNT_NONLM, ",nolock", ",lock" },
-		{ NFS_MOUNT_BROKEN_SUID, ",broken_suid", "" },
-		{ 0, NULL, NULL }
-	};
-	struct proc_nfs_info *nfs_infop;
-	struct nfs_server *nfss = &mnt->mnt_sb->u.nfs_sb.s_server;
-
-	seq_printf(m, ",v%d", nfss->rpc_ops->version);
-	seq_printf(m, ",rsize=%d", nfss->rsize);
-	seq_printf(m, ",wsize=%d", nfss->wsize);
-	if (nfss->acregmin != 3*HZ)
-		seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
-	if (nfss->acregmax != 60*HZ)
-		seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
-	if (nfss->acdirmin != 30*HZ)
-		seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
-	if (nfss->acdirmax != 60*HZ)
-		seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
-	for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
-		if (nfss->flags & nfs_infop->flag)
-			seq_puts(m, nfs_infop->str);
-		else
-			seq_puts(m, nfs_infop->nostr);
-	}
-	seq_puts(m, ",addr=");
-	mangle(m, nfss->hostname);
-}
-
 static int show_vfsmnt(struct seq_file *m, void *v)
 {
 	struct vfsmount *mnt = v;
+	int err = 0;
 	static struct proc_fs_info {
 		int flag;
 		char *str;
@@ -281,10 +238,10 @@
 		if (mnt->mnt_flags & fs_infop->flag)
 			seq_puts(m, fs_infop->str);
 	}
-	if (strcmp("nfs", mnt->mnt_sb->s_type->name) == 0)
-		show_nfs_mount(m, mnt);
+	if (mnt->mnt_sb->s_op->show_options)
+		err = mnt->mnt_sb->s_op->show_options(m, mnt);
 	seq_puts(m, " 0 0\n");
-	return 0;
+	return err;
 }
 
 struct seq_operations mounts_op = {
fs/nfs/inode.c +45 -0

69 lines not on this page · raw message

include/linux/fs.h +3 -0
diff -urN S15-pre6/include/linux/fs.h S15-pre6-mounts/include/linux/fs.h
@@ -853,6 +853,8 @@
 	int (*getattr) (struct dentry *, struct iattr *);
 };
 
+struct seq_file;
+
 /*
  * NOTE: write_inode, delete_inode, clear_inode, put_inode can be called
  * without the big kernel lock held in all filesystems.
@@ -904,6 +906,7 @@
 	 */
 	struct dentry * (*fh_to_dentry)(struct super_block *sb, __u32 *fh, int len, int fhtype, int parent);
 	int (*dentry_to_fh)(struct dentry *, __u32 *fh, int *lenp, int need_parent);
+	int (*show_options)(struct seq_file *, struct vfsmount *);
 };
 
 /* Inode state bits.. */
#5 lore raw
In reply to Urban Widmark
Quoted 9 lines · Urban Widmark
Quoted 4 lines · Sven Vermeulen

As you can see the notail-option of reiserfs isn't listed on /proc/mounts, but it is on "mount".

Does this have any particular reason?

mount writes everything to /etc/mtab and displays that when asked.

Not quite...

~/tmp# strings /bin/mount | grep mounts
Note that one does not really mount a device, one mounts /proc/mounts
~/tmp#

#6 lore raw
Quoted 10 lines · Martin Dalecki
Quoted 2 lines · Urban Widmark

mount writes everything to /etc/mtab and displays that when asked.

Not quite...

~/tmp# strings /bin/mount | grep mounts
Note that one does not really mount a device, one mounts /proc/mounts
~/tmp#

Not sure what you are trying to show with that grep ...

# strace mount |& grep mounts
# strace mount |& grep mtab
open("/etc/mtab", O_RDONLY)             = 3

mount does read /etc/mtab for displaying the mounts. It only looks at /proc/mounts if /etc/mtab can't be read. Or at least, that is what the version(s) I have does. I would say that the common setup is to have a readable mtab.

/Urban