| Message ID | 20250129165803.72138-3-mszeredi@redhat.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | mount notification | expand |
On Wed, Jan 29, 2025 at 11:58 AM Miklos Szeredi <mszeredi@redhat.com> wrote: > > Add notifications for attaching and detaching mounts. The following new > event masks are added: > > FAN_MNT_ATTACH - Mount was attached > FAN_MNT_DETACH - Mount was detached > > If a mount is moved, then the event is reported with (FAN_MNT_ATTACH | > FAN_MNT_DETACH). > > These events add an info record of type FAN_EVENT_INFO_TYPE_MNT containing > these fields identifying the affected mounts: > > __u64 mnt_id - the ID of the mount (see statmount(2)) > > FAN_REPORT_MNT must be supplied to fanotify_init() to receive these events > and no other type of event can be received with this report type. > > Marks are added with FAN_MARK_MNTNS, which records the mount namespace from > an nsfs file (e.g. /proc/self/ns/mnt). > > Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> > --- > fs/mount.h | 2 + > fs/namespace.c | 14 +++-- > fs/notify/fanotify/fanotify.c | 38 +++++++++++-- > fs/notify/fanotify/fanotify.h | 18 +++++++ > fs/notify/fanotify/fanotify_user.c | 87 +++++++++++++++++++++++++----- > fs/notify/fdinfo.c | 5 ++ > include/linux/fanotify.h | 12 +++-- > include/uapi/linux/fanotify.h | 10 ++++ > security/selinux/hooks.c | 4 ++ > 9 files changed, 167 insertions(+), 23 deletions(-) ... > diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c > index 7b867dfec88b..06d073eab53c 100644 > --- a/security/selinux/hooks.c > +++ b/security/selinux/hooks.c > @@ -3395,6 +3395,10 @@ static int selinux_path_notify(const struct path *path, u64 mask, > case FSNOTIFY_OBJ_TYPE_INODE: > perm = FILE__WATCH; > break; > + case FSNOTIFY_OBJ_TYPE_MNTNS: > + /* Maybe introduce FILE__WATCH_MOUNTNS? */ > + perm = FILE__WATCH_MOUNT; > + break; > default: > return -EINVAL; > } Ignoring for a moment that this patch was merged without an explicit ACK for the SELinux changes, let's talk about these SELinux changes ... I understand that you went with the "simpler version" because you didn't believe the discussion was converging, which is fair, however, I believe Daniel's argument is convincing enough to warrant the new permission. Yes, it has taken me approximately two days to find the time to revisit this topic and reply with some clarity, but personally I feel like that is not an unreasonable period of time, especially for a new feature discussion occurring during the merge window. If you need an example on how to add a new SELinux permission, you can look at commit ed5d44d42c95 ("selinux: Implement userns_create hook") for a fairly simple example. In the watch_mountns case things are slightly different due to the existence of the COMMON_FILE_PERMS macro, but you basically want to add "watch_mountns" to the end of the COMMON_FILE_PERMS macro in security/selinux/include/classmap.h. Of course if you aren't sure about something, let me know. As a FYI, my network access will be spotty starting tonight and extending through the weekend, but I will have network/mail access at least once a day. Now back to the merge into the VFS tree ... I was very surprised to open this patchset and see that Christian had merged v5 after less than 24 hours (at least according to the email timestamps that I see) and without an explicit ACK for the SELinux changes. I've mentioned this to you before Christian, please do not merge any SELinux, LSM framework, or audit related patches without an explicit ACK. I recognize that sometimes there are highly critical security issues that need immediate attention, but that is clearly not the case here, and there are other procedures to help deal with those emergency scenarios.
On Thu, 30 Jan 2025 at 22:06, Paul Moore <paul@paul-moore.com> wrote: > > On Wed, Jan 29, 2025 at 11:58 AM Miklos Szeredi <mszeredi@redhat.com> wrote: > > > > Add notifications for attaching and detaching mounts. The following new > > event masks are added: > > > > FAN_MNT_ATTACH - Mount was attached > > FAN_MNT_DETACH - Mount was detached > > > > If a mount is moved, then the event is reported with (FAN_MNT_ATTACH | > > FAN_MNT_DETACH). > > > > These events add an info record of type FAN_EVENT_INFO_TYPE_MNT containing > > these fields identifying the affected mounts: > > > > __u64 mnt_id - the ID of the mount (see statmount(2)) > > > > FAN_REPORT_MNT must be supplied to fanotify_init() to receive these events > > and no other type of event can be received with this report type. > > > > Marks are added with FAN_MARK_MNTNS, which records the mount namespace from > > an nsfs file (e.g. /proc/self/ns/mnt). > > > > Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> > > --- > > fs/mount.h | 2 + > > fs/namespace.c | 14 +++-- > > fs/notify/fanotify/fanotify.c | 38 +++++++++++-- > > fs/notify/fanotify/fanotify.h | 18 +++++++ > > fs/notify/fanotify/fanotify_user.c | 87 +++++++++++++++++++++++++----- > > fs/notify/fdinfo.c | 5 ++ > > include/linux/fanotify.h | 12 +++-- > > include/uapi/linux/fanotify.h | 10 ++++ > > security/selinux/hooks.c | 4 ++ > > 9 files changed, 167 insertions(+), 23 deletions(-) > > ... > > > diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c > > index 7b867dfec88b..06d073eab53c 100644 > > --- a/security/selinux/hooks.c > > +++ b/security/selinux/hooks.c > > @@ -3395,6 +3395,10 @@ static int selinux_path_notify(const struct path *path, u64 mask, > > case FSNOTIFY_OBJ_TYPE_INODE: > > perm = FILE__WATCH; > > break; > > + case FSNOTIFY_OBJ_TYPE_MNTNS: > > + /* Maybe introduce FILE__WATCH_MOUNTNS? */ > > + perm = FILE__WATCH_MOUNT; > > + break; > > default: > > return -EINVAL; > > } > > Ignoring for a moment that this patch was merged without an explicit > ACK for the SELinux changes, let's talk about these SELinux changes > ... > > I understand that you went with the "simpler version" because you > didn't believe the discussion was converging, which is fair, however, > I believe Daniel's argument is convincing enough to warrant the new > permission. Fine, I'll work on this. > Yes, it has taken me approximately two days to find the > time to revisit this topic and reply with some clarity, but personally > I feel like that is not an unreasonable period of time, especially for > a new feature discussion occurring during the merge window. Definitely not. Christian is definitely very responsive and quick to queue things up, and that can have drawbacks. In this he made it clear that he wants to get this queued ASAP regardless of whether there's decision on the SELinux side or not. What I think might be a good thing if Christian could record conditional NAKs such as this one from you, that need to be worked on before sending a feature upstream. That would prevent wrong code being sent upstream due to lack of attention. Thanks, Miklos
On Thu, Jan 30, 2025 at 04:05:53PM -0500, Paul Moore wrote: > On Wed, Jan 29, 2025 at 11:58 AM Miklos Szeredi <mszeredi@redhat.com> wrote: > > > > Add notifications for attaching and detaching mounts. The following new > > event masks are added: > > > > FAN_MNT_ATTACH - Mount was attached > > FAN_MNT_DETACH - Mount was detached > > > > If a mount is moved, then the event is reported with (FAN_MNT_ATTACH | > > FAN_MNT_DETACH). > > > > These events add an info record of type FAN_EVENT_INFO_TYPE_MNT containing > > these fields identifying the affected mounts: > > > > __u64 mnt_id - the ID of the mount (see statmount(2)) > > > > FAN_REPORT_MNT must be supplied to fanotify_init() to receive these events > > and no other type of event can be received with this report type. > > > > Marks are added with FAN_MARK_MNTNS, which records the mount namespace from > > an nsfs file (e.g. /proc/self/ns/mnt). > > > > Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> > > --- > > fs/mount.h | 2 + > > fs/namespace.c | 14 +++-- > > fs/notify/fanotify/fanotify.c | 38 +++++++++++-- > > fs/notify/fanotify/fanotify.h | 18 +++++++ > > fs/notify/fanotify/fanotify_user.c | 87 +++++++++++++++++++++++++----- > > fs/notify/fdinfo.c | 5 ++ > > include/linux/fanotify.h | 12 +++-- > > include/uapi/linux/fanotify.h | 10 ++++ > > security/selinux/hooks.c | 4 ++ > > 9 files changed, 167 insertions(+), 23 deletions(-) > > ... > > > diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c > > index 7b867dfec88b..06d073eab53c 100644 > > --- a/security/selinux/hooks.c > > +++ b/security/selinux/hooks.c > > @@ -3395,6 +3395,10 @@ static int selinux_path_notify(const struct path *path, u64 mask, > > case FSNOTIFY_OBJ_TYPE_INODE: > > perm = FILE__WATCH; > > break; > > + case FSNOTIFY_OBJ_TYPE_MNTNS: > > + /* Maybe introduce FILE__WATCH_MOUNTNS? */ > > + perm = FILE__WATCH_MOUNT; > > + break; > > default: > > return -EINVAL; > > } > > Ignoring for a moment that this patch was merged without an explicit > ACK for the SELinux changes, let's talk about these SELinux changes > ... > > I understand that you went with the "simpler version" because you > didn't believe the discussion was converging, which is fair, however, > I believe Daniel's argument is convincing enough to warrant the new > permission. Yes, it has taken me approximately two days to find the > time to revisit this topic and reply with some clarity, but personally > I feel like that is not an unreasonable period of time, especially for > a new feature discussion occurring during the merge window. > > If you need an example on how to add a new SELinux permission, you can > look at commit ed5d44d42c95 ("selinux: Implement userns_create hook") > for a fairly simple example. In the watch_mountns case things are > slightly different due to the existence of the COMMON_FILE_PERMS > macro, but you basically want to add "watch_mountns" to the end of the > COMMON_FILE_PERMS macro in security/selinux/include/classmap.h. Of > course if you aren't sure about something, let me know. As a FYI, my > network access will be spotty starting tonight and extending through > the weekend, but I will have network/mail access at least once a day. > > Now back to the merge into the VFS tree ... I was very surprised to > open this patchset and see that Christian had merged v5 after less > than 24 hours (at least according to the email timestamps that I see) > and without an explicit ACK for the SELinux changes. I've mentioned > this to you before Christian, please do not merge any SELinux, LSM > framework, or audit related patches without an explicit ACK. I Things go into the tree for testing when the VFS side is ready for testing. We're at v5 and the patchset has gone through four iterations over multiple months. It will go into linux-next and fs-next now for as much expsure as possible. I'm not sure what the confusion between merging things into a tree and sending things upstream is. I have explained this to you before. The application message is also pretty clear about that. > recognize that sometimes there are highly critical security issues > that need immediate attention, but that is clearly not the case here, > and there are other procedures to help deal with those emergency > scenarios. > > -- > paul-moore.com
On Fri, Jan 31, 2025 at 5:53 AM Miklos Szeredi <miklos@szeredi.hu> wrote: > On Thu, 30 Jan 2025 at 22:06, Paul Moore <paul@paul-moore.com> wrote: > > On Wed, Jan 29, 2025 at 11:58 AM Miklos Szeredi <mszeredi@redhat.com> wrote: > > > > > > Add notifications for attaching and detaching mounts. The following new > > > event masks are added: > > > > > > FAN_MNT_ATTACH - Mount was attached > > > FAN_MNT_DETACH - Mount was detached > > > > > > If a mount is moved, then the event is reported with (FAN_MNT_ATTACH | > > > FAN_MNT_DETACH). > > > > > > These events add an info record of type FAN_EVENT_INFO_TYPE_MNT containing > > > these fields identifying the affected mounts: > > > > > > __u64 mnt_id - the ID of the mount (see statmount(2)) > > > > > > FAN_REPORT_MNT must be supplied to fanotify_init() to receive these events > > > and no other type of event can be received with this report type. > > > > > > Marks are added with FAN_MARK_MNTNS, which records the mount namespace from > > > an nsfs file (e.g. /proc/self/ns/mnt). > > > > > > Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> > > > --- > > > fs/mount.h | 2 + > > > fs/namespace.c | 14 +++-- > > > fs/notify/fanotify/fanotify.c | 38 +++++++++++-- > > > fs/notify/fanotify/fanotify.h | 18 +++++++ > > > fs/notify/fanotify/fanotify_user.c | 87 +++++++++++++++++++++++++----- > > > fs/notify/fdinfo.c | 5 ++ > > > include/linux/fanotify.h | 12 +++-- > > > include/uapi/linux/fanotify.h | 10 ++++ > > > security/selinux/hooks.c | 4 ++ > > > 9 files changed, 167 insertions(+), 23 deletions(-) > > > > ... > > > > > diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c > > > index 7b867dfec88b..06d073eab53c 100644 > > > --- a/security/selinux/hooks.c > > > +++ b/security/selinux/hooks.c > > > @@ -3395,6 +3395,10 @@ static int selinux_path_notify(const struct path *path, u64 mask, > > > case FSNOTIFY_OBJ_TYPE_INODE: > > > perm = FILE__WATCH; > > > break; > > > + case FSNOTIFY_OBJ_TYPE_MNTNS: > > > + /* Maybe introduce FILE__WATCH_MOUNTNS? */ > > > + perm = FILE__WATCH_MOUNT; > > > + break; > > > default: > > > return -EINVAL; > > > } > > > > Ignoring for a moment that this patch was merged without an explicit > > ACK for the SELinux changes, let's talk about these SELinux changes > > ... > > > > I understand that you went with the "simpler version" because you > > didn't believe the discussion was converging, which is fair, however, > > I believe Daniel's argument is convincing enough to warrant the new > > permission. > > Fine, I'll work on this. Great, thanks. > > Yes, it has taken me approximately two days to find the > > time to revisit this topic and reply with some clarity, but personally > > I feel like that is not an unreasonable period of time, especially for > > a new feature discussion occurring during the merge window. > > Definitely not. > > Christian is definitely very responsive and quick to queue things up, > and that can have drawbacks. In this he made it clear that he wants > to get this queued ASAP regardless of whether there's decision on the > SELinux side or not. When one merges code that affects another subsystem without an explicit ACK from the affected subsystem when the maintainer has asked for others to clear the code change with an ACK, it's hard to see that as anything but bad behavior at its best and reckless behavior at its worst. It is doubly troubling in cases like this where the code change is user visible. > What I think might be a good thing if Christian could record > conditional NAKs such as this one from you, that need to be worked on > before sending a feature upstream. That would prevent wrong code > being sent upstream due to lack of attention. Christian's merge notification email already has this section: "Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it." ... and to be fair, the vfs-6.15.mount branch mentioned in the notification does appear to be gone.
On Fri, Jan 31, 2025 at 7:09 AM Christian Brauner <brauner@kernel.org> wrote: > On Thu, Jan 30, 2025 at 04:05:53PM -0500, Paul Moore wrote: > > > > Now back to the merge into the VFS tree ... I was very surprised to > > open this patchset and see that Christian had merged v5 after less > > than 24 hours (at least according to the email timestamps that I see) > > and without an explicit ACK for the SELinux changes. I've mentioned > > this to you before Christian, please do not merge any SELinux, LSM > > framework, or audit related patches without an explicit ACK. I > > Things go into the tree for testing when the VFS side is ready for > testing. We're at v5 and the patchset has gone through four iterations > over multiple months. It will go into linux-next and fs-next now for as > much expsure as possible. > > I'm not sure what the confusion between merging things into a tree and > sending things upstream is. I have explained this to you before. The > application message is also pretty clear about that. I'm not sure what the confusion is around my explicit request that you refrain from merging anything that touches the LSM framework, SELinux, or the audit subsystem without an explicit ACK. I have explained this to you before. For the record, your application/merge email makes no statement about only sending patches to Linus that have been ACK'd by all relevant parties. The only statement I can see in your email that remotely relates to ACKs is this: "It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated." ... which once again makes no claims about holding back changes that have not been properly ACK'd.
diff --git a/fs/mount.h b/fs/mount.h index 82aa3bad7cf5..5324a931b403 100644 --- a/fs/mount.h +++ b/fs/mount.h @@ -181,3 +181,5 @@ static inline struct mnt_namespace *to_mnt_ns(struct ns_common *ns) { return container_of(ns, struct mnt_namespace, ns); } + +struct mnt_namespace *mnt_ns_from_dentry(struct dentry *dentry); diff --git a/fs/namespace.c b/fs/namespace.c index 4013fbac354a..d8d70da56e7b 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -2145,16 +2145,24 @@ struct mnt_namespace *get_sequential_mnt_ns(struct mnt_namespace *mntns, bool pr } } +struct mnt_namespace *mnt_ns_from_dentry(struct dentry *dentry) +{ + if (!is_mnt_ns_file(dentry)) + return NULL; + + return to_mnt_ns(get_proc_ns(dentry->d_inode)); +} + static bool mnt_ns_loop(struct dentry *dentry) { /* Could bind mounting the mount namespace inode cause a * mount namespace loop? */ - struct mnt_namespace *mnt_ns; - if (!is_mnt_ns_file(dentry)) + struct mnt_namespace *mnt_ns = mnt_ns_from_dentry(dentry); + + if (!mnt_ns) return false; - mnt_ns = to_mnt_ns(get_proc_ns(dentry->d_inode)); return current->nsproxy->mnt_ns->seq >= mnt_ns->seq; } diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c index 95646f7c46ca..6d386080faf2 100644 --- a/fs/notify/fanotify/fanotify.c +++ b/fs/notify/fanotify/fanotify.c @@ -166,6 +166,8 @@ static bool fanotify_should_merge(struct fanotify_event *old, case FANOTIFY_EVENT_TYPE_FS_ERROR: return fanotify_error_event_equal(FANOTIFY_EE(old), FANOTIFY_EE(new)); + case FANOTIFY_EVENT_TYPE_MNT: + return false; default: WARN_ON_ONCE(1); } @@ -312,7 +314,10 @@ static u32 fanotify_group_event_mask(struct fsnotify_group *group, pr_debug("%s: report_mask=%x mask=%x data=%p data_type=%d\n", __func__, iter_info->report_mask, event_mask, data, data_type); - if (!fid_mode) { + if (FAN_GROUP_FLAG(group, FAN_REPORT_MNT)) { + if (data_type != FSNOTIFY_EVENT_MNT) + return 0; + } else if (!fid_mode) { /* Do we have path to open a file descriptor? */ if (!path) return 0; @@ -557,6 +562,20 @@ static struct fanotify_event *fanotify_alloc_path_event(const struct path *path, return &pevent->fae; } +static struct fanotify_event *fanotify_alloc_mnt_event(u64 mnt_id, gfp_t gfp) +{ + struct fanotify_mnt_event *pevent; + + pevent = kmem_cache_alloc(fanotify_mnt_event_cachep, gfp); + if (!pevent) + return NULL; + + pevent->fae.type = FANOTIFY_EVENT_TYPE_MNT; + pevent->mnt_id = mnt_id; + + return &pevent->fae; +} + static struct fanotify_event *fanotify_alloc_perm_event(const void *data, int data_type, gfp_t gfp) @@ -731,6 +750,7 @@ static struct fanotify_event *fanotify_alloc_event( fid_mode); struct inode *dirid = fanotify_dfid_inode(mask, data, data_type, dir); const struct path *path = fsnotify_data_path(data, data_type); + u64 mnt_id = fsnotify_data_mnt_id(data, data_type); struct mem_cgroup *old_memcg; struct dentry *moved = NULL; struct inode *child = NULL; @@ -826,8 +846,12 @@ static struct fanotify_event *fanotify_alloc_event( moved, &hash, gfp); } else if (fid_mode) { event = fanotify_alloc_fid_event(id, fsid, &hash, gfp); - } else { + } else if (path) { event = fanotify_alloc_path_event(path, &hash, gfp); + } else if (mnt_id) { + event = fanotify_alloc_mnt_event(mnt_id, gfp); + } else { + WARN_ON_ONCE(1); } if (!event) @@ -927,7 +951,7 @@ static int fanotify_handle_event(struct fsnotify_group *group, u32 mask, BUILD_BUG_ON(FAN_RENAME != FS_RENAME); BUILD_BUG_ON(FAN_PRE_ACCESS != FS_PRE_ACCESS); - BUILD_BUG_ON(HWEIGHT32(ALL_FANOTIFY_EVENT_BITS) != 22); + BUILD_BUG_ON(HWEIGHT32(ALL_FANOTIFY_EVENT_BITS) != 24); mask = fanotify_group_event_mask(group, iter_info, &match_mask, mask, data, data_type, dir); @@ -1028,6 +1052,11 @@ static void fanotify_free_error_event(struct fsnotify_group *group, mempool_free(fee, &group->fanotify_data.error_events_pool); } +static void fanotify_free_mnt_event(struct fanotify_event *event) +{ + kmem_cache_free(fanotify_mnt_event_cachep, FANOTIFY_ME(event)); +} + static void fanotify_free_event(struct fsnotify_group *group, struct fsnotify_event *fsn_event) { @@ -1054,6 +1083,9 @@ static void fanotify_free_event(struct fsnotify_group *group, case FANOTIFY_EVENT_TYPE_FS_ERROR: fanotify_free_error_event(group, event); break; + case FANOTIFY_EVENT_TYPE_MNT: + fanotify_free_mnt_event(event); + break; default: WARN_ON_ONCE(1); } diff --git a/fs/notify/fanotify/fanotify.h b/fs/notify/fanotify/fanotify.h index c12cbc270539..b44e70e44be6 100644 --- a/fs/notify/fanotify/fanotify.h +++ b/fs/notify/fanotify/fanotify.h @@ -9,6 +9,7 @@ extern struct kmem_cache *fanotify_mark_cache; extern struct kmem_cache *fanotify_fid_event_cachep; extern struct kmem_cache *fanotify_path_event_cachep; extern struct kmem_cache *fanotify_perm_event_cachep; +extern struct kmem_cache *fanotify_mnt_event_cachep; /* Possible states of the permission event */ enum { @@ -244,6 +245,7 @@ enum fanotify_event_type { FANOTIFY_EVENT_TYPE_PATH_PERM, FANOTIFY_EVENT_TYPE_OVERFLOW, /* struct fanotify_event */ FANOTIFY_EVENT_TYPE_FS_ERROR, /* struct fanotify_error_event */ + FANOTIFY_EVENT_TYPE_MNT, __FANOTIFY_EVENT_TYPE_NUM }; @@ -409,12 +411,23 @@ struct fanotify_path_event { struct path path; }; +struct fanotify_mnt_event { + struct fanotify_event fae; + u64 mnt_id; +}; + static inline struct fanotify_path_event * FANOTIFY_PE(struct fanotify_event *event) { return container_of(event, struct fanotify_path_event, fae); } +static inline struct fanotify_mnt_event * +FANOTIFY_ME(struct fanotify_event *event) +{ + return container_of(event, struct fanotify_mnt_event, fae); +} + /* * Structure for permission fanotify events. It gets allocated and freed in * fanotify_handle_event() since we wait there for user response. When the @@ -466,6 +479,11 @@ static inline bool fanotify_is_error_event(u32 mask) return mask & FAN_FS_ERROR; } +static inline bool fanotify_is_mnt_event(u32 mask) +{ + return mask & (FAN_MNT_ATTACH | FAN_MNT_DETACH); +} + static inline const struct path *fanotify_event_path(struct fanotify_event *event) { if (event->type == FANOTIFY_EVENT_TYPE_PATH) diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c index 6ff94e312232..d0fe7d7880a6 100644 --- a/fs/notify/fanotify/fanotify_user.c +++ b/fs/notify/fanotify/fanotify_user.c @@ -113,6 +113,7 @@ struct kmem_cache *fanotify_mark_cache __ro_after_init; struct kmem_cache *fanotify_fid_event_cachep __ro_after_init; struct kmem_cache *fanotify_path_event_cachep __ro_after_init; struct kmem_cache *fanotify_perm_event_cachep __ro_after_init; +struct kmem_cache *fanotify_mnt_event_cachep __ro_after_init; #define FANOTIFY_EVENT_ALIGN 4 #define FANOTIFY_FID_INFO_HDR_LEN \ @@ -123,6 +124,8 @@ struct kmem_cache *fanotify_perm_event_cachep __ro_after_init; (sizeof(struct fanotify_event_info_error)) #define FANOTIFY_RANGE_INFO_LEN \ (sizeof(struct fanotify_event_info_range)) +#define FANOTIFY_MNT_INFO_LEN \ + (sizeof(struct fanotify_event_info_mnt)) static int fanotify_fid_info_len(int fh_len, int name_len) { @@ -178,6 +181,8 @@ static size_t fanotify_event_len(unsigned int info_mode, fh_len = fanotify_event_object_fh_len(event); event_len += fanotify_fid_info_len(fh_len, dot_len); } + if (fanotify_is_mnt_event(event->mask)) + event_len += FANOTIFY_MNT_INFO_LEN; if (info_mode & FAN_REPORT_PIDFD) event_len += FANOTIFY_PIDFD_INFO_LEN; @@ -405,6 +410,25 @@ static int process_access_response(struct fsnotify_group *group, return -ENOENT; } +static size_t copy_mnt_info_to_user(struct fanotify_event *event, + char __user *buf, int count) +{ + struct fanotify_event_info_mnt info = { }; + + info.hdr.info_type = FAN_EVENT_INFO_TYPE_MNT; + info.hdr.len = FANOTIFY_MNT_INFO_LEN; + + if (WARN_ON(count < info.hdr.len)) + return -EFAULT; + + info.mnt_id = FANOTIFY_ME(event)->mnt_id; + + if (copy_to_user(buf, &info, sizeof(info))) + return -EFAULT; + + return info.hdr.len; +} + static size_t copy_error_info_to_user(struct fanotify_event *event, char __user *buf, int count) { @@ -700,6 +724,15 @@ static int copy_info_records_to_user(struct fanotify_event *event, total_bytes += ret; } + if (fanotify_is_mnt_event(event->mask)) { + ret = copy_mnt_info_to_user(event, buf, count); + if (ret < 0) + return ret; + buf += ret; + count -= ret; + total_bytes += ret; + } + return total_bytes; } @@ -1508,6 +1541,14 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) if ((flags & FAN_REPORT_PIDFD) && (flags & FAN_REPORT_TID)) return -EINVAL; + /* Don't allow mixing mnt events with inode events for now */ + if (flags & FAN_REPORT_MNT) { + if (class != FAN_CLASS_NOTIF) + return -EINVAL; + if (flags & (FANOTIFY_FID_BITS | FAN_REPORT_FD_ERROR)) + return -EINVAL; + } + if (event_f_flags & ~FANOTIFY_INIT_ALL_EVENT_F_BITS) return -EINVAL; @@ -1767,7 +1808,6 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, int dfd, const char __user *pathname) { struct inode *inode = NULL; - struct vfsmount *mnt = NULL; struct fsnotify_group *group; struct path path; struct fan_fsid __fsid, *fsid = NULL; @@ -1800,6 +1840,9 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, case FAN_MARK_FILESYSTEM: obj_type = FSNOTIFY_OBJ_TYPE_SB; break; + case FAN_MARK_MNTNS: + obj_type = FSNOTIFY_OBJ_TYPE_MNTNS; + break; default: return -EINVAL; } @@ -1847,6 +1890,19 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, return -EINVAL; group = fd_file(f)->private_data; + /* Only report mount events on mnt namespace */ + if (FAN_GROUP_FLAG(group, FAN_REPORT_MNT)) { + if (mask & ~FANOTIFY_MOUNT_EVENTS) + return -EINVAL; + if (mark_type != FAN_MARK_MNTNS) + return -EINVAL; + } else { + if (mask & FANOTIFY_MOUNT_EVENTS) + return -EINVAL; + if (mark_type == FAN_MARK_MNTNS) + return -EINVAL; + } + /* * An unprivileged user is not allowed to setup mount nor filesystem * marks. This also includes setting up such marks by a group that @@ -1888,7 +1944,7 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, * point. */ fid_mode = FAN_GROUP_FLAG(group, FANOTIFY_FID_BITS); - if (mask & ~(FANOTIFY_FD_EVENTS|FANOTIFY_EVENT_FLAGS) && + if (mask & ~(FANOTIFY_FD_EVENTS|FANOTIFY_MOUNT_EVENTS|FANOTIFY_EVENT_FLAGS) && (!fid_mode || mark_type == FAN_MARK_MOUNT)) return -EINVAL; @@ -1938,17 +1994,21 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, } /* inode held in place by reference to path; group by fget on fd */ - if (mark_type == FAN_MARK_INODE) { + if (obj_type == FSNOTIFY_OBJ_TYPE_INODE) { inode = path.dentry->d_inode; obj = inode; - } else { - mnt = path.mnt; - if (mark_type == FAN_MARK_MOUNT) - obj = mnt; - else - obj = mnt->mnt_sb; + } else if (obj_type == FSNOTIFY_OBJ_TYPE_VFSMOUNT) { + obj = path.mnt; + } else if (obj_type == FSNOTIFY_OBJ_TYPE_SB) { + obj = path.mnt->mnt_sb; + } else if (obj_type == FSNOTIFY_OBJ_TYPE_MNTNS) { + obj = mnt_ns_from_dentry(path.dentry); } + ret = -EINVAL; + if (!obj) + goto path_put_and_out; + /* * If some other task has this inode open for write we should not add * an ignore mask, unless that ignore mask is supposed to survive @@ -1956,10 +2016,10 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, */ if (mark_cmd == FAN_MARK_ADD && (flags & FANOTIFY_MARK_IGNORE_BITS) && !(flags & FAN_MARK_IGNORED_SURV_MODIFY)) { - ret = mnt ? -EINVAL : -EISDIR; + ret = !inode ? -EINVAL : -EISDIR; /* FAN_MARK_IGNORE requires SURV_MODIFY for sb/mount/dir marks */ if (ignore == FAN_MARK_IGNORE && - (mnt || S_ISDIR(inode->i_mode))) + (!inode || S_ISDIR(inode->i_mode))) goto path_put_and_out; ret = 0; @@ -1968,7 +2028,7 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask, } /* Mask out FAN_EVENT_ON_CHILD flag for sb/mount/non-dir marks */ - if (mnt || !S_ISDIR(inode->i_mode)) { + if (!inode || !S_ISDIR(inode->i_mode)) { mask &= ~FAN_EVENT_ON_CHILD; umask = FAN_EVENT_ON_CHILD; /* @@ -2042,7 +2102,7 @@ static int __init fanotify_user_setup(void) FANOTIFY_DEFAULT_MAX_USER_MARKS); BUILD_BUG_ON(FANOTIFY_INIT_FLAGS & FANOTIFY_INTERNAL_GROUP_FLAGS); - BUILD_BUG_ON(HWEIGHT32(FANOTIFY_INIT_FLAGS) != 13); + BUILD_BUG_ON(HWEIGHT32(FANOTIFY_INIT_FLAGS) != 14); BUILD_BUG_ON(HWEIGHT32(FANOTIFY_MARK_FLAGS) != 11); fanotify_mark_cache = KMEM_CACHE(fanotify_mark, @@ -2055,6 +2115,7 @@ static int __init fanotify_user_setup(void) fanotify_perm_event_cachep = KMEM_CACHE(fanotify_perm_event, SLAB_PANIC); } + fanotify_mnt_event_cachep = KMEM_CACHE(fanotify_mnt_event, SLAB_PANIC); fanotify_max_queued_events = FANOTIFY_DEFAULT_MAX_EVENTS; init_user_ns.ucount_max[UCOUNT_FANOTIFY_GROUPS] = diff --git a/fs/notify/fdinfo.c b/fs/notify/fdinfo.c index e933f9c65d90..1161eabf11ee 100644 --- a/fs/notify/fdinfo.c +++ b/fs/notify/fdinfo.c @@ -121,6 +121,11 @@ static void fanotify_fdinfo(struct seq_file *m, struct fsnotify_mark *mark) seq_printf(m, "fanotify sdev:%x mflags:%x mask:%x ignored_mask:%x\n", sb->s_dev, mflags, mark->mask, mark->ignore_mask); + } else if (mark->connector->type == FSNOTIFY_OBJ_TYPE_MNTNS) { + struct mnt_namespace *mnt_ns = fsnotify_conn_mntns(mark->connector); + + seq_printf(m, "fanotify mnt_ns:%u mflags:%x mask:%x ignored_mask:%x\n", + mnt_ns->ns.inum, mflags, mark->mask, mark->ignore_mask); } } diff --git a/include/linux/fanotify.h b/include/linux/fanotify.h index 78f660ebc318..3c817dc6292e 100644 --- a/include/linux/fanotify.h +++ b/include/linux/fanotify.h @@ -25,7 +25,7 @@ #define FANOTIFY_FID_BITS (FAN_REPORT_DFID_NAME_TARGET) -#define FANOTIFY_INFO_MODES (FANOTIFY_FID_BITS | FAN_REPORT_PIDFD) +#define FANOTIFY_INFO_MODES (FANOTIFY_FID_BITS | FAN_REPORT_PIDFD | FAN_REPORT_MNT) /* * fanotify_init() flags that require CAP_SYS_ADMIN. @@ -38,7 +38,8 @@ FAN_REPORT_PIDFD | \ FAN_REPORT_FD_ERROR | \ FAN_UNLIMITED_QUEUE | \ - FAN_UNLIMITED_MARKS) + FAN_UNLIMITED_MARKS | \ + FAN_REPORT_MNT) /* * fanotify_init() flags that are allowed for user without CAP_SYS_ADMIN. @@ -58,7 +59,7 @@ #define FANOTIFY_INTERNAL_GROUP_FLAGS (FANOTIFY_UNPRIV) #define FANOTIFY_MARK_TYPE_BITS (FAN_MARK_INODE | FAN_MARK_MOUNT | \ - FAN_MARK_FILESYSTEM) + FAN_MARK_FILESYSTEM | FAN_MARK_MNTNS) #define FANOTIFY_MARK_CMD_BITS (FAN_MARK_ADD | FAN_MARK_REMOVE | \ FAN_MARK_FLUSH) @@ -109,10 +110,13 @@ /* Events that can only be reported with data type FSNOTIFY_EVENT_ERROR */ #define FANOTIFY_ERROR_EVENTS (FAN_FS_ERROR) +#define FANOTIFY_MOUNT_EVENTS (FAN_MNT_ATTACH | FAN_MNT_DETACH) + /* Events that user can request to be notified on */ #define FANOTIFY_EVENTS (FANOTIFY_PATH_EVENTS | \ FANOTIFY_INODE_EVENTS | \ - FANOTIFY_ERROR_EVENTS) + FANOTIFY_ERROR_EVENTS | \ + FANOTIFY_MOUNT_EVENTS) /* Extra flags that may be reported with event or control handling of events */ #define FANOTIFY_EVENT_FLAGS (FAN_EVENT_ON_CHILD | FAN_ONDIR) diff --git a/include/uapi/linux/fanotify.h b/include/uapi/linux/fanotify.h index bd8167979707..e710967c7c26 100644 --- a/include/uapi/linux/fanotify.h +++ b/include/uapi/linux/fanotify.h @@ -28,6 +28,8 @@ /* #define FAN_DIR_MODIFY 0x00080000 */ /* Deprecated (reserved) */ #define FAN_PRE_ACCESS 0x00100000 /* Pre-content access hook */ +#define FAN_MNT_ATTACH 0x01000000 /* Mount was attached */ +#define FAN_MNT_DETACH 0x02000000 /* Mount was detached */ #define FAN_EVENT_ON_CHILD 0x08000000 /* Interested in child events */ @@ -64,6 +66,7 @@ #define FAN_REPORT_NAME 0x00000800 /* Report events with name */ #define FAN_REPORT_TARGET_FID 0x00001000 /* Report dirent target id */ #define FAN_REPORT_FD_ERROR 0x00002000 /* event->fd can report error */ +#define FAN_REPORT_MNT 0x00004000 /* Report mount events */ /* Convenience macro - FAN_REPORT_NAME requires FAN_REPORT_DIR_FID */ #define FAN_REPORT_DFID_NAME (FAN_REPORT_DIR_FID | FAN_REPORT_NAME) @@ -94,6 +97,7 @@ #define FAN_MARK_INODE 0x00000000 #define FAN_MARK_MOUNT 0x00000010 #define FAN_MARK_FILESYSTEM 0x00000100 +#define FAN_MARK_MNTNS 0x00000110 /* * Convenience macro - FAN_MARK_IGNORE requires FAN_MARK_IGNORED_SURV_MODIFY @@ -147,6 +151,7 @@ struct fanotify_event_metadata { #define FAN_EVENT_INFO_TYPE_PIDFD 4 #define FAN_EVENT_INFO_TYPE_ERROR 5 #define FAN_EVENT_INFO_TYPE_RANGE 6 +#define FAN_EVENT_INFO_TYPE_MNT 7 /* Special info types for FAN_RENAME */ #define FAN_EVENT_INFO_TYPE_OLD_DFID_NAME 10 @@ -200,6 +205,11 @@ struct fanotify_event_info_range { __u64 count; }; +struct fanotify_event_info_mnt { + struct fanotify_event_info_header hdr; + __u64 mnt_id; +}; + /* * User space may need to record additional information about its decision. * The extra information type records what kind of information is included. diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c index 7b867dfec88b..06d073eab53c 100644 --- a/security/selinux/hooks.c +++ b/security/selinux/hooks.c @@ -3395,6 +3395,10 @@ static int selinux_path_notify(const struct path *path, u64 mask, case FSNOTIFY_OBJ_TYPE_INODE: perm = FILE__WATCH; break; + case FSNOTIFY_OBJ_TYPE_MNTNS: + /* Maybe introduce FILE__WATCH_MOUNTNS? */ + perm = FILE__WATCH_MOUNT; + break; default: return -EINVAL; }
Add notifications for attaching and detaching mounts. The following new event masks are added: FAN_MNT_ATTACH - Mount was attached FAN_MNT_DETACH - Mount was detached If a mount is moved, then the event is reported with (FAN_MNT_ATTACH | FAN_MNT_DETACH). These events add an info record of type FAN_EVENT_INFO_TYPE_MNT containing these fields identifying the affected mounts: __u64 mnt_id - the ID of the mount (see statmount(2)) FAN_REPORT_MNT must be supplied to fanotify_init() to receive these events and no other type of event can be received with this report type. Marks are added with FAN_MARK_MNTNS, which records the mount namespace from an nsfs file (e.g. /proc/self/ns/mnt). Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> --- fs/mount.h | 2 + fs/namespace.c | 14 +++-- fs/notify/fanotify/fanotify.c | 38 +++++++++++-- fs/notify/fanotify/fanotify.h | 18 +++++++ fs/notify/fanotify/fanotify_user.c | 87 +++++++++++++++++++++++++----- fs/notify/fdinfo.c | 5 ++ include/linux/fanotify.h | 12 +++-- include/uapi/linux/fanotify.h | 10 ++++ security/selinux/hooks.c | 4 ++ 9 files changed, 167 insertions(+), 23 deletions(-)