user ns: arbitrary module loading
Subsystems (48)
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least CAP_SETUID along the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).
It might be worth examining GRKERNSEC_MODHARDEN in grsecurity, which examines module symbols to verify that request_module() for a filesystem only loads a module that defines "register_filesystem" (among other things).
-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
--
Kees Cook
Chrome OS Security
Serge E. Hallyn
maintainerQuoted 8 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).
Quoted 8 lines
It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
So the concern is root in a child user namespace doing
mount -t randomfs <...>
in which case do_new_mount() checks ns_capable(), not capable(), before trying to load a module for randomfs.
As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the GRKERNSEC_MODHARDEN won't be acceptable. At least Eric Paris is strongly against it. But how about if we add a check for 'current_user_ns() == &init_user_ns' at that place instead?
Eric Biederman, do you have any objections to that?
thanks,
-serge
Quoted 16 lines · Serge E. Hallyn
Quoted 15 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
Quoted 8 lines
Quoted 1 line
[1] https://twitter.com/grsecurity/status/307473816672665600
So the concern is root in a child user namespace doing
mount -t randomfs <...>in which case
do_new_mount()checksns_capable(), notcapable(), before trying to load a module for randomfs.
Well, not just randomfs. Any module that modprobe in the init ns can find.
Quoted 5 lines
As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the
GRKERNSEC_MODHARDENwon't be acceptable. At least Eric Paris is strongly against it.
I'd be curious to hear the objections. It seems pretty nice to me to add a new argument to every request_module() that specifies the "subsystem" it expects a module to load from. Maybe pass "request_module=filesystem" or "...=netdev" to the modprobe call. And then in init_module(), check the userargs for which subsystem was requested and look up in a table for the entry point module symbol for that subsystem to require. e.g. for "request_module=filesystem", require that the module contains the "register_filesystem" symbol, etc.
Quoted 3 lines
But how about if we
add a check for 'current_user_ns()==&init_user_ns' at that place instead?
Well, we'd need to mostly revert 57eccb830f1cc93d4b506ba306d8dfa685e0c88f ("mount: consolidate permission checks") since get_fs_type() is being called before may_mount() right now. (And then, as you suggest, we should strengthen the test.) I think this will require either more plumbing into get_fs_type (something like "bool load_module_if_missing") or the subsystem verification stuff in request_module. I think the latter is MUCH nicer as it covers this problem in all places, not just this "mount" case.
Quoted 4 lines
Eric Biederman, do you have any objections to that?
thanks,
-serge
-Kees
--
Kees Cook
Chrome OS Security
Serge E. Hallyn
maintainerQuoted 19 lines · Kees Cook
Quoted 18 lines · Serge E. Hallyn
Quoted 16 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
Quoted 8 lines
Quoted 6 lines
So the concern is root in a child user namespace doing
mount -t randomfs <...>in which case
do_new_mount()checksns_capable(), notcapable(), before trying to load a module for randomfs.Well, not just randomfs. Any module that modprobe in the init ns can find.
right
Quoted 7 lines
Quoted 5 lines
As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the
GRKERNSEC_MODHARDENwon't be acceptable. At least Eric Paris is strongly against it.I'd be curious to hear the objections. It seems pretty nice to me to
Wait, sorry, I mis-spoke. The objection would have been to requiring CAP_SYS_MODULE, which is different. Sorry!
Quoted 3 lines
add a new argument to every
request_module()that specifies the "subsystem" it expects a module to load from. Maybe pass "request_module=filesystem" or "...=netdev" to the modprobe call. And
That would be useful for adding to the separation of privileges, i.e. helping contain the leaking of posix caps. It sounds good to me.
Quoted 10 lines
then in
init_module(), check the userargs for which subsystem was requested and look up in a table for the entry point module symbol for that subsystem to require. e.g. for "request_module=filesystem", require that the module contains the "register_filesystem" symbol, etc.Quoted 3 lines
But how about if we
add a check for 'current_user_ns()==&init_user_ns' at that place instead?
Quoted 9 lines
Well, we'd need to mostly revert
57eccb830f1cc93d4b506ba306d8dfa685e0c88f("mount: consolidate permission checks") sinceget_fs_type()is being called beforemay_mount()right now. (And then, as you suggest, we should strengthen the test.) I think this will require either more plumbing intoget_fs_type(something like "boolload_module_if_missing") or the subsystem verification stuff inrequest_module. I think the latter is MUCH nicer as it covers this problem in all places, not just this "mount" case.
My first instinct was to say I'd like to have the kernel 100% belonging to the init_user_ns, with child user namespaces having zero ability to induce loading of any kernel modules, period. So a check for current being in init_user_ns at request_module itself.
However (thinking more) that seems maybe wrong. You don't need privs to induce the loading of a new binfmt module right? The host's /lib/modules and module blacklists should be set up right by the admin (or distro)... If we require that the host admin manually modprobe every module which a task in a child user namespace might need, that goes counter to the goal of kernel modules.
Quoted 1 line
Quoted 1 line
Eric Biederman, do you have any objections to that?
-serge
RFC PATCH 1/1 fs: Limit sys_mount to only loading filesystem modules.
Modify the request_module to prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.
A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing /etc/modprobe.d/*.conf with blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.
This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
--- arch/ia64/kernel/perfmon.c | 1 + arch/powerpc/platforms/cell/spufs/inode.c | 1 + arch/s390/hypfs/inode.c | 1 + drivers/firmware/efivars.c | 1 + drivers/infiniband/hw/ipath/ipath_fs.c | 1 + drivers/infiniband/hw/qib/qib_fs.c | 1 + drivers/misc/ibmasm/ibmasmfs.c | 1 + drivers/mtd/mtdchar.c | 1 + drivers/oprofile/oprofilefs.c | 1 + drivers/staging/ccg/f_fs.c | 1 + drivers/usb/gadget/f_fs.c | 1 + drivers/usb/gadget/inode.c | 1 + drivers/xen/xenfs/super.c | 1 + fs/9p/vfs_super.c | 1 + fs/adfs/super.c | 1 + fs/affs/super.c | 1 + fs/afs/super.c | 1 + fs/autofs4/init.c | 1 + fs/befs/linuxvfs.c | 1 + fs/bfs/inode.c | 1 + fs/binfmt_misc.c | 1 + fs/btrfs/super.c | 1 + fs/ceph/super.c | 1 + fs/coda/inode.c | 1 + fs/configfs/mount.c | 1 + fs/cramfs/inode.c | 1 + fs/debugfs/inode.c | 1 + fs/devpts/inode.c | 1 + fs/ecryptfs/main.c | 1 + fs/efs/super.c | 1 + fs/exofs/super.c | 1 + fs/ext2/super.c | 1 + fs/ext3/super.c | 1 + fs/ext4/super.c | 5 +++-- fs/f2fs/super.c | 1 + fs/fat/namei_msdos.c | 1 + fs/fat/namei_vfat.c | 1 + fs/filesystems.c | 2 +- fs/freevxfs/vxfs_super.c | 2 +- fs/fuse/control.c | 1 + fs/fuse/inode.c | 2 ++ fs/gfs2/ops_fstype.c | 4 +++- fs/hfs/super.c | 1 + fs/hfsplus/super.c | 1 + fs/hppfs/hppfs.c | 1 + fs/hugetlbfs/inode.c | 1 + fs/isofs/inode.c | 3 +-- fs/jffs2/super.c | 1 + fs/jfs/super.c | 1 + fs/logfs/super.c | 1 + fs/minix/inode.c | 1 + fs/ncpfs/inode.c | 1 + fs/nfs/super.c | 3 ++- fs/nfsd/nfsctl.c | 1 + fs/nilfs2/super.c | 1 + fs/ntfs/super.c | 1 + fs/ocfs2/dlmfs/dlmfs.c | 1 + fs/omfs/inode.c | 1 + fs/openpromfs/inode.c | 1 + fs/qnx4/inode.c | 1 + fs/qnx6/inode.c | 1 + fs/reiserfs/super.c | 1 + fs/romfs/super.c | 1 + fs/sysv/super.c | 3 ++- fs/ubifs/super.c | 1 + fs/ufs/super.c | 1 + fs/xfs/xfs_super.c | 1 + include/linux/fs.h | 2 ++ net/sunrpc/rpc_pipe.c | 4 +--- 69 files changed, 77 insertions(+), 12 deletions(-)
arch/ia64/kernel/perfmon.c +1 -0
@@ -619,6 +619,7 @@ static struct file_system_type pfm_fs_type = { .mount = pfmfs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("pfmfs" ); DEFINE_PER_CPU(unsigned long, pfm_syst_info); DEFINE_PER_CPU(struct task_struct *, pmu_owner);
arch/powerpc/platforms/cell/spufs/inode.c +1 -0
@@ -771,6 +771,7 @@ static struct file_system_type spufs_type = { .mount = spufs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("spufs" );static int __init spufs_init (void ) {
arch/s390/hypfs/inode.c +1 -0
@@ -458,6 +458,7 @@ static struct file_system_type hypfs_type = { .mount = hypfs_mount, .kill_sb = hypfs_kill_super }; +MODULE_ALIAS_FS ("s390_hypfs" );static const struct super_operations hypfs_s_ops = { .statfs = simple_statfs,
drivers/firmware/efivars.c +1 -0
@@ -1116,6 +1116,7 @@ static struct file_system_type efivarfs_type = { .mount = efivarfs_mount, .kill_sb = efivarfs_kill_sb, }; +MODULE_ALIAS_FS ("efivarfs" );static const struct inode_operations efivarfs_dir_inode_operations = { .lookup = simple_lookup,
drivers/infiniband/hw/ipath/ipath_fs.c +1 -0
@@ -410,6 +410,7 @@ static struct file_system_type ipathfs_fs_type = { .mount = ipathfs_mount, .kill_sb = ipathfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init ipath_init_ipathfs (void ) {
drivers/infiniband/hw/qib/qib_fs.c +1 -0
@@ -604,6 +604,7 @@ static struct file_system_type qibfs_fs_type = { .mount = qibfs_mount, .kill_sb = qibfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init qib_init_qibfs (void ) {
drivers/misc/ibmasm/ibmasmfs.c +1 -0
@@ -110,6 +110,7 @@ static struct file_system_type ibmasmfs_type = { .mount = ibmasmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ibmasmfs" );static int ibmasmfs_fill_super (struct super_block *sb,void *data,int silent) {
drivers/mtd/mtdchar.c +1 -0
@@ -1238,6 +1238,7 @@ static struct file_system_type mtd_inodefs_type = { .mount = mtd_inodefs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("mtd_inodefs" );static int __init init_mtdchar (void ) {
drivers/oprofile/oprofilefs.c +1 -0
@@ -266,6 +266,7 @@ static struct file_system_type oprofilefs_type = { .mount = oprofilefs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("oprofilefs" );int __init oprofilefs_register (void )
drivers/staging/ccg/f_fs.c +1 -0
@@ -1223,6 +1223,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/
drivers/usb/gadget/f_fs.c +1 -0
@@ -1235,6 +1235,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/
drivers/usb/gadget/inode.c +1 -0
@@ -2105,6 +2105,7 @@ static struct file_system_type gadgetfs_type = { .mount = gadgetfs_mount, .kill_sb = gadgetfs_kill_sb, }; +MODULE_ALIAS_FS ("gadgetfs" );/*----------------------------------------------------------------------*/
drivers/xen/xenfs/super.c +1 -0
@@ -119,6 +119,7 @@ static struct file_system_type xenfs_type = { .mount = xenfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("xenfs" );static int __init xenfs_init (void ) {
fs/9p/vfs_super.c +1 -0
@@ -365,3 +365,4 @@ struct file_system_type v9fs_fs_type = { .owner = THIS_MODULE, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT, }; +MODULE_ALIAS_FS ("9p" );
fs/adfs/super.c +1 -0
@@ -524,6 +524,7 @@ static struct file_system_type adfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("adfs" );static int __init init_adfs_fs (void ) {
fs/affs/super.c +1 -0
@@ -622,6 +622,7 @@ static struct file_system_type affs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("affs" );static int __init init_affs_fs (void ) {
fs/afs/super.c +1 -0
@@ -45,6 +45,7 @@ struct file_system_type afs_fs_type = { .kill_sb = afs_kill_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("afs" );static const struct super_operations afs_super_ops = { .statfs = afs_statfs,
fs/autofs4/init.c +1 -0
@@ -26,6 +26,7 @@ static struct file_system_type autofs_fs_type = { .mount = autofs_mount, .kill_sb = autofs4_kill_sb, }; +MODULE_ALIAS_FS ("autofs" );static int __init init_autofs4_fs (void ) {
fs/befs/linuxvfs.c +1 -0
@@ -951,6 +951,7 @@ static struct file_system_type befs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("befs" );static int __init init_befs_fs (void )
fs/bfs/inode.c +1 -0
@@ -473,6 +473,7 @@ static struct file_system_type bfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("bfs" );static int __init init_bfs_fs (void ) {
fs/binfmt_misc.c +1 -0
@@ -720,6 +720,7 @@ static struct file_system_type bm_fs_type = { .mount = bm_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("binfmt_misc" );static int __init init_misc_binfmt (void ) {
fs/btrfs/super.c +1 -0
@@ -1518,6 +1518,7 @@ static struct file_system_type btrfs_fs_type = { .kill_sb = btrfs_kill_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("btrfs" );/* * used by btrfsctl to scan devices when no FS is mounted
fs/ceph/super.c +1 -0
@@ -947,6 +947,7 @@ static struct file_system_type ceph_fs_type = { .kill_sb = ceph_kill_sb, .fs_flags = FS_RENAME_DOES_D_MOVE, }; +MODULE_ALIAS_FS ("ceph" );#define _STRINGIFY (x) #x#define STRINGIFY (x) _STRINGIFY(x)
fs/coda/inode.c +1 -0
@@ -329,4 +329,5 @@ struct file_system_type coda_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("coda" );
fs/configfs/mount.c +1 -0
@@ -114,6 +114,7 @@ static struct file_system_type configfs_fs_type = { .mount = configfs_do_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("configfs" );struct dentry *configfs_pin_fs (void ) {
fs/cramfs/inode.c +1 -0
@@ -573,6 +573,7 @@ static struct file_system_type cramfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("cramfs" );static int __init init_cramfs_fs (void ) {
fs/debugfs/inode.c +1 -0
@@ -299,6 +299,7 @@ static struct file_system_type debug_fs_type = { .mount = debug_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("debugfs" ); static struct dentry *__create_file(const char *name, umode_t mode, struct dentry *parent, void *data,
fs/devpts/inode.c +1 -0
@@ -380,6 +380,7 @@ static struct file_system_type devpts_fs_type = { .kill_sb = devpts_kill_sb, .fs_flags = FS_USERNS_MOUNT | FS_USERNS_DEV_MOUNT, }; +MODULE_ALIAS_FS ("devpts" );struct vfsmount *devpts_mntget (struct file *filp) {
fs/ecryptfs/main.c +1 -0
@@ -629,6 +629,7 @@ static struct file_system_type ecryptfs_fs_type = { .kill_sb = ecryptfs_kill_block_super, .fs_flags = 0 }; +MODULE_ALIAS_FS ("ecryptfs" );/** * inode_info_init_once
fs/efs/super.c +1 -0
@@ -33,6 +33,7 @@ static struct file_system_type efs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("efs" );static struct pt_types sgi_pt_types[] = { {0x00 ,"SGI vh" },
fs/exofs/super.c +1 -0
@@ -1010,6 +1010,7 @@ static struct file_system_type exofs_type = { .mount = exofs_mount, .kill_sb = generic_shutdown_super, }; +MODULE_ALIAS_FS ("exofs" );static int __init init_exofs (void ) {
fs/ext2/super.c +1 -0
@@ -1542,6 +1542,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT, }; +MODULE_ALIAS_FS ("ext2" );static int __init init_ext2_fs (void ) {
fs/ext3/super.c +1 -0
@@ -3059,6 +3059,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext3" );static int __init init_ext3_fs (void ) {
fs/ext4/super.c +3 -2
@@ -92,6 +92,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext2" );#define IS_EXT2_SB (sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)#else #define IS_EXT2_SB (sb) (0)
@@ -106,6 +107,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext3" );#define IS_EXT3_SB (sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)#else #define IS_EXT3_SB (sb) (0)
@@ -5194,7 +5196,6 @@ static inline int ext2_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext2" );#else static inline void register_as_ext2 (void ) { }static inline void unregister_as_ext2 (void ) { }
@@ -5227,7 +5228,6 @@ static inline int ext3_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext3" );#else static inline void register_as_ext3 (void ) { }static inline void unregister_as_ext3 (void ) { }
@@ -5241,6 +5241,7 @@ static struct file_system_type ext4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext4" );static int __init ext4_init_feat_adverts (void ) {
fs/f2fs/super.c +1 -0
@@ -639,6 +639,7 @@ static struct file_system_type f2fs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("f2fs" );static int __init init_inodecache (void ) {
fs/fat/namei_msdos.c +1 -0
@@ -668,6 +668,7 @@ static struct file_system_type msdos_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("msdos" );static int __init init_msdos_fs (void ) {
fs/fat/namei_vfat.c +1 -0
@@ -1073,6 +1073,7 @@ static struct file_system_type vfat_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vfat" );static int __init init_vfat_fs (void ) {
fs/filesystems.c +1 -1
@@ -274,7 +274,7 @@ struct file_system_type *get_fs_type(const char *name) fs =__get_fs_type (name, len);if (!fs &&capable (CAP_SYS_ADMIN ) && - (request_module ("%.*s" , len, name) ==0 )) + (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {
fs/freevxfs/vxfs_super.c +1 -1
@@ -52,7 +52,6 @@ MODULE_AUTHOR("Christoph Hellwig");MODULE_DESCRIPTION ("Veritas Filesystem (VxFS) driver" );MODULE_LICENSE ("Dual BSD/GPL" ); -MODULE_ALIAS ("vxfs" );/* makes mount -t vxfs autoload the module */ static void vxfs_put_super (struct super_block *);
@@ -258,6 +257,7 @@ static struct file_system_type vxfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vxfs" );/* makes mount -t vxfs autoload the module */ static int __init vxfs_init (void )
fs/fuse/control.c +1 -0
@@ -341,6 +341,7 @@ static struct file_system_type fuse_ctl_fs_type = { .mount = fuse_ctl_mount, .kill_sb = fuse_ctl_kill_sb, }; +MODULE_ALIAS_FS ("fusectl" );int __init fuse_ctl_init (void ) {
fs/fuse/inode.c +2 -0
@@ -1127,6 +1127,7 @@ static struct file_system_type fuse_fs_type = { .mount = fuse_mount, .kill_sb = fuse_kill_sb_anon, }; +MODULE_ALIAS_FS ("fuse" );#ifdef CONFIG_BLOCK static struct dentry *fuse_mount_blk(struct file_system_type *fs_type,
@@ -1156,6 +1157,7 @@ static struct file_system_type fuseblk_fs_type = { .kill_sb = fuse_kill_sb_blk, .fs_flags = FS_REQUIRES_DEV | FS_HAS_SUBTYPE, }; +MODULE_ALIAS_FS ("fuseblk" );static inline int register_fuseblk (void ) {
fs/gfs2/ops_fstype.c +3 -1
@@ -20,6 +20,7 @@#include <linux/gfs2_ondisk.h> #include <linux/quotaops.h> #include <linux/lockdep.h> +#include <linux/module.h> #include "gfs2.h" #include "incore.h"
@@ -1425,6 +1426,7 @@ struct file_system_type gfs2_fs_type = { .kill_sb = gfs2_kill_sb, .owner = THIS_MODULE, }; +MODULE_ALIAS_FS ("gfs2" );struct file_system_type gfs2meta_fs_type = { .name ="gfs2meta" ,
@@ -1432,4 +1434,4 @@ struct file_system_type gfs2meta_fs_type = { .mount = gfs2_mount_meta, .owner = THIS_MODULE, }; - +MODULE_ALIAS_FS ("gfs2meta" );
fs/hfs/super.c +1 -0
@@ -466,6 +466,7 @@ static struct file_system_type hfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfs" );static void hfs_init_once (void *p) {
fs/hfsplus/super.c +1 -0
@@ -618,6 +618,7 @@ static struct file_system_type hfsplus_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfsplus" );static void hfsplus_init_once (void *p) {
fs/hppfs/hppfs.c +1 -0
@@ -748,6 +748,7 @@ static struct file_system_type hppfs_type = { .kill_sb = kill_anon_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("hppfs" );static int __init init_hppfs (void ) {
fs/hugetlbfs/inode.c +1 -0
@@ -896,6 +896,7 @@ static struct file_system_type hugetlbfs_fs_type = { .mount = hugetlbfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("hugetlbfs" );static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE ];
fs/isofs/inode.c +1 -2
@@ -1556,6 +1556,7 @@ static struct file_system_type iso9660_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("iso9660" );static int __init init_iso9660_fs (void ) {
@@ -1593,5 +1594,3 @@ static void __exit exit_iso9660_fs(void) module_init(init_iso9660_fs) module_exit(exit_iso9660_fs) MODULE_LICENSE("GPL"); -/* Actual filesystem name is iso9660, as requested in filesystems.c */ -MODULE_ALIAS ("iso9660" );
fs/jffs2/super.c +1 -0
@@ -356,6 +356,7 @@ static struct file_system_type jffs2_fs_type = { .mount = jffs2_mount, .kill_sb = jffs2_kill_sb, }; +MODULE_ALIAS_FS ("jffs2" );static int __init init_jffs2_fs (void ) {
fs/jfs/super.c +1 -0
@@ -833,6 +833,7 @@ static struct file_system_type jfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("jfs" );static void init_once (void *foo) {
fs/logfs/super.c +1 -0
@@ -608,6 +608,7 @@ static struct file_system_type logfs_fs_type = { .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("logfs" );static int __init logfs_init (void ) {
fs/minix/inode.c +1 -0
@@ -660,6 +660,7 @@ static struct file_system_type minix_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("minix" );static int __init init_minix_fs (void ) {
fs/ncpfs/inode.c +1 -0
@@ -1051,6 +1051,7 @@ static struct file_system_type ncp_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("ncpfs" );static int __init init_ncp_fs (void ) {
fs/nfs/super.c +2 -1
@@ -293,6 +293,7 @@ struct file_system_type nfs_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs" );EXPORT_SYMBOL_GPL (nfs_fs_type);struct file_system_type nfs_xdev_fs_type = {
@@ -332,6 +333,7 @@ struct file_system_type nfs4_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs4" );EXPORT_SYMBOL_GPL (nfs4_fs_type);static int __init register_nfs4_fs (void )
@@ -2716,6 +2718,5 @@ module_param(send_implementation_id, ushort, 0644);MODULE_PARM_DESC (send_implementation_id,"Send implementation ID with NFSv4.1 exchange_id" );MODULE_PARM_DESC (nfs4_unique_id,"nfs_client_id4 uniquifier string" ); -MODULE_ALIAS ("nfs4" );#endif /* CONFIG_NFS_V4 */
fs/nfsd/nfsctl.c +1 -0
@@ -1052,6 +1052,7 @@ static struct file_system_type nfsd_fs_type = { .mount = nfsd_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("nfsd" );#ifdef CONFIG_PROC_FS static int create_proc_exports_entry (void )
fs/nilfs2/super.c +1 -0
@@ -1361,6 +1361,7 @@ struct file_system_type nilfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("nilfs2" );static void nilfs_inode_init_once (void *obj) {
fs/ntfs/super.c +1 -0
@@ -3079,6 +3079,7 @@ static struct file_system_type ntfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ntfs" );/* Stable names for the slab caches. */ static const char ntfs_index_ctx_cache_name[] ="ntfs_index_ctx_cache" ;
fs/ocfs2/dlmfs/dlmfs.c +1 -0
@@ -640,6 +640,7 @@ static struct file_system_type dlmfs_fs_type = { .mount = dlmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ocfs2_dlmfs" );static int __init init_dlmfs_fs (void ) {
fs/omfs/inode.c +1 -0
@@ -572,6 +572,7 @@ static struct file_system_type omfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("omfs" );static int __init init_omfs_fs (void ) {
fs/openpromfs/inode.c +1 -0
@@ -432,6 +432,7 @@ static struct file_system_type openprom_fs_type = { .mount = openprom_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("openpromfs" );static void op_inode_init_once (void *data) {
fs/qnx4/inode.c +1 -0
@@ -412,6 +412,7 @@ static struct file_system_type qnx4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx4" );static int __init init_qnx4_fs (void ) {
fs/qnx6/inode.c +1 -0
@@ -672,6 +672,7 @@ static struct file_system_type qnx6_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx6" );static int __init init_qnx6_fs (void ) {
fs/reiserfs/super.c +1 -0
@@ -2434,6 +2434,7 @@ struct file_system_type reiserfs_fs_type = { .kill_sb = reiserfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("reiserfs" );MODULE_DESCRIPTION ("ReiserFS journaled filesystem" );MODULE_AUTHOR ("Hans Reiser <[email protected]>" );
fs/romfs/super.c +1 -0
@@ -599,6 +599,7 @@ static struct file_system_type romfs_fs_type = { .kill_sb = romfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("romfs" );/* * inode storage initialiser
fs/sysv/super.c +2 -1
@@ -545,6 +545,7 @@ static struct file_system_type sysv_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("sysv" );static struct file_system_type v7_fs_type = { .owner =THIS_MODULE ,
@@ -553,6 +554,7 @@ static struct file_system_type v7_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("v7" );static int __init init_sysv_fs (void ) {
@@ -586,5 +588,4 @@ static void __exit exit_sysv_fs(void) module_init(init_sysv_fs) module_exit(exit_sysv_fs) -MODULE_ALIAS("v7"); MODULE_LICENSE("GPL");
fs/ubifs/super.c +1 -0
@@ -2174,6 +2174,7 @@ static struct file_system_type ubifs_fs_type = { .mount = ubifs_mount, .kill_sb = kill_ubifs_super, }; +MODULE_ALIAS_FS ("ubifs" );/* * Inode slab cache constructor.
fs/ufs/super.c +1 -0
@@ -1500,6 +1500,7 @@ static struct file_system_type ufs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ufs" );static int __init init_ufs_fs (void ) {
fs/xfs/xfs_super.c +1 -0
@@ -1561,6 +1561,7 @@ static struct file_system_type xfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("xfs" ); STATIC int __initxfs_init_zones (void )
include/linux/fs.h +2 -0
@@ -1829,6 +1829,8 @@ struct file_system_type {struct lock_class_key i_mutex_dir_key; }; +#define MODULE_ALIAS_FS (NAME ) MODULE_ALIAS("fs-" NAME) +extern struct dentry *mount_ns (struct file_system_type *fs_type,int flags,void *data,int (*fill_super)(struct super_block *,void *,int )); extern struct dentry *mount_bdev(struct file_system_type *fs_type,
net/sunrpc/rpc_pipe.c +1 -3
@@ -1174,6 +1174,7 @@ static struct file_system_type rpc_pipe_fs_type = { .mount = rpc_mount, .kill_sb = rpc_kill_sb, }; +MODULE_ALIAS_FS ("rpc_pipefs" );static void init_once (void *foo)
@@ -1218,6 +1219,3 @@ void unregister_rpc_pipefs(void)kmem_cache_destroy (rpc_inode_cachep);unregister_filesystem (&rpc_pipe_fs_type); } - -/* Make 'mount -t rpc_pipefs ...' autoload this module. */ -MODULE_ALIAS ("rpc_pipefs" );
Signature · 1 line
1.7.5.4
Re: [RFC][PATCH] fs: Limit sys_mount to only loading filesystem modules.
Quoted 18 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing
/etc/modprobe.d/*.confwith blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.
Quoted 7 lines
This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
Signed-off-by "Eric W. Biederman" <[email protected]>
Quoted 1,003 lines
--- arch/ia64/kernel/perfmon.c | 1 + arch/powerpc/platforms/cell/spufs/inode.c | 1 + arch/s390/hypfs/inode.c | 1 + drivers/firmware/efivars.c | 1 + drivers/infiniband/hw/ipath/ipath_fs.c | 1 + drivers/infiniband/hw/qib/qib_fs.c | 1 + drivers/misc/ibmasm/ibmasmfs.c | 1 + drivers/mtd/mtdchar.c | 1 + drivers/oprofile/oprofilefs.c | 1 + drivers/staging/ccg/f_fs.c | 1 + drivers/usb/gadget/f_fs.c | 1 + drivers/usb/gadget/inode.c | 1 + drivers/xen/xenfs/super.c | 1 + fs/9p/vfs_super.c | 1 + fs/adfs/super.c | 1 + fs/affs/super.c | 1 + fs/afs/super.c | 1 + fs/autofs4/init.c | 1 + fs/befs/linuxvfs.c | 1 + fs/bfs/inode.c | 1 + fs/binfmt_misc.c | 1 + fs/btrfs/super.c | 1 + fs/ceph/super.c | 1 + fs/coda/inode.c | 1 + fs/configfs/mount.c | 1 + fs/cramfs/inode.c | 1 + fs/debugfs/inode.c | 1 + fs/devpts/inode.c | 1 + fs/ecryptfs/main.c | 1 + fs/efs/super.c | 1 + fs/exofs/super.c | 1 + fs/ext2/super.c | 1 + fs/ext3/super.c | 1 + fs/ext4/super.c | 5 +++-- fs/f2fs/super.c | 1 + fs/fat/namei_msdos.c | 1 + fs/fat/namei_vfat.c | 1 + fs/filesystems.c | 2 +- fs/freevxfs/vxfs_super.c | 2 +- fs/fuse/control.c | 1 + fs/fuse/inode.c | 2 ++ fs/gfs2/ops_fstype.c | 4 +++- fs/hfs/super.c | 1 + fs/hfsplus/super.c | 1 + fs/hppfs/hppfs.c | 1 + fs/hugetlbfs/inode.c | 1 + fs/isofs/inode.c | 3 +-- fs/jffs2/super.c | 1 + fs/jfs/super.c | 1 + fs/logfs/super.c | 1 + fs/minix/inode.c | 1 + fs/ncpfs/inode.c | 1 + fs/nfs/super.c | 3 ++- fs/nfsd/nfsctl.c | 1 + fs/nilfs2/super.c | 1 + fs/ntfs/super.c | 1 + fs/ocfs2/dlmfs/dlmfs.c | 1 + fs/omfs/inode.c | 1 + fs/openpromfs/inode.c | 1 + fs/qnx4/inode.c | 1 + fs/qnx6/inode.c | 1 + fs/reiserfs/super.c | 1 + fs/romfs/super.c | 1 + fs/sysv/super.c | 3 ++- fs/ubifs/super.c | 1 + fs/ufs/super.c | 1 + fs/xfs/xfs_super.c | 1 + include/linux/fs.h | 2 ++ net/sunrpc/rpc_pipe.c | 4 +--- 69 files changed, 77 insertions(+), 12 deletions(-)arch/ia64/kernel/perfmon.c +1
-0@@ -619,6 +619,7 @@ static struct file_system_type pfm_fs_type = { .mount = pfmfs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("pfmfs" ); DEFINE_PER_CPU(unsigned long, pfm_syst_info); DEFINE_PER_CPU(struct task_struct *, pmu_owner);arch/powerpc/platforms/cell/spufs/inode.c +1
-0@@ -771,6 +771,7 @@ static struct file_system_type spufs_type = { .mount = spufs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("spufs" );static int __init spufs_init (void ) {arch/s390/hypfs/inode.c +1
-0@@ -458,6 +458,7 @@ static struct file_system_type hypfs_type = { .mount = hypfs_mount, .kill_sb = hypfs_kill_super }; +MODULE_ALIAS_FS ("s390_hypfs" );static const struct super_operations hypfs_s_ops = { .statfs = simple_statfs,drivers/firmware/efivars.c +1
-0@@ -1116,6 +1116,7 @@ static struct file_system_type efivarfs_type = { .mount = efivarfs_mount, .kill_sb = efivarfs_kill_sb, }; +MODULE_ALIAS_FS ("efivarfs" );static const struct inode_operations efivarfs_dir_inode_operations = { .lookup = simple_lookup,drivers/infiniband/hw/ipath/ipath_fs.c +1
-0@@ -410,6 +410,7 @@ static struct file_system_type ipathfs_fs_type = { .mount = ipathfs_mount, .kill_sb = ipathfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init ipath_init_ipathfs (void ) {drivers/infiniband/hw/qib/qib_fs.c +1
-0@@ -604,6 +604,7 @@ static struct file_system_type qibfs_fs_type = { .mount = qibfs_mount, .kill_sb = qibfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init qib_init_qibfs (void ) {drivers/misc/ibmasm/ibmasmfs.c +1
-0@@ -110,6 +110,7 @@ static struct file_system_type ibmasmfs_type = { .mount = ibmasmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ibmasmfs" );static int ibmasmfs_fill_super (struct super_block *sb,void *data,int silent) {drivers/mtd/mtdchar.c +1
-0@@ -1238,6 +1238,7 @@ static struct file_system_type mtd_inodefs_type = { .mount = mtd_inodefs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("mtd_inodefs" );static int __init init_mtdchar (void ) {drivers/oprofile/oprofilefs.c +1
-0@@ -266,6 +266,7 @@ static struct file_system_type oprofilefs_type = { .mount = oprofilefs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("oprofilefs" );int __init oprofilefs_register (void )drivers/staging/ccg/f_fs.c +1
-0@@ -1223,6 +1223,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/ drivers/usb/gadget/f_fs.c +1
-0@@ -1235,6 +1235,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/ drivers/usb/gadget/inode.c +1
-0@@ -2105,6 +2105,7 @@ static struct file_system_type gadgetfs_type = { .mount = gadgetfs_mount, .kill_sb = gadgetfs_kill_sb, }; +MODULE_ALIAS_FS ("gadgetfs" );/*----------------------------------------------------------------------*/ drivers/xen/xenfs/super.c +1
-0@@ -119,6 +119,7 @@ static struct file_system_type xenfs_type = { .mount = xenfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("xenfs" );static int __init xenfs_init (void ) {fs/9p/vfs_super.c +1
-0@@ -365,3 +365,4 @@ struct file_system_type v9fs_fs_type = { .owner = THIS_MODULE, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT, }; +MODULE_ALIAS_FS ("9p" );fs/adfs/super.c +1
-0@@ -524,6 +524,7 @@ static struct file_system_type adfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("adfs" );static int __init init_adfs_fs (void ) {fs/affs/super.c +1
-0@@ -622,6 +622,7 @@ static struct file_system_type affs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("affs" );static int __init init_affs_fs (void ) {fs/afs/super.c +1
-0@@ -45,6 +45,7 @@ struct file_system_type afs_fs_type = { .kill_sb = afs_kill_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("afs" );static const struct super_operations afs_super_ops = { .statfs = afs_statfs,fs/autofs4/init.c +1
-0@@ -26,6 +26,7 @@ static struct file_system_type autofs_fs_type = { .mount = autofs_mount, .kill_sb = autofs4_kill_sb, }; +MODULE_ALIAS_FS ("autofs" );static int __init init_autofs4_fs (void ) {fs/befs/linuxvfs.c +1
-0@@ -951,6 +951,7 @@ static struct file_system_type befs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("befs" );static int __init init_befs_fs (void )fs/bfs/inode.c +1
-0@@ -473,6 +473,7 @@ static struct file_system_type bfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("bfs" );static int __init init_bfs_fs (void ) {fs/binfmt_misc.c +1
-0@@ -720,6 +720,7 @@ static struct file_system_type bm_fs_type = { .mount = bm_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("binfmt_misc" );static int __init init_misc_binfmt (void ) {fs/btrfs/super.c +1
-0@@ -1518,6 +1518,7 @@ static struct file_system_type btrfs_fs_type = { .kill_sb = btrfs_kill_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("btrfs" );/* * used by btrfsctl to scan devices when no FS is mounted fs/ceph/super.c +1
-0@@ -947,6 +947,7 @@ static struct file_system_type ceph_fs_type = { .kill_sb = ceph_kill_sb, .fs_flags = FS_RENAME_DOES_D_MOVE, }; +MODULE_ALIAS_FS ("ceph" );#define _STRINGIFY (x) #x#define STRINGIFY (x) _STRINGIFY(x)fs/coda/inode.c +1
-0@@ -329,4 +329,5 @@ struct file_system_type coda_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("coda" );fs/configfs/mount.c +1
-0@@ -114,6 +114,7 @@ static struct file_system_type configfs_fs_type = { .mount = configfs_do_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("configfs" );struct dentry *configfs_pin_fs (void ) {fs/cramfs/inode.c +1
-0@@ -573,6 +573,7 @@ static struct file_system_type cramfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("cramfs" );static int __init init_cramfs_fs (void ) {fs/debugfs/inode.c +1
-0@@ -299,6 +299,7 @@ static struct file_system_type debug_fs_type = { .mount = debug_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("debugfs" ); static struct dentry *__create_file(const char *name, umode_t mode, struct dentry *parent, void *data,fs/devpts/inode.c +1
-0@@ -380,6 +380,7 @@ static struct file_system_type devpts_fs_type = { .kill_sb = devpts_kill_sb, .fs_flags = FS_USERNS_MOUNT | FS_USERNS_DEV_MOUNT, }; +MODULE_ALIAS_FS ("devpts" );struct vfsmount *devpts_mntget (struct file *filp) {fs/ecryptfs/main.c +1
-0@@ -629,6 +629,7 @@ static struct file_system_type ecryptfs_fs_type = { .kill_sb = ecryptfs_kill_block_super, .fs_flags = 0 }; +MODULE_ALIAS_FS ("ecryptfs" );/** * inode_info_init_once fs/efs/super.c +1
-0@@ -33,6 +33,7 @@ static struct file_system_type efs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("efs" );static struct pt_types sgi_pt_types[] = { {0x00 ,"SGI vh" },fs/exofs/super.c +1
-0@@ -1010,6 +1010,7 @@ static struct file_system_type exofs_type = { .mount = exofs_mount, .kill_sb = generic_shutdown_super, }; +MODULE_ALIAS_FS ("exofs" );static int __init init_exofs (void ) {fs/ext2/super.c +1
-0@@ -1542,6 +1542,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT, }; +MODULE_ALIAS_FS ("ext2" );static int __init init_ext2_fs (void ) {fs/ext3/super.c +1
-0@@ -3059,6 +3059,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext3" );static int __init init_ext3_fs (void ) {fs/ext4/super.c +3
-2@@ -92,6 +92,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext2" );#define IS_EXT2_SB (sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)#else #define IS_EXT2_SB (sb) (0)@@ -106,6 +107,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext3" );#define IS_EXT3_SB (sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)#else #define IS_EXT3_SB (sb) (0)@@ -5194,7 +5196,6 @@ static inline int ext2_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext2" );#else static inline void register_as_ext2 (void ) { }static inline void unregister_as_ext2 (void ) { }@@ -5227,7 +5228,6 @@ static inline int ext3_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext3" );#else static inline void register_as_ext3 (void ) { }static inline void unregister_as_ext3 (void ) { }@@ -5241,6 +5241,7 @@ static struct file_system_type ext4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext4" );static int __init ext4_init_feat_adverts (void ) {fs/f2fs/super.c +1
-0@@ -639,6 +639,7 @@ static struct file_system_type f2fs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("f2fs" );static int __init init_inodecache (void ) {fs/fat/namei_msdos.c +1
-0@@ -668,6 +668,7 @@ static struct file_system_type msdos_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("msdos" );static int __init init_msdos_fs (void ) {fs/fat/namei_vfat.c +1
-0@@ -1073,6 +1073,7 @@ static struct file_system_type vfat_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vfat" );static int __init init_vfat_fs (void ) {fs/filesystems.c +1
-1@@ -274,7 +274,7 @@ struct file_system_type *get_fs_type(const char *name) fs =__get_fs_type (name, len);if (!fs &&capable (CAP_SYS_ADMIN ) && - (request_module ("%.*s" , len, name) ==0 )) + (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {fs/freevxfs/vxfs_super.c +1
-1@@ -52,7 +52,6 @@ MODULE_AUTHOR("Christoph Hellwig");MODULE_DESCRIPTION ("Veritas Filesystem (VxFS) driver" );MODULE_LICENSE ("Dual BSD/GPL" ); -MODULE_ALIAS ("vxfs" );/* makes mount -t vxfs autoload the module */ static void vxfs_put_super (struct super_block *);@@ -258,6 +257,7 @@ static struct file_system_type vxfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vxfs" );/* makes mount -t vxfs autoload the module */ static int __init vxfs_init (void )fs/fuse/control.c +1
-0@@ -341,6 +341,7 @@ static struct file_system_type fuse_ctl_fs_type = { .mount = fuse_ctl_mount, .kill_sb = fuse_ctl_kill_sb, }; +MODULE_ALIAS_FS ("fusectl" );int __init fuse_ctl_init (void ) {fs/fuse/inode.c +2
-0@@ -1127,6 +1127,7 @@ static struct file_system_type fuse_fs_type = { .mount = fuse_mount, .kill_sb = fuse_kill_sb_anon, }; +MODULE_ALIAS_FS ("fuse" );#ifdef CONFIG_BLOCK static struct dentry *fuse_mount_blk(struct file_system_type *fs_type,@@ -1156,6 +1157,7 @@ static struct file_system_type fuseblk_fs_type = { .kill_sb = fuse_kill_sb_blk, .fs_flags = FS_REQUIRES_DEV | FS_HAS_SUBTYPE, }; +MODULE_ALIAS_FS ("fuseblk" );static inline int register_fuseblk (void ) {fs/gfs2/ops_fstype.c +3
-1@@ -20,6 +20,7 @@#include <linux/gfs2_ondisk.h> #include <linux/quotaops.h> #include <linux/lockdep.h> +#include <linux/module.h> #include "gfs2.h" #include "incore.h" @@ -1425,6 +1426,7 @@ struct file_system_type gfs2_fs_type = { .kill_sb = gfs2_kill_sb, .owner = THIS_MODULE, }; +MODULE_ALIAS_FS ("gfs2" );struct file_system_type gfs2meta_fs_type = { .name ="gfs2meta" ,@@ -1432,4 +1434,4 @@ struct file_system_type gfs2meta_fs_type = { .mount = gfs2_mount_meta, .owner = THIS_MODULE, }; - +MODULE_ALIAS_FS ("gfs2meta" );fs/hfs/super.c +1
-0@@ -466,6 +466,7 @@ static struct file_system_type hfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfs" );static void hfs_init_once (void *p) {fs/hfsplus/super.c +1
-0@@ -618,6 +618,7 @@ static struct file_system_type hfsplus_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfsplus" );static void hfsplus_init_once (void *p) {fs/hppfs/hppfs.c +1
-0@@ -748,6 +748,7 @@ static struct file_system_type hppfs_type = { .kill_sb = kill_anon_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("hppfs" );static int __init init_hppfs (void ) {fs/hugetlbfs/inode.c +1
-0@@ -896,6 +896,7 @@ static struct file_system_type hugetlbfs_fs_type = { .mount = hugetlbfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("hugetlbfs" );static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE ];fs/isofs/inode.c +1
-2@@ -1556,6 +1556,7 @@ static struct file_system_type iso9660_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("iso9660" );static int __init init_iso9660_fs (void ) {@@ -1593,5 +1594,3 @@ static void __exit exit_iso9660_fs(void) module_init(init_iso9660_fs) module_exit(exit_iso9660_fs) MODULE_LICENSE("GPL"); -/* Actual filesystem name is iso9660, as requested in filesystems.c */ -MODULE_ALIAS ("iso9660" );fs/jffs2/super.c +1
-0@@ -356,6 +356,7 @@ static struct file_system_type jffs2_fs_type = { .mount = jffs2_mount, .kill_sb = jffs2_kill_sb, }; +MODULE_ALIAS_FS ("jffs2" );static int __init init_jffs2_fs (void ) {fs/jfs/super.c +1
-0@@ -833,6 +833,7 @@ static struct file_system_type jfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("jfs" );static void init_once (void *foo) {fs/logfs/super.c +1
-0@@ -608,6 +608,7 @@ static struct file_system_type logfs_fs_type = { .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("logfs" );static int __init logfs_init (void ) {fs/minix/inode.c +1
-0@@ -660,6 +660,7 @@ static struct file_system_type minix_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("minix" );static int __init init_minix_fs (void ) {fs/ncpfs/inode.c +1
-0@@ -1051,6 +1051,7 @@ static struct file_system_type ncp_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("ncpfs" );static int __init init_ncp_fs (void ) {fs/nfs/super.c +2
-1@@ -293,6 +293,7 @@ struct file_system_type nfs_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs" );EXPORT_SYMBOL_GPL (nfs_fs_type);struct file_system_type nfs_xdev_fs_type = {@@ -332,6 +333,7 @@ struct file_system_type nfs4_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs4" );EXPORT_SYMBOL_GPL (nfs4_fs_type);static int __init register_nfs4_fs (void )@@ -2716,6 +2718,5 @@ module_param(send_implementation_id, ushort, 0644);MODULE_PARM_DESC (send_implementation_id,"Send implementation ID with NFSv4.1 exchange_id" );MODULE_PARM_DESC (nfs4_unique_id,"nfs_client_id4 uniquifier string" ); -MODULE_ALIAS ("nfs4" );#endif /* CONFIG_NFS_V4 */ fs/nfsd/nfsctl.c +1
-0@@ -1052,6 +1052,7 @@ static struct file_system_type nfsd_fs_type = { .mount = nfsd_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("nfsd" );#ifdef CONFIG_PROC_FS static int create_proc_exports_entry (void )fs/nilfs2/super.c +1
-0@@ -1361,6 +1361,7 @@ struct file_system_type nilfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("nilfs2" );static void nilfs_inode_init_once (void *obj) {fs/ntfs/super.c +1
-0@@ -3079,6 +3079,7 @@ static struct file_system_type ntfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ntfs" );/* Stable names for the slab caches. */ static const char ntfs_index_ctx_cache_name[] ="ntfs_index_ctx_cache" ;fs/ocfs2/dlmfs/dlmfs.c +1
-0@@ -640,6 +640,7 @@ static struct file_system_type dlmfs_fs_type = { .mount = dlmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ocfs2_dlmfs" );static int __init init_dlmfs_fs (void ) {fs/omfs/inode.c +1
-0@@ -572,6 +572,7 @@ static struct file_system_type omfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("omfs" );static int __init init_omfs_fs (void ) {fs/openpromfs/inode.c +1
-0@@ -432,6 +432,7 @@ static struct file_system_type openprom_fs_type = { .mount = openprom_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("openpromfs" );static void op_inode_init_once (void *data) {fs/qnx4/inode.c +1
-0@@ -412,6 +412,7 @@ static struct file_system_type qnx4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx4" );static int __init init_qnx4_fs (void ) {fs/qnx6/inode.c +1
-0@@ -672,6 +672,7 @@ static struct file_system_type qnx6_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx6" );static int __init init_qnx6_fs (void ) {fs/reiserfs/super.c +1
-0@@ -2434,6 +2434,7 @@ struct file_system_type reiserfs_fs_type = { .kill_sb = reiserfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("reiserfs" );MODULE_DESCRIPTION ("ReiserFS journaled filesystem" );MODULE_AUTHOR ("Hans Reiser <[email protected]>" );fs/romfs/super.c +1
-0@@ -599,6 +599,7 @@ static struct file_system_type romfs_fs_type = { .kill_sb = romfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("romfs" );/* * inode storage initialiser fs/sysv/super.c +2
-1@@ -545,6 +545,7 @@ static struct file_system_type sysv_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("sysv" );static struct file_system_type v7_fs_type = { .owner =THIS_MODULE ,@@ -553,6 +554,7 @@ static struct file_system_type v7_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("v7" );static int __init init_sysv_fs (void ) {@@ -586,5 +588,4 @@ static void __exit exit_sysv_fs(void) module_init(init_sysv_fs) module_exit(exit_sysv_fs) -MODULE_ALIAS("v7"); MODULE_LICENSE("GPL");fs/ubifs/super.c +1
-0@@ -2174,6 +2174,7 @@ static struct file_system_type ubifs_fs_type = { .mount = ubifs_mount, .kill_sb = kill_ubifs_super, }; +MODULE_ALIAS_FS ("ubifs" );/* * Inode slab cache constructor. fs/ufs/super.c +1
-0@@ -1500,6 +1500,7 @@ static struct file_system_type ufs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ufs" );static int __init init_ufs_fs (void ) {fs/xfs/xfs_super.c +1
-0@@ -1561,6 +1561,7 @@ static struct file_system_type xfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("xfs" ); STATIC int __initxfs_init_zones (void )include/linux/fs.h +2
-0@@ -1829,6 +1829,8 @@ struct file_system_type {struct lock_class_key i_mutex_dir_key; }; +#define MODULE_ALIAS_FS (NAME ) MODULE_ALIAS("fs-" NAME) +extern struct dentry *mount_ns (struct file_system_type *fs_type,int flags,void *data,int (*fill_super)(struct super_block *,void *,int )); extern struct dentry *mount_bdev(struct file_system_type *fs_type,net/sunrpc/rpc_pipe.c +1
-3@@ -1174,6 +1174,7 @@ static struct file_system_type rpc_pipe_fs_type = { .mount = rpc_mount, .kill_sb = rpc_kill_sb, }; +MODULE_ALIAS_FS ("rpc_pipefs" );static void init_once (void *foo)@@ -1218,6 +1219,3 @@ void unregister_rpc_pipefs(void)kmem_cache_destroy (rpc_inode_cachep);unregister_filesystem (&rpc_pipe_fs_type); } - -/* Make 'mount -t rpc_pipefs ...' autoload this module. */ -MODULE_ALIAS ("rpc_pipefs" );Signature · 1 line
1.7.5.4
Re: [RFC][PATCH] fs: Limit sys_mount to only loading filesystem modules.
In reply to Eric W. BiedermanQuoted 18 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing
/etc/modprobe.d/*.confwith blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.
Quoted 7 lines
This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
Signed-off-by "Eric W. Biederman" <[email protected]>
Signature · 2 lines
Kees Cook Chrome OS Security
Quoted 75 lines · Serge E. Hallyn
Quoted 19 lines · Kees Cook
Quoted 18 lines · Serge E. Hallyn
Quoted 16 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
Quoted 8 lines
Quoted 6 lines
So the concern is root in a child user namespace doing
mount -t randomfs <...>in which case
do_new_mount()checksns_capable(), notcapable(), before trying to load a module for randomfs.Well, not just randomfs. Any module that modprobe in the init ns can find.
right
Quoted 7 lines
Quoted 5 lines
As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the
GRKERNSEC_MODHARDENwon't be acceptable. At least Eric Paris is strongly against it.I'd be curious to hear the objections. It seems pretty nice to me to
Wait, sorry, I mis-spoke. The objection would have been to requiring
CAP_SYS_MODULE, which is different. Sorry!Quoted 3 lines
add a new argument to every
request_module()that specifies the "subsystem" it expects a module to load from. Maybe pass "request_module=filesystem" or "...=netdev" to the modprobe call. AndThat would be useful for adding to the separation of privileges, i.e. helping contain the leaking of posix caps. It sounds good to me.
Quoted 10 lines
then in
init_module(), check the userargs for which subsystem was requested and look up in a table for the entry point module symbol for that subsystem to require. e.g. for "request_module=filesystem", require that the module contains the "register_filesystem" symbol, etc.Quoted 3 lines
But how about if we
add a check for 'current_user_ns()==&init_user_ns' at that place instead?Quoted 9 lines
Well, we'd need to mostly revert
57eccb830f1cc93d4b506ba306d8dfa685e0c88f("mount: consolidate permission checks") sinceget_fs_type()is being called beforemay_mount()right now. (And then, as you suggest, we should strengthen the test.) I think this will require either more plumbing intoget_fs_type(something like "boolload_module_if_missing") or the subsystem verification stuff inrequest_module. I think the latter is MUCH nicer as it covers this problem in all places, not just this "mount" case.My first instinct was to say I'd like to have the kernel 100% belonging to the
init_user_ns, with child user namespaces having zero ability to induce loading of any kernel modules, period. So a check for current being ininit_user_nsatrequest_moduleitself.
Quoted 6 lines
However (thinking more) that seems maybe wrong. You don't need privs to induce the loading of a new binfmt module right? The host's
/lib/modulesand module blacklists should be set up right by the admin (or distro)... If we require that the host admin manually modprobe every module which a task in a child user namespace might need, that goes counter to the goal of kernel modules.
Several subsystems already have an implicit subsystem restriction because they load with aliases. (e.g. binfmt-XXXX, net-pf=NNN, snd-card-NNN, FOO-iosched, etc). This isn't the case for filesystems and a few others, unfortunately:
$ git grep 'request_module("%.*s"' | grep -vi prefix
crypto/api.c: request_module("%s", name);
drivers/mtd/chips/chipreg.c: if (!drv && !request_module("%s", name))
drivers/mtd/mtdpart.c: if (!parser && !request_module("%s", *types))
drivers/net/wireless/iwlwifi/iwl-drv.c: request_module("%s", op->name);
drivers/staging/rtl8192e/rtllib_wx.c: request_module("%s", tempbuf);
fs/filesystems.c: if (!fs && (request_module("%.*s", len, name) == 0))
net/core/dev_ioctl.c: if (!request_module("%s", name))Several of these come from hardcoded values, though (e.g. crypto, chipreg).
Quoted 3 lines
Quoted 1 line
Quoted 1 line
Eric Biederman, do you have any objections to that?
-serge
-Kees
--
Kees Cook
Chrome OS Security
Quoted 11 lines · Kees Cook
Several subsystems already have an implicit subsystem restriction because they load with aliases. (e.g. binfmt-XXXX, net-pf=NNN, snd-card-NNN, FOO-iosched, etc). This isn't the case for filesystems and a few others, unfortunately:
$ git grep 'request_module("%.*s"' | grep -vi prefix crypto/api.c: request_module("%s", name);[...]
Several of these come from hardcoded values, though (e.g. crypto, chipreg).
Well, crypto does not. Try the code snippet below on a system with CONFIG_CRYPTO_USER_API=y. It'll abuse the above request_module() call to load any module the user requests -- iregardless of being contained in a user ns or not.
---8<---
/* Loading arbitrary modules using crypto api since v2.6.38
*
* - minipli
*/
#include <linux/if_alg.h>
#include <sys/socket.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#ifndef AF_ALG
#define AF_ALG 38
#endif
int main(int argc, char **argv) {
struct sockaddr_alg sa_alg = {
.salg_family = AF_ALG,
.salg_type = "hash",
};
int sock;
if (argc != 2) {
printf("usage: %s MODULE_NAME\n", argv[0]);
exit(1);
}
sock = socket(AF_ALG, SOCK_SEQPACKET, 0);
if (sock < 0) {
perror("socket(AF_ALG)");
exit(1);
}
strncpy((char *) sa_alg.salg_name, argv[1], sizeof(sa_alg.salg_name));
bind(sock, (struct sockaddr *) &sa_alg, sizeof(sa_alg));
close(sock);
return 0;
}--->8---
If people care about unprivileged users not being able to load arbitrary modules, could someone please fix this in crypto API, then? Herbert?
Thanks,
Mathias
Quoted 9 lines · Mathias Krause
Quoted 8 lines · Kees Cook
Several subsystems already have an implicit subsystem restriction because they load with aliases. (e.g. binfmt-XXXX, net-pf=NNN, snd-card-NNN, FOO-iosched, etc). This isn't the case for filesystems and a few others, unfortunately:
$ git grep 'request_module("%.*s"' | grep -vi prefix crypto/api.c: request_module("%s", name);
Quoted 8 lines
Quoted 3 lines
[...]
Several of these come from hardcoded values, though (e.g. crypto, chipreg).
Well, crypto does not. Try the code snippet below on a system with
CONFIG_CRYPTO_USER_API=y. It'll abuse the aboverequest_module()call to load any module the user requests -- iregardless of being contained in a user ns or not.
Oh ew. Yeah, I must have missed the path through the user api. Arg.
Quoted 41 lines
---8<---
/* Loading arbitrary modules using crypto api since v2.6.38
*
* - minipli
*/
#include <linux/if_alg.h>
#include <sys/socket.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>#ifndef
AF_ALG
#defineAF_ALG38
#endifint main(int argc, char **argv) { struct sockaddr_alg sa_alg = { .salg_family = AF_ALG, .salg_type = "hash", }; int sock; if (argc != 2) { printf("usage: %s MODULE_NAME\n", argv[0]); exit(1); } sock = socket(AF_ALG, SOCK_SEQPACKET, 0); if (sock < 0) { perror("socket(AF_ALG)"); exit(1); } strncpy((char *) sa_alg.salg_name, argv[1], sizeof(sa_alg.salg_name)); bind(sock, (struct sockaddr *) &sa_alg, sizeof(sa_alg)); close(sock); return 0; }
Quoted 4 lines
--->8---
If people care about unprivileged users not being able to load arbitrary modules, could someone please fix this in crypto API, then? Herbert?
So, should this get a prefix too? Maybe we need to change the request_module primitive to request_module(prefix, fmt, args) to stop these request_module("%s", name) things from continuing to exist...
-Kees
Signature · 2 lines
Kees Cook Chrome OS Security
Quoted 12 lines · Kees Cook
Quoted 11 lines · Mathias Krause
Quoted 10 lines · Kees Cook
Several subsystems already have an implicit subsystem restriction because they load with aliases. (e.g. binfmt-XXXX, net-pf=NNN, snd-card-NNN, FOO-iosched, etc). This isn't the case for filesystems and a few others, unfortunately:
$ git grep 'request_module("%.*s"' | grep -vi prefix crypto/api.c: request_module("%s", name);[...]
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Several of these come from hardcoded values, though (e.g. crypto, chipreg).
Well, crypto does not. Try the code snippet below on a system with
CONFIG_CRYPTO_USER_API=y. It'll abuse the aboverequest_module()call to load any module the user requests -- iregardless of being contained in a user ns or not.Oh ew. Yeah, I must have missed the path through the user api. Arg.
I will let someone else write the patch that adds the module aliases to crypto.
It seems worth doing even outside of any security concerns as it just makes the reqest to modprobe make more sense, and allows the existing modprobe policy controls to work.
Whereas an ill-formed string just doesn't tell modprobe enough to really act intelligently.
Quoted 43 lines
Quoted 43 lines
---8<---
/* Loading arbitrary modules using crypto api since v2.6.38
*
* - minipli
*/
#include <linux/if_alg.h>
#include <sys/socket.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>#ifndef
AF_ALG
#defineAF_ALG38
#endifint main(int argc, char **argv) { struct sockaddr_alg sa_alg = { .salg_family = AF_ALG, .salg_type = "hash", }; int sock; if (argc != 2) { printf("usage: %s MODULE_NAME\n", argv[0]); exit(1); } sock = socket(AF_ALG, SOCK_SEQPACKET, 0); if (sock < 0) { perror("socket(AF_ALG)"); exit(1); } strncpy((char *) sa_alg.salg_name, argv[1], sizeof(sa_alg.salg_name)); bind(sock, (struct sockaddr *) &sa_alg, sizeof(sa_alg)); close(sock); return 0; }--->8---
Quoted 6 lines
Quoted 2 lines
If people care about unprivileged users not being able to load arbitrary modules, could someone please fix this in crypto API, then? Herbert?
So, should this get a prefix too? Maybe we need to change the
request_moduleprimitive torequest_module(prefix, fmt, args)to stop theserequest_module("%s", name)things from continuing to exist...
Something like the patch below?
kernel/kmod.c +4 -0
@@ -131,6 +131,10 @@ int __request_module(bool wait, const char *fmt, ...)#define MAX_KMOD_CONCURRENT 50/* Completely arbitrary value - KAO */ static int kmod_loop_msg; +/* Require that calls to request module have a little structure */ +if (fmt[0 ] =='%' ) +return -EINVAL ; +/* * We don't allow synchronous module loading from async. Module * init may invoke async_synchronize_full() which will end up
Quoted 90 lines · Eric W. Biederman
Quoted 12 lines · Kees Cook
Quoted 11 lines · Mathias Krause
Quoted 10 lines · Kees Cook
Several subsystems already have an implicit subsystem restriction because they load with aliases. (e.g. binfmt-XXXX, net-pf=NNN, snd-card-NNN, FOO-iosched, etc). This isn't the case for filesystems and a few others, unfortunately:
$ git grep 'request_module("%.*s"' | grep -vi prefix crypto/api.c: request_module("%s", name);[...]
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Several of these come from hardcoded values, though (e.g. crypto, chipreg).
Well, crypto does not. Try the code snippet below on a system with
CONFIG_CRYPTO_USER_API=y. It'll abuse the aboverequest_module()call to load any module the user requests -- iregardless of being contained in a user ns or not.Oh ew. Yeah, I must have missed the path through the user api. Arg.
I will let someone else write the patch that adds the module aliases to crypto.
It seems worth doing even outside of any security concerns as it just makes the reqest to modprobe make more sense, and allows the existing modprobe policy controls to work.
Whereas an ill-formed string just doesn't tell modprobe enough to really act intelligently.
Quoted 43 lines
Quoted 43 lines
---8<---
/* Loading arbitrary modules using crypto api since v2.6.38
*
* - minipli
*/
#include <linux/if_alg.h>
#include <sys/socket.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>#ifndef
AF_ALG
#defineAF_ALG38
#endifint main(int argc, char **argv) { struct sockaddr_alg sa_alg = { .salg_family = AF_ALG, .salg_type = "hash", }; int sock; if (argc != 2) { printf("usage: %s MODULE_NAME\n", argv[0]); exit(1); } sock = socket(AF_ALG, SOCK_SEQPACKET, 0); if (sock < 0) { perror("socket(AF_ALG)"); exit(1); } strncpy((char *) sa_alg.salg_name, argv[1], sizeof(sa_alg.salg_name)); bind(sock, (struct sockaddr *) &sa_alg, sizeof(sa_alg)); close(sock); return 0; }--->8---
Quoted 6 lines
Quoted 2 lines
If people care about unprivileged users not being able to load arbitrary modules, could someone please fix this in crypto API, then? Herbert?
So, should this get a prefix too? Maybe we need to change the
request_moduleprimitive torequest_module(prefix, fmt, args)to stop theserequest_module("%s", name)things from continuing to exist...Something like the patch below?
kernel/kmod.c +0
-0
kernel/kmod.c int __request_module(bool wait, const char *fmt, ...) · line 140
Quoted 10 lines
kernel/kmod.c +4
-0@@ -131,6 +131,10 @@ int __request_module(bool wait, const char *fmt, ...)#define MAX_KMOD_CONCURRENT 50/* Completely arbitrary value - KAO */ static int kmod_loop_msg; +/* Require that calls to request module have a little structure */ +if (fmt[0 ] =='%' ) +return -EINVAL ; +/* * We don't allow synchronous module loading from async. Module * init may invoke async_synchronize_full() which will end up
Something like that, but that'll break some things that do stuff like %s-suffix.
-Kees
Signature · 2 lines
Kees Cook Chrome OS Security
Quoted 8 lines · Serge E. Hallyn
Quoted 7 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).
CAP_SETUID is not needed.
Quoted 13 lines
Quoted 8 lines
It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
So the concern is root in a child user namespace doing
mount -t randomfs <...>
Quoted 8 lines
in which case
do_new_mount()checksns_capable(), notcapable(), before trying to load a module for randomfs.As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the
GRKERNSEC_MODHARDENwon't be acceptable. At least Eric Paris is strongly against it.
What is wrong with GRKERNSEC_MODHARDEN? It took a quick look and the code is far from clean. But that would not be a fundamental objection from keeping code like that out of the kernel.
It is also entertaining to read security code that won't even build with CONFIG_UIDGID_STRICT_TYPE_CHECKS enabled.
Quoted 5 lines
But how about if we
add a check for 'current_user_ns()==&init_user_ns' at that place instead?Eric Biederman, do you have any objections to that?
The obvious solution here is to test for CAP_SYS_ADMIN rather than current_user_ns == &init_user_ns before we request the module here. As that is what was previously required on this path.
Reading the comments the concerns are.
- Non-root users are allowed to load obscure and possibly kernel modules.
- get_fs_type can trigger the load of any kernel module.
At a practical level I don't see adding a capalbe(CAP_SYS_ADMIN) check as having much effect for the functionality currently present in user namespaces today as the filesystems that an legal to mount in a user namespace (ramfs, tmpfs, mqueuefs, sysfs, proc, devpts) are so common most of them can not even be built as modules and even if they are modules the modules will already be loaded. So I will see about adding a capable(CAP_SYS_ADMIN) check to shore things up for the short term.
In the longer term I very much would like to get loopback devices and mounts of filesystems on those loopback devices working, and being able to mount filesystems from usb sticks that people commonly plug in, and remove the need for privileged daemons to do that work. At that point manually having to do something that was automatic before will either mean a regression in functionality or bugs as people manually load things.
So I am wondering what I a good policy should be. Should we trust kernel modules to not be buggy (especially if they were signed as part of the build process)? Do we add some defense in depth and add filesystem registration magic? Thinking...
We can limit the request_module in get_fs_type to just filesystems fairly easily.
In include/linux/fs.h:
#define MODULE_ALIAS_FS(type) MODULE_ALIAS("fs-" __stringify(type))In fs/filesystems.c:
if (request_moudle("fs-%.*s", len, name) == 0)Then just add the appropriate MODULE_ALIAS_FS lines in all of the filesystems. This also allows user space to say set the module loading policy for filesystems using the blacklist and the alias keywords in /etc/modprobe.d/*.conf.
That seems a whole lot simpler, more powerful and more maintainable than what little I saw in GRKERNSEC_MODHARDEN to prevent loading of non-filesystem modules from get_fs_type.
Eric
p.s. This is the patch I am looking at pushing to Linus in the near future.
fs/filesystems.c +2 -1
@@ -273,7 +273,8 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs &&capable (CAP_SYS_ADMIN ) && + (request_module ("%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {
Quoted 7 lines · Eric W. Biederman
Quoted 5 lines · Serge E. Hallyn
Quoted 4 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
Quoted 5 lines
Quoted 3 lines
Quoted 3 lines
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).
CAP_SETUIDis not needed.
Do you have an example? I wasn't able to gain capabilities within the userns until I had a uid map set that allowed my uid to map to root. (I needed CAP_SYS_ADMIN in the mnt ns to get past may_mount() before I could touch get_fs_type().)
Quoted 35 lines
Quoted 13 lines
Quoted 8 lines
It might be worth examining
GRKERNSEC_MODHARDENin grsecurity, which examines module symbols to verify thatrequest_module()for a filesystem only loads a module that defines "register_filesystem" (among other things).-Kees
[1] https://twitter.com/grsecurity/status/307473816672665600
So the concern is root in a child user namespace doing
mount -t randomfs <...>Quoted 8 lines
in which case
do_new_mount()checksns_capable(), notcapable(), before trying to load a module for randomfs.As well as (secondly) the fact that there is no enforcement on the format of the module names (i.e. fs-*).
Kees, from what I've seen the
GRKERNSEC_MODHARDENwon't be acceptable. At least Eric Paris is strongly against it.What is wrong with
GRKERNSEC_MODHARDEN? It took a quick look and the code is far from clean. But that would not be a fundamental objection from keeping code like that out of the kernel.It is also entertaining to read security code that won't even build with
CONFIG_UIDGID_STRICT_TYPE_CHECKSenabled.Quoted 5 lines
But how about if we
add a check for 'current_user_ns()==&init_user_ns' at that place instead?Eric Biederman, do you have any objections to that?
Quoted 8 lines
The obvious solution here is to test for
CAP_SYS_ADMINrather thancurrent_user_ns==&init_user_nsbefore we request the module here. As that is what was previously required on this path.Reading the comments the concerns are.
- Non-root users are allowed to load obscure and possibly kernel modules.
-get_fs_typecan trigger the load of any kernel module.
Yes, though I think you meant "... possibly vulnerable kernel modules ..." There has been a history of weird stuff living in kernel modules that gets built by distros. If we can raise the bar on being able to force the kernel to load those things, it'd be nice. As mentioned in my other email, this is basically one of the few remaining places were arbitrary module names can get loaded.
Quoted 7 lines
At a practical level I don't see adding a
capalbe(CAP_SYS_ADMIN)check as having much effect for the functionality currently present in user namespaces today as the filesystems that an legal to mount in a user namespace (ramfs, tmpfs, mqueuefs, sysfs, proc, devpts) are so common most of them can not even be built as modules and even if they are modules the modules will already be loaded. So I will see about adding acapable(CAP_SYS_ADMIN)check to shore things up for the short term.
For the short-term, yes, this solves the "regression". But it obviously isn't what is desired for userns in the long term.
Quoted 9 lines
In the longer term I very much would like to get loopback devices and mounts of filesystems on those loopback devices working, and being able to mount filesystems from usb sticks that people commonly plug in, and remove the need for privileged daemons to do that work. At that point manually having to do something that was automatic before will either mean a regression in functionality or bugs as people manually load things.
Quoted 4 lines
So I am wondering what I a good policy should be. Should we trust kernel modules to not be buggy (especially if they were signed as part of the build process)? Do we add some defense in depth and add filesystem registration magic? Thinking...
If we can produce a mechanism that provides some defensive design, we should do it. There will always be bugs, so we should always try to make them harder to get to or harder to exploit. Reducing the available attack surface to "just filesystems" would be a win here, and it would be consistent with the _intent_ of existing kernel code.
Quoted 7 lines
We can limit the
request_moduleinget_fs_typeto just filesystems fairly easily.In
include/linux/fs.h:#define MODULE_ALIAS_FS(type) MODULE_ALIAS("fs-" __stringify(type))
Quoted 8 lines
In
fs/filesystems.c:if (request_moudle("fs-%.*s", len, name) == 0)Then just add the appropriate
MODULE_ALIAS_FSlines in all of the filesystems. This also allows user space to say set the module loading policy for filesystems using the blacklist and the alias keywords in/etc/modprobe.d/*.conf.
This was the solution for netdev. The backward compat situation is this:
if (no_module && capable(CAP_NET_ADMIN))
no_module = request_module("netdev-%s", name);
if (no_module && capable(CAP_SYS_MODULE)) {
if (!request_module("%s", name))
pr_warn("Loading kernel module for a network
device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias
netdev-%s instead.\n",
name);
}Those aren't ns_capable, though, so right now userns can't trigger loading new network drivers via ifconfig, but this seems like a reasonable approach to take.
Quoted 14 lines
That seems a whole lot simpler, more powerful and more maintainable than what little I saw in
GRKERNSEC_MODHARDENto prevent loading of non-filesystem modules fromget_fs_type.Eric
p.s. This is the patch I am looking at pushing to Linus in the near future.
fs/filesystems.c +0
-0
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 280
Quoted 9 lines
fs/filesystems.c +2
-1@@ -273,7 +273,8 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs &&capable (CAP_SYS_ADMIN ) && + (request_module ("%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {
Will this break other users of get_fs_type()? I think it's okay; it looks like all the other users are expected to be privileged.
Should this look like netdev in the future? Something like this, after making may_mount() available to fs/filesystems.c:
- if (!fs && (request_module("%.*s", len, name) == 0))
+ if (!fs && ((may_mount() && request_module("fs-%.*s", len,
name) == 0) ||
+ (capable(CAP_SYS_ADMIN) && request_module("%.*s",
len, name) == 0)))
and adding all the filesystem module aliases.
At some point maybe we can add a flag to kill the old unqualified paths available to CAP_SYS_ADMIN for netdev and mount...
-Kees
--
Kees Cook
Chrome OS Security
Quoted 12 lines · Kees Cook
Quoted 11 lines · Eric W. Biederman
Quoted 8 lines · Serge E. Hallyn
Quoted 7 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).
Quoted 6 lines
Quoted 1 line
CAP_SETUIDis not needed.Do you have an example? I wasn't able to gain capabilities within the userns until I had a uid map set that allowed my uid to map to root. (I needed
CAP_SYS_ADMINin the mnt ns to get pastmay_mount()before I could touchget_fs_type().)
Likely because exec drops capbilities if your uid isn't 0.
Without CAP_SETUID you can still map uid to 0 to your current uid.
The following shell script is an easy basis for testing and playing around with things. It does assume a recent version of unshare from util-linux.
#!/bin/sh
export IFIFO=/tmp/userns-test-$$-in
export OFIFO=/tmp/userns-test-$$-out
rm -f $IFIFO $OFIFO
mkfifo $IFIFO
mkfifo $OFIFO
unshare --user -- /bin/bash -s <<'EOF' &
echo waiting-for-uid-and-gid-maps > $OFIFO read LINE < $IFIFO
exec unshare --mount --net -- /bin/sh -s <<'EOF2' mount --bind $HOME /root/
mount --bind /dev/null /dev/log
# Start a shell to keep the namespace reference alive $SHELL -i < /dev/tty > /dev/tty 2> /dev/tty EOF2
EOF
child=$!
read LINE < $OFIFO
uid=$(id --user)
gid=$(id --group)
echo "0 $uid 1" > /proc/$child/uid_map
echo "0 $gid 1" > /proc/$child/gid_map
echo uid-and-gid-maps > $IFIFO
wait $child
Quoted 17 lines
Quoted 7 lines
At a practical level I don't see adding a
capalbe(CAP_SYS_ADMIN)check as having much effect for the functionality currently present in user namespaces today as the filesystems that an legal to mount in a user namespace (ramfs, tmpfs, mqueuefs, sysfs, proc, devpts) are so common most of them can not even be built as modules and even if they are modules the modules will already be loaded. So I will see about adding acapable(CAP_SYS_ADMIN)check to shore things up for the short term.For the short-term, yes, this solves the "regression". But it obviously isn't what is desired for userns in the long term.
Quoted 4 lines
So I am wondering what I a good policy should be. Should we trust kernel modules to not be buggy (especially if they were signed as part of the build process)? Do we add some defense in depth and add filesystem registration magic? Thinking...
Quoted 5 lines
If we can produce a mechanism that provides some defensive design, we should do it. There will always be bugs, so we should always try to make them harder to get to or harder to exploit. Reducing the available attack surface to "just filesystems" would be a win here, and it would be consistent with the _intent_ of existing kernel code.
Agreed.
Quoted 18 lines
Quoted 7 lines
We can limit the
request_moduleinget_fs_typeto just filesystems fairly easily.In
include/linux/fs.h:#define MODULE_ALIAS_FS(type) MODULE_ALIAS("fs-" __stringify(type))Quoted 8 lines
In
fs/filesystems.c:if (request_moudle("fs-%.*s", len, name) == 0)Then just add the appropriate
MODULE_ALIAS_FSlines in all of the filesystems. This also allows user space to say set the module loading policy for filesystems using the blacklist and the alias keywords in/etc/modprobe.d/*.conf.This was the solution for netdev. The backward compat situation is this:
Quoted 9 lines
if (no_module && capable(CAP_NET_ADMIN))
no_module = request_module("netdev-%s", name);
if (no_module && capable(CAP_SYS_MODULE)) {
if (!request_module("%s", name))
pr_warn("Loading kernel module for a network
device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias
netdev-%s instead.\n",
name);
}If I have read it right the backwards compat case is there to be compatibile with existing peoples configuration who had created specific aliases like eth0 for specific network devices in /etc/modprobe.d/*.conf. Where the new netdev-* would break userspace.
I don't believe anyone does anything like that for filesystems, and even if they tried it wouldn't work because the registered filesystem type would have a different name. So I don't see needing to worry about that kind of backwards compatibility for filesystems.
Quoted 3 lines
Those aren't
ns_capable, though, so right now userns can't trigger loading new network drivers via ifconfig, but this seems like a reasonable approach to take.
Yes. I saw those and deliberately didn't make those ns_capable, to be on the safe side when I made the rest of the networking ns_capable. That and ns_capabable is not meaningful as a check for loading modules. The only interesting question is can the global root user trigger them or can everyone trigger module loading.
Quoted 22 lines
Quoted 22 lines
That seems a whole lot simpler, more powerful and more maintainable than what little I saw in
GRKERNSEC_MODHARDENto prevent loading of non-filesystem modules fromget_fs_type.Eric
p.s. This is the patch I am looking at pushing to Linus in the near future.
fs/filesystems.c +2
-1@@ -273,7 +273,8 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs &&capable (CAP_SYS_ADMIN ) && + (request_module ("%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);
Quoted 4 lines
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 280
Quoted 1 line
fs/filesystems.c +0
-0if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {Will this break other users of
get_fs_type()? I think it's okay; it looks like all the other users are expected to be privileged.
I read through them all, and I didn't see problems. The most concerning one is tomoyo's use of get_fs_type before capabilities are checked in do_mount. But that is neither here nor there.
Quoted 9 lines
Should this look like netdev in the future? Something like this, after making
may_mount()available tofs/filesystems.c:(quoted diff) +1
-1- if (!fs && (request_module("%.*s", len, name) == 0)) + if (!fs && ((may_mount() && request_module("fs-%.*s", len,name) == 0) ||
+ (capable(CAP_SYS_ADMIN)&& request_module("%.*s",
len, name) == 0)))
Quoted 4 lines
and adding all the filesystem module aliases.
At some point maybe we can add a flag to kill the old unqualified paths available to
CAP_SYS_ADMINfor netdev and mount...
For netdev at least I don't see it as being particularly interesting. The case for tunnels is already as unprivileged as you can reasonably get with "request_module("rtnl-link-%s", kind);" in net/core/rtnetlink.c:rtnl_newlink(). For real physical devices there is both greater chance of a buggy module and no realy need as udev will load the module based on hardware auto-discovery.
This leads to the fundamental question: Should we require privilege to request the load filesystem modules?
I have looked at GRKERNSEC_MODHARDEN to see if that could give me some guidance. Unfortunately GRKERNSEC_MODHARDEN takes the position that fs kernel modules are the only kernel modules that should ever auto-load and only in very specific situations. So I can't see that being the normal kernel policy, especially since there are the sysctls /proc/sys/kernel/modprobe and /proc/sys/kernel/modules_disabled.
Overall the basic policy building blocks for controlling which modules are loaded seem solid and in use. So I don't see any particular reason why the kernel's default policy should not be to allow any users actions to request modules.
So I think I am going to scrap the change sitting in my development tree to require capalbe(CAP_SYS_ADMIN) to load a filesystem module and just go with my request_module("fs-%.*",...); change. That is simple and seems to match the rest of the kernel.
Does anyone see a reason why we should need CAP_SYS_ADMIN or be in the initial user namespace to trigger a request of filesystem modules?
Eric
Quoted 44 lines · Eric W. Biederman
Quoted 12 lines · Kees Cook
Quoted 11 lines · Eric W. Biederman
Quoted 8 lines · Serge E. Hallyn
Quoted 7 lines · Kees Cook
The rearranging done for user ns has resulted in allowing arbitrary kernel module loading[1] (i.e. re-introducing a form of CVE-2011-1019) by what is assumed to be an unprivileged process.
At present, it does look to require at least
CAP_SETUIDalong the way to set up the uidmap (but things like the setuid helper newuidmap might soon start providing such a thing by default).Quoted 6 lines
Quoted 1 line
CAP_SETUIDis not needed.Do you have an example? I wasn't able to gain capabilities within the userns until I had a uid map set that allowed my uid to map to root. (I needed
CAP_SYS_ADMINin the mnt ns to get pastmay_mount()before I could touchget_fs_type().)Likely because exec drops capbilities if your uid isn't 0.
Without
CAP_SETUIDyou can still map uid to 0 to your current uid.The following shell script is an easy basis for testing and playing around with things. It does assume a recent version of unshare from util-linux.
#!/bin/sh
export IFIFO=/tmp/userns-test-$$-in
export OFIFO=/tmp/userns-test-$$-out
rm -f $IFIFO $OFIFO
mkfifo $IFIFO
mkfifo $OFIFO
unshare --user --/bin/bash-s <<'EOF' &
echo waiting-for-uid-and-gid-maps > $OFIFO read LINE < $IFIFO
exec unshare --mount --net --/bin/sh-s <<'EOF2' mount --bind $HOME/root/
mount --bind/dev/null/dev/log
# Start a shell to keep the namespace reference alive $SHELL -i </dev/tty>/dev/tty2>/dev/ttyEOF2
EOF
Quoted 8 lines
child=$!
read LINE < $OFIFO
uid=$(id --user)
gid=$(id --group)
echo "0 $uid 1" >/proc/$child/uid_map
echo "0 $gid 1" >/proc/$child/gid_map
echo uid-and-gid-maps > $IFIFO
wait $child
Ah-ha, thanks! Yes, that worked great. I think map_write()'s cap_valid/ns_capable calls confused me. :)
Quoted 58 lines
Quoted 17 lines
Quoted 7 lines
At a practical level I don't see adding a
capalbe(CAP_SYS_ADMIN)check as having much effect for the functionality currently present in user namespaces today as the filesystems that an legal to mount in a user namespace (ramfs, tmpfs, mqueuefs, sysfs, proc, devpts) are so common most of them can not even be built as modules and even if they are modules the modules will already be loaded. So I will see about adding acapable(CAP_SYS_ADMIN)check to shore things up for the short term.For the short-term, yes, this solves the "regression". But it obviously isn't what is desired for userns in the long term.
Quoted 4 lines
So I am wondering what I a good policy should be. Should we trust kernel modules to not be buggy (especially if they were signed as part of the build process)? Do we add some defense in depth and add filesystem registration magic? Thinking...
Quoted 5 lines
If we can produce a mechanism that provides some defensive design, we should do it. There will always be bugs, so we should always try to make them harder to get to or harder to exploit. Reducing the available attack surface to "just filesystems" would be a win here, and it would be consistent with the _intent_ of existing kernel code.
Agreed.
Quoted 18 lines
Quoted 7 lines
We can limit the
request_moduleinget_fs_typeto just filesystems fairly easily.In
include/linux/fs.h:#define MODULE_ALIAS_FS(type) MODULE_ALIAS("fs-" __stringify(type))Quoted 8 lines
In
fs/filesystems.c:if (request_moudle("fs-%.*s", len, name) == 0)Then just add the appropriate
MODULE_ALIAS_FSlines in all of the filesystems. This also allows user space to say set the module loading policy for filesystems using the blacklist and the alias keywords in/etc/modprobe.d/*.conf.This was the solution for netdev. The backward compat situation is this:
Quoted 9 lines
if (no_module && capable(CAP_NET_ADMIN)) no_module = request_module("netdev-%s", name); if (no_module && capable(CAP_SYS_MODULE)) { if (!request_module("%s", name)) pr_warn("Loading kernel module for a network device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias netdev-%s instead.\n", name); }If I have read it right the backwards compat case is there to be compatibile with existing peoples configuration who had created specific aliases like eth0 for specific network devices in
/etc/modprobe.d/*.conf. Where the new netdev-* would break userspace.
Quoted 4 lines
I don't believe anyone does anything like that for filesystems, and even if they tried it wouldn't work because the registered filesystem type would have a different name. So I don't see needing to worry about that kind of backwards compatibility for filesystems.
Yeah, on re-reading the thread that introduced the backward compat mode, I agree. There is no need for this with filesystems AFAICT.
Quoted 62 lines
Quoted 3 lines
Those aren't
ns_capable, though, so right now userns can't trigger loading new network drivers via ifconfig, but this seems like a reasonable approach to take.Yes. I saw those and deliberately didn't make those
ns_capable, to be on the safe side when I made the rest of the networkingns_capable. That andns_capabableis not meaningful as a check for loading modules. The only interesting question is can the global root user trigger them or can everyone trigger module loading.Quoted 22 lines
Quoted 22 lines
That seems a whole lot simpler, more powerful and more maintainable than what little I saw in
GRKERNSEC_MODHARDENto prevent loading of non-filesystem modules fromget_fs_type.Eric
p.s. This is the patch I am looking at pushing to Linus in the near future.
fs/filesystems.c +2
-1@@ -273,7 +273,8 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs &&capable (CAP_SYS_ADMIN ) && + (request_module ("%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);Quoted 4 lines
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 280
Quoted 1 line
fs/filesystems.c +0
-0if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {Will this break other users of
get_fs_type()? I think it's okay; it looks like all the other users are expected to be privileged.I read through them all, and I didn't see problems. The most concerning one is tomoyo's use of
get_fs_typebefore capabilities are checked indo_mount. But that is neither here nor there.Quoted 9 lines
Should this look like netdev in the future? Something like this, after making
may_mount()available tofs/filesystems.c:(quoted diff) +1
-1- if (!fs && (request_module("%.*s", len, name) == 0)) + if (!fs && ((may_mount() && request_module("fs-%.*s", len,name) == 0) ||
+ (capable(CAP_SYS_ADMIN)&& request_module("%.*s",
len, name) == 0)))Quoted 4 lines
and adding all the filesystem module aliases.
At some point maybe we can add a flag to kill the old unqualified paths available to
CAP_SYS_ADMINfor netdev and mount...For netdev at least I don't see it as being particularly interesting. The case for tunnels is already as unprivileged as you can reasonably get with "
request_module("rtnl-link-%s", kind);" in net/core/rtnetlink.c:rtnl_newlink(). For real physical devices there is both greater chance of a buggy module and no realy need as udev will load the module based on hardware auto-discovery.
Quoted 2 lines
This leads to the fundamental question: Should we require privilege to request the load filesystem modules?
I think that the past dictated the need for privilege due to it being tied to mounting. With userns, this is weakened, but it seems like the privilege should at least attempt to segregate caps to certain classes of modules.
Quoted 5 lines
I have looked at
GRKERNSEC_MODHARDENto see if that could give me some guidance. UnfortunatelyGRKERNSEC_MODHARDENtakes the position that fs kernel modules are the only kernel modules that should ever auto-load and only in very specific situations. So I can't see that being the normal kernel policy, especially since there are the sysctls
Right -- modharden's basic goal is to block all non-root autoloading. It had to work around some corner-cases (mount being setuid, etc).
Quoted 1 line
/proc/sys/kernel/modprobeand/proc/sys/kernel/modules_disabled.
Right -- this gives the granularity of "autoloading" and "loading" respectively, but there is no concept of "only privileged autoloading" in the current kernel.
It seems to me that unpriv users shouldn't be able to arbitrarily load kernel modules, but if userns continues in this direction, there will always be a path to doing autoloading for each different subsystem's modules, ultimately leading to unpriv loading. Still, I think it's worth creating obvious subsystem aliases so userspace can more easily blacklist/whitelist things.
Quoted 9 lines
Overall the basic policy building blocks for controlling which modules are loaded seem solid and in use. So I don't see any particular reason why the kernel's default policy should not be to allow any users actions to request modules.
So I think I am going to scrap the change sitting in my development tree to require
capalbe(CAP_SYS_ADMIN)to load a filesystem module and just go with myrequest_module("fs-%.*",...); change. That is simple and seems to match the rest of the kernel.
Agreed.
Quoted 4 lines
Does anyone see a reason why we should need
CAP_SYS_ADMINor be in the initial user namespace to trigger a request of filesystem modules?Eric
-Kees
Signature · 2 lines
Kees Cook Chrome OS Security
Quoted 4 lines · Kees Cook
On Sun, Mar 3, 2013 at 1:58 PM, Eric W. Biederman <[email protected]> wrote:
Ah-ha, thanks! Yes, that worked great. I think map_write()'s cap_valid/ns_capable calls confused me. :)
Yes permissions across user namespaces can be a little weird. But mostly if you are their creator, you are their all powerful god and they must watch out.
Quoted 7 lines
Quoted 7 lines
For netdev at least I don't see it as being particularly interesting. The case for tunnels is already as unprivileged as you can reasonably get with "
request_module("rtnl-link-%s", kind);" in net/core/rtnetlink.c:rtnl_newlink(). For real physical devices there is both greater chance of a buggy module and no realy need as udev will load the module based on hardware auto-discovery.
Quoted 7 lines
Quoted 2 lines
This leads to the fundamental question: Should we require privilege to request the load filesystem modules?
I think that the past dictated the need for privilege due to it being tied to mounting. With userns, this is weakened, but it seems like the privilege should at least attempt to segregate caps to certain classes of modules.
With filesystems in particular the attack surface is practically non-existent if the filesystem is not mounted, and FS_USERNS_MOUNT prevents most filesystems from being mounted. So right now I do not see any real danger in allowing filesystem modules to be auto-loaded.
Quoted 8 lines
Quoted 5 lines
I have looked at
GRKERNSEC_MODHARDENto see if that could give me some guidance. UnfortunatelyGRKERNSEC_MODHARDENtakes the position that fs kernel modules are the only kernel modules that should ever auto-load and only in very specific situations. So I can't see that being the normal kernel policy, especially since there are the sysctlsRight -- modharden's basic goal is to block all non-root autoloading. It had to work around some corner-cases (mount being setuid, etc).
The mail goal appears to be disable module auto-loading and to log a message.
Quoted 5 lines
Quoted 1 line
/proc/sys/kernel/modprobeand/proc/sys/kernel/modules_disabled.Right -- this gives the granularity of "autoloading" and "loading" respectively, but there is no concept of "only privileged autoloading" in the current kernel.
As of the 3.8 patch when a module is loaded it does:
+#ifdef CONFIG_GRKERNSEC_MODHARDEN
+ {
+ char *p, *p2;
+
+ if (strstr(mod->args, "grsec_modharden_netdev")) {
+ printk(KERN_ALERT "grsec: denied auto-loading kernel module for a network device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias netdev-%.64s instead.", mod->name);
+ err = -EPERM;
+ goto free_modinfo;
+ } else if ((p = strstr(mod->args, "grsec_modharden_normal"))) {
+ p += sizeof("grsec_modharden_normal") - 1;
+ p2 = strstr(p, "_");
+ if (p2) {
+ *p2 = '\0';
+ printk(KERN_ALERT "grsec: denied kernel module auto-load of %.64s by uid %.9s\n", mod->name, p);
+ *p2 = '_';
+ }
+ err = -EPERM;
+ goto free_modinfo;
+ }
+ }
+#endif
Which simplifies to
if (strstr(mod->args, "grsec_modharden_fs") != 0) {
err = -EPERM;
goto free_modinfo;
}So perhaps it tries but there is not really a concept of privileged processes being able to auto-load anything except fs modules.
In general what I see GRKERNSEC_MODHARDEN implementing is a policy of printing a logging a message instead of auto-loading any kernel modules.
Quoted 6 lines
It seems to me that unpriv users shouldn't be able to arbitrarily load kernel modules, but if userns continues in this direction, there will always be a path to doing autoloading for each different subsystem's modules, ultimately leading to unpriv loading. Still, I think it's worth creating obvious subsystem aliases so userspace can more easily blacklist/whitelist things.
But I don't see any problem with uprivileged users being able to cause the kernel to request kernel modules, as long as the request is clear enough that modprobe can implement the desired policy.
Quoted 11 lines
Quoted 9 lines
Overall the basic policy building blocks for controlling which modules are loaded seem solid and in use. So I don't see any particular reason why the kernel's default policy should not be to allow any users actions to request modules.
So I think I am going to scrap the change sitting in my development tree to require
capalbe(CAP_SYS_ADMIN)to load a filesystem module and just go with myrequest_module("fs-%.*",...); change. That is simple and seems to match the rest of the kernel.Agreed.
Eric
cover userns bug fixes for v3.9-rc2 for review
In reply to Kees CookBaring problems these are the changes I intend to put in linux-next and then send to Linus for v3.9-rc2.
The first is a trivial oops fix.
The second reworks how mount -t triggers module loading to make it harder to abuse.
Eric W. Biederman (2):
userns: Stop oopsing in key_change_session_keyring
fs: Limit sys_mount to only request filesystem modules.
arch/ia64/kernel/perfmon.c | 1 + arch/powerpc/platforms/cell/spufs/inode.c | 1 + arch/s390/hypfs/inode.c | 1 + drivers/firmware/efivars.c | 1 + drivers/infiniband/hw/ipath/ipath_fs.c | 1 + drivers/infiniband/hw/qib/qib_fs.c | 1 + drivers/misc/ibmasm/ibmasmfs.c | 1 + drivers/mtd/mtdchar.c | 1 + drivers/oprofile/oprofilefs.c | 1 + drivers/staging/ccg/f_fs.c | 1 + drivers/usb/gadget/f_fs.c | 1 + drivers/usb/gadget/inode.c | 1 + drivers/xen/xenfs/super.c | 1 + fs/9p/vfs_super.c | 1 + fs/adfs/super.c | 1 + fs/affs/super.c | 1 + fs/afs/super.c | 1 + fs/autofs4/init.c | 1 + fs/befs/linuxvfs.c | 1 + fs/bfs/inode.c | 1 + fs/binfmt_misc.c | 1 + fs/btrfs/super.c | 1 + fs/ceph/super.c | 1 + fs/coda/inode.c | 1 + fs/configfs/mount.c | 1 + fs/cramfs/inode.c | 1 + fs/debugfs/inode.c | 1 + fs/devpts/inode.c | 1 + fs/ecryptfs/main.c | 1 + fs/efs/super.c | 1 + fs/exofs/super.c | 1 + fs/ext2/super.c | 1 + fs/ext3/super.c | 1 + fs/ext4/super.c | 5 +++-- fs/f2fs/super.c | 1 + fs/fat/namei_msdos.c | 1 + fs/fat/namei_vfat.c | 1 + fs/filesystems.c | 2 +- fs/freevxfs/vxfs_super.c | 2 +- fs/fuse/control.c | 1 + fs/fuse/inode.c | 2 ++ fs/gfs2/ops_fstype.c | 4 +++- fs/hfs/super.c | 1 + fs/hfsplus/super.c | 1 + fs/hppfs/hppfs.c | 1 + fs/hugetlbfs/inode.c | 1 + fs/isofs/inode.c | 3 +-- fs/jffs2/super.c | 1 + fs/jfs/super.c | 1 + fs/logfs/super.c | 1 + fs/minix/inode.c | 1 + fs/ncpfs/inode.c | 1 + fs/nfs/super.c | 3 ++- fs/nfsd/nfsctl.c | 1 + fs/nilfs2/super.c | 1 + fs/ntfs/super.c | 1 + fs/ocfs2/dlmfs/dlmfs.c | 1 + fs/omfs/inode.c | 1 + fs/openpromfs/inode.c | 1 + fs/qnx4/inode.c | 1 + fs/qnx6/inode.c | 1 + fs/reiserfs/super.c | 1 + fs/romfs/super.c | 1 + fs/sysv/super.c | 3 ++- fs/ubifs/super.c | 1 + fs/ufs/super.c | 1 + fs/xfs/xfs_super.c | 1 + include/linux/fs.h | 2 ++ net/sunrpc/rpc_pipe.c | 4 +--- security/keys/process_keys.c | 2 +- 70 files changed, 78 insertions(+), 13 deletions(-)
PATCH 1/2 userns: Stop oopsing in key_change_session_keyring
Quoted 39 lines · Dave Jones
Just hit this on Linus' current tree.
[ 89.621770] BUG: unable to handle kernel NULL pointer dereference at 00000000000000c8 [ 89.623111] IP: [<ffffffff810784b0>] commit_creds+0x250/0x2f0 [ 89.624062] PGD 122bfd067 PUD 122bfe067 PMD 0 [ 89.624901] Oops: 0000 [#1] PREEMPT SMP [ 89.625678] Modules linked in: caif_socket caif netrom bridge hidp 8021q garp stp mrp rose llc2 af_rxrpc phonet af_key binfmt_misc bnep l2tp_ppp can_bcm l2tp_core pppoe pppox can_raw scsi_transport_iscsi ppp_generic slhc nfnetlink can ipt_ULOG ax25 decnet irda nfc rds x25 crc_ccitt appletalk atm ipx p8023 psnap p8022 llc lockd sunrpc ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 xt_conntrack nf_conntrack ip6table_filter ip6_tables btusb bluetooth snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_pcm vhost_net snd_page_alloc snd_timer tun macvtap usb_debug snd rfkill microcode macvlan edac_core pcspkr serio_raw kvm_amd soundcore kvm r8169 mii [ 89.637846] CPU 2 [ 89.638175] Pid: 782, comm: trinity-main Not tainted 3.8.0+ #63 Gigabyte Technology Co., Ltd. GA-MA78GM-S2H/GA-MA78GM-S2H [ 89.639850] RIP: 0010:[<ffffffff810784b0>] [<ffffffff810784b0>] commit_creds+0x250/0x2f0 [ 89.641161] RSP: 0018:ffff880115657eb8 EFLAGS: 00010207 [ 89.641984] RAX: 00000000000003e8 RBX: ffff88012688b000 RCX: 0000000000000000 [ 89.643069] RDX: 0000000000000000 RSI: ffffffff81c32960 RDI: ffff880105839600 [ 89.644167] RBP: ffff880115657ed8 R08: 0000000000000000 R09: 0000000000000000 [ 89.645254] R10: 0000000000000001 R11: 0000000000000246 R12: ffff880105839600 [ 89.646340] R13: ffff88011beea490 R14: ffff88011beea490 R15: 0000000000000000 [ 89.647431] FS: 00007f3ac063b740(0000) GS:ffff88012b200000(0000) knlGS:0000000000000000 [ 89.648660] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [ 89.649548] CR2: 00000000000000c8 CR3: 0000000122bfc000 CR4: 00000000000007e0 [ 89.650635] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 89.651723] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 [ 89.652812] Process trinity-main (pid: 782, threadinfo ffff880115656000, task ffff88011beea490) [ 89.654128] Stack: [ 89.654433] 0000000000000000 ffff8801058396a0 ffff880105839600 ffff88011beeaa78 [ 89.655769] ffff880115657ef8 ffffffff812c7d9b ffffffff82079be0 0000000000000000 [ 89.657073] ffff880115657f28 ffffffff8106c665 0000000000000002 ffff880115657f58 [ 89.658399] Call Trace: [ 89.658822] [<ffffffff812c7d9b>] key_change_session_keyring+0xfb/0x140 [ 89.659845] [<ffffffff8106c665>] task_work_run+0xa5/0xd0 [ 89.660698] [<ffffffff81002911>] do_notify_resume+0x71/0xb0 [ 89.661581] [<ffffffff816c9a4a>] int_signal+0x12/0x17 [ 89.662385] Code: 24 90 00 00 00 48 8b b3 90 00 00 00 49 8b 4c 24 40 48 39 f2 75 08 e9 83 00 00 00 48 89 ca 48 81 fa 60 29 c3 81 0f 84 41 fe ff ff <48> 8b 8a c8 00 00 00 48 39 ce 75 e4 3b 82 d0 00 00 00 0f 84 4b [ 89.667778] RIP [<ffffffff810784b0>] commit_creds+0x250/0x2f0 [ 89.668733] RSP <ffff880115657eb8> [ 89.669301] CR2: 00000000000000c8My fastest trinity induced oops yet!
Quoted 6 lines
Appears to be..
if ((set_ns == subset_ns->parent) && 850: 48 8b 8a c8 00 00 00 mov 0xc8(%rdx),%rcxfrom the inlined
cred_cap_issubset
By historical accident we have been reading trying to set new->user_ns from new->user_ns. Which is totally silly as new->user_ns is NULL (as is every other field in new except session_keyring at that point).
The intent is clearly to copy all of the fields from old to new so copy old->user_ns into into new->user_ns.
--- security/keys/process_keys.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
security/keys/process_keys.c +1 -1
@@ -839,7 +839,7 @@ void key_change_session_keyring(struct callback_head *twork) new-> sgid = old-> sgid; new->fsgid = old->fsgid; new->user =get_uid (old->user); - new->user_ns =get_user_ns (new->user_ns); + new->user_ns =get_user_ns (old->user_ns); new->group_info =get_group_info (old->group_info); new->securebits = old->securebits;
Signature · 1 line
1.7.5.4
PATCH 2/2 fs: Limit sys_mount to only request filesystem modules.
In reply to Eric W. BiedermanModify the request_module to prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.
A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing /etc/modprobe.d/*.conf with blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.
This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
This is relevant to user namespaces because we can reach the request module in get_fs_type() without having any special permissions, and people get uncomfortable when a user specified string (in this case the filesystem type) goes all of the way to request_module.
After having looked at this issue I don't think there is any particular reason to perform any filtering or permission checks beyond making it clear in the module request that we want a filesystem module. The common pattern in the kernel is to call request_module() without regards to the users permissions. In general all a filesystem module does once loaded is call register_filesystem() and go to sleep. Which means there is not much attack surface exposed by loading a filesytem module unless the filesystem is mounted. In a user namespace filesystems are not mounted unless .fs_flags = FS_USERNS_MOUNT, which most filesystems do not set today.
--- arch/ia64/kernel/perfmon.c | 1 + arch/powerpc/platforms/cell/spufs/inode.c | 1 + arch/s390/hypfs/inode.c | 1 + drivers/firmware/efivars.c | 1 + drivers/infiniband/hw/ipath/ipath_fs.c | 1 + drivers/infiniband/hw/qib/qib_fs.c | 1 + drivers/misc/ibmasm/ibmasmfs.c | 1 + drivers/mtd/mtdchar.c | 1 + drivers/oprofile/oprofilefs.c | 1 + drivers/staging/ccg/f_fs.c | 1 + drivers/usb/gadget/f_fs.c | 1 + drivers/usb/gadget/inode.c | 1 + drivers/xen/xenfs/super.c | 1 + fs/9p/vfs_super.c | 1 + fs/adfs/super.c | 1 + fs/affs/super.c | 1 + fs/afs/super.c | 1 + fs/autofs4/init.c | 1 + fs/befs/linuxvfs.c | 1 + fs/bfs/inode.c | 1 + fs/binfmt_misc.c | 1 + fs/btrfs/super.c | 1 + fs/ceph/super.c | 1 + fs/coda/inode.c | 1 + fs/configfs/mount.c | 1 + fs/cramfs/inode.c | 1 + fs/debugfs/inode.c | 1 + fs/devpts/inode.c | 1 + fs/ecryptfs/main.c | 1 + fs/efs/super.c | 1 + fs/exofs/super.c | 1 + fs/ext2/super.c | 1 + fs/ext3/super.c | 1 + fs/ext4/super.c | 5 +++-- fs/f2fs/super.c | 1 + fs/fat/namei_msdos.c | 1 + fs/fat/namei_vfat.c | 1 + fs/filesystems.c | 2 +- fs/freevxfs/vxfs_super.c | 2 +- fs/fuse/control.c | 1 + fs/fuse/inode.c | 2 ++ fs/gfs2/ops_fstype.c | 4 +++- fs/hfs/super.c | 1 + fs/hfsplus/super.c | 1 + fs/hppfs/hppfs.c | 1 + fs/hugetlbfs/inode.c | 1 + fs/isofs/inode.c | 3 +-- fs/jffs2/super.c | 1 + fs/jfs/super.c | 1 + fs/logfs/super.c | 1 + fs/minix/inode.c | 1 + fs/ncpfs/inode.c | 1 + fs/nfs/super.c | 3 ++- fs/nfsd/nfsctl.c | 1 + fs/nilfs2/super.c | 1 + fs/ntfs/super.c | 1 + fs/ocfs2/dlmfs/dlmfs.c | 1 + fs/omfs/inode.c | 1 + fs/openpromfs/inode.c | 1 + fs/qnx4/inode.c | 1 + fs/qnx6/inode.c | 1 + fs/reiserfs/super.c | 1 + fs/romfs/super.c | 1 + fs/sysv/super.c | 3 ++- fs/ubifs/super.c | 1 + fs/ufs/super.c | 1 + fs/xfs/xfs_super.c | 1 + include/linux/fs.h | 2 ++ net/sunrpc/rpc_pipe.c | 4 +--- 69 files changed, 77 insertions(+), 12 deletions(-)
arch/ia64/kernel/perfmon.c +1 -0
@@ -619,6 +619,7 @@ static struct file_system_type pfm_fs_type = { .mount = pfmfs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("pfmfs" ); DEFINE_PER_CPU(unsigned long, pfm_syst_info); DEFINE_PER_CPU(struct task_struct *, pmu_owner);
arch/powerpc/platforms/cell/spufs/inode.c +1 -0
@@ -749,6 +749,7 @@ static struct file_system_type spufs_type = { .mount = spufs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("spufs" );static int __init spufs_init (void ) {
arch/s390/hypfs/inode.c +1 -0
@@ -456,6 +456,7 @@ static struct file_system_type hypfs_type = { .mount = hypfs_mount, .kill_sb = hypfs_kill_super }; +MODULE_ALIAS_FS ("s390_hypfs" );static const struct super_operations hypfs_s_ops = { .statfs = simple_statfs,
drivers/firmware/efivars.c +1 -0
@@ -1234,6 +1234,7 @@ static struct file_system_type efivarfs_type = { .mount = efivarfs_mount, .kill_sb = efivarfs_kill_sb, }; +MODULE_ALIAS_FS ("efivarfs" );/* * Handle negative dentry.
drivers/infiniband/hw/ipath/ipath_fs.c +1 -0
@@ -410,6 +410,7 @@ static struct file_system_type ipathfs_fs_type = { .mount = ipathfs_mount, .kill_sb = ipathfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init ipath_init_ipathfs (void ) {
drivers/infiniband/hw/qib/qib_fs.c +1 -0
@@ -604,6 +604,7 @@ static struct file_system_type qibfs_fs_type = { .mount = qibfs_mount, .kill_sb = qibfs_kill_super, }; +MODULE_ALIAS_FS ("ipathfs" );int __init qib_init_qibfs (void ) {
drivers/misc/ibmasm/ibmasmfs.c +1 -0
@@ -110,6 +110,7 @@ static struct file_system_type ibmasmfs_type = { .mount = ibmasmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ibmasmfs" );static int ibmasmfs_fill_super (struct super_block *sb,void *data,int silent) {
drivers/mtd/mtdchar.c +1 -0
@@ -1238,6 +1238,7 @@ static struct file_system_type mtd_inodefs_type = { .mount = mtd_inodefs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("mtd_inodefs" );static int __init init_mtdchar (void ) {
drivers/oprofile/oprofilefs.c +1 -0
@@ -276,6 +276,7 @@ static struct file_system_type oprofilefs_type = { .mount = oprofilefs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("oprofilefs" );int __init oprofilefs_register (void )
drivers/staging/ccg/f_fs.c +1 -0
@@ -1223,6 +1223,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/
drivers/usb/gadget/f_fs.c +1 -0
@@ -1235,6 +1235,7 @@ static struct file_system_type ffs_fs_type = { .mount = ffs_fs_mount, .kill_sb = ffs_fs_kill_sb, }; +MODULE_ALIAS_FS ("functionfs" );/* Driver's main init/cleanup functions *************************************/
drivers/usb/gadget/inode.c +1 -0
@@ -2105,6 +2105,7 @@ static struct file_system_type gadgetfs_type = { .mount = gadgetfs_mount, .kill_sb = gadgetfs_kill_sb, }; +MODULE_ALIAS_FS ("gadgetfs" );/*----------------------------------------------------------------------*/
drivers/xen/xenfs/super.c +1 -0
@@ -75,6 +75,7 @@ static struct file_system_type xenfs_type = { .mount = xenfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("xenfs" );static int __init xenfs_init (void ) {
fs/9p/vfs_super.c +1 -0
@@ -365,3 +365,4 @@ struct file_system_type v9fs_fs_type = { .owner = THIS_MODULE, .fs_flags = FS_RENAME_DOES_D_MOVE, }; +MODULE_ALIAS_FS ("9p" );
fs/adfs/super.c +1 -0
@@ -524,6 +524,7 @@ static struct file_system_type adfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("adfs" );static int __init init_adfs_fs (void ) {
fs/affs/super.c +1 -0
@@ -622,6 +622,7 @@ static struct file_system_type affs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("affs" );static int __init init_affs_fs (void ) {
fs/afs/super.c +1 -0
@@ -45,6 +45,7 @@ struct file_system_type afs_fs_type = { .kill_sb = afs_kill_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("afs" );static const struct super_operations afs_super_ops = { .statfs = afs_statfs,
fs/autofs4/init.c +1 -0
@@ -26,6 +26,7 @@ static struct file_system_type autofs_fs_type = { .mount = autofs_mount, .kill_sb = autofs4_kill_sb, }; +MODULE_ALIAS_FS ("autofs" );static int __init init_autofs4_fs (void ) {
fs/befs/linuxvfs.c +1 -0
@@ -951,6 +951,7 @@ static struct file_system_type befs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("befs" );static int __init init_befs_fs (void )
fs/bfs/inode.c +1 -0
@@ -473,6 +473,7 @@ static struct file_system_type bfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("bfs" );static int __init init_bfs_fs (void ) {
fs/binfmt_misc.c +1 -0
@@ -720,6 +720,7 @@ static struct file_system_type bm_fs_type = { .mount = bm_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("binfmt_misc" );static int __init init_misc_binfmt (void ) {
fs/btrfs/super.c +1 -0
@@ -1558,6 +1558,7 @@ static struct file_system_type btrfs_fs_type = { .kill_sb = btrfs_kill_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("btrfs" );/* * used by btrfsctl to scan devices when no FS is mounted
fs/ceph/super.c +1 -0
@@ -952,6 +952,7 @@ static struct file_system_type ceph_fs_type = { .kill_sb = ceph_kill_sb, .fs_flags = FS_RENAME_DOES_D_MOVE, }; +MODULE_ALIAS_FS ("ceph" );#define _STRINGIFY (x) #x#define STRINGIFY (x) _STRINGIFY(x)
fs/coda/inode.c +1 -0
@@ -329,4 +329,5 @@ struct file_system_type coda_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("coda" );
fs/configfs/mount.c +1 -0
@@ -114,6 +114,7 @@ static struct file_system_type configfs_fs_type = { .mount = configfs_do_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("configfs" );struct dentry *configfs_pin_fs (void ) {
fs/cramfs/inode.c +1 -0
@@ -573,6 +573,7 @@ static struct file_system_type cramfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("cramfs" );static int __init init_cramfs_fs (void ) {
fs/debugfs/inode.c +1 -0
@@ -299,6 +299,7 @@ static struct file_system_type debug_fs_type = { .mount = debug_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("debugfs" ); static struct dentry *__create_file(const char *name, umode_t mode, struct dentry *parent, void *data,
fs/devpts/inode.c +1 -0
@@ -510,6 +510,7 @@ static struct file_system_type devpts_fs_type = { .fs_flags = FS_USERNS_MOUNT | FS_USERNS_DEV_MOUNT,#endif }; +MODULE_ALIAS_FS ("devpts" );/* * The normal naming convention is simply /dev/pts/<number>; this conforms
fs/ecryptfs/main.c +1 -0
@@ -629,6 +629,7 @@ static struct file_system_type ecryptfs_fs_type = { .kill_sb = ecryptfs_kill_block_super, .fs_flags = 0 }; +MODULE_ALIAS_FS ("ecryptfs" );/** * inode_info_init_once
fs/efs/super.c +1 -0
@@ -33,6 +33,7 @@ static struct file_system_type efs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("efs" );static struct pt_types sgi_pt_types[] = { {0x00 ,"SGI vh" },
fs/exofs/super.c +1 -0
@@ -1010,6 +1010,7 @@ static struct file_system_type exofs_type = { .mount = exofs_mount, .kill_sb = generic_shutdown_super, }; +MODULE_ALIAS_FS ("exofs" );static int __init init_exofs (void ) {
fs/ext2/super.c +1 -0
@@ -1536,6 +1536,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext2" );static int __init init_ext2_fs (void ) {
fs/ext3/super.c +1 -0
@@ -3068,6 +3068,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext3" );static int __init init_ext3_fs (void ) {
fs/ext4/super.c +3 -2
@@ -90,6 +90,7 @@ static struct file_system_type ext2_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext2" );#define IS_EXT2_SB (sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)#else #define IS_EXT2_SB (sb) (0)
@@ -104,6 +105,7 @@ static struct file_system_type ext3_fs_type = { .kill_sb = kill_block_super, .fs_flags =FS_REQUIRES_DEV , }; +MODULE_ALIAS_FS ("ext3" );#define IS_EXT3_SB (sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)#else #define IS_EXT3_SB (sb) (0)
@@ -5152,7 +5154,6 @@ static inline int ext2_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext2" );#else static inline void register_as_ext2 (void ) { }static inline void unregister_as_ext2 (void ) { }
@@ -5185,7 +5186,6 @@ static inline int ext3_feature_set_ok(struct super_block *sb)return 0 ; return 1; } -MODULE_ALIAS ("ext3" );#else static inline void register_as_ext3 (void ) { }static inline void unregister_as_ext3 (void ) { }
@@ -5199,6 +5199,7 @@ static struct file_system_type ext4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ext4" );static int __init ext4_init_feat_adverts (void ) {
fs/f2fs/super.c +1 -0
@@ -687,6 +687,7 @@ static struct file_system_type f2fs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("f2fs" );static int __init init_inodecache (void ) {
fs/fat/namei_msdos.c +1 -0
@@ -668,6 +668,7 @@ static struct file_system_type msdos_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("msdos" );static int __init init_msdos_fs (void ) {
fs/fat/namei_vfat.c +1 -0
@@ -1073,6 +1073,7 @@ static struct file_system_type vfat_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vfat" );static int __init init_vfat_fs (void ) {
fs/filesystems.c +1 -1
@@ -273,7 +273,7 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs && (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {
fs/freevxfs/vxfs_super.c +1 -1
@@ -52,7 +52,6 @@ MODULE_AUTHOR("Christoph Hellwig");MODULE_DESCRIPTION ("Veritas Filesystem (VxFS) driver" );MODULE_LICENSE ("Dual BSD/GPL" ); -MODULE_ALIAS ("vxfs" );/* makes mount -t vxfs autoload the module */ static void vxfs_put_super (struct super_block *);
@@ -258,6 +257,7 @@ static struct file_system_type vxfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("vxfs" );/* makes mount -t vxfs autoload the module */ static int __init vxfs_init (void )
fs/fuse/control.c +1 -0
@@ -341,6 +341,7 @@ static struct file_system_type fuse_ctl_fs_type = { .mount = fuse_ctl_mount, .kill_sb = fuse_ctl_kill_sb, }; +MODULE_ALIAS_FS ("fusectl" );int __init fuse_ctl_init (void ) {
fs/fuse/inode.c +2 -0
@@ -1117,6 +1117,7 @@ static struct file_system_type fuse_fs_type = { .mount = fuse_mount, .kill_sb = fuse_kill_sb_anon, }; +MODULE_ALIAS_FS ("fuse" );#ifdef CONFIG_BLOCK static struct dentry *fuse_mount_blk(struct file_system_type *fs_type,
@@ -1146,6 +1147,7 @@ static struct file_system_type fuseblk_fs_type = { .kill_sb = fuse_kill_sb_blk, .fs_flags = FS_REQUIRES_DEV | FS_HAS_SUBTYPE, }; +MODULE_ALIAS_FS ("fuseblk" );static inline int register_fuseblk (void ) {
fs/gfs2/ops_fstype.c +3 -1
@@ -20,6 +20,7 @@#include <linux/gfs2_ondisk.h> #include <linux/quotaops.h> #include <linux/lockdep.h> +#include <linux/module.h> #include "gfs2.h" #include "incore.h"
@@ -1425,6 +1426,7 @@ struct file_system_type gfs2_fs_type = { .kill_sb = gfs2_kill_sb, .owner = THIS_MODULE, }; +MODULE_ALIAS_FS ("gfs2" );struct file_system_type gfs2meta_fs_type = { .name ="gfs2meta" ,
@@ -1432,4 +1434,4 @@ struct file_system_type gfs2meta_fs_type = { .mount = gfs2_mount_meta, .owner = THIS_MODULE, }; - +MODULE_ALIAS_FS ("gfs2meta" );
fs/hfs/super.c +1 -0
@@ -466,6 +466,7 @@ static struct file_system_type hfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfs" );static void hfs_init_once (void *p) {
fs/hfsplus/super.c +1 -0
@@ -654,6 +654,7 @@ static struct file_system_type hfsplus_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("hfsplus" );static void hfsplus_init_once (void *p) {
fs/hppfs/hppfs.c +1 -0
@@ -748,6 +748,7 @@ static struct file_system_type hppfs_type = { .kill_sb = kill_anon_super, .fs_flags = 0, }; +MODULE_ALIAS_FS ("hppfs" );static int __init init_hppfs (void ) {
fs/hugetlbfs/inode.c +1 -0
@@ -896,6 +896,7 @@ static struct file_system_type hugetlbfs_fs_type = { .mount = hugetlbfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("hugetlbfs" );static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE ];
fs/isofs/inode.c +1 -2
@@ -1556,6 +1556,7 @@ static struct file_system_type iso9660_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("iso9660" );static int __init init_iso9660_fs (void ) {
@@ -1593,5 +1594,3 @@ static void __exit exit_iso9660_fs(void) module_init(init_iso9660_fs) module_exit(exit_iso9660_fs) MODULE_LICENSE("GPL"); -/* Actual filesystem name is iso9660, as requested in filesystems.c */ -MODULE_ALIAS ("iso9660" );
fs/jffs2/super.c +1 -0
@@ -356,6 +356,7 @@ static struct file_system_type jffs2_fs_type = { .mount = jffs2_mount, .kill_sb = jffs2_kill_sb, }; +MODULE_ALIAS_FS ("jffs2" );static int __init init_jffs2_fs (void ) {
fs/jfs/super.c +1 -0
@@ -833,6 +833,7 @@ static struct file_system_type jfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("jfs" );static void init_once (void *foo) {
fs/logfs/super.c +1 -0
@@ -608,6 +608,7 @@ static struct file_system_type logfs_fs_type = { .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("logfs" );static int __init logfs_init (void ) {
fs/minix/inode.c +1 -0
@@ -660,6 +660,7 @@ static struct file_system_type minix_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("minix" );static int __init init_minix_fs (void ) {
fs/ncpfs/inode.c +1 -0
@@ -1051,6 +1051,7 @@ static struct file_system_type ncp_fs_type = { .kill_sb = kill_anon_super, .fs_flags = FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("ncpfs" );static int __init init_ncp_fs (void ) {
fs/nfs/super.c +2 -1
@@ -294,6 +294,7 @@ struct file_system_type nfs_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs" );EXPORT_SYMBOL_GPL (nfs_fs_type);struct file_system_type nfs_xdev_fs_type = {
@@ -333,6 +334,7 @@ struct file_system_type nfs4_fs_type = { .kill_sb = nfs_kill_super, .fs_flags = FS_RENAME_DOES_D_MOVE|FS_BINARY_MOUNTDATA, }; +MODULE_ALIAS_FS ("nfs4" );EXPORT_SYMBOL_GPL (nfs4_fs_type);static int __init register_nfs4_fs (void )
@@ -2717,6 +2719,5 @@ module_param(send_implementation_id, ushort, 0644);MODULE_PARM_DESC (send_implementation_id,"Send implementation ID with NFSv4.1 exchange_id" );MODULE_PARM_DESC (nfs4_unique_id,"nfs_client_id4 uniquifier string" ); -MODULE_ALIAS ("nfs4" );#endif /* CONFIG_NFS_V4 */
fs/nfsd/nfsctl.c +1 -0
@@ -1090,6 +1090,7 @@ static struct file_system_type nfsd_fs_type = { .mount = nfsd_mount, .kill_sb = nfsd_umount, }; +MODULE_ALIAS_FS ("nfsd" );#ifdef CONFIG_PROC_FS static int create_proc_exports_entry (void )
fs/nilfs2/super.c +1 -0
@@ -1361,6 +1361,7 @@ struct file_system_type nilfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("nilfs2" );static void nilfs_inode_init_once (void *obj) {
fs/ntfs/super.c +1 -0
@@ -3079,6 +3079,7 @@ static struct file_system_type ntfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ntfs" );/* Stable names for the slab caches. */ static const char ntfs_index_ctx_cache_name[] ="ntfs_index_ctx_cache" ;
fs/ocfs2/dlmfs/dlmfs.c +1 -0
@@ -640,6 +640,7 @@ static struct file_system_type dlmfs_fs_type = { .mount = dlmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS ("ocfs2_dlmfs" );static int __init init_dlmfs_fs (void ) {
fs/omfs/inode.c +1 -0
@@ -572,6 +572,7 @@ static struct file_system_type omfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("omfs" );static int __init init_omfs_fs (void ) {
fs/openpromfs/inode.c +1 -0
@@ -432,6 +432,7 @@ static struct file_system_type openprom_fs_type = { .mount = openprom_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS ("openpromfs" );static void op_inode_init_once (void *data) {
fs/qnx4/inode.c +1 -0
@@ -412,6 +412,7 @@ static struct file_system_type qnx4_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx4" );static int __init init_qnx4_fs (void ) {
fs/qnx6/inode.c +1 -0
@@ -672,6 +672,7 @@ static struct file_system_type qnx6_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("qnx6" );static int __init init_qnx6_fs (void ) {
fs/reiserfs/super.c +1 -0
@@ -2434,6 +2434,7 @@ struct file_system_type reiserfs_fs_type = { .kill_sb = reiserfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("reiserfs" );MODULE_DESCRIPTION ("ReiserFS journaled filesystem" );MODULE_AUTHOR ("Hans Reiser <[email protected]>" );
fs/romfs/super.c +1 -0
@@ -599,6 +599,7 @@ static struct file_system_type romfs_fs_type = { .kill_sb = romfs_kill_sb, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("romfs" );/* * inode storage initialiser
fs/sysv/super.c +2 -1
@@ -545,6 +545,7 @@ static struct file_system_type sysv_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("sysv" );static struct file_system_type v7_fs_type = { .owner =THIS_MODULE ,
@@ -553,6 +554,7 @@ static struct file_system_type v7_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("v7" );static int __init init_sysv_fs (void ) {
@@ -586,5 +588,4 @@ static void __exit exit_sysv_fs(void) module_init(init_sysv_fs) module_exit(exit_sysv_fs) -MODULE_ALIAS("v7"); MODULE_LICENSE("GPL");
fs/ubifs/super.c +1 -0
@@ -2174,6 +2174,7 @@ static struct file_system_type ubifs_fs_type = { .mount = ubifs_mount, .kill_sb = kill_ubifs_super, }; +MODULE_ALIAS_FS ("ubifs" );/* * Inode slab cache constructor.
fs/ufs/super.c +1 -0
@@ -1500,6 +1500,7 @@ static struct file_system_type ufs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("ufs" );static int __init init_ufs_fs (void ) {
fs/xfs/xfs_super.c +1 -0
@@ -1561,6 +1561,7 @@ static struct file_system_type xfs_fs_type = { .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV, }; +MODULE_ALIAS_FS ("xfs" ); STATIC int __initxfs_init_zones (void )
include/linux/fs.h +2 -0
@@ -1825,6 +1825,8 @@ struct file_system_type {struct lock_class_key i_mutex_dir_key; }; +#define MODULE_ALIAS_FS (NAME ) MODULE_ALIAS("fs-" NAME) +extern struct dentry *mount_ns (struct file_system_type *fs_type,int flags,void *data,int (*fill_super)(struct super_block *,void *,int )); extern struct dentry *mount_bdev(struct file_system_type *fs_type,
net/sunrpc/rpc_pipe.c +1 -3
@@ -1174,6 +1174,7 @@ static struct file_system_type rpc_pipe_fs_type = { .mount = rpc_mount, .kill_sb = rpc_kill_sb, }; +MODULE_ALIAS_FS ("rpc_pipefs" );static void init_once (void *foo)
@@ -1218,6 +1219,3 @@ void unregister_rpc_pipefs(void)kmem_cache_destroy (rpc_inode_cachep);unregister_filesystem (&rpc_pipe_fs_type); } - -/* Make 'mount -t rpc_pipefs ...' autoload this module. */ -MODULE_ALIAS ("rpc_pipefs" );
Signature · 1 line
1.7.5.4
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
(cc'ed kernel-hardening)
Quoted 43 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing
/etc/modprobe.d/*.confwith blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
This is relevant to user namespaces because we can reach the request module in
get_fs_type()without having any special permissions, and people get uncomfortable when a user specified string (in this case the filesystem type) goes all of the way torequest_module.After having looked at this issue I don't think there is any particular reason to perform any filtering or permission checks beyond making it clear in the module request that we want a filesystem module. The common pattern in the kernel is to call
request_module()without regards to the users permissions. In general all a filesystem module does once loaded is callregister_filesystem()and go to sleep. Which means there is not much attack surface exposed by loading a filesytem module unless the filesystem is mounted. In a user namespace filesystems are not mounted unless .fs_flags =FS_USERNS_MOUNT, which most filesystems do not set today.Acked-by Serge Hallyn <[email protected]>Acked-by Kees Cook <[email protected]>Reported-by Kees Cook <[email protected]>
...
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 279
Quoted 13 lines
fs/filesystems.c +1
-1@@ -273,7 +273,7 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs && (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {
Maybe we should divide request_module() into several functions regarding expected caller's privileges?
- request_module() for CAP_SYS_MODULE in init_ns
- request_module_relaxed() for everybody
request_module_relaxed() is used in get_fs_type(), dev_load() and all places where the safety of module loading is manually checked. All old not yet checked users of request_module() will not be triggerable from user_ns. That's the same scheme as with capable() and ns_capable().
Thanks,
Signature · 2 lines
Vasily Kulikov http://www.openwall.com - bringing security into open computing environments
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
Quoted 64 lines · Vasily Kulikov
(cc'ed kernel-hardening)
Quoted 43 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing
/etc/modprobe.d/*.confwith blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
This is relevant to user namespaces because we can reach the request module in
get_fs_type()without having any special permissions, and people get uncomfortable when a user specified string (in this case the filesystem type) goes all of the way torequest_module.After having looked at this issue I don't think there is any particular reason to perform any filtering or permission checks beyond making it clear in the module request that we want a filesystem module. The common pattern in the kernel is to call
request_module()without regards to the users permissions. In general all a filesystem module does once loaded is callregister_filesystem()and go to sleep. Which means there is not much attack surface exposed by loading a filesytem module unless the filesystem is mounted. In a user namespace filesystems are not mounted unless .fs_flags =FS_USERNS_MOUNT, which most filesystems do not set today.Acked-by Serge Hallyn <[email protected]>Acked-by Kees Cook <[email protected]>Reported-by Kees Cook <[email protected]>Signed-off-by "Eric W. Biederman" <[email protected]>...
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 279
Quoted 13 lines
fs/filesystems.c +1
-1@@ -273,7 +273,7 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs && (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {Maybe we should divide
request_module()into several functions regarding expected caller's privileges?
Quoted 7 lines
-
request_module()forCAP_SYS_MODULEininit_ns
-request_module_relaxed()for everybody
request_module_relaxed()is used inget_fs_type(),dev_load()and all places where the safety of module loading is manually checked. All old not yet checked users ofrequest_module()will not be triggerable fromuser_ns. That's the same scheme as withcapable()andns_capable().
User namespaces in this discussion are pretty much a red-herring. You can already reach most request_module callers without having any capabilities. And honestly that seems fine.
It never ever hurts to request a module.
It only sometimes when something else has already gone wrong hurts to get the module.
It makes sense to add a prefix when sending the module request to make it clear what kind of module we are looking for. That makes it easy to tell why we are requesting the module and makes it easy to implement policy controls in userspace.
I don't see any reason to limit request_module to people with some capability or other. The filesystem module_request just happened to be after a capable(CAP_SYS_AMDIN) in this case which is the case people noticed was a little fishy.
But if I have overlooked something I am happy to hear it.
Eric
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
In reply to Vasily KulikovQuoted 64 lines · Vasily Kulikov
(cc'ed kernel-hardening)
Quoted 43 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Using aliases means user space can control the policy of which filesystem modules are auto-loaded by editing
/etc/modprobe.d/*.confwith blacklist and alias directives. Allowing simple, safe, well understood work-arounds to known problematic software.This also addresses a rare but unfortunate problem where the filesystem name is not the same as it's module name and module auto-loading would not work. While writing this patch I saw a handful of such cases. The most significant being autofs that lives in the module autofs4.
This is relevant to user namespaces because we can reach the request module in
get_fs_type()without having any special permissions, and people get uncomfortable when a user specified string (in this case the filesystem type) goes all of the way torequest_module.After having looked at this issue I don't think there is any particular reason to perform any filtering or permission checks beyond making it clear in the module request that we want a filesystem module. The common pattern in the kernel is to call
request_module()without regards to the users permissions. In general all a filesystem module does once loaded is callregister_filesystem()and go to sleep. Which means there is not much attack surface exposed by loading a filesytem module unless the filesystem is mounted. In a user namespace filesystems are not mounted unless .fs_flags =FS_USERNS_MOUNT, which most filesystems do not set today.Acked-by Serge Hallyn <[email protected]>Acked-by Kees Cook <[email protected]>Reported-by Kees Cook <[email protected]>...
fs/filesystems.c struct file_system_type *get_fs_type(const char *name) · line 279
Quoted 13 lines
fs/filesystems.c +1
-1@@ -273,7 +273,7 @@ struct file_system_type *get_fs_type(const char *name)int len = dot ? dot - name :strlen (name); fs =__get_fs_type (name, len); -if (!fs && (request_module ("%.*s" , len, name) ==0 )) +if (!fs && (request_module ("fs-%.*s" , len, name) ==0 )) fs =__get_fs_type (name, len);if (dot && fs && !(fs->fs_flags &FS_HAS_SUBTYPE )) {Maybe we should divide
request_module()into several functions regarding expected caller's privileges?
Quoted 7 lines
-
request_module()forCAP_SYS_MODULEininit_ns
-request_module_relaxed()for everybody
request_module_relaxed()is used inget_fs_type(),dev_load()and all places where the safety of module loading is manually checked. All old not yet checked users ofrequest_module()will not be triggerable fromuser_ns. That's the same scheme as withcapable()andns_capable().
User namespaces in this discussion are pretty much a red-herring. You can already reach most request_module callers without having any capabilities. And honestly that seems fine.
It never ever hurts to request a module.
It only sometimes when something else has already gone wrong hurts to get the module.
It makes sense to add a prefix when sending the module request to make it clear what kind of module we are looking for. That makes it easy to tell why we are requesting the module and makes it easy to implement policy controls in userspace.
I don't see any reason to limit request_module to people with some capability or other. The filesystem module_request just happened to be after a capable(CAP_SYS_AMDIN) in this case which is the case people noticed was a little fishy.
But if I have overlooked something I am happy to hear it.
Eric
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
In reply to Eric W. BiedermanQuoted 5 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.
Quoted 7 lines
A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
'-' is a commonly used part of a module name, and does not mix well with ramdom user provided names.
We usually use ':' as the prefix separator for modaliases, when user-supplied strings are prefixed with the subsystem.
I think it would be nicer to change that, and I'm sure some creative guy calls the next filesystem of the month fs-$something :)
Kay
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
Quoted 14 lines · Kay Sievers
Quoted 5 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.Quoted 7 lines
A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
Quoted 8 lines
'-' is a commonly used part of a module name, and does not mix well with ramdom user provided names.
We usually use ':' as the prefix separator for modaliases, when user-supplied strings are prefixed with the subsystem.
I think it would be nicer to change that, and I'm sure some creative guy calls the next filesystem of the month fs-$something :)
The precedent is "-". "netdev-" "pf-net-" etc. Naming something fs-$something is fine as long as it's actually a filesystem. :)
-Kees
Signature · 2 lines
Kees Cook Chrome OS Security
Re: [PATCH 2/2] fs: Limit sys_mount to only request filesystem modules.
In reply to Kay SieversQuoted 10 lines · Kay Sievers
Quoted 9 lines · Eric W. Biederman
Modify the
request_moduleto prefix the file system type with "fs-" and add aliases to all of the filesystems that can be built as modules to match.A common practice is to build all of the kernel code and leave code that is not commonly needed as modules, with the result that many users are exposed to any bug anywhere in the kernel.
Quoted 6 lines
Quoted 3 lines
Looking for filesystems with a fs- prefix limits the pool of possible modules that can be loaded by mount to just filesystems trivially making things safer with no real cost.
'-' is a commonly used part of a module name, and does not mix well with ramdom user provided names.
The symbols '-' and '_' occur in 2382 out of 3968 modules from an allmodconfig build, and modprobe ignores the difference between the two. However only three of those modules begin with fs and none of them begin with fs-.
Furthermore if it actually becomes a concern to ensure we are talking about an alias rather than a real module name, the solution is to change how we call modprobe. As long as we are in the same namespace something can go wrong.
fs- seems sufficiently unique for the purpose.
Quoted 2 lines
We usually use ':' as the prefix separator for modaliases, when user-supplied strings are prefixed with the subsystem.
There are at least two different conventions in use. For software subsystems like the networking stack '-' is the commonly used to separate the prefix. For hardware specific subsystems ':' is commonly used. What I really don't want to load here are hardware modules so using a hardware module style convention does not seem like the right way to go.
Quoted 2 lines
I think it would be nicer to change that, and I'm sure some creative guy calls the next filesystem of the month fs-$something :)
If it is a filesystem it simply does not matter. The goal is to only load filesystems.
If it is not a filesystem someone has choosen a confusing naming convention.
If it turns out I am wrong it is a two line change.
Eric