Skip to contentSkip to conversation

findmnt: add option to list all statvfs f_flags

PATCH In review Roberto Bergantinos Corpas 2 participants 3 messages Nov 25, 2020 Util-Linux package

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 findmnt: add option to list all statvfs f_flags

It might be useful for security auditing purposes list all possible mount flags/options including default set which are normally not listed.

Given that we already call statvfs to retrieve i.e. filesystem size, add an extra column to list all possible f_flags, default or not.

---
 misc-utils/findmnt.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
misc-utils/findmnt.c +23 -0
index 43b4dc7d6..da9f42020 100644
@@ -56,6 +56,7 @@ enum {
 	COL_FSTYPE,
 	COL_OPTIONS,
 	COL_VFS_OPTIONS,
+	COL_MNT_OPTIONS,
 	COL_FS_OPTIONS,
 	COL_LABEL,
 	COL_UUID,
@@ -101,6 +102,7 @@ static struct colinfo infos[] = {
 	[COL_FSTYPE]       = { "FSTYPE",       0.10, SCOLS_FL_TRUNC, N_("filesystem type") },
 	[COL_OPTIONS]      = { "OPTIONS",      0.10, SCOLS_FL_TRUNC, N_("all mount options") },
 	[COL_VFS_OPTIONS]  = { "VFS-OPTIONS",  0.20, SCOLS_FL_TRUNC, N_("VFS specific mount options") },
+	[COL_MNT_OPTIONS]  = { "MNT-OPTIONS",  0.20, SCOLS_FL_TRUNC, N_("All statvfs f_flags from mount") },
 	[COL_FS_OPTIONS]   = { "FS-OPTIONS",   0.10, SCOLS_FL_TRUNC, N_("FS specific mount options") },
 	[COL_LABEL]        = { "LABEL",        0.10, 0, N_("filesystem label") },
 	[COL_UUID]         = { "UUID",           36, 0, N_("filesystem UUID") },
@@ -490,6 +492,26 @@ static char *get_vfs_attr(struct libmnt_fs *fs, int sizetype)
 				(double)(buf.f_blocks - buf.f_bfree) /
 				buf.f_blocks * 100);
 		return sizestr;
+	case COL_MNT_OPTIONS:
+		if (mnt_fs_get_vfs_options(fs)) {
+			sizestr = xstrdup(mnt_fs_get_vfs_options(fs));
+			if (!(buf.f_flag & MS_NOEXEC))
+				mnt_optstr_append_option(&sizestr, "exec", NULL);
+			if (!(buf.f_flag & MS_NOSUID))
+				mnt_optstr_append_option(&sizestr, "suid", NULL);
+			if (!(buf.f_flag & MS_NODEV))
+				mnt_optstr_append_option(&sizestr, "dev", NULL);
+			if (!(buf.f_flag & MS_SYNCHRONOUS))
+				mnt_optstr_append_option(&sizestr, "async", NULL);
+			if (!(buf.f_flag & MS_MANDLOCK))
+				mnt_optstr_append_option(&sizestr, "nomand", NULL);
+			if (!(buf.f_flag & MS_NOATIME))
+				mnt_optstr_append_option(&sizestr, "atime", NULL);
+			if (!(buf.f_flag & MS_NODIRATIME))
+				mnt_optstr_append_option(&sizestr, "diratime", NULL);
+			return sizestr;
+		} else
+			return NULL;
 	}
 
 	if (!vfs_attr)
@@ -585,6 +607,7 @@ static char *get_data(struct libmnt_fs *fs, int num)
 	case COL_AVAIL:
 	case COL_USED:
 	case COL_USEPERC:
+	case COL_MNT_OPTIONS:
 		str = get_vfs_attr(fs, col_id);
 		break;
 	case COL_FSROOT:
Signature · 1 line
2.21.0
maintainer
#1 lore raw
Quoted 5 lines · Roberto Bergantinos Corpas

It might be useful for security auditing purposes list all possible mount flags/options including default set which are normally not listed.

Given that we already call statvfs to retrieve i.e. filesystem size, add an extra column to list all possible f_flags, default or not.

(Sorry Roberto I forgot reply to our private email.)

I understand the goal, but I'm not sure with implementation ;-)

* statvfs() requires access to the mount point, it's better to avoid it if possible due to issue with permissions for non-root users and due to inaccessible NFS, etc.

The mount options are already accessible for everyone in /proc/self/mountinfo. It's possible to use mnt_optstr_get_flags() to convert the string from mnt_fs_get_vfs_options() to flags.

* your implementation hardcodeds the flags to findmnt.c. This is fragile in terms of future development. And for example right now your list does not include dirsync, sub, silent, strictatime, lazytime, relatime and symfollow options.

It would be better to use libmount options map where we maintain the option flags and strings, something like

     const struct libmnt_optmap *ent;
     const struct libmnt_optmap *map;

     map = mnt_get_builtin_optmap(MNT_LINUX_MAP);

     for (ent = map; ent && ent->name; ent++) {
        if (flags & ent->id)
            ...
     }

Maybe we can add a function to return all (including default) options to libmount to make it easy to use for all applications (something like mnt_fs_get_vfs_alloptions()).

misc-utils/findmnt.c static struct colinfo infos[] = { · line 105

Quoted 1 line
misc-utils/findmnt.c +1 -0
+	[COL_MNT_OPTIONS]  = { "MNT-OPTIONS",  0.20, SCOLS_FL_TRUNC, N_("All statvfs f_flags from mount") },

Not sure, but do we really need a new column for this purpose?

It's still about VFS-OPTIONS, the difference is that you need default build-in options too. What about to add --vfs-show-default (or so) command line options to modify VFS-OPTIONS output?

I can implement it next week (if you don't want play with libmount, etc.).

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com
#2 lore raw

Hi Karel!

thanks for inputs, appreciate it.

Yes i agree with your objections, in fact first i thought about extending VFS-OPTIONS via some flag,
and check agains options map, thought i guess i took easy path at some point ;-)

I'll try to see if i can make some work based on your suggestions before next week.

Quoted 60 lines · Karel Zak
Quoted 5 lines · Roberto Bergantinos Corpas

It might be useful for security auditing purposes list all possible mount flags/options including default set which are normally not listed.

Given that we already call statvfs to retrieve i.e. filesystem size, add an extra column to list all possible f_flags, default or not.

(Sorry Roberto I forgot reply to our private email.)

I understand the goal, but I'm not sure with implementation ;-)

* statvfs() requires access to the mount point, it's better to avoid it if possible due to issue with permissions for non-root users and due to inaccessible NFS, etc.

The mount options are already accessible for everyone in /proc/self/mountinfo. It's possible to use mnt_optstr_get_flags() to convert the string from mnt_fs_get_vfs_options() to flags.

* your implementation hardcodeds the flags to findmnt.c. This is fragile in terms of future development. And for example right now your list does not include dirsync, sub, silent, strictatime, lazytime, relatime and symfollow options.

It would be better to use libmount options map where we maintain the option flags and strings, something like

     const struct libmnt_optmap *ent;
     const struct libmnt_optmap *map;

     map = mnt_get_builtin_optmap(MNT_LINUX_MAP);

     for (ent = map; ent && ent->name; ent++) {
        if (flags & ent->id)
            ...
     }

Maybe we can add a function to return all (including default) options to libmount to make it easy to use for all applications (something like mnt_fs_get_vfs_alloptions()).

Quoted 1 line

+ [COL_MNT_OPTIONS] = { "MNT-OPTIONS", 0.20, SCOLS_FL_TRUNC, N_("All statvfs f_flags from mount") },

Not sure, but do we really need a new column for this purpose?

It's still about VFS-OPTIONS, the difference is that you need default build-in options too. What about to add --vfs-show-default (or so) command line options to modify VFS-OPTIONS output?

I can implement it next week (if you don't want play with libmount, etc.).

Karel

--
Karel Zak <[email protected]>
http://karelzak.blogspot.com