linux.git/fs/notify, branch v5.10

fanotify: fix logic of reporting name info with watched parent

2020-11-09T14:03:08+00:00

The victim inode's parent and name info is required when an event
needs to be delivered to a group interested in filename info OR
when the inode's parent is interested in an event on its children.

Let us call the first condition 'parent_needed' and the second
condition 'parent_interested'.

In fsnotify_parent(), the condition where the inode's parent is
interested in some events on its children, but not necessarily
interested the specific event is called 'parent_watched'.

fsnotify_parent() tests the condition (!parent_watched && !parent_needed)
for sending the event without parent and name info, which is correct.

It then wrongly assumes that parent_watched implies !parent_needed
and tests the condition (parent_watched && !parent_interested)
for sending the event without parent and name info, which is wrong,
because parent may still be needed by some group.

For example, after initializing a group with FAN_REPORT_DFID_NAME and
adding a FAN_MARK_MOUNT with FAN_OPEN mask, open events on non-directory
children of "testdir" are delivered with file name info.

After adding another mark to the same group on the parent "testdir"
with FAN_CLOSE|FAN_EVENT_ON_CHILD mask, open events on non-directory
children of "testdir" are no longer delivered with file name info.

Fix the logic and use auxiliary variables to clarify the conditions.

Fixes: 9b93f33105f5 ("fsnotify: send event with parent/name info to sb/mount/non-dir marks")
Cc: stable@vger.kernel.org#v5.9
Link: https://lore.kernel.org/r/20201108105906.8493-1-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

mm, memcg: rework remote charging API to support nesting

2020-10-18T16:27:09+00:00

Currently the remote memcg charging API consists of two functions:
memalloc_use_memcg() and memalloc_unuse_memcg(), which set and clear the
memcg value, which overwrites the memcg of the current task.

  memalloc_use_memcg(target_memcg);
  <...>
  memalloc_unuse_memcg();

It works perfectly for allocations performed from a normal context,
however an attempt to call it from an interrupt context or just nest two
remote charging blocks will lead to an incorrect accounting.  On exit from
the inner block the active memcg will be cleared instead of being
restored.

  memalloc_use_memcg(target_memcg);

  memalloc_use_memcg(target_memcg_2);
    <...>
    memalloc_unuse_memcg();

    Error: allocation here are charged to the memcg of the current
    process instead of target_memcg.

  memalloc_unuse_memcg();

This patch extends the remote charging API by switching to a single
function: struct mem_cgroup *set_active_memcg(struct mem_cgroup *memcg),
which sets the new value and returns the old one.  So a remote charging
block will look like:

  old_memcg = set_active_memcg(target_memcg);
  <...>
  set_active_memcg(old_memcg);

This patch is heavily based on the patch by Johannes Weiner, which can be
found here: https://lkml.org/lkml/2020/5/28/806 .

Signed-off-by: Roman Gushchin 
Signed-off-by: Andrew Morton 
Reviewed-by: Shakeel Butt 
Cc: Johannes Weiner 
Cc: Dan Schatzberg 
Link: https://lkml.kernel.org/r/20200821212056.3769116-1-guro@fb.com
Signed-off-by: Linus Torvalds

treewide: Use fallthrough pseudo-keyword

2020-08-23T22:36:59+00:00

Replace the existing /* fall through */ comments and its variants with
the new pseudo-keyword macro fallthrough[1]. Also, remove unnecessary
fall-through markings when it is the case.

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Signed-off-by: Gustavo A. R. Silva

fanotify: compare fsid when merging name event

2020-07-28T08:58:07+00:00

When merging name events, fsids of the two involved events have to
match. Otherwise we could merge events from two different filesystems
and thus effectively loose the second event.

Backporting note: Although the commit cacfb956d46e introducing this bug
was merged for 5.7, the relevant code didn't get used in the end until
7e8283af6ede ("fanotify: report parent fid + name + child fid") which
will be merged with this patch. So there's no need for backporting this.

Fixes: cacfb956d46e ("fanotify: record name info for FAN_DIR_MODIFY event")
Reported-by: Amir Goldstein 
Signed-off-by: Jan Kara

fsnotify: create method handle_inode_event() in fsnotify_operations

2020-07-27T21:25:50+00:00

The method handle_event() grew a lot of complexity due to the design of
fanotify and merging of ignore masks.

Most backends do not care about this complex functionality, so we can hide
this complexity from them.

Introduce a method handle_inode_event() that serves those backends and
passes a single inode mark and less arguments.

This change converts all backends except fanotify and inotify to use the
simplified handle_inode_event() method.  In pricipal, inotify could have
also used the new method, but that would require passing more arguments
on the simple helper (data, data_type, cookie), so we leave it with the
handle_event() method.

Link: https://lore.kernel.org/r/20200722125849.17418-9-amir73il@gmail.com
Suggested-by: Jan Kara 
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

fanotify: report parent fid + child fid

2020-07-27T21:24:01+00:00

Add support for FAN_REPORT_FID | FAN_REPORT_DIR_FID.
Internally, it is implemented as a private case of reporting both
parent and child fids and name, the parent and child fids are recorded
in a variable length fanotify_name_event, but there is no name.

It should be noted that directory modification events are recorded
in fixed size fanotify_fid_event when not reporting name, just like
with group flags FAN_REPORT_FID.

Link: https://lore.kernel.org/r/20200716084230.30611-23-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

fanotify: report parent fid + name + child fid

2020-07-27T21:24:00+00:00

For a group with fanotify_init() flag FAN_REPORT_DFID_NAME, the parent
fid and name are reported for events on non-directory objects with an
info record of type FAN_EVENT_INFO_TYPE_DFID_NAME.

If the group also has the init flag FAN_REPORT_FID, the child fid
is also reported with another info record that follows the first info
record. The second info record is the same info record that would have
been reported to a group with only FAN_REPORT_FID flag.

When the child fid needs to be recorded, the variable size struct
fanotify_name_event is preallocated with enough space to store the
child fh between the dir fh and the name.

Link: https://lore.kernel.org/r/20200716084230.30611-22-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

fanotify: add support for FAN_REPORT_NAME

2020-07-27T21:24:00+00:00

Introduce a new fanotify_init() flag FAN_REPORT_NAME.  It requires the
flag FAN_REPORT_DIR_FID and there is a constant for setting both flags
named FAN_REPORT_DFID_NAME.

For a group with flag FAN_REPORT_NAME, the parent fid and name are
reported for directory entry modification events (create/detete/move)
and for events on non-directory objects.

Events on directories themselves are reported with their own fid and
"." as the name.

The parent fid and name are reported with an info record of type
FAN_EVENT_INFO_TYPE_DFID_NAME, similar to the way that parent fid is
reported with into type FAN_EVENT_INFO_TYPE_DFID, but with an appended
null terminated name string.

Link: https://lore.kernel.org/r/20200716084230.30611-21-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

fanotify: report events with parent dir fid to sb/mount/non-dir marks

2020-07-27T21:24:00+00:00

In a group with flag FAN_REPORT_DIR_FID, when adding an inode mark with
FAN_EVENT_ON_CHILD, events on non-directory children are reported with
the fid of the parent.

When adding a filesystem or mount mark or mark on a non-dir inode, we
want to report events that are "possible on child" (e.g. open/close)
also with fid of the parent, as if the victim inode's parent is
interested in events "on child".

Some events, currently only FAN_MOVE_SELF, should be reported to a
sb/mount/non-dir mark with parent fid even though they are not
reported to a watching parent.

To get the desired behavior we set the flag FAN_EVENT_ON_CHILD on
all the sb/mount/non-dir mark masks in a group with FAN_REPORT_DIR_FID.

Link: https://lore.kernel.org/r/20200716084230.30611-20-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara

fanotify: add basic support for FAN_REPORT_DIR_FID

2020-07-27T21:24:00+00:00

For now, the flag is mutually exclusive with FAN_REPORT_FID.
Events include a single info record of type FAN_EVENT_INFO_TYPE_DFID
with a directory file handle.

For now, events are only reported for:
- Directory modification events
- Events on children of a watching directory
- Events on directory objects

Soon, we will add support for reporting the parent directory fid
for events on non-directories with filesystem/mount mark and
support for reporting both parent directory fid and child fid.

Link: https://lore.kernel.org/r/20200716084230.30611-19-amir73il@gmail.com
Signed-off-by: Amir Goldstein 
Signed-off-by: Jan Kara