Implement mount querying syscalls agreed on at LSF/MM 2023.
Features:
- statx-like want/got mask
- allows returning ascii strings (fs type, root, mount point)
- returned buffer is relocatable (no pointers)
Still missing:
- man pages
- kselftest
- syscalls on non-x86 archs
Please find the test utility at the end of this mail.
Usage: statmnt [-l] (mnt_id|path)
Git tree:
git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs.git#statmount-v3
Changes v1..v3:
- rename statmnt(2) -> statmount(2)
- rename listmnt(2) -> listmount(2)
- make ABI 32bit compatible by passing 64bit args in a struct (tested on
i386 and x32)
- only accept new 64bit mount IDs
- fix compile on !CONFIG_PROC_FS
- call security_sb_statfs() in both syscalls
- make lookup_mnt_in_ns() static
- add LISTMOUNT_UNREACHABLE flag to listmnt() to explicitly ask for
listing unreachable mounts
- remove .sb_opts
- remove subtype from .fs_type
- return the number of bytes used (including strings) in .size
- rename .mountpoint -> .mnt_point
- point strings by an offset against char[] VLA at the end of the struct.
E.g. printf("fs_type: %s\n", st->str + st->fs_type);
- don't save string lengths
- extend spare space in struct statmnt (complete size is now 512 bytes)
---
Miklos Szeredi (4):
add unique mount ID
namespace: extract show_path() helper
add statmount(2) syscall
add listmount(2) syscall
arch/x86/entry/syscalls/syscall_32.tbl | 2 +
arch/x86/entry/syscalls/syscall_64.tbl | 2 +
fs/internal.h | 2 +
fs/mount.h | 3 +-
fs/namespace.c | 365 +++++++++++++++++++++++++
fs/proc_namespace.c | 10 +-
fs/stat.c | 9 +-
fs/statfs.c | 1 +
include/linux/syscalls.h | 8 +
include/uapi/asm-generic/unistd.h | 8 +-
include/uapi/linux/mount.h | 59 ++++
include/uapi/linux/stat.h | 1 +
12 files changed, 459 insertions(+), 11 deletions(-)
--
2.41.0
=== statmnt.c ===
#define _GNU_SOURCE
#include <unistd.h>
#include <stdio.h>
#include <fcntl.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/param.h>
#include <err.h>
/*
* Structure for getting mount/superblock/filesystem info with statmount(2).
*
* The interface is similar to statx(2): individual fields or groups can be
* selected with the @mask argument of statmount(). Kernel will set the @mask
* field according to the supported fields.
*
* If string fields are selected, then the caller needs to pass a buffer that
* has space after the fixed part of the structure. Nul terminated strings are
* copied there and offsets relative to @str are stored in the relevant fields.
* If the buffer is too small, then EOVERFLOW is returned. The actually used
* size is returned in @size.
*/
struct statmnt {
__u32 size; /* Total size, including strings */
__u32 __spare1;
__u64 mask; /* What results were written */
__u32 sb_dev_major; /* Device ID */
__u32 sb_dev_minor;
__u64 sb_magic; /* ..._SUPER_MAGIC */
__u32 sb_flags; /* MS_{RDONLY,SYNCHRONOUS,DIRSYNC,LAZYTIME} */
__u32 fs_type; /* [str] Filesystem type */
__u64 mnt_id; /* Unique ID of mount */
__u64 mnt_parent_id; /* Unique ID of parent (for root == mnt_id) */
__u32 mnt_id_old; /* Reused IDs used in proc/.../mountinfo */
__u32 mnt_parent_id_old;
__u64 mnt_attr; /* MOUNT_ATTR_... */
__u64 mnt_propagation; /* MS_{SHARED,SLAVE,PRIVATE,UNBINDABLE} */
__u64 mnt_peer_group; /* ID of shared peer group */
__u64 mnt_master; /* Mount receives propagation from this ID */
__u64 propagate_from; /* Propagation from in current namespace */
__u32 mnt_root; /* [str] Root of mount relative to root of fs */
__u32 mnt_point; /* [str] Mountpoint relative to current root */
__u64 __spare2[50];
char str[]; /* Variable size part containing strings */
};
/*
* To be used on the kernel ABI only for passing 64bit arguments to statmount(2)
*/
struct __mount_arg {
__u64 mnt_id;
__u64 request_mask;
};
/*
* @mask bits for statmount(2)
*/
#define STMT_SB_BASIC 0x00000001U /* Want/got sb_... */
#define STMT_MNT_BASIC 0x00000002U /* Want/got mnt_... */
#define STMT_PROPAGATE_FROM 0x00000004U /* Want/got propagate_from */
#define STMT_MNT_ROOT 0x00000008U /* Want/got mnt_root */
#define STMT_MNT_POINT 0x00000010U /* Want/got mnt_point */
#define STMT_FS_TYPE 0x00000020U /* Want/got fs_type */
/* listmount(2) flags */
#define LISTMOUNT_UNREACHABLE 0x01 /* List unreachable mounts too */
#define __NR_statmount 454
#define __NR_listmount 455
#define STATX_MNT_ID_UNIQUE 0x00004000U /* Want/got extended stx_mount_id */
static void free_if_neq(void *p, const void *q)
{
if (p != q)
free(p);
}
static struct statmnt *statmount(uint64_t mnt_id, uint64_t mask, unsigned int flags)
{
struct __mount_arg arg = {
.mnt_id = mnt_id,
.request_mask = mask,
};
union {
struct statmnt m;
char s[4096];
} buf;
struct statmnt *ret, *mm = &buf.m;
size_t bufsize = sizeof(buf);
while (syscall(__NR_statmount, &arg, mm, bufsize, flags) == -1) {
free_if_neq(mm, &buf.m);
if (errno != EOVERFLOW)
return NULL;
bufsize = MAX(1 << 15, bufsize << 1);
mm = malloc(bufsize);
if (!mm)
return NULL;
}
ret = malloc(mm->size);
if (ret)
memcpy(ret, mm, mm->size);
free_if_neq(mm, &buf.m);
return ret;
}
static int listmount(uint64_t mnt_id, uint64_t **listp, unsigned int flags)
{
struct __mount_arg arg = {
.mnt_id = mnt_id,
};
uint64_t buf[512];
size_t bufsize = sizeof(buf);
uint64_t *ret, *ll = buf;
long len;
while ((len = syscall(__NR_listmount, &arg, ll, bufsize / sizeof(buf[0]), flags)) == -1) {
free_if_neq(ll, buf);
if (errno != EOVERFLOW)
return -1;
bufsize = MAX(1 << 15, bufsize << 1);
ll = malloc(bufsize);
if (!ll)
return -1;
}
bufsize = len * sizeof(buf[0]);
ret = malloc(bufsize);
if (!ret)
return -1;
*listp = ret;
memcpy(ret, ll, bufsize);
free_if_neq(ll, buf);
return len;
}
int main(int argc, char *argv[])
{
struct statmnt *st;
char *end;
const char *arg = argv[1];
int res;
int list = 0;
uint64_t mask = STMT_SB_BASIC | STMT_MNT_BASIC | STMT_PROPAGATE_FROM | STMT_MNT_ROOT | STMT_MNT_POINT | STMT_FS_TYPE;
uint64_t mnt_id;
if (arg && strcmp(arg, "-l") == 0) {
list = 1;
arg = argv[2];
}
if (argc != list + 2)
errx(1, "usage: %s [-l] (mnt_id|path)", argv[0]);
mnt_id = strtoll(arg, &end, 0);
if (!mnt_id || *end != '\0') {
struct statx sx;
res = statx(AT_FDCWD, arg, 0, STATX_MNT_ID_UNIQUE, &sx);
if (res == -1)
err(1, "%s", arg);
if (!(sx.stx_mask & (STATX_MNT_ID | STATX_MNT_ID_UNIQUE)))
errx(1, "Sorry, no mount ID");
mnt_id = sx.stx_mnt_id;
}
if (list) {
uint64_t *list;
int num, i;
res = listmount(mnt_id, &list, LISTMOUNT_UNREACHABLE);
if (res == -1)
err(1, "listmnt(%llu)", mnt_id);
num = res;
for (i = 0; i < num; i++) {
printf("0x%llx", list[i]);
st = statmount(list[i], STMT_MNT_POINT, 0);
if (!st) {
printf("\t[%s]\n", strerror(errno));
} else {
printf("\t%s\n", (st->mask & STMT_MNT_POINT) ? st->str + st->mnt_point : "???");
}
free(st);
}
free(list);
return 0;
}
st = statmount(mnt_id, mask, 0);
if (!st)
err(1, "statmnt(%llu)", mnt_id);
printf("size: %u\n", st->size);
printf("mask: 0x%llx\n", st->mask);
if (st->mask & STMT_SB_BASIC) {
printf("sb_dev_major: %u\n", st->sb_dev_major);
printf("sb_dev_minor: %u\n", st->sb_dev_minor);
printf("sb_magic: 0x%llx\n", st->sb_magic);
printf("sb_flags: 0x%08x\n", st->sb_flags);
}
if (st->mask & STMT_MNT_BASIC) {
printf("mnt_id: 0x%llx\n", st->mnt_id);
printf("mnt_parent_id: 0x%llx\n", st->mnt_parent_id);
printf("mnt_id_old: %u\n", st->mnt_id_old);
printf("mnt_parent_id_old: %u\n", st->mnt_parent_id_old);
printf("mnt_attr: 0x%08llx\n", st->mnt_attr);
printf("mnt_propagation: %s%s%s%s\n",
st->mnt_propagation & MS_SHARED ? "shared," : "",
st->mnt_propagation & MS_SLAVE ? "slave," : "",
st->mnt_propagation & MS_UNBINDABLE ? "unbindable," : "",
st->mnt_propagation & MS_PRIVATE ? "private" : "");
printf("mnt_peer_group: %llu\n", st->mnt_peer_group);
printf("mnt_master: %llu\n", st->mnt_master);
}
if (st->mask & STMT_PROPAGATE_FROM)
printf("propagate_from: %llu\n", st->propagate_from);
if (st->mask & STMT_MNT_ROOT)
printf("mnt_root: %u <%s>\n", st->mnt_root, st->str + st->mnt_root);
if (st->mask & STMT_MNT_POINT)
printf("mnt_point: %u <%s>\n", st->mnt_point, st->str + st->mnt_point);
if (st->mask & STMT_FS_TYPE)
printf("fs_type: %u <%s>\n", st->fs_type, st->str + st->fs_type);
free(st);
return 0;
}
Add a way to query attributes of a single mount instead of having to parse
the complete /proc/$PID/mountinfo, which might be huge.
Lookup the mount the new 64bit mount ID. If a mount needs to be queried
based on path, then statx(2) can be used to first query the mount ID
belonging to the path.
Design is based on a suggestion by Linus:
"So I'd suggest something that is very much like "statfsat()", which gets
a buffer and a length, and returns an extended "struct statfs" *AND*
just a string description at the end."
The interface closely mimics that of statx.
Handle ASCII attributes by appending after the end of the structure (as per
above suggestion). Pointers to strings are stored in u64 members to make
the structure the same regardless of pointer size. Strings are nul
terminated.
Link: https://lore.kernel.org/all/CAHk-=wh5YifP7hzKSbwJj94+DZ2czjrZsczy6GBimiogZws=rg@mail.gmail.com/
Signed-off-by: Miklos Szeredi <[email protected]>
---
arch/x86/entry/syscalls/syscall_32.tbl | 1 +
arch/x86/entry/syscalls/syscall_64.tbl | 1 +
fs/namespace.c | 283 +++++++++++++++++++++++++
fs/statfs.c | 1 +
include/linux/syscalls.h | 5 +
include/uapi/asm-generic/unistd.h | 5 +-
include/uapi/linux/mount.h | 56 +++++
7 files changed, 351 insertions(+), 1 deletion(-)
diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 2d0b1bd866ea..317b1320ad18 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -457,3 +457,4 @@
450 i386 set_mempolicy_home_node sys_set_mempolicy_home_node
451 i386 cachestat sys_cachestat
452 i386 fchmodat2 sys_fchmodat2
+454 i386 statmount sys_statmount
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index 1d6eee30eceb..7312c440978f 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -375,6 +375,7 @@
451 common cachestat sys_cachestat
452 common fchmodat2 sys_fchmodat2
453 64 map_shadow_stack sys_map_shadow_stack
+454 common statmount sys_statmount
#
# Due to a historical design error, certain syscalls are numbered differently
diff --git a/fs/namespace.c b/fs/namespace.c
index c3a41200fe70..3326ba2b2810 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -4687,6 +4687,289 @@ int show_path(struct seq_file *m, struct dentry *root)
return 0;
}
+static struct vfsmount *lookup_mnt_in_ns(u64 id, struct mnt_namespace *ns)
+{
+ struct mount *mnt;
+ struct vfsmount *res = NULL;
+
+ lock_ns_list(ns);
+ list_for_each_entry(mnt, &ns->list, mnt_list) {
+ if (!mnt_is_cursor(mnt) && id == mnt->mnt_id_unique) {
+ res = &mnt->mnt;
+ break;
+ }
+ }
+ unlock_ns_list(ns);
+ return res;
+}
+
+struct stmt_state {
+ struct statmnt __user *const buf;
+ size_t const bufsize;
+ struct vfsmount *const mnt;
+ u64 const mask;
+ struct seq_file seq;
+ struct path root;
+ struct statmnt sm;
+ size_t pos;
+ int err;
+};
+
+typedef int (*stmt_func_t)(struct stmt_state *);
+
+static int stmt_string_seq(struct stmt_state *s, stmt_func_t func)
+{
+ size_t rem = s->bufsize - s->pos - sizeof(s->sm);
+ struct seq_file *seq = &s->seq;
+ int ret;
+
+ seq->count = 0;
+ seq->size = min(seq->size, rem);
+ seq->buf = kvmalloc(seq->size, GFP_KERNEL_ACCOUNT);
+ if (!seq->buf)
+ return -ENOMEM;
+
+ ret = func(s);
+ if (ret)
+ return ret;
+
+ if (seq_has_overflowed(seq)) {
+ if (seq->size == rem)
+ return -EOVERFLOW;
+ seq->size *= 2;
+ if (seq->size > MAX_RW_COUNT)
+ return -ENOMEM;
+ kvfree(seq->buf);
+ return 0;
+ }
+
+ /* Done */
+ return 1;
+}
+
+static void stmt_string(struct stmt_state *s, u64 mask, stmt_func_t func,
+ u32 *str)
+{
+ int ret = s->pos + sizeof(s->sm) >= s->bufsize ? -EOVERFLOW : 0;
+ struct statmnt *sm = &s->sm;
+ struct seq_file *seq = &s->seq;
+
+ if (s->err || !(s->mask & mask))
+ return;
+
+ seq->size = PAGE_SIZE;
+ while (!ret)
+ ret = stmt_string_seq(s, func);
+
+ if (ret < 0) {
+ s->err = ret;
+ } else {
+ seq->buf[seq->count++] = '\0';
+ if (copy_to_user(s->buf->str + s->pos, seq->buf, seq->count)) {
+ s->err = -EFAULT;
+ } else {
+ *str = s->pos;
+ s->pos += seq->count;
+ }
+ }
+ kvfree(seq->buf);
+ sm->mask |= mask;
+}
+
+static void stmt_numeric(struct stmt_state *s, u64 mask, stmt_func_t func)
+{
+ if (s->err || !(s->mask & mask))
+ return;
+
+ s->err = func(s);
+ s->sm.mask |= mask;
+}
+
+static u64 mnt_to_attr_flags(struct vfsmount *mnt)
+{
+ unsigned int mnt_flags = READ_ONCE(mnt->mnt_flags);
+ u64 attr_flags = 0;
+
+ if (mnt_flags & MNT_READONLY)
+ attr_flags |= MOUNT_ATTR_RDONLY;
+ if (mnt_flags & MNT_NOSUID)
+ attr_flags |= MOUNT_ATTR_NOSUID;
+ if (mnt_flags & MNT_NODEV)
+ attr_flags |= MOUNT_ATTR_NODEV;
+ if (mnt_flags & MNT_NOEXEC)
+ attr_flags |= MOUNT_ATTR_NOEXEC;
+ if (mnt_flags & MNT_NODIRATIME)
+ attr_flags |= MOUNT_ATTR_NODIRATIME;
+ if (mnt_flags & MNT_NOSYMFOLLOW)
+ attr_flags |= MOUNT_ATTR_NOSYMFOLLOW;
+
+ if (mnt_flags & MNT_NOATIME)
+ attr_flags |= MOUNT_ATTR_NOATIME;
+ else if (mnt_flags & MNT_RELATIME)
+ attr_flags |= MOUNT_ATTR_RELATIME;
+ else
+ attr_flags |= MOUNT_ATTR_STRICTATIME;
+
+ if (is_idmapped_mnt(mnt))
+ attr_flags |= MOUNT_ATTR_IDMAP;
+
+ return attr_flags;
+}
+
+static u64 mnt_to_propagation_flags(struct mount *m)
+{
+ u64 propagation = 0;
+
+ if (IS_MNT_SHARED(m))
+ propagation |= MS_SHARED;
+ if (IS_MNT_SLAVE(m))
+ propagation |= MS_SLAVE;
+ if (IS_MNT_UNBINDABLE(m))
+ propagation |= MS_UNBINDABLE;
+ if (!propagation)
+ propagation |= MS_PRIVATE;
+
+ return propagation;
+}
+
+static int stmt_sb_basic(struct stmt_state *s)
+{
+ struct super_block *sb = s->mnt->mnt_sb;
+
+ s->sm.sb_dev_major = MAJOR(sb->s_dev);
+ s->sm.sb_dev_minor = MINOR(sb->s_dev);
+ s->sm.sb_magic = sb->s_magic;
+ s->sm.sb_flags = sb->s_flags & (SB_RDONLY|SB_SYNCHRONOUS|SB_DIRSYNC|SB_LAZYTIME);
+
+ return 0;
+}
+
+static int stmt_mnt_basic(struct stmt_state *s)
+{
+ struct mount *m = real_mount(s->mnt);
+
+ s->sm.mnt_id = m->mnt_id_unique;
+ s->sm.mnt_parent_id = m->mnt_parent->mnt_id_unique;
+ s->sm.mnt_id_old = m->mnt_id;
+ s->sm.mnt_parent_id_old = m->mnt_parent->mnt_id;
+ s->sm.mnt_attr = mnt_to_attr_flags(&m->mnt);
+ s->sm.mnt_propagation = mnt_to_propagation_flags(m);
+ s->sm.mnt_peer_group = IS_MNT_SHARED(m) ? m->mnt_group_id : 0;
+ s->sm.mnt_master = IS_MNT_SLAVE(m) ? m->mnt_master->mnt_group_id : 0;
+
+ return 0;
+}
+
+static int stmt_propagate_from(struct stmt_state *s)
+{
+ struct mount *m = real_mount(s->mnt);
+
+ if (!IS_MNT_SLAVE(m))
+ return 0;
+
+ s->sm.propagate_from = get_dominating_id(m, ¤t->fs->root);
+
+ return 0;
+}
+
+static int stmt_mnt_root(struct stmt_state *s)
+{
+ struct seq_file *seq = &s->seq;
+ int err = show_path(seq, s->mnt->mnt_root);
+
+ if (!err && !seq_has_overflowed(seq)) {
+ seq->buf[seq->count] = '\0';
+ seq->count = string_unescape_inplace(seq->buf, UNESCAPE_OCTAL);
+ }
+ return err;
+}
+
+static int stmt_mnt_point(struct stmt_state *s)
+{
+ struct vfsmount *mnt = s->mnt;
+ struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
+ int err = seq_path_root(&s->seq, &mnt_path, &s->root, "");
+
+ return err == SEQ_SKIP ? 0 : err;
+}
+
+static int stmt_fs_type(struct stmt_state *s)
+{
+ struct seq_file *seq = &s->seq;
+ struct super_block *sb = s->mnt->mnt_sb;
+
+ seq_puts(seq, sb->s_type->name);
+ return 0;
+}
+
+static int do_statmount(struct stmt_state *s)
+{
+ struct statmnt *sm = &s->sm;
+ struct mount *m = real_mount(s->mnt);
+ size_t copysize = min_t(size_t, s->bufsize, sizeof(*sm));
+ int err;
+
+ err = security_sb_statfs(s->mnt->mnt_root);
+ if (err)
+ return err;
+
+ if (!capable(CAP_SYS_ADMIN) &&
+ !is_path_reachable(m, m->mnt.mnt_root, &s->root))
+ return -EPERM;
+
+ stmt_numeric(s, STMT_SB_BASIC, stmt_sb_basic);
+ stmt_numeric(s, STMT_MNT_BASIC, stmt_mnt_basic);
+ stmt_numeric(s, STMT_PROPAGATE_FROM, stmt_propagate_from);
+ stmt_string(s, STMT_FS_TYPE, stmt_fs_type, &sm->fs_type);
+ stmt_string(s, STMT_MNT_ROOT, stmt_mnt_root, &sm->mnt_root);
+ stmt_string(s, STMT_MNT_POINT, stmt_mnt_point, &sm->mnt_point);
+
+ if (s->err)
+ return s->err;
+
+ /* Return the number of bytes copied to the buffer */
+ sm->size = copysize + s->pos;
+
+ if (copy_to_user(s->buf, sm, copysize))
+ return -EFAULT;
+
+ return 0;
+}
+
+SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
+ struct statmnt __user *, buf, size_t, bufsize,
+ unsigned int, flags)
+{
+ struct vfsmount *mnt;
+ struct __mount_arg kreq;
+ int ret;
+
+ if (flags)
+ return -EINVAL;
+
+ if (copy_from_user(&kreq, req, sizeof(kreq)))
+ return -EFAULT;
+
+ down_read(&namespace_sem);
+ mnt = lookup_mnt_in_ns(kreq.mnt_id, current->nsproxy->mnt_ns);
+ ret = -ENOENT;
+ if (mnt) {
+ struct stmt_state s = {
+ .mask = kreq.request_mask,
+ .buf = buf,
+ .bufsize = bufsize,
+ .mnt = mnt,
+ };
+
+ get_fs_root(current->fs, &s.root);
+ ret = do_statmount(&s);
+ path_put(&s.root);
+ }
+ up_read(&namespace_sem);
+
+ return ret;
+}
+
static void __init init_mount_tree(void)
{
struct vfsmount *mnt;
diff --git a/fs/statfs.c b/fs/statfs.c
index 96d1c3edf289..cc774c2e2c9a 100644
--- a/fs/statfs.c
+++ b/fs/statfs.c
@@ -9,6 +9,7 @@
#include <linux/security.h>
#include <linux/uaccess.h>
#include <linux/compat.h>
+#include <uapi/linux/mount.h>
#include "internal.h"
static int flags_by_mnt(int mnt_flags)
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 22bc6bc147f8..ba371024d902 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -74,6 +74,8 @@ struct landlock_ruleset_attr;
enum landlock_rule_type;
struct cachestat_range;
struct cachestat;
+struct statmnt;
+struct __mount_arg;
#include <linux/types.h>
#include <linux/aio_abi.h>
@@ -408,6 +410,9 @@ asmlinkage long sys_statfs64(const char __user *path, size_t sz,
asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user *buf);
asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz,
struct statfs64 __user *buf);
+asmlinkage long sys_statmount(const struct __mount_arg __user *req,
+ struct statmnt __user *buf, size_t bufsize,
+ unsigned int flags);
asmlinkage long sys_truncate(const char __user *path, long length);
asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length);
#if BITS_PER_LONG == 32
diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h
index abe087c53b4b..8f034e934a2e 100644
--- a/include/uapi/asm-generic/unistd.h
+++ b/include/uapi/asm-generic/unistd.h
@@ -823,8 +823,11 @@ __SYSCALL(__NR_cachestat, sys_cachestat)
#define __NR_fchmodat2 452
__SYSCALL(__NR_fchmodat2, sys_fchmodat2)
+#define __NR_statmount 454
+__SYSCALL(__NR_statmount, sys_statmount)
+
#undef __NR_syscalls
-#define __NR_syscalls 453
+#define __NR_syscalls 455
/*
* 32 bit systems traditionally used different
diff --git a/include/uapi/linux/mount.h b/include/uapi/linux/mount.h
index bb242fdcfe6b..d2c988ab526b 100644
--- a/include/uapi/linux/mount.h
+++ b/include/uapi/linux/mount.h
@@ -138,4 +138,60 @@ struct mount_attr {
/* List of all mount_attr versions. */
#define MOUNT_ATTR_SIZE_VER0 32 /* sizeof first published struct */
+
+/*
+ * Structure for getting mount/superblock/filesystem info with statmount(2).
+ *
+ * The interface is similar to statx(2): individual fields or groups can be
+ * selected with the @mask argument of statmount(). Kernel will set the @mask
+ * field according to the supported fields.
+ *
+ * If string fields are selected, then the caller needs to pass a buffer that
+ * has space after the fixed part of the structure. Nul terminated strings are
+ * copied there and offsets relative to @str are stored in the relevant fields.
+ * If the buffer is too small, then EOVERFLOW is returned. The actually used
+ * size is returned in @size.
+ */
+struct statmnt {
+ __u32 size; /* Total size, including strings */
+ __u32 __spare1;
+ __u64 mask; /* What results were written */
+ __u32 sb_dev_major; /* Device ID */
+ __u32 sb_dev_minor;
+ __u64 sb_magic; /* ..._SUPER_MAGIC */
+ __u32 sb_flags; /* MS_{RDONLY,SYNCHRONOUS,DIRSYNC,LAZYTIME} */
+ __u32 fs_type; /* [str] Filesystem type */
+ __u64 mnt_id; /* Unique ID of mount */
+ __u64 mnt_parent_id; /* Unique ID of parent (for root == mnt_id) */
+ __u32 mnt_id_old; /* Reused IDs used in proc/.../mountinfo */
+ __u32 mnt_parent_id_old;
+ __u64 mnt_attr; /* MOUNT_ATTR_... */
+ __u64 mnt_propagation; /* MS_{SHARED,SLAVE,PRIVATE,UNBINDABLE} */
+ __u64 mnt_peer_group; /* ID of shared peer group */
+ __u64 mnt_master; /* Mount receives propagation from this ID */
+ __u64 propagate_from; /* Propagation from in current namespace */
+ __u32 mnt_root; /* [str] Root of mount relative to root of fs */
+ __u32 mnt_point; /* [str] Mountpoint relative to current root */
+ __u64 __spare2[50];
+ char str[]; /* Variable size part containing strings */
+};
+
+/*
+ * To be used on the kernel ABI only for passing 64bit arguments to statmount(2)
+ */
+struct __mount_arg {
+ __u64 mnt_id;
+ __u64 request_mask;
+};
+
+/*
+ * @mask bits for statmount(2)
+ */
+#define STMT_SB_BASIC 0x00000001U /* Want/got sb_... */
+#define STMT_MNT_BASIC 0x00000002U /* Want/got mnt_... */
+#define STMT_PROPAGATE_FROM 0x00000004U /* Want/got propagate_from */
+#define STMT_MNT_ROOT 0x00000008U /* Want/got mnt_root */
+#define STMT_MNT_POINT 0x00000010U /* Want/got mnt_point */
+#define STMT_FS_TYPE 0x00000020U /* Want/got fs_type */
+
#endif /* _UAPI_LINUX_MOUNT_H */
--
2.41.0
Add way to query the children of a particular mount. This is a more
flexible way to iterate the mount tree than having to parse the complete
/proc/self/mountinfo.
Lookup the mount by the new 64bit mount ID. If a mount needs to be queried
based on path, then statx(2) can be used to first query the mount ID
belonging to the path.
Return an array of new (64bit) mount ID's. Without privileges only mounts
are listed which are reachable from the task's root.
Signed-off-by: Miklos Szeredi <[email protected]>
---
arch/x86/entry/syscalls/syscall_32.tbl | 1 +
arch/x86/entry/syscalls/syscall_64.tbl | 1 +
fs/namespace.c | 69 ++++++++++++++++++++++++++
include/linux/syscalls.h | 3 ++
include/uapi/asm-generic/unistd.h | 5 +-
include/uapi/linux/mount.h | 3 ++
6 files changed, 81 insertions(+), 1 deletion(-)
diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 317b1320ad18..65e0185b47a9 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -458,3 +458,4 @@
451 i386 cachestat sys_cachestat
452 i386 fchmodat2 sys_fchmodat2
454 i386 statmount sys_statmount
+455 i386 listmount sys_listmount
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index 7312c440978f..a1b3ce7d22cc 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -376,6 +376,7 @@
452 common fchmodat2 sys_fchmodat2
453 64 map_shadow_stack sys_map_shadow_stack
454 common statmount sys_statmount
+455 common listmount sys_listmount
#
# Due to a historical design error, certain syscalls are numbered differently
diff --git a/fs/namespace.c b/fs/namespace.c
index 3326ba2b2810..050e2d2af110 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -4970,6 +4970,75 @@ SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
return ret;
}
+static long do_listmount(struct vfsmount *mnt, u64 __user *buf, size_t bufsize,
+ const struct path *root, unsigned int flags)
+{
+ struct mount *r, *m = real_mount(mnt);
+ struct path rootmnt = {
+ .mnt = root->mnt,
+ .dentry = root->mnt->mnt_root
+ };
+ long ctr = 0;
+ bool reachable_only = true;
+ int err;
+
+ err = security_sb_statfs(mnt->mnt_root);
+ if (err)
+ return err;
+
+ if (flags & LISTMOUNT_UNREACHABLE) {
+ if (!capable(CAP_SYS_ADMIN))
+ return -EPERM;
+ reachable_only = false;
+ }
+
+ if (reachable_only && !is_path_reachable(m, mnt->mnt_root, &rootmnt))
+ return capable(CAP_SYS_ADMIN) ? 0 : -EPERM;
+
+ list_for_each_entry(r, &m->mnt_mounts, mnt_child) {
+ if (reachable_only &&
+ !is_path_reachable(r, r->mnt.mnt_root, root))
+ continue;
+
+ if (ctr >= bufsize)
+ return -EOVERFLOW;
+ if (put_user(r->mnt_id_unique, buf + ctr))
+ return -EFAULT;
+ ctr++;
+ if (ctr < 0)
+ return -ERANGE;
+ }
+ return ctr;
+}
+
+SYSCALL_DEFINE4(listmount, const struct __mount_arg __user *, req,
+ u64 __user *, buf, size_t, bufsize, unsigned int, flags)
+{
+ struct __mount_arg kreq;
+ struct vfsmount *mnt;
+ struct path root;
+ long err;
+
+ if (flags & ~LISTMOUNT_UNREACHABLE)
+ return -EINVAL;
+
+ if (copy_from_user(&kreq, req, sizeof(kreq)))
+ return -EFAULT;
+
+ down_read(&namespace_sem);
+ mnt = lookup_mnt_in_ns(kreq.mnt_id, current->nsproxy->mnt_ns);
+ err = -ENOENT;
+ if (mnt) {
+ get_fs_root(current->fs, &root);
+ err = do_listmount(mnt, buf, bufsize, &root, flags);
+ path_put(&root);
+ }
+ up_read(&namespace_sem);
+
+ return err;
+}
+
+
static void __init init_mount_tree(void)
{
struct vfsmount *mnt;
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index ba371024d902..38f3da7e04d1 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -413,6 +413,9 @@ asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz,
asmlinkage long sys_statmount(const struct __mount_arg __user *req,
struct statmnt __user *buf, size_t bufsize,
unsigned int flags);
+asmlinkage long sys_listmount(const struct __mount_arg __user *req,
+ u64 __user *buf, size_t bufsize,
+ unsigned int flags);
asmlinkage long sys_truncate(const char __user *path, long length);
asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length);
#if BITS_PER_LONG == 32
diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h
index 8f034e934a2e..8df6a747e21a 100644
--- a/include/uapi/asm-generic/unistd.h
+++ b/include/uapi/asm-generic/unistd.h
@@ -826,8 +826,11 @@ __SYSCALL(__NR_fchmodat2, sys_fchmodat2)
#define __NR_statmount 454
__SYSCALL(__NR_statmount, sys_statmount)
+#define __NR_listmount 455
+__SYSCALL(__NR_listmount, sys_listmount)
+
#undef __NR_syscalls
-#define __NR_syscalls 455
+#define __NR_syscalls 456
/*
* 32 bit systems traditionally used different
diff --git a/include/uapi/linux/mount.h b/include/uapi/linux/mount.h
index d2c988ab526b..7aa9916659d2 100644
--- a/include/uapi/linux/mount.h
+++ b/include/uapi/linux/mount.h
@@ -194,4 +194,7 @@ struct __mount_arg {
#define STMT_MNT_POINT 0x00000010U /* Want/got mnt_point */
#define STMT_FS_TYPE 0x00000020U /* Want/got fs_type */
+/* listmount(2) flags */
+#define LISTMOUNT_UNREACHABLE 0x01 /* List unreachable mounts too */
+
#endif /* _UAPI_LINUX_MOUNT_H */
--
2.41.0
To be used by the statmount(2) syscall as well.
Signed-off-by: Miklos Szeredi <[email protected]>
---
fs/internal.h | 2 ++
fs/namespace.c | 9 +++++++++
fs/proc_namespace.c | 10 +++-------
3 files changed, 14 insertions(+), 7 deletions(-)
diff --git a/fs/internal.h b/fs/internal.h
index d64ae03998cc..0c4f4cf2ff5a 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -83,6 +83,8 @@ int path_mount(const char *dev_name, struct path *path,
const char *type_page, unsigned long flags, void *data_page);
int path_umount(struct path *path, int flags);
+int show_path(struct seq_file *m, struct dentry *root);
+
/*
* fs_struct.c
*/
diff --git a/fs/namespace.c b/fs/namespace.c
index e02bc5f41c7b..c3a41200fe70 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -4678,6 +4678,15 @@ SYSCALL_DEFINE5(mount_setattr, int, dfd, const char __user *, path,
return err;
}
+int show_path(struct seq_file *m, struct dentry *root)
+{
+ if (root->d_sb->s_op->show_path)
+ return root->d_sb->s_op->show_path(m, root);
+
+ seq_dentry(m, root, " \t\n\\");
+ return 0;
+}
+
static void __init init_mount_tree(void)
{
struct vfsmount *mnt;
diff --git a/fs/proc_namespace.c b/fs/proc_namespace.c
index 250eb5bf7b52..5638ad419f52 100644
--- a/fs/proc_namespace.c
+++ b/fs/proc_namespace.c
@@ -142,13 +142,9 @@ static int show_mountinfo(struct seq_file *m, struct vfsmount *mnt)
seq_printf(m, "%i %i %u:%u ", r->mnt_id, r->mnt_parent->mnt_id,
MAJOR(sb->s_dev), MINOR(sb->s_dev));
- if (sb->s_op->show_path) {
- err = sb->s_op->show_path(m, mnt->mnt_root);
- if (err)
- goto out;
- } else {
- seq_dentry(m, mnt->mnt_root, " \t\n\\");
- }
+ err = show_path(m, mnt->mnt_root);
+ if (err)
+ goto out;
seq_putc(m, ' ');
/* mountpoints outside of chroot jail will give SEQ_SKIP on this */
--
2.41.0
On 28/9/23 21:01, Miklos Szeredi wrote:
> Add a way to query attributes of a single mount instead of having to parse
> the complete /proc/$PID/mountinfo, which might be huge.
>
> Lookup the mount the new 64bit mount ID. If a mount needs to be queried
> based on path, then statx(2) can be used to first query the mount ID
> belonging to the path.
>
> Design is based on a suggestion by Linus:
>
> "So I'd suggest something that is very much like "statfsat()", which gets
> a buffer and a length, and returns an extended "struct statfs" *AND*
> just a string description at the end."
>
> The interface closely mimics that of statx.
>
> Handle ASCII attributes by appending after the end of the structure (as per
> above suggestion). Pointers to strings are stored in u64 members to make
> the structure the same regardless of pointer size. Strings are nul
> terminated.
>
> Link: https://lore.kernel.org/all/CAHk-=wh5YifP7hzKSbwJj94+DZ2czjrZsczy6GBimiogZws=rg@mail.gmail.com/
> Signed-off-by: Miklos Szeredi <[email protected]>
> ---
> arch/x86/entry/syscalls/syscall_32.tbl | 1 +
> arch/x86/entry/syscalls/syscall_64.tbl | 1 +
> fs/namespace.c | 283 +++++++++++++++++++++++++
> fs/statfs.c | 1 +
> include/linux/syscalls.h | 5 +
> include/uapi/asm-generic/unistd.h | 5 +-
> include/uapi/linux/mount.h | 56 +++++
> 7 files changed, 351 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
> index 2d0b1bd866ea..317b1320ad18 100644
> --- a/arch/x86/entry/syscalls/syscall_32.tbl
> +++ b/arch/x86/entry/syscalls/syscall_32.tbl
> @@ -457,3 +457,4 @@
> 450 i386 set_mempolicy_home_node sys_set_mempolicy_home_node
> 451 i386 cachestat sys_cachestat
> 452 i386 fchmodat2 sys_fchmodat2
> +454 i386 statmount sys_statmount
> diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
> index 1d6eee30eceb..7312c440978f 100644
> --- a/arch/x86/entry/syscalls/syscall_64.tbl
> +++ b/arch/x86/entry/syscalls/syscall_64.tbl
> @@ -375,6 +375,7 @@
> 451 common cachestat sys_cachestat
> 452 common fchmodat2 sys_fchmodat2
> 453 64 map_shadow_stack sys_map_shadow_stack
> +454 common statmount sys_statmount
>
> #
> # Due to a historical design error, certain syscalls are numbered differently
> diff --git a/fs/namespace.c b/fs/namespace.c
> index c3a41200fe70..3326ba2b2810 100644
> --- a/fs/namespace.c
> +++ b/fs/namespace.c
> @@ -4687,6 +4687,289 @@ int show_path(struct seq_file *m, struct dentry *root)
> return 0;
> }
>
> +static struct vfsmount *lookup_mnt_in_ns(u64 id, struct mnt_namespace *ns)
> +{
> + struct mount *mnt;
> + struct vfsmount *res = NULL;
> +
> + lock_ns_list(ns);
> + list_for_each_entry(mnt, &ns->list, mnt_list) {
> + if (!mnt_is_cursor(mnt) && id == mnt->mnt_id_unique) {
> + res = &mnt->mnt;
> + break;
> + }
> + }
> + unlock_ns_list(ns);
> + return res;
> +}
Seems like we might need to consider making (struct mnt_namespace)->list
a hashed list.
The number of mounts could be large, for example people using autofs direct
mount setups.
It's not common for people to have of the order of 8k map entries (for
which there is a trigger mount per entry plus any mounts that have been
automounted) but it does happen. A small setup would be of the order of
1k map entries plus automounted mounts so the benefit is likely still
there to some extent.
Ian
> +
> +struct stmt_state {
> + struct statmnt __user *const buf;
> + size_t const bufsize;
> + struct vfsmount *const mnt;
> + u64 const mask;
> + struct seq_file seq;
> + struct path root;
> + struct statmnt sm;
> + size_t pos;
> + int err;
> +};
> +
> +typedef int (*stmt_func_t)(struct stmt_state *);
> +
> +static int stmt_string_seq(struct stmt_state *s, stmt_func_t func)
> +{
> + size_t rem = s->bufsize - s->pos - sizeof(s->sm);
> + struct seq_file *seq = &s->seq;
> + int ret;
> +
> + seq->count = 0;
> + seq->size = min(seq->size, rem);
> + seq->buf = kvmalloc(seq->size, GFP_KERNEL_ACCOUNT);
> + if (!seq->buf)
> + return -ENOMEM;
> +
> + ret = func(s);
> + if (ret)
> + return ret;
> +
> + if (seq_has_overflowed(seq)) {
> + if (seq->size == rem)
> + return -EOVERFLOW;
> + seq->size *= 2;
> + if (seq->size > MAX_RW_COUNT)
> + return -ENOMEM;
> + kvfree(seq->buf);
> + return 0;
> + }
> +
> + /* Done */
> + return 1;
> +}
> +
> +static void stmt_string(struct stmt_state *s, u64 mask, stmt_func_t func,
> + u32 *str)
> +{
> + int ret = s->pos + sizeof(s->sm) >= s->bufsize ? -EOVERFLOW : 0;
> + struct statmnt *sm = &s->sm;
> + struct seq_file *seq = &s->seq;
> +
> + if (s->err || !(s->mask & mask))
> + return;
> +
> + seq->size = PAGE_SIZE;
> + while (!ret)
> + ret = stmt_string_seq(s, func);
> +
> + if (ret < 0) {
> + s->err = ret;
> + } else {
> + seq->buf[seq->count++] = '\0';
> + if (copy_to_user(s->buf->str + s->pos, seq->buf, seq->count)) {
> + s->err = -EFAULT;
> + } else {
> + *str = s->pos;
> + s->pos += seq->count;
> + }
> + }
> + kvfree(seq->buf);
> + sm->mask |= mask;
> +}
> +
> +static void stmt_numeric(struct stmt_state *s, u64 mask, stmt_func_t func)
> +{
> + if (s->err || !(s->mask & mask))
> + return;
> +
> + s->err = func(s);
> + s->sm.mask |= mask;
> +}
> +
> +static u64 mnt_to_attr_flags(struct vfsmount *mnt)
> +{
> + unsigned int mnt_flags = READ_ONCE(mnt->mnt_flags);
> + u64 attr_flags = 0;
> +
> + if (mnt_flags & MNT_READONLY)
> + attr_flags |= MOUNT_ATTR_RDONLY;
> + if (mnt_flags & MNT_NOSUID)
> + attr_flags |= MOUNT_ATTR_NOSUID;
> + if (mnt_flags & MNT_NODEV)
> + attr_flags |= MOUNT_ATTR_NODEV;
> + if (mnt_flags & MNT_NOEXEC)
> + attr_flags |= MOUNT_ATTR_NOEXEC;
> + if (mnt_flags & MNT_NODIRATIME)
> + attr_flags |= MOUNT_ATTR_NODIRATIME;
> + if (mnt_flags & MNT_NOSYMFOLLOW)
> + attr_flags |= MOUNT_ATTR_NOSYMFOLLOW;
> +
> + if (mnt_flags & MNT_NOATIME)
> + attr_flags |= MOUNT_ATTR_NOATIME;
> + else if (mnt_flags & MNT_RELATIME)
> + attr_flags |= MOUNT_ATTR_RELATIME;
> + else
> + attr_flags |= MOUNT_ATTR_STRICTATIME;
> +
> + if (is_idmapped_mnt(mnt))
> + attr_flags |= MOUNT_ATTR_IDMAP;
> +
> + return attr_flags;
> +}
> +
> +static u64 mnt_to_propagation_flags(struct mount *m)
> +{
> + u64 propagation = 0;
> +
> + if (IS_MNT_SHARED(m))
> + propagation |= MS_SHARED;
> + if (IS_MNT_SLAVE(m))
> + propagation |= MS_SLAVE;
> + if (IS_MNT_UNBINDABLE(m))
> + propagation |= MS_UNBINDABLE;
> + if (!propagation)
> + propagation |= MS_PRIVATE;
> +
> + return propagation;
> +}
> +
> +static int stmt_sb_basic(struct stmt_state *s)
> +{
> + struct super_block *sb = s->mnt->mnt_sb;
> +
> + s->sm.sb_dev_major = MAJOR(sb->s_dev);
> + s->sm.sb_dev_minor = MINOR(sb->s_dev);
> + s->sm.sb_magic = sb->s_magic;
> + s->sm.sb_flags = sb->s_flags & (SB_RDONLY|SB_SYNCHRONOUS|SB_DIRSYNC|SB_LAZYTIME);
> +
> + return 0;
> +}
> +
> +static int stmt_mnt_basic(struct stmt_state *s)
> +{
> + struct mount *m = real_mount(s->mnt);
> +
> + s->sm.mnt_id = m->mnt_id_unique;
> + s->sm.mnt_parent_id = m->mnt_parent->mnt_id_unique;
> + s->sm.mnt_id_old = m->mnt_id;
> + s->sm.mnt_parent_id_old = m->mnt_parent->mnt_id;
> + s->sm.mnt_attr = mnt_to_attr_flags(&m->mnt);
> + s->sm.mnt_propagation = mnt_to_propagation_flags(m);
> + s->sm.mnt_peer_group = IS_MNT_SHARED(m) ? m->mnt_group_id : 0;
> + s->sm.mnt_master = IS_MNT_SLAVE(m) ? m->mnt_master->mnt_group_id : 0;
> +
> + return 0;
> +}
> +
> +static int stmt_propagate_from(struct stmt_state *s)
> +{
> + struct mount *m = real_mount(s->mnt);
> +
> + if (!IS_MNT_SLAVE(m))
> + return 0;
> +
> + s->sm.propagate_from = get_dominating_id(m, ¤t->fs->root);
> +
> + return 0;
> +}
> +
> +static int stmt_mnt_root(struct stmt_state *s)
> +{
> + struct seq_file *seq = &s->seq;
> + int err = show_path(seq, s->mnt->mnt_root);
> +
> + if (!err && !seq_has_overflowed(seq)) {
> + seq->buf[seq->count] = '\0';
> + seq->count = string_unescape_inplace(seq->buf, UNESCAPE_OCTAL);
> + }
> + return err;
> +}
> +
> +static int stmt_mnt_point(struct stmt_state *s)
> +{
> + struct vfsmount *mnt = s->mnt;
> + struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
> + int err = seq_path_root(&s->seq, &mnt_path, &s->root, "");
> +
> + return err == SEQ_SKIP ? 0 : err;
> +}
> +
> +static int stmt_fs_type(struct stmt_state *s)
> +{
> + struct seq_file *seq = &s->seq;
> + struct super_block *sb = s->mnt->mnt_sb;
> +
> + seq_puts(seq, sb->s_type->name);
> + return 0;
> +}
> +
> +static int do_statmount(struct stmt_state *s)
> +{
> + struct statmnt *sm = &s->sm;
> + struct mount *m = real_mount(s->mnt);
> + size_t copysize = min_t(size_t, s->bufsize, sizeof(*sm));
> + int err;
> +
> + err = security_sb_statfs(s->mnt->mnt_root);
> + if (err)
> + return err;
> +
> + if (!capable(CAP_SYS_ADMIN) &&
> + !is_path_reachable(m, m->mnt.mnt_root, &s->root))
> + return -EPERM;
> +
> + stmt_numeric(s, STMT_SB_BASIC, stmt_sb_basic);
> + stmt_numeric(s, STMT_MNT_BASIC, stmt_mnt_basic);
> + stmt_numeric(s, STMT_PROPAGATE_FROM, stmt_propagate_from);
> + stmt_string(s, STMT_FS_TYPE, stmt_fs_type, &sm->fs_type);
> + stmt_string(s, STMT_MNT_ROOT, stmt_mnt_root, &sm->mnt_root);
> + stmt_string(s, STMT_MNT_POINT, stmt_mnt_point, &sm->mnt_point);
> +
> + if (s->err)
> + return s->err;
> +
> + /* Return the number of bytes copied to the buffer */
> + sm->size = copysize + s->pos;
> +
> + if (copy_to_user(s->buf, sm, copysize))
> + return -EFAULT;
> +
> + return 0;
> +}
> +
> +SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
> + struct statmnt __user *, buf, size_t, bufsize,
> + unsigned int, flags)
> +{
> + struct vfsmount *mnt;
> + struct __mount_arg kreq;
> + int ret;
> +
> + if (flags)
> + return -EINVAL;
> +
> + if (copy_from_user(&kreq, req, sizeof(kreq)))
> + return -EFAULT;
> +
> + down_read(&namespace_sem);
> + mnt = lookup_mnt_in_ns(kreq.mnt_id, current->nsproxy->mnt_ns);
> + ret = -ENOENT;
> + if (mnt) {
> + struct stmt_state s = {
> + .mask = kreq.request_mask,
> + .buf = buf,
> + .bufsize = bufsize,
> + .mnt = mnt,
> + };
> +
> + get_fs_root(current->fs, &s.root);
> + ret = do_statmount(&s);
> + path_put(&s.root);
> + }
> + up_read(&namespace_sem);
> +
> + return ret;
> +}
> +
> static void __init init_mount_tree(void)
> {
> struct vfsmount *mnt;
> diff --git a/fs/statfs.c b/fs/statfs.c
> index 96d1c3edf289..cc774c2e2c9a 100644
> --- a/fs/statfs.c
> +++ b/fs/statfs.c
> @@ -9,6 +9,7 @@
> #include <linux/security.h>
> #include <linux/uaccess.h>
> #include <linux/compat.h>
> +#include <uapi/linux/mount.h>
> #include "internal.h"
>
> static int flags_by_mnt(int mnt_flags)
> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
> index 22bc6bc147f8..ba371024d902 100644
> --- a/include/linux/syscalls.h
> +++ b/include/linux/syscalls.h
> @@ -74,6 +74,8 @@ struct landlock_ruleset_attr;
> enum landlock_rule_type;
> struct cachestat_range;
> struct cachestat;
> +struct statmnt;
> +struct __mount_arg;
>
> #include <linux/types.h>
> #include <linux/aio_abi.h>
> @@ -408,6 +410,9 @@ asmlinkage long sys_statfs64(const char __user *path, size_t sz,
> asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user *buf);
> asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz,
> struct statfs64 __user *buf);
> +asmlinkage long sys_statmount(const struct __mount_arg __user *req,
> + struct statmnt __user *buf, size_t bufsize,
> + unsigned int flags);
> asmlinkage long sys_truncate(const char __user *path, long length);
> asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length);
> #if BITS_PER_LONG == 32
> diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h
> index abe087c53b4b..8f034e934a2e 100644
> --- a/include/uapi/asm-generic/unistd.h
> +++ b/include/uapi/asm-generic/unistd.h
> @@ -823,8 +823,11 @@ __SYSCALL(__NR_cachestat, sys_cachestat)
> #define __NR_fchmodat2 452
> __SYSCALL(__NR_fchmodat2, sys_fchmodat2)
>
> +#define __NR_statmount 454
> +__SYSCALL(__NR_statmount, sys_statmount)
> +
> #undef __NR_syscalls
> -#define __NR_syscalls 453
> +#define __NR_syscalls 455
>
> /*
> * 32 bit systems traditionally used different
> diff --git a/include/uapi/linux/mount.h b/include/uapi/linux/mount.h
> index bb242fdcfe6b..d2c988ab526b 100644
> --- a/include/uapi/linux/mount.h
> +++ b/include/uapi/linux/mount.h
> @@ -138,4 +138,60 @@ struct mount_attr {
> /* List of all mount_attr versions. */
> #define MOUNT_ATTR_SIZE_VER0 32 /* sizeof first published struct */
>
> +
> +/*
> + * Structure for getting mount/superblock/filesystem info with statmount(2).
> + *
> + * The interface is similar to statx(2): individual fields or groups can be
> + * selected with the @mask argument of statmount(). Kernel will set the @mask
> + * field according to the supported fields.
> + *
> + * If string fields are selected, then the caller needs to pass a buffer that
> + * has space after the fixed part of the structure. Nul terminated strings are
> + * copied there and offsets relative to @str are stored in the relevant fields.
> + * If the buffer is too small, then EOVERFLOW is returned. The actually used
> + * size is returned in @size.
> + */
> +struct statmnt {
> + __u32 size; /* Total size, including strings */
> + __u32 __spare1;
> + __u64 mask; /* What results were written */
> + __u32 sb_dev_major; /* Device ID */
> + __u32 sb_dev_minor;
> + __u64 sb_magic; /* ..._SUPER_MAGIC */
> + __u32 sb_flags; /* MS_{RDONLY,SYNCHRONOUS,DIRSYNC,LAZYTIME} */
> + __u32 fs_type; /* [str] Filesystem type */
> + __u64 mnt_id; /* Unique ID of mount */
> + __u64 mnt_parent_id; /* Unique ID of parent (for root == mnt_id) */
> + __u32 mnt_id_old; /* Reused IDs used in proc/.../mountinfo */
> + __u32 mnt_parent_id_old;
> + __u64 mnt_attr; /* MOUNT_ATTR_... */
> + __u64 mnt_propagation; /* MS_{SHARED,SLAVE,PRIVATE,UNBINDABLE} */
> + __u64 mnt_peer_group; /* ID of shared peer group */
> + __u64 mnt_master; /* Mount receives propagation from this ID */
> + __u64 propagate_from; /* Propagation from in current namespace */
> + __u32 mnt_root; /* [str] Root of mount relative to root of fs */
> + __u32 mnt_point; /* [str] Mountpoint relative to current root */
> + __u64 __spare2[50];
> + char str[]; /* Variable size part containing strings */
> +};
> +
> +/*
> + * To be used on the kernel ABI only for passing 64bit arguments to statmount(2)
> + */
> +struct __mount_arg {
> + __u64 mnt_id;
> + __u64 request_mask;
> +};
> +
> +/*
> + * @mask bits for statmount(2)
> + */
> +#define STMT_SB_BASIC 0x00000001U /* Want/got sb_... */
> +#define STMT_MNT_BASIC 0x00000002U /* Want/got mnt_... */
> +#define STMT_PROPAGATE_FROM 0x00000004U /* Want/got propagate_from */
> +#define STMT_MNT_ROOT 0x00000008U /* Want/got mnt_root */
> +#define STMT_MNT_POINT 0x00000010U /* Want/got mnt_point */
> +#define STMT_FS_TYPE 0x00000020U /* Want/got fs_type */
> +
> #endif /* _UAPI_LINUX_MOUNT_H */
On Fri, 29 Sept 2023 at 02:42, Ian Kent <[email protected]> wrote:
>
> On 28/9/23 21:01, Miklos Szeredi wrote:
> > +static struct vfsmount *lookup_mnt_in_ns(u64 id, struct mnt_namespace *ns)
> > +{
> > + struct mount *mnt;
> > + struct vfsmount *res = NULL;
> > +
> > + lock_ns_list(ns);
> > + list_for_each_entry(mnt, &ns->list, mnt_list) {
> > + if (!mnt_is_cursor(mnt) && id == mnt->mnt_id_unique) {
> > + res = &mnt->mnt;
> > + break;
> > + }
> > + }
> > + unlock_ns_list(ns);
> > + return res;
> > +}
>
> Seems like we might need to consider making (struct mnt_namespace)->list
>
> a hashed list.
Yes, linear search needs to go. A hash table is probably the easiest solution.
But I'd also consider replacing ns->list with an rbtree. Not as
trivial as adding a system hash table and probably also slightly
slower, but it would have some advantages:
- most space efficient (no overhead of hash buckets)
- cursor can go away (f_pos can just contain last ID)
Thanks,
Miklos
On 29/9/23 17:10, Miklos Szeredi wrote:
> On Fri, 29 Sept 2023 at 02:42, Ian Kent <[email protected]> wrote:
>> On 28/9/23 21:01, Miklos Szeredi wrote:
>>> +static struct vfsmount *lookup_mnt_in_ns(u64 id, struct mnt_namespace *ns)
>>> +{
>>> + struct mount *mnt;
>>> + struct vfsmount *res = NULL;
>>> +
>>> + lock_ns_list(ns);
>>> + list_for_each_entry(mnt, &ns->list, mnt_list) {
>>> + if (!mnt_is_cursor(mnt) && id == mnt->mnt_id_unique) {
>>> + res = &mnt->mnt;
>>> + break;
>>> + }
>>> + }
>>> + unlock_ns_list(ns);
>>> + return res;
>>> +}
>> Seems like we might need to consider making (struct mnt_namespace)->list
>>
>> a hashed list.
> Yes, linear search needs to go. A hash table is probably the easiest solution.
>
> But I'd also consider replacing ns->list with an rbtree. Not as
> trivial as adding a system hash table and probably also slightly
> slower, but it would have some advantages:
>
> - most space efficient (no overhead of hash buckets)
>
> - cursor can go away (f_pos can just contain last ID)
I guess that would be ok.
Avoiding the cursor is a big plus.
An rbtree is used in kernfs and its readdir function is rather painful so
I wonder what the implications might be for other enumeration needs.
Ian
On Thu, Sep 28, 2023 at 9:03 AM Miklos Szeredi <[email protected]> wrote:
>
> Add a way to query attributes of a single mount instead of having to parse
> the complete /proc/$PID/mountinfo, which might be huge.
>
> Lookup the mount the new 64bit mount ID. If a mount needs to be queried
> based on path, then statx(2) can be used to first query the mount ID
> belonging to the path.
>
> Design is based on a suggestion by Linus:
>
> "So I'd suggest something that is very much like "statfsat()", which gets
> a buffer and a length, and returns an extended "struct statfs" *AND*
> just a string description at the end."
>
> The interface closely mimics that of statx.
>
> Handle ASCII attributes by appending after the end of the structure (as per
> above suggestion). Pointers to strings are stored in u64 members to make
> the structure the same regardless of pointer size. Strings are nul
> terminated.
>
> Link: https://lore.kernel.org/all/CAHk-=wh5YifP7hzKSbwJj94+DZ2czjrZsczy6GBimiogZws=rg@mail.gmail.com/
> Signed-off-by: Miklos Szeredi <[email protected]>
> ---
> arch/x86/entry/syscalls/syscall_32.tbl | 1 +
> arch/x86/entry/syscalls/syscall_64.tbl | 1 +
> fs/namespace.c | 283 +++++++++++++++++++++++++
> fs/statfs.c | 1 +
> include/linux/syscalls.h | 5 +
> include/uapi/asm-generic/unistd.h | 5 +-
> include/uapi/linux/mount.h | 56 +++++
> 7 files changed, 351 insertions(+), 1 deletion(-)
...
> diff --git a/fs/namespace.c b/fs/namespace.c
> index c3a41200fe70..3326ba2b2810 100644
> --- a/fs/namespace.c
> +++ b/fs/namespace.c
...
> +static int do_statmount(struct stmt_state *s)
> +{
> + struct statmnt *sm = &s->sm;
> + struct mount *m = real_mount(s->mnt);
> + size_t copysize = min_t(size_t, s->bufsize, sizeof(*sm));
> + int err;
> +
> + err = security_sb_statfs(s->mnt->mnt_root);
> + if (err)
> + return err;
Thank you for adding the security_sb_statfs() call to this operation,
however I believe we want to place it *after* the capability check to
be consistent with other LSM calls.
> + if (!capable(CAP_SYS_ADMIN) &&
> + !is_path_reachable(m, m->mnt.mnt_root, &s->root))
> + return -EPERM;
> +
> + stmt_numeric(s, STMT_SB_BASIC, stmt_sb_basic);
> + stmt_numeric(s, STMT_MNT_BASIC, stmt_mnt_basic);
> + stmt_numeric(s, STMT_PROPAGATE_FROM, stmt_propagate_from);
> + stmt_string(s, STMT_FS_TYPE, stmt_fs_type, &sm->fs_type);
> + stmt_string(s, STMT_MNT_ROOT, stmt_mnt_root, &sm->mnt_root);
> + stmt_string(s, STMT_MNT_POINT, stmt_mnt_point, &sm->mnt_point);
> +
> + if (s->err)
> + return s->err;
> +
> + /* Return the number of bytes copied to the buffer */
> + sm->size = copysize + s->pos;
> +
> + if (copy_to_user(s->buf, sm, copysize))
> + return -EFAULT;
> +
> + return 0;
> +}
--
paul-moore.com
On Thu, Sep 28, 2023 at 9:04 AM Miklos Szeredi <[email protected]> wrote:
>
> Add way to query the children of a particular mount. This is a more
> flexible way to iterate the mount tree than having to parse the complete
> /proc/self/mountinfo.
>
> Lookup the mount by the new 64bit mount ID. If a mount needs to be queried
> based on path, then statx(2) can be used to first query the mount ID
> belonging to the path.
>
> Return an array of new (64bit) mount ID's. Without privileges only mounts
> are listed which are reachable from the task's root.
>
> Signed-off-by: Miklos Szeredi <[email protected]>
> ---
> arch/x86/entry/syscalls/syscall_32.tbl | 1 +
> arch/x86/entry/syscalls/syscall_64.tbl | 1 +
> fs/namespace.c | 69 ++++++++++++++++++++++++++
> include/linux/syscalls.h | 3 ++
> include/uapi/asm-generic/unistd.h | 5 +-
> include/uapi/linux/mount.h | 3 ++
> 6 files changed, 81 insertions(+), 1 deletion(-)
...
> diff --git a/fs/namespace.c b/fs/namespace.c
> index 3326ba2b2810..050e2d2af110 100644
> --- a/fs/namespace.c
> +++ b/fs/namespace.c
> @@ -4970,6 +4970,75 @@ SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
> return ret;
> }
>
> +static long do_listmount(struct vfsmount *mnt, u64 __user *buf, size_t bufsize,
> + const struct path *root, unsigned int flags)
> +{
> + struct mount *r, *m = real_mount(mnt);
> + struct path rootmnt = {
> + .mnt = root->mnt,
> + .dentry = root->mnt->mnt_root
> + };
> + long ctr = 0;
> + bool reachable_only = true;
> + int err;
> +
> + err = security_sb_statfs(mnt->mnt_root);
> + if (err)
> + return err;
> +
> + if (flags & LISTMOUNT_UNREACHABLE) {
> + if (!capable(CAP_SYS_ADMIN))
> + return -EPERM;
> + reachable_only = false;
> + }
> +
> + if (reachable_only && !is_path_reachable(m, mnt->mnt_root, &rootmnt))
> + return capable(CAP_SYS_ADMIN) ? 0 : -EPERM;
> +
> + list_for_each_entry(r, &m->mnt_mounts, mnt_child) {
> + if (reachable_only &&
> + !is_path_reachable(r, r->mnt.mnt_root, root))
> + continue;
I believe we would want to move the security_sb_statfs() call from
above to down here; something like this I think ...
err = security_sb_statfs(r->mnt.mnt_root);
if (err)
/* if we can't access the mount, pretend it doesn't exist */
continue;
> + if (ctr >= bufsize)
> + return -EOVERFLOW;
> + if (put_user(r->mnt_id_unique, buf + ctr))
> + return -EFAULT;
> + ctr++;
> + if (ctr < 0)
> + return -ERANGE;
> + }
> + return ctr;
> +}
--
paul-moore.com
On Wed, 4 Oct 2023 at 21:38, Paul Moore <[email protected]> wrote:
>
> On Thu, Sep 28, 2023 at 9:04 AM Miklos Szeredi <[email protected]> wrote:
> >
> > Add way to query the children of a particular mount. This is a more
> > flexible way to iterate the mount tree than having to parse the complete
> > /proc/self/mountinfo.
> >
> > Lookup the mount by the new 64bit mount ID. If a mount needs to be queried
> > based on path, then statx(2) can be used to first query the mount ID
> > belonging to the path.
> >
> > Return an array of new (64bit) mount ID's. Without privileges only mounts
> > are listed which are reachable from the task's root.
> >
> > Signed-off-by: Miklos Szeredi <[email protected]>
> > ---
> > arch/x86/entry/syscalls/syscall_32.tbl | 1 +
> > arch/x86/entry/syscalls/syscall_64.tbl | 1 +
> > fs/namespace.c | 69 ++++++++++++++++++++++++++
> > include/linux/syscalls.h | 3 ++
> > include/uapi/asm-generic/unistd.h | 5 +-
> > include/uapi/linux/mount.h | 3 ++
> > 6 files changed, 81 insertions(+), 1 deletion(-)
>
> ...
>
> > diff --git a/fs/namespace.c b/fs/namespace.c
> > index 3326ba2b2810..050e2d2af110 100644
> > --- a/fs/namespace.c
> > +++ b/fs/namespace.c
> > @@ -4970,6 +4970,75 @@ SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
> > return ret;
> > }
> >
> > +static long do_listmount(struct vfsmount *mnt, u64 __user *buf, size_t bufsize,
> > + const struct path *root, unsigned int flags)
> > +{
> > + struct mount *r, *m = real_mount(mnt);
> > + struct path rootmnt = {
> > + .mnt = root->mnt,
> > + .dentry = root->mnt->mnt_root
> > + };
> > + long ctr = 0;
> > + bool reachable_only = true;
> > + int err;
> > +
> > + err = security_sb_statfs(mnt->mnt_root);
> > + if (err)
> > + return err;
> > +
> > + if (flags & LISTMOUNT_UNREACHABLE) {
> > + if (!capable(CAP_SYS_ADMIN))
> > + return -EPERM;
> > + reachable_only = false;
> > + }
> > +
> > + if (reachable_only && !is_path_reachable(m, mnt->mnt_root, &rootmnt))
> > + return capable(CAP_SYS_ADMIN) ? 0 : -EPERM;
> > +
> > + list_for_each_entry(r, &m->mnt_mounts, mnt_child) {
> > + if (reachable_only &&
> > + !is_path_reachable(r, r->mnt.mnt_root, root))
> > + continue;
>
> I believe we would want to move the security_sb_statfs() call from
> above to down here; something like this I think ...
>
> err = security_sb_statfs(r->mnt.mnt_root);
> if (err)
> /* if we can't access the mount, pretend it doesn't exist */
> continue;
Hmm. Why is this specific to listing mounts (i.e. why doesn't readdir
have a similar filter)?
Also why hasn't this come up with regards to the proc interfaces that
list mounts?
I just want to understand the big picture here.
Thanks,
Miklos
On 5/10/23 12:01, Miklos Szeredi wrote:
> On Wed, 4 Oct 2023 at 21:38, Paul Moore <[email protected]> wrote:
>> On Thu, Sep 28, 2023 at 9:04 AM Miklos Szeredi <[email protected]> wrote:
>>> Add way to query the children of a particular mount. This is a more
>>> flexible way to iterate the mount tree than having to parse the complete
>>> /proc/self/mountinfo.
>>>
>>> Lookup the mount by the new 64bit mount ID. If a mount needs to be queried
>>> based on path, then statx(2) can be used to first query the mount ID
>>> belonging to the path.
>>>
>>> Return an array of new (64bit) mount ID's. Without privileges only mounts
>>> are listed which are reachable from the task's root.
>>>
>>> Signed-off-by: Miklos Szeredi <[email protected]>
>>> ---
>>> arch/x86/entry/syscalls/syscall_32.tbl | 1 +
>>> arch/x86/entry/syscalls/syscall_64.tbl | 1 +
>>> fs/namespace.c | 69 ++++++++++++++++++++++++++
>>> include/linux/syscalls.h | 3 ++
>>> include/uapi/asm-generic/unistd.h | 5 +-
>>> include/uapi/linux/mount.h | 3 ++
>>> 6 files changed, 81 insertions(+), 1 deletion(-)
>> ...
>>
>>> diff --git a/fs/namespace.c b/fs/namespace.c
>>> index 3326ba2b2810..050e2d2af110 100644
>>> --- a/fs/namespace.c
>>> +++ b/fs/namespace.c
>>> @@ -4970,6 +4970,75 @@ SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
>>> return ret;
>>> }
>>>
>>> +static long do_listmount(struct vfsmount *mnt, u64 __user *buf, size_t bufsize,
>>> + const struct path *root, unsigned int flags)
>>> +{
>>> + struct mount *r, *m = real_mount(mnt);
>>> + struct path rootmnt = {
>>> + .mnt = root->mnt,
>>> + .dentry = root->mnt->mnt_root
>>> + };
>>> + long ctr = 0;
>>> + bool reachable_only = true;
>>> + int err;
>>> +
>>> + err = security_sb_statfs(mnt->mnt_root);
>>> + if (err)
>>> + return err;
>>> +
>>> + if (flags & LISTMOUNT_UNREACHABLE) {
>>> + if (!capable(CAP_SYS_ADMIN))
>>> + return -EPERM;
>>> + reachable_only = false;
>>> + }
>>> +
>>> + if (reachable_only && !is_path_reachable(m, mnt->mnt_root, &rootmnt))
>>> + return capable(CAP_SYS_ADMIN) ? 0 : -EPERM;
>>> +
>>> + list_for_each_entry(r, &m->mnt_mounts, mnt_child) {
>>> + if (reachable_only &&
>>> + !is_path_reachable(r, r->mnt.mnt_root, root))
>>> + continue;
>> I believe we would want to move the security_sb_statfs() call from
>> above to down here; something like this I think ...
>>
>> err = security_sb_statfs(r->mnt.mnt_root);
>> if (err)
>> /* if we can't access the mount, pretend it doesn't exist */
>> continue;
> Hmm. Why is this specific to listing mounts (i.e. why doesn't readdir
> have a similar filter)?
>
> Also why hasn't this come up with regards to the proc interfaces that
> list mounts?
The proc interfaces essentially use <mount namespace>->list to provide
the mounts that can be seen so it's filtered by mount namespace of the
task that's doing the open().
See fs/namespace.c:mnt_list_next() and just below the m_start(), m_next(),
etc.
Ian
>
> I just want to understand the big picture here.
>
> Thanks,
> Miklos
On Thu, 5 Oct 2023 at 06:23, Ian Kent <[email protected]> wrote:
> The proc interfaces essentially use <mount namespace>->list to provide
>
> the mounts that can be seen so it's filtered by mount namespace of the
>
> task that's doing the open().
>
>
> See fs/namespace.c:mnt_list_next() and just below the m_start(), m_next(),
/proc/$PID/mountinfo will list the mount namespace of $PID. Whether
current task has permission to do so is decided at open time.
listmount() will list the children of the given mount ID. The mount
ID is looked up in the task's mount namespace, so this cannot be used
to list mounts of other namespaces. It's a more limited interface.
I sort of understand the reasoning behind calling into a security hook
on entry to statmount() and listmount(). And BTW I also think that if
statmount() and listmount() is limited in this way, then the same
limitation should be applied to the proc interfaces. But that needs
to be done real carefully because it might cause regressions. OTOH if
it's only done on the new interfaces, then what is the point, since
the old interfaces will be available indefinitely?
Also I cannot see the point in hiding some mount ID's from the list.
It seems to me that the list is just an array of numbers that in
itself doesn't carry any information.
Thanks,
Miklos
On 5/10/23 23:47, Miklos Szeredi wrote:
> On Thu, 5 Oct 2023 at 06:23, Ian Kent <[email protected]> wrote:
>
>> The proc interfaces essentially use <mount namespace>->list to provide
>>
>> the mounts that can be seen so it's filtered by mount namespace of the
>>
>> task that's doing the open().
>>
>>
>> See fs/namespace.c:mnt_list_next() and just below the m_start(), m_next(),
> /proc/$PID/mountinfo will list the mount namespace of $PID. Whether
> current task has permission to do so is decided at open time.
>
> listmount() will list the children of the given mount ID. The mount
> ID is looked up in the task's mount namespace, so this cannot be used
> to list mounts of other namespaces. It's a more limited interface.
Yep. But isn't the ability to see these based on task privilege?
Is the proc style restriction actually what we need here (or some variation
of that implementation)?
An privileged task typically has the init namespace as its mount namespace
and mounts should propagate from there so it should be able to see all
mounts.
If the file handle has been opened in a task that is using some other mount
namespace then presumably that's what the program author wants the task
to see.
So I'm not sure I see a problem obeying the namespace of a given task.
Ian
>
> I sort of understand the reasoning behind calling into a security hook
> on entry to statmount() and listmount(). And BTW I also think that if
> statmount() and listmount() is limited in this way, then the same
> limitation should be applied to the proc interfaces. But that needs
> to be done real carefully because it might cause regressions. OTOH if
> it's only done on the new interfaces, then what is the point, since
> the old interfaces will be available indefinitely?
>
> Also I cannot see the point in hiding some mount ID's from the list.
> It seems to me that the list is just an array of numbers that in
> itself doesn't carry any information.
>
> Thanks,
> Miklos
On Thu, Oct 5, 2023 at 11:47 AM Miklos Szeredi <[email protected]> wrote:
> On Thu, 5 Oct 2023 at 06:23, Ian Kent <[email protected]> wrote:
> > The proc interfaces essentially use <mount namespace>->list to provide
> >
> > the mounts that can be seen so it's filtered by mount namespace of the
> >
> > task that's doing the open().
> >
> >
> > See fs/namespace.c:mnt_list_next() and just below the m_start(), m_next(),
>
> /proc/$PID/mountinfo will list the mount namespace of $PID. Whether
> current task has permission to do so is decided at open time.
>
> listmount() will list the children of the given mount ID. The mount
> ID is looked up in the task's mount namespace, so this cannot be used
> to list mounts of other namespaces. It's a more limited interface.
>
> I sort of understand the reasoning behind calling into a security hook
> on entry to statmount() and listmount(). And BTW I also think that if
> statmount() and listmount() is limited in this way, then the same
> limitation should be applied to the proc interfaces. But that needs
> to be done real carefully because it might cause regressions. OTOH if
> it's only done on the new interfaces, then what is the point, since
> the old interfaces will be available indefinitely?
LSMs that are designed to enforce access controls on procfs interfaces
typically leverage the fact that the procfs interfaces are file based
and the normal file I/O access controls can be used. In some cases,
e.g. /proc/self/attr, there may also be additional access controls
implemented via a dedicated set of LSM hooks.
> Also I cannot see the point in hiding some mount ID's from the list.
> It seems to me that the list is just an array of numbers that in
> itself doesn't carry any information.
I think it really comes down to the significance of the mount ID, and
I can't say I know enough of the details here to be entirely
comfortable taking a hard stance on this. Can you help me understand
the mount ID concept a bit better?
While I'm reasonably confident that we want a security_sb_statfs()
control point in statmount(), it may turn out that we don't want/need
a call in the listmount() case. Perhaps your original patch was
correct in the sense that we only want a single security_sb_statfs()
call for the root (implying that the child mount IDs are attributes of
the root/parent mount)? Maybe it's something else entirely?
--
paul-moore.com
On Fri, 6 Oct 2023 at 04:56, Paul Moore <[email protected]> wrote:
> > Also I cannot see the point in hiding some mount ID's from the list.
> > It seems to me that the list is just an array of numbers that in
> > itself doesn't carry any information.
>
> I think it really comes down to the significance of the mount ID, and
> I can't say I know enough of the details here to be entirely
> comfortable taking a hard stance on this. Can you help me understand
> the mount ID concept a bit better?
Mount ID is a descriptor that allows referring to a specific struct
mount from userspace.
The old 32 bit mount id is allocated with IDA from a global pool.
Because it's non-referencing it doesn't allow uniquely identifying a
mount. That was a design mistake that I made back in 2008, thinking
that the same sort of dense descriptor space as used for file
descriptors would work. Originally it was used to identify the mount
and the parent mount in /proc/PID/mountinfo. Later it was also added
to the following interfaces:
- name_to_handle_at(2) returns 32 bit value
- /proc/PID/FD/fdinfo
- statx(2) returns 64 bit value
It was never used on the kernel interfaces as an input argument.
statmount(2) and listmount(2) require the mount to be identified by
userspace, so having a unique ID is important. So the "[1/4] add
unique mount ID" adds a new 64 bit ID (still global) that is allocated
sequentially and only reused after reboot. It is used as an input to
these syscalls. It is returned by statx(2) if requested by
STATX_MNT_ID_UNIQUE and as an array of ID's by listmount(2).
I can see mild security problems with the global allocation, since a
task can observe mounts being done in other namespaces. This doesn't
sound too serious, and the old ID has similar issues. But I think
making the new ID be local to the mount namespace is also feasible.
> While I'm reasonably confident that we want a security_sb_statfs()
> control point in statmount(), it may turn out that we don't want/need
> a call in the listmount() case. Perhaps your original patch was
> correct in the sense that we only want a single security_sb_statfs()
> call for the root (implying that the child mount IDs are attributes of
> the root/parent mount)? Maybe it's something else entirely?
Mounts are arranged in a tree (I think it obvious how) and
listmount(2) just lists the IDs of the immediate children of a mount.
I don't see ID being an attribute of a mount, it's a descriptor.
Thanks,
Miklos
On Fri, Oct 6, 2023 at 4:53 AM Miklos Szeredi <[email protected]> wrote:
> On Fri, 6 Oct 2023 at 04:56, Paul Moore <[email protected]> wrote:
>
> > > Also I cannot see the point in hiding some mount ID's from the list.
> > > It seems to me that the list is just an array of numbers that in
> > > itself doesn't carry any information.
> >
> > I think it really comes down to the significance of the mount ID, and
> > I can't say I know enough of the details here to be entirely
> > comfortable taking a hard stance on this. Can you help me understand
> > the mount ID concept a bit better?
>
> Mount ID is a descriptor that allows referring to a specific struct
> mount from userspace.
>
> The old 32 bit mount id is allocated with IDA from a global pool.
> Because it's non-referencing it doesn't allow uniquely identifying a
> mount. That was a design mistake that I made back in 2008, thinking
> that the same sort of dense descriptor space as used for file
> descriptors would work. Originally it was used to identify the mount
> and the parent mount in /proc/PID/mountinfo. Later it was also added
> to the following interfaces:
>
> - name_to_handle_at(2) returns 32 bit value
> - /proc/PID/FD/fdinfo
> - statx(2) returns 64 bit value
>
> It was never used on the kernel interfaces as an input argument.
Thanks for the background.
> statmount(2) and listmount(2) require the mount to be identified by
> userspace, so having a unique ID is important. So the "[1/4] add
> unique mount ID" adds a new 64 bit ID (still global) that is allocated
> sequentially and only reused after reboot. It is used as an input to
> these syscalls. It is returned by statx(2) if requested by
> STATX_MNT_ID_UNIQUE and as an array of ID's by listmount(2).
>
> I can see mild security problems with the global allocation, since a
> task can observe mounts being done in other namespaces. This doesn't
> sound too serious, and the old ID has similar issues. But I think
> making the new ID be local to the mount namespace is also feasible.
The LSM hook API is designed to operate independently from any of the
kernel namespaces; while some LSMs may choose to be aware of
namespaces and adjust their controls accordingly, there is no
requirement that they do so. For that reason, I'm not too bothered
either way if the mount ID is global or tied to a namespace.
> > While I'm reasonably confident that we want a security_sb_statfs()
> > control point in statmount(), it may turn out that we don't want/need
> > a call in the listmount() case. Perhaps your original patch was
> > correct in the sense that we only want a single security_sb_statfs()
> > call for the root (implying that the child mount IDs are attributes of
> > the root/parent mount)? Maybe it's something else entirely?
>
> Mounts are arranged in a tree (I think it obvious how) and
> listmount(2) just lists the IDs of the immediate children of a mount.
>
> I don't see ID being an attribute of a mount, it's a descriptor.
In this case I think the approach you took originally in this thread
is likely what we want, call security_sb_statfs() against the root
mount in listmount(). Just please move it after the capability
checks.
If you look at the two LSMs which implement the security_sb_statfs(),
Smack and SELinux, you see that Smack treats this as a read operation
between the current process and the specified mount and SELinux treats
this as a filesystem:getattr operations between the current process
and the specified mount. In both cases I can see that being the right
approach for reading a list of child mounts off of a root mount.
Does that sound good? I'm guessing that's okay since that was how you
wrote it in your original patch, but there has been a lot of
discussion since then :)
--
paul-moore.com
On 6/10/23 08:27, Ian Kent wrote:
> On 5/10/23 23:47, Miklos Szeredi wrote:
>> On Thu, 5 Oct 2023 at 06:23, Ian Kent <[email protected]> wrote:
>>
>>> The proc interfaces essentially use <mount namespace>->list to provide
>>>
>>> the mounts that can be seen so it's filtered by mount namespace of the
>>>
>>> task that's doing the open().
>>>
>>>
>>> See fs/namespace.c:mnt_list_next() and just below the m_start(),
>>> m_next(),
>> /proc/$PID/mountinfo will list the mount namespace of $PID. Whether
>> current task has permission to do so is decided at open time.
>>
>> listmount() will list the children of the given mount ID. The mount
>> ID is looked up in the task's mount namespace, so this cannot be used
>> to list mounts of other namespaces. It's a more limited interface.
>
> Yep. But isn't the ability to see these based on task privilege?
>
>
> Is the proc style restriction actually what we need here (or some
> variation
>
> of that implementation)?
>
>
> An privileged task typically has the init namespace as its mount
> namespace
>
> and mounts should propagate from there so it should be able to see all
> mounts.
>
>
> If the file handle has been opened in a task that is using some other
> mount
>
> namespace then presumably that's what the program author wants the
> task to see.
>
> So I'm not sure I see a problem obeying the namespace of a given task.
I've had a look through the code we had in the old fsinfo() proposal
because I think we need to consider the use cases that are needed.
IIRC initially we had a flag FSINFO_ATTR_MOUNT_CHILDREN that essentially
enumerated the children of the given mount in much the same way as is
done now in this system call.
But because we needed to enumerate mounts in the same way as the proc file
system mount tables a flag FSINFO_ATTR_MOUNT_ALL was added that essentially
used the mount namespace mounts list in a similar way to the proc file
system so that a list of mounts for a mount namespace could be retrieved.
This later use case is what is used by processes that monitor mounts and
is what's needed more so than enumerating the children as we do now.
I'm still looking at the mount id lookup.
Ian
>
>
> Ian
>
>>
>> I sort of understand the reasoning behind calling into a security hook
>> on entry to statmount() and listmount(). And BTW I also think that if
>> statmount() and listmount() is limited in this way, then the same
>> limitation should be applied to the proc interfaces. But that needs
>> to be done real carefully because it might cause regressions. OTOH if
>> it's only done on the new interfaces, then what is the point, since
>> the old interfaces will be available indefinitely?
>>
>> Also I cannot see the point in hiding some mount ID's from the list.
>> It seems to me that the list is just an array of numbers that in
>> itself doesn't carry any information.
>>
>> Thanks,
>> Miklos
On Fri, 13 Oct 2023 at 04:40, Ian Kent <[email protected]> wrote:
> But because we needed to enumerate mounts in the same way as the proc file
>
> system mount tables a flag FSINFO_ATTR_MOUNT_ALL was added that essentially
>
> used the mount namespace mounts list in a similar way to the proc file
>
> system so that a list of mounts for a mount namespace could be retrieved.
Makes sense.
Added a LISTMOUNT_RECURSIVE flag.
Thanks,
Miklos