Since Ulrich is OK with getting rid of dup3(fd, fd, flags) completely,
to hell the damn thing goes.  Corner case for dup2() is handled in
sys_dup2() (complete with -EBADF if dup2(fd, fd) is called with fd
that is not open), the rest is done in dup3().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fs.h needs path.h, not namei.h; nfs_fs.h doesn't need it at all.
Several places in the tree needed direct include.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

make it atomic_long_t; while we are at it, get rid of useless checks in affs,
hfs and hpfs - ->open() always has it equal to 1, ->release() - to 0.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Al Viro notice one cornercase that the new dup3() code.  The dup2()
function, as a special case, handles dup-ing to the same file
descriptor.  In this case the current dup3() code does nothing at
all.  I.e., it ingnores the flags parameter.  This shouldn't happen,
the close-on-exec flag should be set if requested.

In case the O_CLOEXEC bit in the flags parameter is not set the
dup3() function should behave in this respect identical to dup2().
This means dup3(fd, fd, 0) should not actively reset the c-o-e
flag.

The patch below implements this minor change.

[AV: credits to Artur Grabowski for bringing that up as potential subtle point
in dup2() behaviour]
Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Preparation to untangling intents mess: reduce the number of do_path_lookup()
callers.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* do not pass nameidata; struct path is all the callers want.
* switch to new helpers:
	user_path_at(dfd, pathname, flags, &path)
	user_path(pathname, &path)
	user_lpath(pathname, &path)
	user_path_dir(pathname, &path)  (fail if not a directory)
  The last 3 are trivial macro wrappers for the first one.
* remove nameidata in callers.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Almost all users __user_walk_fd() and friends care only about struct path.
Get rid of the few that do not.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Incidentally, the name that gives hundreds of false positives on grep
is not a good idea...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

On Mon, May 19, 2008 at 12:01:49AM +0200, Marcin Slusarz wrote:
> open_exec is needlessly indented, calls ERR_PTR with 0 argument
> (which is not valid errno) and jumps into middle of function
> just to return value.
> So clean it up a bit.

Still looks rather messy.  See below for a better version.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Move the immutable and append-only checks from chmod, chown and utimes
into notify_change().  Checks for immutable and append-only files are
always performed by the VFS and not by the filesystem (see
permission() and may_...() in namei.c), so these belong in
notify_change(), and not in inode_change_ok().

This should be completely equivalent.

CC: Ulrich Drepper <drepper@redhat.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

The FAT_IOCTL_SET_ATTRIBUTES ioctl() calls notify_change() to change
the file mode before changing the inode attributes.  Replace with
explicit calls to security_inode_setattr(), fat_setattr() and
fsnotify_change().

This is equivalent to the original.  The reason it is needed, is that
later in the series we move the immutable check into notify_change().
That would break the FAT_IOCTL_SET_ATTRIBUTES ioctl, as it needs to
perform the mode change regardless of the immutability of the file.

[Fix error if fat is built as a module.  Thanks to OGAWA Hirofumi for
noticing.]
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Untange the mess that is do_utimes().  Add kerneldoc comment to
do_utimes().

CC: Ulrich Drepper <drepper@redhat.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Add a new ia_valid flag: ATTR_TIMES_SET, to handle the
UTIMES_OMIT/UTIMES_NOW and UTIMES_NOW/UTIMES_OMIT cases.  In these
cases neither ATTR_MTIME_SET nor ATTR_ATIME_SET is in the flags, yet
the POSIX draft specifies that permission checking is performed the
same way as if one or both of the times was explicitly set to a
timestamp.

See the path "vfs: utimensat(): fix error checking for
{UTIME_NOW,UTIME_OMIT} case" by Michael Kerrisk for the patch
introducing this behavior.

This is a cleanup, as well as allowing filesystems (NFS/fuse/...) to
perform their own permission checking instead of the default.

CC: Ulrich Drepper <drepper@redhat.com>
CC: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

- use kstrdup() instead of kmalloc() + memcpy()
- return NULL if allocating ->mnt_devname failed
- mnt_devname should be const
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... and get rid of the last "let's deduce mask from nameidata->flags"
bit.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* MAY_CHDIR is redundant - it's an equivalent of MAY_ACCESS
* MAY_ACCESS on fuse should affect only the last step of pathname resolution
* fchdir() and chroot() should pass MAY_ACCESS, for the same reason why
  chdir() needs that.
* now that we pass MAY_ACCESS explicitly in all cases, LOOKUP_ACCESS can be
  removed; it has no business being in nameidata.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

long overdue...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... so we ought to pass MAY_CHDIR to vfs_permission() instead of having
it triggered on every step of preceding pathname resolution.  LOOKUP_CHDIR
is killed by that.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Remove the unused mode parameter from vfs_symlink and callers.

Thanks to Tetsuo Handa for noticing.

CC: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>

Why not reuse "inode" which is assigned as

  struct inode *inode = old_dentry->d_inode;

in the beginning of vfs_link() ?
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>

All calls to remove_suid() are made with a file pointer, because
(similarly to file_update_time) it is called when the file is written.

Clean up callers by passing in a file instead of a dentry.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>

vfs_permission(MAY_WRITE) already checked for the inode being
immutable, so no need to repeat it.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Christoph Hellwig <hch@infradead.org>

* kill nameidata * argument; map the 3 bits in ->flags anybody cares
  about to new MAY_... ones and pass with the mask.
* kill redundant gfs2_iop_permission()
* sanitize ecryptfs_permission()
* fix remaining places where ->permission() instances might barf on new
  MAY_... found in mask.

The obvious next target in that direction is permission(9)

folded fix for nfs_permission() breakage from Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

hpfs_unlink() calls permission() prior to truncating the file.  HPFS
doesn't define a .permission method, so replace with explicit call to
generic_permission().

This is equivalent, except that devcgroup_inode_permission() and
security_inode_permission() are not called.

The truncation is just an implementation detail of the unlink, so
these security checks are unnecessary.

I suspect that even calling generic_permission() is unnecessary, since
we shouldn't mind if the file isn't writable.  But I leave that to the
maintainer to decide.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>

* keep references to ctl_table_head and ctl_table in /proc/sys inodes
* grab the former during operations, use the latter for access to
  entry if that succeeds
* have ->d_compare() check if table should be seen for one who does lookup;
  that allows us to avoid flipping inodes - if we have the same name resolve
  to different things, we'll just keep several dentries and ->d_compare()
  will reject the wrong ones.
* have ->lookup() and ->readdir() scan the table of our inode first, then
  walk all ctl_table_header and scan ->attached_by for those that are
  attached to our directory.
* implement ->getattr().
* get rid of insane amounts of tree-walking
* get rid of the need to know dentry in ->permission() and of the contortions
  induced by that.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

In a sense, that's the heart of the series.  It's based on the following
property of the trees we are actually asked to add: they can be split into
stem that is already covered by registered trees and crown that is entirely
new.  IOW, if a/b and a/c/d are introduced by our tree, then a/c is also
introduced by it.

That allows to associate tree and table entry with each node in the union;
while directory nodes might be covered by many trees, only one will cover
the node by its crown.  And that will allow much saner logics for /proc/sys
in the next patches.  This patch introduces the data structures needed to
keep track of that.

When adding a sysctl table, we find a "parent" one.  Which is to say,
find the deepest node on its stem that already is present in one of the
tables from our table set or its ancestor sets.  That table will be our
parent and that node in it - attachment point.  Add our table to list
anchored in parent, have it refer the parent and contents of attachment
point.  Also remember where its crown lives.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Massage ipv4 initialization - make sure that net.ipv4 appears as
non-per-net-namespace before it shows up in per-net-namespace sysctls.
That's the only change outside of sysctl.c needed to get sane ordering
rules and data structures for sysctls (esp. for procfs side of that
mess).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Refcount the sucker; instead of freeing it by the end of unregistration
just drop the refcount and free only when it hits zero.  Make sure that
we _always_ make ->unregistering non-NULL in start_unregistering().

That allows anybody to get a reference to such puppy, preventing its
freeing and reuse.  It does *not* block unregistration.  Anybody who
holds such a reference can
	* try to grab a "use" reference (ctl_head_grab()); that will
succeeds if and only if it hadn't entered unregistration yet.  If it
succeeds, we can use it in all normal ways until we release the "use"
reference (with ctl_head_finish()).  Note that this relies on having
->unregistering become non-NULL in all cases when one starts to unregister
the sucker.
	* keep pointers to ctl_table entries; they *can* be freed if
the entire thing is unregistered.  However, if ctl_head_grab() succeeds,
we know that unregistration had not happened (and will not happen until
ctl_head_finish()) and such pointers can be used safely.

IOW, now we can have inodes under /proc/sys keep references to ctl_table
entries, protecting them with references to ctl_table_header and
grabbing the latter for the duration of operations that require access
to ctl_table.  That won't cause deadlocks, since unregistration will not
be stopped by mere keeping a reference to ctl_table_header.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

New object: set of sysctls [currently - root and per-net-ns].
Contains: pointer to parent set, list of tables and "should I see this set?"
method (->is_seen(set)).
Current lists of tables are subsumed by that; net-ns contains such a beast.
->lookup() for ctl_table_root returns pointer to ctl_table_set instead of
that to ->list of that ctl_table_set.

[folded compile fixes by rdd for configs without sysctl]
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

hppfs_permission() is equivalent to the '.permission == NULL' case.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Merge fifo and pipe file_operations.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Lookup can install a child dentry for a deleted directory.  This keeps
the directory dentry alive, and the inode pinned in the cache and on
disk, even after all external references have gone away.

This isn't a big problem normally, since memory pressure or umount
will clear out the directory dentry and its children, releasing the
inode.  But for UBIFS this causes problems because its orphan area can
overflow.

Fix this by returning ENOENT for all lookups on a S_DEAD directory
before creating a child dentry.

Thanks to Zoltan Sogor for noticing this while testing UBIFS, and
Artem for the excellent analysis of the problem and testing.
Reported-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Tested-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  ftrace: fix modular build
  ftrace: disable tracing on acpi idle calls
  ftrace: remove latency-tracer leftover
  ftrace: only trace preempt off with preempt tracer
  ftrace: fix 4d3702b6 (post-v2.6.26): WARNING: at kernel/lockdep.c:2731 check_flags (ftrace)

Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, AMD IOMMU: include amd_iommu_last_bdf in device initialization
  x86: fix IBM Summit based systems' phys_cpu_present_map on 32-bit kernels
  x86, RDC321x: remove gpio.h complications
  x86, RDC321x: add to mach-default
  crashdump: fix undefined reference to `elfcorehdr_addr'
  flag parameters: fix compile error of sys_epoll_create1