Message ID | 158204553565.3299825.3864357054582488949.stgit@warthog.procyon.org.uk (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | VFS: Filesystem information and notifications [ver #16] | expand |
On Tue, Feb 18, 2020 at 05:05:35PM +0000, David Howells wrote: > Introduce an (effectively) non-repeating system-unique superblock ID that > can be used to determine that two object are in the same superblock without > risking reuse of the ID in the meantime (as is possible with device IDs). > > The ID is time-based to make it harder to use it as a covert communications > channel. > > Also make it so that this ID can be fetched by the fsinfo() system call. > The ID added so that superblock notification messages will also be able to > be tagged with it. > > Signed-off-by: David Howells <dhowells@redhat.com> > --- > > fs/fsinfo.c | 1 + > fs/super.c | 24 ++++++++++++++++++++++++ > include/linux/fs.h | 3 +++ > samples/vfs/test-fsinfo.c | 1 + > 4 files changed, 29 insertions(+) > > diff --git a/fs/fsinfo.c b/fs/fsinfo.c > index 55710d6da327..f8e85762fc47 100644 > --- a/fs/fsinfo.c > +++ b/fs/fsinfo.c > @@ -92,6 +92,7 @@ static int fsinfo_generic_ids(struct path *path, struct fsinfo_context *ctx) > p->f_fstype = sb->s_magic; > p->f_dev_major = MAJOR(sb->s_dev); > p->f_dev_minor = MINOR(sb->s_dev); > + p->f_sb_id = sb->s_unique_id; Ahah, this is what the f_sb_id field is for. I noticed a few patches ago that it was in a header file but was never set. I'm losing track of which IDs do what... * f_fsid is that old int[2] thing that we used for statfs. It sucks but we can't remove it because it's been in statfs since the beginning of time. * f_fs_name is a string coming from s_type, which is the name of the fs (e.g. "XFS")? * f_fstype comes from s_magic, which (for XFS) is 0x58465342. * f_sb_id is basically an incore u64 cookie that one can use with the mount events thing that comes later in this patchset? * FSINFO_ATTR_VOLUME_ID comes from s_id, which tends to be the block device name (at least for local filesystems) * FSINFO_ATTR_VOLUME_UUID comes from s_uuid, which some filesystems fill in at mount time. * FSINFO_ATTR_VOLUME_NAME is ... left to individual filesystems to implement, and (AFAICT) can be the label that one uses for things like: "mount LABEL=foo /home" ? Assuming I got all of that right, can we please capture what all of these "IDs" mean in the documentation? (Assuming I got all that right, the code looks ok.) --D > > memcpy(&p->f_fsid, &buf.f_fsid, sizeof(p->f_fsid)); > strlcpy(p->f_fs_name, path->dentry->d_sb->s_type->name, > diff --git a/fs/super.c b/fs/super.c > index cd352530eca9..a63073e6127e 100644 > --- a/fs/super.c > +++ b/fs/super.c > @@ -44,6 +44,8 @@ static int thaw_super_locked(struct super_block *sb); > > static LIST_HEAD(super_blocks); > static DEFINE_SPINLOCK(sb_lock); > +static u64 sb_last_identifier; > +static u64 sb_identifier_offset; > > static char *sb_writers_name[SB_FREEZE_LEVELS] = { > "sb_writers", > @@ -188,6 +190,27 @@ static void destroy_unused_super(struct super_block *s) > destroy_super_work(&s->destroy_work); > } > > +/* > + * Generate a unique identifier for a superblock. > + */ > +static void generate_super_id(struct super_block *s) > +{ > + u64 id = ktime_to_ns(ktime_get()); > + > + spin_lock(&sb_lock); > + > + id += sb_identifier_offset; > + if (id <= sb_last_identifier) { > + id = sb_last_identifier + 1; > + sb_identifier_offset = sb_last_identifier - id; > + } > + > + sb_last_identifier = id; > + spin_unlock(&sb_lock); > + > + s->s_unique_id = id; > +} > + > /** > * alloc_super - create new superblock > * @type: filesystem type superblock should belong to > @@ -273,6 +296,7 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags, > goto fail; > if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) > goto fail; > + generate_super_id(s); > return s; > > fail: > diff --git a/include/linux/fs.h b/include/linux/fs.h > index f74a4ee36eb3..e5db22d536a3 100644 > --- a/include/linux/fs.h > +++ b/include/linux/fs.h > @@ -1550,6 +1550,9 @@ struct super_block { > > spinlock_t s_inode_wblist_lock; > struct list_head s_inodes_wb; /* writeback inodes */ > + > + /* Superblock event notifications */ > + u64 s_unique_id; > } __randomize_layout; > > /* Helper functions so that in most cases filesystems will > diff --git a/samples/vfs/test-fsinfo.c b/samples/vfs/test-fsinfo.c > index 6fbf0ce099b2..d6ec5713364f 100644 > --- a/samples/vfs/test-fsinfo.c > +++ b/samples/vfs/test-fsinfo.c > @@ -140,6 +140,7 @@ static void dump_fsinfo_generic_ids(void *reply, unsigned int size) > printf("\tdev : %02x:%02x\n", f->f_dev_major, f->f_dev_minor); > printf("\tfs : type=%x name=%s\n", f->f_fstype, f->f_fs_name); > printf("\tfsid : %llx\n", (unsigned long long)f->f_fsid); > + printf("\tsbid : %llx\n", (unsigned long long)f->f_sb_id); > } > > static void dump_fsinfo_generic_limits(void *reply, unsigned int size) > >
Darrick J. Wong <darrick.wong@oracle.com> wrote: > Ahah, this is what the f_sb_id field is for. I noticed a few patches > ago that it was in a header file but was never set. > > I'm losing track of which IDs do what... > > * f_fsid is that old int[2] thing that we used for statfs. It sucks but > we can't remove it because it's been in statfs since the beginning of > time. > > * f_fs_name is a string coming from s_type, which is the name of the fs > (e.g. "XFS")? > > * f_fstype comes from s_magic, which (for XFS) is 0x58465342. > > * f_sb_id is basically an incore u64 cookie that one can use with the > mount events thing that comes later in this patchset? > > * FSINFO_ATTR_VOLUME_ID comes from s_id, which tends to be the block > device name (at least for local filesystems) > > * FSINFO_ATTR_VOLUME_UUID comes from s_uuid, which some filesystems fill > in at mount time. > > * FSINFO_ATTR_VOLUME_NAME is ... left to individual filesystems to > implement, and (AFAICT) can be the label that one uses for things > like: "mount LABEL=foo /home" ? > > Assuming I got all of that right, can we please capture what all of > these "IDs" mean in the documentation? Basically, yes. Would it help if I: (1) Put the ID generation into its own patch, first. (2) Put the notification counter patches right after that. (3) Renamed the fields a bit, say: f_fsid -> fsid f_fs_name -> filesystem_name f_fstype -> filesystem_magic f_sb_id -> superblock_id f_dev_* -> backing_dev_* David
diff --git a/fs/fsinfo.c b/fs/fsinfo.c index 55710d6da327..f8e85762fc47 100644 --- a/fs/fsinfo.c +++ b/fs/fsinfo.c @@ -92,6 +92,7 @@ static int fsinfo_generic_ids(struct path *path, struct fsinfo_context *ctx) p->f_fstype = sb->s_magic; p->f_dev_major = MAJOR(sb->s_dev); p->f_dev_minor = MINOR(sb->s_dev); + p->f_sb_id = sb->s_unique_id; memcpy(&p->f_fsid, &buf.f_fsid, sizeof(p->f_fsid)); strlcpy(p->f_fs_name, path->dentry->d_sb->s_type->name, diff --git a/fs/super.c b/fs/super.c index cd352530eca9..a63073e6127e 100644 --- a/fs/super.c +++ b/fs/super.c @@ -44,6 +44,8 @@ static int thaw_super_locked(struct super_block *sb); static LIST_HEAD(super_blocks); static DEFINE_SPINLOCK(sb_lock); +static u64 sb_last_identifier; +static u64 sb_identifier_offset; static char *sb_writers_name[SB_FREEZE_LEVELS] = { "sb_writers", @@ -188,6 +190,27 @@ static void destroy_unused_super(struct super_block *s) destroy_super_work(&s->destroy_work); } +/* + * Generate a unique identifier for a superblock. + */ +static void generate_super_id(struct super_block *s) +{ + u64 id = ktime_to_ns(ktime_get()); + + spin_lock(&sb_lock); + + id += sb_identifier_offset; + if (id <= sb_last_identifier) { + id = sb_last_identifier + 1; + sb_identifier_offset = sb_last_identifier - id; + } + + sb_last_identifier = id; + spin_unlock(&sb_lock); + + s->s_unique_id = id; +} + /** * alloc_super - create new superblock * @type: filesystem type superblock should belong to @@ -273,6 +296,7 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags, goto fail; if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) goto fail; + generate_super_id(s); return s; fail: diff --git a/include/linux/fs.h b/include/linux/fs.h index f74a4ee36eb3..e5db22d536a3 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -1550,6 +1550,9 @@ struct super_block { spinlock_t s_inode_wblist_lock; struct list_head s_inodes_wb; /* writeback inodes */ + + /* Superblock event notifications */ + u64 s_unique_id; } __randomize_layout; /* Helper functions so that in most cases filesystems will diff --git a/samples/vfs/test-fsinfo.c b/samples/vfs/test-fsinfo.c index 6fbf0ce099b2..d6ec5713364f 100644 --- a/samples/vfs/test-fsinfo.c +++ b/samples/vfs/test-fsinfo.c @@ -140,6 +140,7 @@ static void dump_fsinfo_generic_ids(void *reply, unsigned int size) printf("\tdev : %02x:%02x\n", f->f_dev_major, f->f_dev_minor); printf("\tfs : type=%x name=%s\n", f->f_fstype, f->f_fs_name); printf("\tfsid : %llx\n", (unsigned long long)f->f_fsid); + printf("\tsbid : %llx\n", (unsigned long long)f->f_sb_id); } static void dump_fsinfo_generic_limits(void *reply, unsigned int size)
Introduce an (effectively) non-repeating system-unique superblock ID that can be used to determine that two object are in the same superblock without risking reuse of the ID in the meantime (as is possible with device IDs). The ID is time-based to make it harder to use it as a covert communications channel. Also make it so that this ID can be fetched by the fsinfo() system call. The ID added so that superblock notification messages will also be able to be tagged with it. Signed-off-by: David Howells <dhowells@redhat.com> --- fs/fsinfo.c | 1 + fs/super.c | 24 ++++++++++++++++++++++++ include/linux/fs.h | 3 +++ samples/vfs/test-fsinfo.c | 1 + 4 files changed, 29 insertions(+)