All file_operations should get a .llseek operation so we can make
nonseekable_open the default for future file operations without a
.llseek pointer.
The three cases that we can automatically detect are no_llseek, seq_lseek
and default_llseek. For cases where we can we can automatically prove that
the file offset is always ignored, we use noop_llseek, which maintains
the current behavior of not returning an error from a seek.
New drivers should normally not use noop_llseek but instead use no_llseek
and call nonseekable_open at open time. Existing drivers can be converted
to do the same when the maintainer knows for certain that no user code
relies on calling seek on the device file.
The generated code is often incorrectly indented and right now contains
comments that clarify for each added line why a specific variant was
chosen. In the version that gets submitted upstream, the comments will
be gone and I will manually fix the indentation, because there does not
seem to be a way to do that using coccinelle.
Some amount of new code is currently sitting in linux-next that should get
the same modifications, which I will do at the end of the merge window.
Many thanks to Julia Lawall for helping me learn to write a semantic
patch that does all this.
===== begin semantic patch =====
// This adds an llseek= method to all file operations,
// as a preparation for making no_llseek the default.
//
// The rules are
// - use no_llseek explicitly if we do nonseekable_open
// - use seq_lseek for sequential files
// - use default_llseek if we know we access f_pos
// - use noop_llseek if we know we don't access f_pos,
// but we still want to allow users to call lseek
//
@ open1 exists @
identifier nested_open;
@@
nested_open(...)
{
<+...
nonseekable_open(...)
...+>
}
@ open exists@
identifier open_f;
identifier i, f;
identifier open1.nested_open;
@@
int open_f(struct inode *i, struct file *f)
{
<+...
(
nonseekable_open(...)
|
nested_open(...)
)
...+>
}
@ read disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ read_no_fpos disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
... when != off
}
@ write @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ write_no_fpos @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
... when != off
}
@ fops0 @
identifier fops;
@@
struct file_operations fops = {
...
};
@ has_llseek depends on fops0 @
identifier fops0.fops;
identifier llseek_f;
@@
struct file_operations fops = {
...
.llseek = llseek_f,
...
};
@ has_read depends on fops0 @
identifier fops0.fops;
identifier read_f;
@@
struct file_operations fops = {
...
.read = read_f,
...
};
@ has_write depends on fops0 @
identifier fops0.fops;
identifier write_f;
@@
struct file_operations fops = {
...
.write = write_f,
...
};
@ has_open depends on fops0 @
identifier fops0.fops;
identifier open_f;
@@
struct file_operations fops = {
...
.open = open_f,
...
};
// use no_llseek if we call nonseekable_open
////////////////////////////////////////////
@ nonseekable1 depends on !has_llseek && has_open @
identifier fops0.fops;
identifier nso ~= "nonseekable_open";
@@
struct file_operations fops = {
... .open = nso, ...
+.llseek = no_llseek, /* nonseekable */
};
@ nonseekable2 depends on !has_llseek @
identifier fops0.fops;
identifier open.open_f;
@@
struct file_operations fops = {
... .open = open_f, ...
+.llseek = no_llseek, /* open uses nonseekable */
};
// use seq_lseek for sequential files
/////////////////////////////////////
@ seq depends on !has_llseek @
identifier fops0.fops;
identifier sr ~= "seq_read";
@@
struct file_operations fops = {
... .read = sr, ...
+.llseek = seq_lseek, /* we have seq_read */
};
// use default_llseek if there is a readdir
///////////////////////////////////////////
@ fops1 depends on !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier readdir_e;
@@
// any other fop is used that changes pos
struct file_operations fops = {
... .readdir = readdir_e, ...
+.llseek = default_llseek, /* readdir is present */
};
// use default_llseek if at least one of read/write touches f_pos
/////////////////////////////////////////////////////////////////
@ fops2 depends on !fops1 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read.read_f;
@@
// read fops use offset
struct file_operations fops = {
... .read = read_f, ...
+.llseek = default_llseek, /* read accesses f_pos */
};
@ fops3 depends on !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write.write_f;
@@
// write fops use offset
struct file_operations fops = {
... .write = write_f, ...
+ .llseek = default_llseek, /* write accesses f_pos */
};
// Use noop_llseek if neither read nor write accesses f_pos
///////////////////////////////////////////////////////////
@ fops4 depends on !fops1 && !fops2 && !fops3 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
identifier write_no_fpos.write_f;
@@
// write fops use offset
struct file_operations fops = {
...
.write = write_f,
.read = read_f,
...
+.llseek = noop_llseek, /* read and write both use no f_pos */
};
@ depends on has_write && !has_read && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write_no_fpos.write_f;
@@
struct file_operations fops = {
... .write = write_f, ...
+.llseek = noop_llseek, /* write uses no f_pos */
};
@ depends on has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
@@
struct file_operations fops = {
... .read = read_f, ...
+.llseek = noop_llseek, /* read uses no f_pos */
};
@ depends on !has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
@@
struct file_operations fops = {
...
+.llseek = noop_llseek, /* no read or write fn */
};
===== End semantic patch =====
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Julia Lawall <julia@diku.dk>
Cc: Christoph Hellwig <hch@infradead.org>
This reverts commit 3bcf3860a4 (and the
accompanying commit c1e5c95402 "vfs/fsnotify: fsnotify_close can delay
the final work in fput" that was a horribly ugly hack to make it work at
all).
The 'struct file' approach not only causes that disgusting hack, it
somehow breaks pulseaudio, probably due to some other subtlety with
f_count handling.
Fix up various conflicts due to later fsnotify work.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fanotify currently, when given a vfsmount_mark will look up (if it exists)
the corresponding inode mark. This patch drops that lookup and uses the
mark provided.
Signed-off-by: Eric Paris <eparis@redhat.com>
should_send_event() and handle_event() will both need to look up the inode
event if they get a vfsmount event. Lets just pass both at the same time
since we have them both after walking the lists in lockstep.
Signed-off-by: Eric Paris <eparis@redhat.com>
group->mask is now useless. It was originally a shortcut for fsnotify to
save on performance. These checks are now redundant, so we remove them.
Signed-off-by: Eric Paris <eparis@redhat.com>
The change to use srcu and walk the object list rather than the global
fsnotify_group list means that should_send_event is no longer needed for a
number of groups and can be simplified for others. Do that.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify now gets a mark in the should_send_event and handle_event
functions. Rather than look up the mark themselves inotify should just use
the mark it was handed.
Signed-off-by: Eric Paris <eparis@redhat.com>
With the change of fsnotify to use srcu walking the marks list instead of
walking the global groups list we now know the mark in question. The code can
send the mark to the group's handling functions and the groups won't have to
find those marks themselves.
Signed-off-by: Eric Paris <eparis@redhat.com>
Al explains that calling dentry_open() with a mnt/dentry pair is only
garunteed to be safe if they are already used in an open struct file. To
make sure this is the case don't store and use a struct path in fsnotify,
always use a struct file.
Signed-off-by: Eric Paris <eparis@redhat.com>
Rather than the horrific void ** argument and such just to pass the
fanotify_merge event back to the caller of fsnotify_add_notify_event() have
those things return an event if it was different than the event suggusted to
be added.
Signed-off-by: Eric Paris <eparis@redhat.com>
It can be hard to debug fsnotify since there are so few printks. Use
pr_debug to allow for dynamic debugging.
Signed-off-by: Eric Paris <eparis@redhat.com>
The mask checks in inotify_update_existing_watch() and
inotify_new_watch() are useless because inotify_arg_to_mask() sets
FS_IN_IGNORED and FS_EVENT_ON_CHILD bits anyway.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify uses bits called IN_* and fsnotify uses bits called FS_*. These
need to line up. This patch adds build time checks to make sure noone can
change these bits so they are not the same.
Signed-off-by: Eric Paris <eparis@redhat.com>
An inotify watch on a directory will send events for children even if those
children have been unlinked. This patch add a new inotify flag IN_EXCL_UNLINK
which allows a watch to specificy they don't care about unlinked children.
This should fix performance problems seen by tasks which add a watch to
/tmp and then are overrun with events when other processes are reading and
writing to unlinked files they created in /tmp.
https://bugzilla.kernel.org/show_bug.cgi?id=16296
Requested-by: Matthias Clasen <mclasen@redhat.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Since the .31 or so notify rewrite inotify has not sent events about
inodes which are unmounted. This patch restores those events.
Signed-off-by: Eric Paris <eparis@redhat.com>
During the large inotify rewrite to fsnotify I completely dropped support
for IN_ONESHOT. Reimplement that support.
Signed-off-by: Eric Paris <eparis@redhat.com>
The symbol inotify_max_user_watches is not used outside this
file and should be static.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: John McCutchan <john@johnmccutchan.com>
Cc: Robert Love <rlove@rlove.org>
Cc: Eric Paris <eparis@parisplace.org>
Signed-off-by: Eric Paris <eparis@redhat.com>
Each group can define their own notification (and secondary_q) merge
function. Inotify does tail drop, fanotify does matching and drop which
can actually allocate a completely new event. But for fanotify to properly
deal with permissions events it needs to know the new event which was
ultimately added to the notification queue. This patch just implements a
void ** argument which is passed to the merge function. fanotify can use
this field to pass the new event back to higher layers.
Signed-off-by: Eric Paris <eparis@redhat.com>
for fanotify to properly deal with permissions events
inotify marks must pin inodes in core. dnotify doesn't technically need to
since they are closed when the directory is closed. fanotify also need to
pin inodes in core as it works today. But the next step is to introduce
the concept of 'ignored masks' which is actually a mask of events for an
inode of no interest. I claim that these should be liberally sent to the
kernel and should not pin the inode in core. If the inode is brought back
in the listener will get an event it may have thought excluded, but this is
not a serious situation and one any listener should deal with.
This patch lays the ground work for non-pinning inode marks by using lazy
inode pinning. We do not pin a mark until it has a non-zero mask entry. If a
listener new sets a mask we never pin the inode.
Signed-off-by: Eric Paris <eparis@redhat.com>
currently all marking is done by functions in inode-mark.c. Some of this
is pretty generic and should be instead done in a generic function and we
should only put the inode specific code in inode-mark.c
Signed-off-by: Eric Paris <eparis@redhat.com>
All callers to fsnotify_find_mark_entry() except one take and
release inode->i_lock around the call. Take the lock inside
fsnotify_find_mark_entry() instead.
Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Signed-off-by: Eric Paris <eparis@redhat.com>
The addition of marks on vfs mounts will be simplified if the inode
specific parts of a mark and the vfsmnt specific parts of a mark are
actually in a union so naming can be easy. This patch just implements the
inode struct and the union.
Signed-off-by: Eric Paris <eparis@redhat.com>
To ensure that a group will not duplicate events when it receives it based
on the vfsmount and the inode should_send_event test we should distinguish
those two cases. We pass a vfsmount to this function so groups can make
their own determinations.
Signed-off-by: Eric Paris <eparis@redhat.com>
fsnotify_obtain_group was intended to be able to find an already existing
group. Nothing uses that functionality. This just renames it to
fsnotify_alloc_group so it is clear what it is doing.
Signed-off-by: Eric Paris <eparis@redhat.com>
The original fsnotify interface has a group-num which was intended to be
able to find a group after it was added. I no longer think this is a
necessary thing to do and so we remove the group_num.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify only wishes to merge a new event with the last event on the
notification fifo. fanotify is willing to merge any events including by
means of bitwise OR masks of multiple events together. This patch moves
the inotify event merging logic out of the generic fsnotify notification.c
and into the inotify code. This allows each use of fsnotify to provide
their own merge functionality.
Signed-off-by: Eric Paris <eparis@redhat.com>
fanotify is going to need to look at file->private_data to know if an event
should be sent or not. This passes the data (which might be a file,
dentry, inode, or none) to the should_send function calls so fanotify can
get that information when available
Signed-off-by: Eric Paris <eparis@redhat.com>
fanotify is only interested in event types which contain enough information
to open the original file in the context of the fanotify listener. Since
fanotify may not want to send events if that data isn't present we pass
the data type to the should_send_event function call so fanotify can express
its lack of interest.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify was supposed to have a dmesg printk ratelimitor which would cause
inotify to only emit one message per boot. The static bool was never set
so it kept firing messages. This patch correctly limits warnings in multiple
places.
Signed-off-by: Eric Paris <eparis@redhat.com>
Prior to 2.6.31 inotify would not reuse watch descriptors until all of
them had been used at least once. After the rewrite inotify would reuse
watch descriptors. The selinux utility 'restorecond' was found to have
problems when watch descriptors were reused. This patch reverts to the
pre inotify rewrite behavior to not reuse watch descriptors.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify_free_mark casts directly from an fsnotify_mark_entry to an
inotify_inode_mark_entry. This works, but should use container_of instead
for future proofing.
Signed-off-by: Eric Paris <eparis@redhat.com>
This patch allows a task to add a second fsnotify mark to an inode for the
same group. This mark will be added to the end of the inode's list and
this will never be found by the stand fsnotify_find_mark() function. This
is useful if a user wants to add a new mark before removing the old one.
Signed-off-by: Eric Paris <eparis@redhat.com>
This patch moves all of the idr editing operations into their own idr
functions. It makes it easier to prove locking correctness and to to
understand the code flow.
Signed-off-by: Eric Paris <eparis@redhat.com>
Make sure that s_umount is acquired *before* we drop the final
active reference; we still have the fast path (atomic_dec_unless)
and we have gotten rid of the window between the moment when
s_active hits zero and s_umount is acquired. Which simplifies
the living hell out of grab_super() and inotify pin_to_kill()
stuff.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* 'for-linus' of git://git.infradead.org/users/eparis/notify:
inotify: don't leak user struct on inotify release
inotify: race use after free/double free in inotify inode marks
inotify: clean up the inotify_add_watch out path
Inotify: undefined reference to `anon_inode_getfd'
Manual merge to remove duplicate "select ANON_INODES" from Kconfig file
inotify_new_group() receives a get_uid-ed user_struct and saves the
reference on group->inotify_data.user. The problem is that free_uid() is
never called on it.
Issue seem to be introduced by 63c882a0 (inotify: reimplement inotify
using fsnotify) after 2.6.30.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Eric Paris <eparis@parisplace.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Eric Paris <eparis@redhat.com>
There is a race in the inotify add/rm watch code. A task can find and
remove a mark which doesn't have all of it's references. This can
result in a use after free/double free situation.
Task A Task B
------------ -----------
inotify_new_watch()
allocate a mark (refcnt == 1)
add it to the idr
inotify_rm_watch()
inotify_remove_from_idr()
fsnotify_put_mark()
refcnt hits 0, free
take reference because we are on idr
[at this point it is a use after free]
[time goes on]
refcnt may hit 0 again, double free
The fix is to take the reference BEFORE the object can be found in the
idr.
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: <stable@kernel.org>
inotify_add_watch explictly frees the unused inode mark, but it can just
use the generic code. Just do that.
Signed-off-by: Eric Paris <eparis@redhat.com>
Fix:
fs/built-in.o: In function `sys_inotify_init1':
summary.c:(.text+0x347a4): undefined reference to `anon_inode_getfd'
found by kautobuild with arms bcmring_defconfig, which ends up with
INOTIFY_USER enabled (through the 'default y') but leaves ANON_INODES
unset. However, inotify_user.c uses anon_inode_getfd().
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Eric Paris <eparis@redhat.com>
CONFIG_INOTIFY_USER defined but CONFIG_ANON_INODES undefined will result
in the following build failure:
LD vmlinux
fs/built-in.o: In function 'sys_inotify_init1':
(.text.sys_inotify_init1+0x22c): undefined reference to 'anon_inode_getfd'
fs/built-in.o: In function `sys_inotify_init1':
(.text.sys_inotify_init1+0x22c): relocation truncated to fit: R_MIPS_26 against 'anon_inode_getfd'
make[2]: *** [vmlinux] Error 1
make[1]: *** [sub-make] Error 2
make: *** [all] Error 2
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
inotify will WARN() if it finds that the idr and the fsnotify internals
somehow got out of sync. It was only supposed to do this once but due
to this stupid bug it would warn every single time a problem was
detected.
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since commit 7e790dd5fc ("inotify: fix
error paths in inotify_update_watch") inotify changed the manor in which
it gave watch descriptors back to userspace. Previous to this commit
inotify acted like the following:
inotify_add_watch(X, Y, Z) = 1
inotify_rm_watch(X, 1);
inotify_add_watch(X, Y, Z) = 2
but after this patch inotify would return watch descriptors like so:
inotify_add_watch(X, Y, Z) = 1
inotify_rm_watch(X, 1);
inotify_add_watch(X, Y, Z) = 1
which I saw as equivalent to opening an fd where
open(file) = 1;
close(1);
open(file) = 1;
seemed perfectly reasonable. The issue is that quite a bit of userspace
apparently relies on the behavior in which watch descriptors will not be
quickly reused. KDE relies on it, I know some selinux packages rely on
it, and I have heard complaints from other random sources such as debian
bug 558981.
Although the man page implies what we do is ok, we broke userspace so
this patch almost reverts us to the old behavior. It is still slightly
racey and I have patches that would fix that, but they are rather large
and this will fix it for all real world cases. The race is as follows:
- task1 creates a watch and blocks in idr_new_watch() before it updates
the hint.
- task2 creates a watch and updates the hint.
- task1 updates the hint with it's older wd
- task removes the watch created by task2
- task adds a new watch and will reuse the wd originally given to task2
it requires moving some locking around the hint (last_wd) but this should
solve it for the real world and be -stable safe.
As a side effect this patch papers over a bug in the lib/idr code which
is causing a large number WARN's to pop on people's system and many
reports in kerneloops.org. I'm working on the root cause of that idr
bug seperately but this should make inotify immune to that issue.
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For consistency drop & in front of every proc_handler. Explicity
taking the address is unnecessary and it prevents optimizations
like stubbing the proc_handlers to NULL.
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Now that sys_sysctl is a generic wrapper around /proc/sys .ctl_name
and .strategy members of sysctl tables are dead code. Remove them.
Cc: Jan Harkes <jaharkes@cs.cmu.edu>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Seperating the addition and update of marks in inotify resulted in a
regression in that inotify never gets events. The inotify group mask is
always 0. This mask should be updated any time a new mark is added.
Signed-off-by: Eric Paris <eparis@redhat.com>
0db501bd06 introduced a regresion in that it now sends a nul
terminator but the length accounting when checking for space or
reporting to userspace did not take this into account. This corrects
all of the rounding logic.
Signed-off-by: Eric Paris <eparis@redhat.com>
When an event has no pathname, there's no need to pad it with a null byte and
therefore generate an inotify_event sized block of zeros. This fixes a
regression introduced by commit 0db501bd06 where
my system wouldn't finish booting because some process was being confused by
this.
Signed-off-by: Brian Rogers <brian@xyzw.org>
Signed-off-by: Eric Paris <eparis@redhat.com>
Before the rewrite copy_event_to_user always wrote a terqminating '\0'
byte to user space after the filename. Since the rewrite that
terminating byte was skipped if your filename is exactly a multiple of
event_size. Ouch!
So add one byte to name_size before we round up and use clear_user to
set userspace to zero like /dev/zero does instead of copying the
strange nul_inotify_event. I can't quite convince myself len_to_zero
will never exceed 16 and even if it doesn't clear_user should be more
efficient and a more accurate reflection of what the code is trying to
do.
Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
The are races around the idr storage of inotify watches. It's possible
that a watch could be found from sys_inotify_rm_watch() in the idr, but it
could be removed from the idr before that code does it's removal. Move the
locking and the refcnt'ing so that these have to happen atomically.
Signed-off-by: Eric Paris <eparis@redhat.com>
If an inotify watch is left in the idr when an fsnotify group is destroyed
this will lead to a BUG. This is not a dangerous situation and really
indicates a programming bug and leak of memory. This patch changes it to
use a WARN and a printk rather than killing people's boxes.
Signed-off-by: Eric Paris <eparis@redhat.com>
There is nothing known wrong with the inotify watch addition/modification
but this patch seperates the two code paths to make them each easy to
verify as correct.
Signed-off-by: Eric Paris <eparis@redhat.com>
The inotify_add_watch man page specifies that inotify_add_watch() will
return a non-negative integer. However, historically the inotify
watches started at 1, not at 0.
Turns out that the inotifywait program provided by the inotify-tools
package doesn't properly handle a 0 watch descriptor. In 7e790dd5 we
changed from starting at 1 to starting at 0. This patch starts at 1,
just like in previous kernels, but also just like in previous kernels
it's possible for it to wrap back to 0. This preserves the kernel
functionality exactly like it was before the patch (neither method broke
the spec)
Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
inotify decides if private data it passed to get added to an event was
used by checking list_empty(). But it's possible that the event may
have been dequeued and the private event removed so it would look empty.
The fix is to use the return code from fsnotify_add_notify_event rather
than looking at the list.
Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
inotify can have a watchs removed under filesystem reclaim.
=================================
[ INFO: inconsistent lock state ]
2.6.31-rc2 #16
---------------------------------
inconsistent {IN-RECLAIM_FS-W} -> {RECLAIM_FS-ON-W} usage.
khubd/217 [HC0[0]:SC0[0]:HE1:SE1] takes:
(iprune_mutex){+.+.?.}, at: [<c10ba899>] invalidate_inodes+0x20/0xe3
{IN-RECLAIM_FS-W} state was registered at:
[<c10536ab>] __lock_acquire+0x2c9/0xac4
[<c1053f45>] lock_acquire+0x9f/0xc2
[<c1308872>] __mutex_lock_common+0x2d/0x323
[<c1308c00>] mutex_lock_nested+0x2e/0x36
[<c10ba6ff>] shrink_icache_memory+0x38/0x1b2
[<c108bfb6>] shrink_slab+0xe2/0x13c
[<c108c3e1>] kswapd+0x3d1/0x55d
[<c10449b5>] kthread+0x66/0x6b
[<c1003fdf>] kernel_thread_helper+0x7/0x10
[<ffffffff>] 0xffffffff
Two things are needed to fix this. First we need a method to tell
fsnotify_create_event() to use GFP_NOFS and second we need to stop using
one global IN_IGNORED event and allocate them one at a time. This solves
current issues with multiple IN_IGNORED on a queue having tail drop
problems and simplifies the allocations since we don't have to worry about
two tasks opperating on the IGNORED event concurrently.
Signed-off-by: Eric Paris <eparis@redhat.com>
fsnotify doens't give the user anything. If someone chooses inotify or
dnotify it should build fsnotify, if they don't select one it shouldn't be
built. This patch changes fsnotify to be a def_bool=n and makes everything
else select it. Also fixes the issue people complained about on lwn where
gdm hung because they didn't have inotify and they didn't get the inotify
build option.....
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify_update_watch could leave things in a horrid state on a number of
error paths. We could try to remove idr entries that didn't exist, we
could send an IN_IGNORED to userspace for watches that don't exist, and a
bit of other stupidity. Clean these up by doing the idr addition before we
put the mark on the inode since we can clean that up on error and getting
off the inode's mark list is hard.
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify_add_watch had a couple of problems. The biggest being that if
inotify_add_watch was called on the same inode twice (to update or change the
event mask) a refence was taken on the original inode mark by
fsnotify_find_mark_entry but was not being dropped at the end of the
inotify_add_watch call. Thus if inotify_rm_watch was called although the mark
was removed from the inode, the refcnt wouldn't hit zero and we would leak
memory.
Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
The inotify rewrite forgot to drop the inotify watch use cound when a watch
was removed. This means that a single inotify fd can only ever register a
maximum of /proc/sys/fs/max_user_watches even if some of those had been
freed.
Signed-off-by: Eric Paris <eparis@redhat.com>
The per-user inotify_devs value is incremented each time a new file is
allocated, but never decremented. This led to inotify_init failing after a
limited number of calls.
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
inotify_destroy_mark_entry could get called twice for the same mark since it
is called directly in inotify_rm_watch and when the mark is being destroyed for
another reason. As an example assume that the file being watched was just
deleted so inotify_destroy_mark_entry would get called from the path
fsnotify_inoderemove() -> fsnotify_destroy_marks_by_inode() ->
fsnotify_destroy_mark_entry() -> inotify_destroy_mark_entry(). If this
happened at the same time as userspace tried to remove a watch via
inotify_rm_watch we could attempt to remove the mark from the idr twice and
could thus double dec the ref cnt and potentially could be in a use after
free/double free situation. The fix is to have inotify_rm_watch use the
generic recursive safe fsnotify_destroy_mark_by_entry() so we are sure the
inotify_destroy_mark_entry() function can only be called one.
This patch also renames the function to inotify_ingored_remove_idr() so it is
clear what is actually going on in the function.
Hopefully this fixes:
[ 20.342058] idr_remove called for id=20 which is not allocated.
[ 20.348000] Pid: 1860, comm: udevd Not tainted 2.6.30-tip #1077
[ 20.353933] Call Trace:
[ 20.356410] [<ffffffff811a82b7>] idr_remove+0x115/0x18f
[ 20.361737] [<ffffffff8134259d>] ? _spin_lock+0x6d/0x75
[ 20.367061] [<ffffffff8111640a>] ? inotify_destroy_mark_entry+0xa3/0xcf
[ 20.373771] [<ffffffff8111641e>] inotify_destroy_mark_entry+0xb7/0xcf
[ 20.380306] [<ffffffff81115913>] inotify_freeing_mark+0xe/0x10
[ 20.386238] [<ffffffff8111410d>] fsnotify_destroy_mark_by_entry+0x143/0x170
[ 20.393293] [<ffffffff811163a3>] inotify_destroy_mark_entry+0x3c/0xcf
[ 20.399829] [<ffffffff811164d1>] sys_inotify_rm_watch+0x9b/0xc6
[ 20.405850] [<ffffffff8100bcdb>] system_call_fastpath+0x16/0x1b
Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Eric Paris <eparis@redhat.com>
Tested-by: Peter Ziljlstra <peterz@infradead.org>
inotify and dnotify will both indicate that they want any event which came
from a child inode. The fix is to mask off FS_EVENT_ON_CHILD when deciding
if inotify or dnotify is interested in a given event.
Signed-off-by: Eric Paris <eparis@redhat.com>
Reimplement inotify_user using fsnotify. This should be feature for feature
exactly the same as the original inotify_user. This does not make any changes
to the in kernel inotify feature used by audit. Those patches (and the eventual
removal of in kernel inotify) will come after the new inotify_user proves to be
working correctly.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
fsnotify is a backend for filesystem notification. fsnotify does
not provide any userspace interface but does provide the basis
needed for other notification schemes such as dnotify. fsnotify
can be extended to be the backend for inotify or the upcoming
fanotify. fsnotify provides a mechanism for "groups" to register for
some set of filesystem events and to then deliver those events to
those groups for processing.
fsnotify has a number of benefits, the first being actually shrinking the size
of an inode. Before fsnotify to support both dnotify and inotify an inode had
unsigned long i_dnotify_mask; /* Directory notify events */
struct dnotify_struct *i_dnotify; /* for directory notifications */
struct list_head inotify_watches; /* watches on this inode */
struct mutex inotify_mutex; /* protects the watches list
But with fsnotify this same functionallity (and more) is done with just
__u32 i_fsnotify_mask; /* all events for this inode */
struct hlist_head i_fsnotify_mark_entries; /* marks on this inode */
That's right, inotify, dnotify, and fanotify all in 64 bits. We used that
much space just in inotify_watches alone, before this patch set.
fsnotify object lifetime and locking is MUCH better than what we have today.
inotify locking is incredibly complex. See 8f7b0ba1c8 as an example of
what's been busted since inception. inotify needs to know internal semantics
of superblock destruction and unmounting to function. The inode pinning and
vfs contortions are horrible.
no fsnotify implementers do allocation under locks. This means things like
f04b30de3 which (due to an overabundance of caution) changes GFP_KERNEL to
GFP_NOFS can be reverted. There are no longer any allocation rules when using
or implementing your own fsnotify listener.
fsnotify paves the way for fanotify. In brief fanotify is a notification
mechanism that delivers the lisener both an 'event' and an open file descriptor
to the object in question. This means that fanotify is pathname agnostic.
Some on lkml may not care for the original companies or users that pushed for
TALPA, but fanotify was designed with flexibility and input for other users in
mind. The readahead group expressed interest in fanotify as it could be used
to profile disk access on boot without breaking the audit system. The desktop
search groups have also expressed interest in fanotify as it solves a number
of the race conditions and problems present with managing inotify when more
than a limited number of specific files are of interest. fanotify can provide
for a userspace access control system which makes it a clean interface for AV
vendors to hook without trying to do binary patching on the syscall table,
LSM, and everywhere else they do their things today. With this patch series
fanotify can be implemented in less than 1200 lines of easy to review code.
Almost all of which is the socket based user interface.
This patch series builds fsnotify to the point that it can implement
dnotify and inotify_user. Patches exist and will be sent soon after
acceptance to finish the in kernel inotify conversion (audit) and implement
fanotify.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
There is what we believe to be a false positive reported by lockdep.
inotify_inode_queue_event() => take inotify_mutex => kernel_event() =>
kmalloc() => SLOB => alloc_pages_node() => page reclaim => slab reclaim =>
dcache reclaim => inotify_inode_is_dead => take inotify_mutex => deadlock
The plan is to fix this via lockdep annotation, but that is proving to be
quite involved.
The patch flips the allocation over to GFP_NFS to shut the warning up, for
the 2.6.30 release.
Hopefully we will fix this for real in 2.6.31. I'll queue a patch in -mm
to switch it back to GFP_KERNEL so we don't forget.
=================================
[ INFO: inconsistent lock state ]
2.6.30-rc2-next-20090417 #203
---------------------------------
inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage.
kswapd0/380 [HC0[0]:SC0[0]:HE1:SE1] takes:
(&inode->inotify_mutex){+.+.?.}, at: [<ffffffff8112f1b5>] inotify_inode_is_dead+0x35/0xb0
{RECLAIM_FS-ON-W} state was registered at:
[<ffffffff81079188>] mark_held_locks+0x68/0x90
[<ffffffff810792a5>] lockdep_trace_alloc+0xf5/0x100
[<ffffffff810f5261>] __kmalloc_node+0x31/0x1e0
[<ffffffff81130652>] kernel_event+0xe2/0x190
[<ffffffff81130826>] inotify_dev_queue_event+0x126/0x230
[<ffffffff8112f096>] inotify_inode_queue_event+0xc6/0x110
[<ffffffff8110444d>] vfs_create+0xcd/0x140
[<ffffffff8110825d>] do_filp_open+0x88d/0xa20
[<ffffffff810f6b68>] do_sys_open+0x98/0x140
[<ffffffff810f6c50>] sys_open+0x20/0x30
[<ffffffff8100c272>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
irq event stamp: 690455
hardirqs last enabled at (690455): [<ffffffff81564fe4>] _spin_unlock_irqrestore+0x44/0x80
hardirqs last disabled at (690454): [<ffffffff81565372>] _spin_lock_irqsave+0x32/0xa0
softirqs last enabled at (690178): [<ffffffff81052282>] __do_softirq+0x202/0x220
softirqs last disabled at (690157): [<ffffffff8100d50c>] call_softirq+0x1c/0x50
other info that might help us debug this:
2 locks held by kswapd0/380:
#0: (shrinker_rwsem){++++..}, at: [<ffffffff810d0bd7>] shrink_slab+0x37/0x180
#1: (&type->s_umount_key#17){++++..}, at: [<ffffffff8110cfbf>] shrink_dcache_memory+0x11f/0x1e0
stack backtrace:
Pid: 380, comm: kswapd0 Not tainted 2.6.30-rc2-next-20090417 #203
Call Trace:
[<ffffffff810789ef>] print_usage_bug+0x19f/0x200
[<ffffffff81018bff>] ? save_stack_trace+0x2f/0x50
[<ffffffff81078f0b>] mark_lock+0x4bb/0x6d0
[<ffffffff810799e0>] ? check_usage_forwards+0x0/0xc0
[<ffffffff8107b142>] __lock_acquire+0xc62/0x1ae0
[<ffffffff810f478c>] ? slob_free+0x10c/0x370
[<ffffffff8107c0a1>] lock_acquire+0xe1/0x120
[<ffffffff8112f1b5>] ? inotify_inode_is_dead+0x35/0xb0
[<ffffffff81562d43>] mutex_lock_nested+0x63/0x420
[<ffffffff8112f1b5>] ? inotify_inode_is_dead+0x35/0xb0
[<ffffffff8112f1b5>] ? inotify_inode_is_dead+0x35/0xb0
[<ffffffff81012fe9>] ? sched_clock+0x9/0x10
[<ffffffff81077165>] ? lock_release_holdtime+0x35/0x1c0
[<ffffffff8112f1b5>] inotify_inode_is_dead+0x35/0xb0
[<ffffffff8110c9dc>] dentry_iput+0xbc/0xe0
[<ffffffff8110cb23>] d_kill+0x33/0x60
[<ffffffff8110ce23>] __shrink_dcache_sb+0x2d3/0x350
[<ffffffff8110cffa>] shrink_dcache_memory+0x15a/0x1e0
[<ffffffff810d0cc5>] shrink_slab+0x125/0x180
[<ffffffff810d1540>] kswapd+0x560/0x7a0
[<ffffffff810ce160>] ? isolate_pages_global+0x0/0x2c0
[<ffffffff81065a30>] ? autoremove_wake_function+0x0/0x40
[<ffffffff8107953d>] ? trace_hardirqs_on+0xd/0x10
[<ffffffff810d0fe0>] ? kswapd+0x0/0x7a0
[<ffffffff8106555b>] kthread+0x5b/0xa0
[<ffffffff8100d40a>] child_rip+0xa/0x20
[<ffffffff8100cdd0>] ? restore_args+0x0/0x30
[<ffffffff81065500>] ? kthread+0x0/0xa0
[<ffffffff8100d400>] ? child_rip+0x0/0x20
[eparis@redhat.com: fix audit too]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
To be on the safe side, it should be less fragile to exclude I_NEW inodes
from inode list scans by default (unless there is an important reason to
have them).
Normally they will get excluded (eg. by zero refcount or writecount etc),
however it is a bit fragile for list walkers to know exactly what parts of
the inode state is set up and valid to test when in I_NEW. So along these
lines, move I_NEW checks upward as well (sometimes taking I_FREEING etc
checks with them too -- this shouldn't be a problem should it?)
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If userspace supplies an invalid pointer to a read() of an inotify
instance, the inotify device's event list mutex is unlocked twice.
This causes an unbalance which effectively leaves the data structure
unprotected, and we can trigger oopses by accessing the inotify
instance from different tasks concurrently.
The best fix (contributed largely by Linus) is a total rewrite
of the function in question:
On Thu, Jan 22, 2009 at 7:05 AM, Linus Torvalds wrote:
> The thing to notice is that:
>
> - locking is done in just one place, and there is no question about it
> not having an unlock.
>
> - that whole double-while(1)-loop thing is gone.
>
> - use multiple functions to make nesting and error handling sane
>
> - do error testing after doing the things you always need to do, ie do
> this:
>
> mutex_lock(..)
> ret = function_call();
> mutex_unlock(..)
>
> .. test ret here ..
>
> instead of doing conditional exits with unlocking or freeing.
>
> So if the code is written in this way, it may still be buggy, but at least
> it's not buggy because of subtle "forgot to unlock" or "forgot to free"
> issues.
>
> This _always_ unlocks if it locked, and it always frees if it got a
> non-error kevent.
Cc: John McCutchan <ttb@tentacle.dhs.org>
Cc: Robert Love <rlove@google.com>
Cc: <stable@kernel.org>
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The problems lie in the types used for some inotify interfaces, both at the kernel level and at the glibc level. This mail addresses the kernel problem. I will follow up with some suggestions for glibc changes.
For the sys_inotify_rm_watch() interface, the type of the 'wd' argument is
currently 'u32', it should be '__s32' . That is Robert's suggestion, and
is consistent with the other declarations of watch descriptors in the
kernel source, in particular, the inotify_event structure in
include/linux/inotify.h:
struct inotify_event {
__s32 wd; /* watch descriptor */
__u32 mask; /* watch mask */
__u32 cookie; /* cookie to synchronize two events */
__u32 len; /* length (including nulls) of name */
char name[0]; /* stub for possible name */
};
The patch makes the changes needed for inotify_rm_watch().
Signed-off-by: Michael Kerrisk <mtk.manpages@googlemail.com>
Cc: Robert Love <rlove@google.com>
Cc: Vegard Nossum <vegard.nossum@gmail.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Creating a generic filesystem notification interface, fsnotify, which will be
used by inotify, dnotify, and eventually fanotify is really starting to
clutter the fs directory. This patch simply moves inotify and dnotify into
fs/notify/inotify and fs/notify/dnotify respectively to make both current fs/
and future notification tidier.
Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
