GitHub/LineageOS/android_kernel_motorola_exynos9610.git
13 years agotidy the trailing symlinks traversal up
Al Viro [Tue, 15 Mar 2011 02:20:34 +0000 (22:20 -0400)]
tidy the trailing symlinks traversal up

* pull the handling of current->total_link_count into
__do_follow_link()
* put the common "do ->put_link() if needed and path_put() the link"
  stuff into a helper (put_link(nd, link, cookie))
* rename __do_follow_link() to follow_link(), while we are at it

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoTurn resolution of trailing symlinks iterative everywhere
Al Viro [Tue, 15 Mar 2011 01:54:55 +0000 (21:54 -0400)]
Turn resolution of trailing symlinks iterative everywhere

The last remaining place (resolution of nested symlink) converted
to the loop of the same kind we have in path_lookupat() and
path_openat().

Note that we still *do* have a recursion in pathname resolution;
can't avoid it, really.  However, it's strictly for nested symlinks
now - i.e. ones in the middle of a pathname.

link_path_walk() has lost the tail now - it always walks everything
except the last component.

do_follow_link() renamed to nested_symlink() and moved down.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agosimplify link_path_walk() tail
Al Viro [Tue, 15 Mar 2011 01:28:04 +0000 (21:28 -0400)]
simplify link_path_walk() tail

Now that link_path_walk() is called without LOOKUP_PARENT
only from do_follow_link(), we can simplify the checks in
last component handling.  First of all, checking if we'd
arrived to a directory is not needed - the caller will check
it anyway.  And LOOKUP_FOLLOW is guaranteed to be there,
since we only get to that place with nd->depth > 0.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoMake trailing symlink resolution in path_lookupat() iterative
Al Viro [Mon, 14 Mar 2011 23:54:59 +0000 (19:54 -0400)]
Make trailing symlink resolution in path_lookupat() iterative

Now the only caller of link_path_walk() that does *not* pass
LOOKUP_PARENT is do_follow_link()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoupdate nd->inode in __do_follow_link() instead of after do_follow_link()
Al Viro [Tue, 15 Mar 2011 00:01:51 +0000 (20:01 -0400)]
update nd->inode in __do_follow_link() instead of after do_follow_link()

... and note that we only need to do it for LAST_BIND symlinks

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agopull handling of one pathname component into a helper
Al Viro [Sun, 13 Mar 2011 23:58:58 +0000 (19:58 -0400)]
pull handling of one pathname component into a helper

new helper: walk_component().  Handles everything except symlinks;
returns negative on error, 0 on success and 1 on symlinks we decided
to follow.  Drops out of RCU mode on such symlinks.

link_path_walk() and do_last() switched to using that.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofs: allow AT_EMPTY_PATH in linkat(), limit that to CAP_DAC_READ_SEARCH
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:42 +0000 (18:43 +0530)]
fs: allow AT_EMPTY_PATH in linkat(), limit that to CAP_DAC_READ_SEARCH

We don't want to allow creation of private hardlinks by different application
using the fd passed to them via SCM_RIGHTS. So limit the null relative name
usage in linkat syscall to CAP_DAC_READ_SEARCH

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
13 years agoAllow passing O_PATH descriptors via SCM_RIGHTS datagrams
Al Viro [Sun, 13 Mar 2011 21:08:22 +0000 (17:08 -0400)]
Allow passing O_PATH descriptors via SCM_RIGHTS datagrams

Just need to make sure that AF_UNIX garbage collector won't
confuse O_PATHed socket on filesystem for real AF_UNIX opened
socket.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoreadlinkat(), fchownat() and fstatat() with empty relative pathnames
Al Viro [Sun, 13 Mar 2011 19:56:26 +0000 (15:56 -0400)]
readlinkat(), fchownat() and fstatat() with empty relative pathnames

For readlinkat() we simply allow empty pathname; it will fail unless
we have dfd equal to O_PATH-opened symlink, so we are outside of
POSIX scope here.  For fchownat() and fstatat() we allow AT_EMPTY_PATH;
let the caller explicitly ask for such behaviour.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoAllow O_PATH for symlinks
Al Viro [Sun, 13 Mar 2011 20:42:14 +0000 (16:42 -0400)]
Allow O_PATH for symlinks

At that point we can't do almost nothing with them.  They can be opened
with O_PATH, we can manipulate such descriptors with dup(), etc. and
we can see them in /proc/*/{fd,fdinfo}/*.

We can't (and won't be able to) follow /proc/*/fd/* symlinks for those;
there's simply not enough information for pathname resolution to go on
from such point - to resolve a symlink we need to know which directory
does it live in.

We will be able to do useful things with them after the next commit, though -
readlinkat() and fchownat() will be possible to use with dfd being an
O_PATH-opened symlink and empty relative pathname.  Combined with
open_by_handle() it'll give us a way to do realink-by-handle and
lchown-by-handle without messing with more redundant syscalls.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoNew kind of open files - "location only".
Al Viro [Sun, 13 Mar 2011 07:51:11 +0000 (03:51 -0400)]
New kind of open files - "location only".

New flag for open(2) - O_PATH.  Semantics:
* pathname is resolved, but the file itself is _NOT_ opened
as far as filesystem is concerned.
* almost all operations on the resulting descriptors shall
fail with -EBADF.  Exceptions are:
1) operations on descriptors themselves (i.e.
close(), dup(), dup2(), dup3(), fcntl(fd, F_DUPFD),
fcntl(fd, F_DUPFD_CLOEXEC, ...), fcntl(fd, F_GETFD),
fcntl(fd, F_SETFD, ...))
2) fcntl(fd, F_GETFL), for a common non-destructive way to
check if descriptor is open
3) "dfd" arguments of ...at(2) syscalls, i.e. the starting
points of pathname resolution
* closing such descriptor does *NOT* affect dnotify or
posix locks.
* permissions are checked as usual along the way to file;
no permission checks are applied to the file itself.  Of course,
giving such thing to syscall will result in permission checks (at
the moment it means checking that starting point of ....at() is
a directory and caller has exec permissions on it).

fget() and fget_light() return NULL on such descriptors; use of
fget_raw() and fget_raw_light() is needed to get them.  That protects
existing code from dealing with those things.

There are two things still missing (they come in the next commits):
one is handling of symlinks (right now we refuse to open them that
way; see the next commit for semantics related to those) and another
is descriptor passing via SCM_RIGHTS datagrams.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoext4: Copy fs UUID to superblock
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:40 +0000 (18:43 +0530)]
ext4: Copy fs UUID to superblock

File system UUID is made available to application
via  /proc/<pid>/mountinfo

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoext3: Copy fs UUID to superblock.
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:39 +0000 (18:43 +0530)]
ext3: Copy fs UUID to superblock.

File system UUID is made available to application
via  /proc/<pid>/mountinfo

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agovfs: Export file system uuid via /proc/<pid>/mountinfo
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:38 +0000 (18:43 +0530)]
vfs: Export file system uuid via /proc/<pid>/mountinfo

We add a per superblock uuid field. File systems should
update the uuid in the fill_super callback

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agounistd.h: Add new syscalls numbers to asm-generic
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:38 +0000 (18:43 +0530)]
unistd.h: Add new syscalls numbers to asm-generic

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agox86: Add new syscalls for x86_64
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:37 +0000 (18:43 +0530)]
x86: Add new syscalls for x86_64

This patch add new syscalls to x86_64

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agox86: Add new syscalls for x86_32
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:35 +0000 (18:43 +0530)]
x86: Add new syscalls for x86_32

This patch adds new syscalls to x86_32

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofs: Remove i_nlink check from file system link callback
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:30 +0000 (18:43 +0530)]
fs: Remove i_nlink check from file system link callback

Now that VFS check for inode->i_nlink == 0 and returns proper
error, remove similar check from file system

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofs: Don't allow to create hardlink for deleted file
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:27 +0000 (18:43 +0530)]
fs: Don't allow to create hardlink for deleted file

Add inode->i_nlink == 0 check in VFS. Some of the file systems
do this internally. A followup patch will remove those instance.
This is needed to ensure that with link by handle we don't allow
to create hardlink of an unlinked file. The check also prevent a race
between unlink and link

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agovfs: Add open by file handle support
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:26 +0000 (18:43 +0530)]
vfs: Add open by file handle support

[AV: duplicate of open() guts removed; file_open_root() used instead]

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agovfs: Add name to file handle conversion support
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:26 +0000 (18:43 +0530)]
vfs: Add name to file handle conversion support

The syscall also return mount id which can be used
to lookup file system specific information such as uuid
in /proc/<pid>/mountinfo

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoNew AT_... flag: AT_EMPTY_PATH
Al Viro [Mon, 14 Mar 2011 22:56:51 +0000 (18:56 -0400)]
New AT_... flag: AT_EMPTY_PATH

For name_to_handle_at(2) we'll want both ...at()-style syscall that
would be usable for non-directory descriptors (with empty relative
pathname).  Introduce new flag (AT_EMPTY_PATH) to deal with that and
corresponding LOOKUP_EMPTY; teach user_path_at() and path_init() to
deal with the latter.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoexportfs: Return the minimum required handle size
Aneesh Kumar K.V [Sat, 29 Jan 2011 13:13:25 +0000 (18:43 +0530)]
exportfs: Return the minimum required handle size

The exportfs encode handle function should return the minimum required
handle size. This helps user to find out the handle size by passing 0
handle size in the first step and then redoing to the call again with
the returned handle size value.

Acked-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoclean statfs-like syscalls up
Al Viro [Sat, 12 Mar 2011 15:41:39 +0000 (10:41 -0500)]
clean statfs-like syscalls up

New helpers: user_statfs() and fd_statfs(), taking userland pathname and
descriptor resp. and filling struct kstatfs.  Syscalls of statfs family
(native, compat and foreign - osf and hpux on alpha and parisc resp.)
switched to those.  Removes some boilerplate code, simplifies cleanup
on errors...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoopen-style analog of vfs_path_lookup()
Al Viro [Fri, 11 Mar 2011 17:08:24 +0000 (12:08 -0500)]
open-style analog of vfs_path_lookup()

new function: file_open_root(dentry, mnt, name, flags) opens the file
vfs_path_lookup would arrive to.

Note that name can be empty; in that case the usual requirement that
dentry should be a directory is lifted.

open-coded equivalents switched to it, may_open() got down exactly
one caller and became static.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoreduce vfs_path_lookup() to do_path_lookup()
Al Viro [Thu, 10 Mar 2011 04:04:47 +0000 (23:04 -0500)]
reduce vfs_path_lookup() to do_path_lookup()

New lookup flag: LOOKUP_ROOT.  nd->root is set (and held) by caller,
path_init() starts walking from that place and all pathname resolution
machinery never drops nd->root if that flag is set.  That turns
vfs_path_lookup() into a special case of do_path_lookup() *and*
gets us down to 3 callers of link_path_walk(), making it finally
feasible to rip the handling of trailing symlink out of link_path_walk().
That will not only simply the living hell out of it, but make life
much simpler for unionfs merge.  Trailing symlink handling will
become iterative, which is a good thing for stack footprint in
a lot of situations as well.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agountangle do_lookup()
Al Viro [Fri, 11 Mar 2011 09:44:53 +0000 (04:44 -0500)]
untangle do_lookup()

That thing has devolved into rats nest of gotos; sane use of unlikely()
gets rid of that horror and gives much more readable structure:
* make a fast attempt to find a dentry; false negatives are OK.
In RCU mode if everything went fine, we are done, otherwise just drop
out of RCU.  If we'd done (RCU) ->d_revalidate() and it had not refused
outright (i.e. didn't give us -ECHILD), remember its result.
* now we are not in RCU mode and hopefully have a dentry.  If we
do not, lock parent, do full d_lookup() and if that has not found anything,
allocate and call ->lookup().  If we'd done that ->lookup(), remember that
dentry is good and we don't need to revalidate it.
* now we have a dentry.  If it has ->d_revalidate() and we can't
skip it, call it.
* hopefully dentry is good; if not, either fail (in case of error)
or try to invalidate it.  If d_invalidate() has succeeded, drop it and
retry everything as if original attempt had not found a dentry.
* now we can finish it up - deal with mountpoint crossing and
automount.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agopath_openat: clean ELOOP handling a bit
Al Viro [Wed, 9 Mar 2011 21:22:18 +0000 (16:22 -0500)]
path_openat: clean ELOOP handling a bit

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agodo_last: kill a rudiment of old ->d_revalidate() workaround
Al Viro [Wed, 9 Mar 2011 06:34:45 +0000 (01:34 -0500)]
do_last: kill a rudiment of old ->d_revalidate() workaround

There used to be time when ->d_revalidate() couldn't return an error.
So intents code had lookup_instantiate_filp() stash ERR_PTR(error)
in nd->intent.open.filp and had it checked after lookup_hash(), to
catch the otherwise silent failures.  That had been introduced by
commit 4af4c52f34606bdaab6930a845550c6fb02078a4.  These days
->d_revalidate() can and does propagate errors back to callers
explicitly, so this check isn't needed anymore.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofold __open_namei_create() and open_will_truncate() into do_last()
Al Viro [Wed, 9 Mar 2011 05:59:59 +0000 (00:59 -0500)]
fold __open_namei_create() and open_will_truncate() into do_last()

... and clean up a bit more

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agodo_last: unify may_open() call and everyting after it
Al Viro [Wed, 9 Mar 2011 05:36:45 +0000 (00:36 -0500)]
do_last: unify may_open() call and everyting after it

We have a bunch of diverging codepaths in do_last(); some of
them converge, but the case of having to create a new file
duplicates large part of common tail of the rest and exits
separately.  Massage them so that they could be merged.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agomove may_open() from __open_name_create() to do_last()
Al Viro [Wed, 9 Mar 2011 05:17:27 +0000 (00:17 -0500)]
move may_open() from __open_name_create() to do_last()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoexpand finish_open() in its only caller
Al Viro [Wed, 9 Mar 2011 05:13:14 +0000 (00:13 -0500)]
expand finish_open() in its only caller

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agosanitize pathname component hash calculation
Al Viro [Tue, 8 Mar 2011 19:17:44 +0000 (14:17 -0500)]
sanitize pathname component hash calculation

Lift it to lookup_one_len() and link_path_walk() resp. into the
same place where we calculated default hash function of the same
name.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agokill __lookup_one_len()
Al Viro [Tue, 8 Mar 2011 04:49:20 +0000 (23:49 -0500)]
kill __lookup_one_len()

only one caller left

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoswitch non-create side of open() to use of do_last()
Al Viro [Sun, 6 Mar 2011 03:58:25 +0000 (22:58 -0500)]
switch non-create side of open() to use of do_last()

Instead of path_lookupat() doing trailing symlink resolution,
use the same scheme as on the O_CREAT side.  Walk with
LOOKUP_PARENT, then (in do_last()) look the final component
up, then either open it or return error or, if it's a symlink,
give the symlink back to path_openat() to be resolved there.

The really messy complication here is RCU.  We don't want to drop
out of RCU mode before the final lookup, since we don't want to
bounce parent directory ->d_count without a good reason.

Result is _not_ pretty; later in the series we'll clean it up.
For now we are roughly back where we'd been before the revert
done by Nick's series - top-level logics of path_openat() is
cleaned up, do_last() does actual opening, symlink resolution is
done uniformly.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoget rid of nd->file
Al Viro [Sun, 6 Mar 2011 02:12:22 +0000 (21:12 -0500)]
get rid of nd->file

Don't stash the struct file * used as starting point of walk in nameidata;
pass file ** to path_init() instead.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoget rid of the last LOOKUP_RCU dependencies in link_path_walk()
Al Viro [Fri, 4 Mar 2011 19:44:37 +0000 (14:44 -0500)]
get rid of the last LOOKUP_RCU dependencies in link_path_walk()

New helper: terminate_walk().  An error has happened during pathname
resolution and we either drop nd->path or terminate RCU, depending
the mode we had been in.  After that, nd is essentially empty.
Switch link_path_walk() to using that for cleanup.

Now the top-level logics in link_path_walk() is back to sanity.  RCU
dependencies are in the lower-level functions.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agomake nameidata_dentry_drop_rcu_maybe() always leave RCU mode
Al Viro [Fri, 4 Mar 2011 19:39:30 +0000 (14:39 -0500)]
make nameidata_dentry_drop_rcu_maybe() always leave RCU mode

Now we have do_follow_link() guaranteed to leave without dangling RCU
and the next step will get LOOKUP_RCU logics completely out of
link_path_walk().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agomake handle_dots() leave RCU mode on error
Al Viro [Fri, 4 Mar 2011 19:35:59 +0000 (14:35 -0500)]
make handle_dots() leave RCU mode on error

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoclear RCU on all failure exits from link_path_walk()
Al Viro [Fri, 4 Mar 2011 19:28:10 +0000 (14:28 -0500)]
clear RCU on all failure exits from link_path_walk()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agopull handling of . and .. into inlined helper
Al Viro [Fri, 4 Mar 2011 19:22:06 +0000 (14:22 -0500)]
pull handling of . and .. into inlined helper

getting LOOKUP_RCU checks out of link_path_walk()...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agokill out_dput: in link_path_walk()
Al Viro [Thu, 24 Feb 2011 00:41:31 +0000 (19:41 -0500)]
kill out_dput: in link_path_walk()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoseparate -ESTALE/-ECHILD retries in do_filp_open() from real work
Al Viro [Wed, 23 Feb 2011 22:54:08 +0000 (17:54 -0500)]
separate -ESTALE/-ECHILD retries in do_filp_open() from real work

new helper: path_openat().  Does what do_filp_open() does, except
that it tries only the walk mode (RCU/normal/force revalidation)
it had been told to.

Both create and non-create branches are using path_lookupat() now.
Fixed the double audit_inode() in non-create branch.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoswitch do_filp_open() to struct open_flags
Al Viro [Wed, 23 Feb 2011 22:44:09 +0000 (17:44 -0500)]
switch do_filp_open() to struct open_flags

take calculation of open_flags by open(2) arguments into new helper
in fs/open.c, move filp_open() over there, have it and do_sys_open()
use that helper, switch exec.c callers of do_filp_open() to explicit
(and constant) struct open_flags.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoCollect "operation mode" arguments of do_last() into a structure
Al Viro [Wed, 23 Feb 2011 18:39:45 +0000 (13:39 -0500)]
Collect "operation mode" arguments of do_last() into a structure

No point messing with passing shitloads of "operation mode" arguments
to do_open() one by one, especially since they are not going to change
during do_filp_open().  Collect them into a struct, fill it and pass
to do_last() by reference.

Make sure that lookup intent flags are correctly set and removed - we
want them for do_last(), but they make no sense for __do_follow_link().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoclean up the failure exits after __do_follow_link() in do_filp_open()
Al Viro [Wed, 23 Feb 2011 03:27:28 +0000 (22:27 -0500)]
clean up the failure exits after __do_follow_link() in do_filp_open()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agopull security_inode_follow_link() into __do_follow_link()
Al Viro [Wed, 23 Feb 2011 02:24:38 +0000 (21:24 -0500)]
pull security_inode_follow_link() into __do_follow_link()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agopull dropping RCU on success of link_path_walk() into path_lookupat()
Al Viro [Wed, 23 Feb 2011 01:56:27 +0000 (20:56 -0500)]
pull dropping RCU on success of link_path_walk() into path_lookupat()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agountangle the "need_reval_dot" mess
Al Viro [Tue, 22 Feb 2011 20:50:10 +0000 (15:50 -0500)]
untangle the "need_reval_dot" mess

instead of ad-hackery around need_reval_dot(), do the following:
set a flag (LOOKUP_JUMPED) in the beginning of path, on absolute
symlink traversal, on ".." and on procfs-style symlinks.  Clear on
normal components, leave unchanged on ".".  Non-nested callers of
link_path_walk() call handle_reval_path(), which checks that flag
is set and that fs does want the final revalidate thing, then does
->d_revalidate().  In link_path_walk() all the return_reval stuff
is gone.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agomerge component type recognition
Al Viro [Tue, 22 Feb 2011 20:10:03 +0000 (15:10 -0500)]
merge component type recognition

no need to do it in three places...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agomerge path_init and path_init_rcu
Al Viro [Tue, 22 Feb 2011 19:02:58 +0000 (14:02 -0500)]
merge path_init and path_init_rcu

Actual dependency on whether we want RCU or not is in 3 small areas
(as it ought to be) and everything around those is the same in both
versions.  Since each function has only one caller and those callers
are on two sides of if (flags & LOOKUP_RCU), it's easier and cleaner
to merge them and pull the checks inside.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agosanitize path_walk() mess
Al Viro [Tue, 22 Feb 2011 04:38:09 +0000 (23:38 -0500)]
sanitize path_walk() mess

New helper: path_lookupat().  Basically, what do_path_lookup() boils to
modulo -ECHILD/-ESTALE handler.  path_walk* family is gone; vfs_path_lookup()
is using link_path_walk() directly, do_path_lookup() and do_filp_open()
are using path_lookupat().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agotake RCU-dependent stuff around exec_permission() into a new helper
Al Viro [Tue, 22 Feb 2011 02:34:47 +0000 (21:34 -0500)]
take RCU-dependent stuff around exec_permission() into a new helper

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agokill path_lookup()
Al Viro [Wed, 16 Feb 2011 20:15:47 +0000 (15:15 -0500)]
kill path_lookup()

all remaining callers pass LOOKUP_PARENT to it, so
flags argument can die; renamed to kern_path_parent()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofix race in audit_get_nd()
Al Viro [Wed, 16 Feb 2011 20:08:54 +0000 (15:08 -0500)]
fix race in audit_get_nd()

don't rely on pathname resolution ending up twice at the same point...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agocompat breakage in preadv() and pwritev()
Al Viro [Sun, 13 Mar 2011 06:50:58 +0000 (01:50 -0500)]
compat breakage in preadv() and pwritev()

Fix for a dumb preadv()/pwritev() compat bug - unlike the native
variants, compat_... ones forget to check FMODE_P{READ,WRITE}, so e.g.
on pipe the native preadv() will fail with -ESPIPE and compat one will
act as readv() and succeed.  Not critical, but it's a clear bug with trivial
fix.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoMerge branch 'media_fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Thu, 10 Mar 2011 21:22:10 +0000 (13:22 -0800)]
Merge branch 'media_fixes' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'media_fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
  [media] mantis_pci: remove asm/pgtable.h include
  [media] tda829x: fix regression in probe functions
  [media] mceusb: don't claim multifunction device non-IR parts
  [media] nuvoton-cir: fix wake from suspend
  [media] cx18: Add support for Hauppauge HVR-1600 models with s5h1411
  [media] ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang
  [media] cx25840: fix probing of cx2583x chips
  [media] cx23885: Remove unused 'err:' labels to quiet compiler warning
  [media] cx23885: Revert "Check for slave nack on all transactions"
  [media] DiB7000M: add pid filtering
  [media] Fix sysfs rc protocol lookup for rc-5-sz
  [media] au0828: fix VBI handling when in V4L2 streaming mode
  [media] ir-raw: Properly initialize the IR event (BZ#27202)
  [media] s2255drv: firmware re-loading changes
  [media] Fix double free of video_device in mem2mem_testdev
  [media] DM04/QQBOX memcpy to const char fix

13 years agoipmi: Fix IPMI errors due to timing problems
Doe, YiCheng [Thu, 10 Mar 2011 20:00:21 +0000 (14:00 -0600)]
ipmi: Fix IPMI errors due to timing problems

This patch fixes an issue in OpenIPMI module where sometimes an ABORT command
is sent after sending an IPMI request to BMC causing the IPMI request to fail.

Signed-off-by: YiCheng Doe <yicheng.doe@hp.com>
Signed-off-by: Corey Minyard <cminyard@mvista.com>
Acked-by: Tom Mingarelli <thomas.mingarelli@hp.com>
Tested-by: Andy Cress <andy.cress@us.kontron.com>
Tested-by: Mika Lansirine <Mika.Lansirinne@stonesoft.com>
Tested-by: Brian De Wolf <bldewolf@csupomona.edu>
Cc: Jean Michel Audet <Jean-Michel.Audet@ca.Kontron.com>
Cc: Jozef Sudelsky <jozef.sudolsky@elbiahosting.sk>
Acked-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Thu, 10 Mar 2011 21:16:01 +0000 (13:16 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  fs/dcache: allow d_obtain_alias() to return unhashed dentries
  Check for immutable/append flag in fallocate path
  sysctl: the include of rcupdate.h is only needed in the kernel
  fat: fix d_revalidate oopsen on NFS exports
  jfs: fix d_revalidate oopsen on NFS exports
  ocfs2: fix d_revalidate oopsen on NFS exports
  gfs2: fix d_revalidate oopsen on NFS exports
  fuse: fix d_revalidate oopsen on NFS exports
  ceph: fix d_revalidate oopsen on NFS exports
  reiserfs xattr ->d_revalidate() shouldn't care about RCU
  /proc/self is never going to be invalidated...

13 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 10 Mar 2011 21:09:26 +0000 (13:09 -0800)]
Merge branch 'x86-fixes-for-linus' of git://git./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, UV: Initialize the broadcast assist unit base destination node id properly
  x86, numa: Fix numa_emulation code with memory-less node0
  x86, build: Make sure mkpiggy fails on read error

13 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 10 Mar 2011 21:08:59 +0000 (13:08 -0800)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: Fix sched rt group scheduling when hierachy is enabled

13 years agoMerge branch 'perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux...
Linus Torvalds [Thu, 10 Mar 2011 21:07:38 +0000 (13:07 -0800)]
Merge branch 'perf/urgent' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  perf symbols: Avoid resolving [kernel.kallsyms] to real path for buildid cache
  perf symbols: Fix vmlinux path when not using --symfs

13 years agodrm/i915: Revive combination mode for backlight control
Takashi Iwai [Thu, 10 Mar 2011 13:02:12 +0000 (14:02 +0100)]
drm/i915: Revive combination mode for backlight control

This reverts commit 951f3512dba5bd44cda3e5ee22b4b522e4bb09fb

    drm/i915: Do not handle backlight combination mode specially

since this commit introduced other regressions due to untouched LBPC
register, e.g. the backlight dimmed after resume.

In addition to the revert, this patch includes a fix for the original
issue (weird backlight levels) by removing the wrong bit shift for
computing the current backlight level.
Also, including typo fixes (lpbc -> lbpc).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=34524
Acked-by: Indan Zupancic <indan@nul.nu>
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agofs/dcache: allow d_obtain_alias() to return unhashed dentries
J. Bruce Fields [Tue, 18 Jan 2011 20:45:09 +0000 (15:45 -0500)]
fs/dcache: allow d_obtain_alias() to return unhashed dentries

Without this patch, inodes are not promptly freed on last close of an
unlinked file by an nfs client:

client$ mount -tnfs4 server:/export/ /mnt/
client$ tail -f /mnt/FOO
...
server$ df -i /export
server$ rm /export/FOO
(^C the tail -f)
server$ df -i /export
server$ echo 2 >/proc/sys/vm/drop_caches
server$ df -i /export

the df's will show that the inode is not freed on the filesystem until
the last step, when it could have been freed after killing the client's
tail -f. On-disk data won't be deallocated either, leading to possible
spurious ENOSPC.

This occurs because when the client does the close, it arrives in a
compound with a putfh and a close, processed like:

- putfh: look up the filehandle. Â The only alias found for the
  inode will be DCACHE_UNHASHED alias referenced by the filp
  this, so it creates a new DCACHE_DISCONECTED dentry and
  returns that instead.
- close: closes the existing filp, which is destroyed
  immediately by dput() since it's DCACHE_UNHASHED.
- end of the compound: release the reference
  to the current filehandle, and dput() the new
  DCACHE_DISCONECTED dentry, which gets put on the
  unused list instead of being destroyed immediately.

Nick Piggin suggested fixing this by allowing d_obtain_alias to return
the unhashed dentry that is referenced by the filp, instead of making it
create a new dentry.

Leave __d_find_alias() alone to avoid changing behavior of other
callers.

Also nfsd doesn't need all the checks of __d_find_alias(); any dentry,
hashed or unhashed, disconnected or not, should work.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoCheck for immutable/append flag in fallocate path
Marco Stornelli [Sat, 5 Mar 2011 10:10:19 +0000 (11:10 +0100)]
Check for immutable/append flag in fallocate path

In the fallocate path the kernel doesn't check for the immutable/append
flag. It's possible to have a race condition in this scenario: an
application open a file in read/write and it does something, meanwhile
root set the immutable flag on the file, the application at that point
can call fallocate with success. In addition, we don't allow to do any
unreserve operation on an append only file but only the reserve one.

Signed-off-by: Marco Stornelli <marco.stornelli@gmail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agosysctl: the include of rcupdate.h is only needed in the kernel
Stephen Rothwell [Thu, 10 Mar 2011 00:25:43 +0000 (11:25 +1100)]
sysctl: the include of rcupdate.h is only needed in the kernel

Fixes this built error:

include/linux/sysctl.h:28: included file 'linux/rcupdate.h' is not exported

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofat: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:45:49 +0000 (03:45 -0500)]
fat: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agojfs: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:45:28 +0000 (03:45 -0500)]
jfs: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoocfs2: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:45:07 +0000 (03:45 -0500)]
ocfs2: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agogfs2: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:44:48 +0000 (03:44 -0500)]
gfs2: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agofuse: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:44:31 +0000 (03:44 -0500)]
fuse: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoceph: fix d_revalidate oopsen on NFS exports
Al Viro [Thu, 10 Mar 2011 08:44:05 +0000 (03:44 -0500)]
ceph: fix d_revalidate oopsen on NFS exports

can't blindly check nd->flags in ->d_revalidate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoreiserfs xattr ->d_revalidate() shouldn't care about RCU
Al Viro [Wed, 16 Feb 2011 15:14:56 +0000 (10:14 -0500)]
reiserfs xattr ->d_revalidate() shouldn't care about RCU

... it returns an error unconditionally

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years ago/proc/self is never going to be invalidated...
Al Viro [Wed, 16 Feb 2011 15:04:50 +0000 (10:04 -0500)]
/proc/self is never going to be invalidated...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Thu, 10 Mar 2011 00:46:06 +0000 (16:46 -0800)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/pseries: Disable VPNH feature
  powerpc/iseries: Fix early init access to lppaca

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Thu, 10 Mar 2011 00:45:02 +0000 (16:45 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6:
  net: don't allow CAP_NET_ADMIN to load non-netdev kernel modules

13 years agosysctl: the include of rcupdate.h is only needed in the kernel
Stephen Rothwell [Thu, 10 Mar 2011 00:14:17 +0000 (11:14 +1100)]
sysctl: the include of rcupdate.h is only needed in the kernel

Fixes this build-check error:

  include/linux/sysctl.h:28: included file 'linux/rcupdate.h' is not exported

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agonet: don't allow CAP_NET_ADMIN to load non-netdev kernel modules
Vasiliy Kulikov [Tue, 1 Mar 2011 21:33:13 +0000 (00:33 +0300)]
net: don't allow CAP_NET_ADMIN to load non-netdev kernel modules

Since a8f80e8ff94ecba629542d9b4b5f5a8ee3eb565c any process with
CAP_NET_ADMIN may load any module from /lib/modules/.  This doesn't mean
that CAP_NET_ADMIN is a superset of CAP_SYS_MODULE as modules are
limited to /lib/modules/**.  However, CAP_NET_ADMIN capability shouldn't
allow anybody load any module not related to networking.

This patch restricts an ability of autoloading modules to netdev modules
with explicit aliases.  This fixes CVE-2011-1019.

Arnd Bergmann suggested to leave untouched the old pre-v2.6.32 behavior
of loading netdev modules by name (without any prefix) for processes
with CAP_SYS_MODULE to maintain the compatibility with network scripts
that use autoloading netdev modules by aliases like "eth0", "wlan0".

Currently there are only three users of the feature in the upstream
kernel: ipip, ip_gre and sit.

    root@albatros:~# capsh --drop=$(seq -s, 0 11),$(seq -s, 13 34) --
    root@albatros:~# grep Cap /proc/$$/status
    CapInh: 0000000000000000
    CapPrm: fffffff800001000
    CapEff: fffffff800001000
    CapBnd: fffffff800001000
    root@albatros:~# modprobe xfs
    FATAL: Error inserting xfs
    (/lib/modules/2.6.38-rc6-00001-g2bf4ca3/kernel/fs/xfs/xfs.ko): Operation not permitted
    root@albatros:~# lsmod | grep xfs
    root@albatros:~# ifconfig xfs
    xfs: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep xfs
    root@albatros:~# lsmod | grep sit
    root@albatros:~# ifconfig sit
    sit: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep sit
    root@albatros:~# ifconfig sit0
    sit0      Link encap:IPv6-in-IPv4
      NOARP  MTU:1480  Metric:1

    root@albatros:~# lsmod | grep sit
    sit                    10457  0
    tunnel4                 2957  1 sit

For CAP_SYS_MODULE module loading is still relaxed:

    root@albatros:~# grep Cap /proc/$$/status
    CapInh: 0000000000000000
    CapPrm: ffffffffffffffff
    CapEff: ffffffffffffffff
    CapBnd: ffffffffffffffff
    root@albatros:~# ifconfig xfs
    xfs: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep xfs
    xfs                   745319  0

Reference: https://lkml.org/lkml/2011/2/24/203

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Kees Cook <kees.cook@canonical.com>
Signed-off-by: James Morris <jmorris@namei.org>
13 years agopowerpc/pseries: Disable VPNH feature
Benjamin Herrenschmidt [Wed, 9 Mar 2011 13:00:14 +0000 (13:00 +0000)]
powerpc/pseries: Disable VPNH feature

This feature triggers nasty races in the scheduler between the
rebuilding of the topology and the load balancing code, causing
the machine to hang.

Disable it for now until the races are fixed.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/iseries: Fix early init access to lppaca
Benjamin Herrenschmidt [Sun, 6 Mar 2011 18:02:31 +0000 (18:02 +0000)]
powerpc/iseries: Fix early init access to lppaca

The combination of commit

8154c5d22d91cd16bd9985b0638c8957e4688d0e and
93c22703efa72c7527dbd586d1951c1f4a85fd70

Broke boot on iSeries.

The problem is that iSeries very early boot code, which generates
the device-tree and runs before our normal early initializations
does need access the lppaca's very early, before the PACA array is
initialized, and in fact even before the boot PACA has been
initialized (it contains all 0's at this stage).

However, the first patch above makes that code use the new
llpaca_of(cpu) accessor, which itself is changed by the second patch to
use the PACA array.

We fix that by reverting iSeries to directly dereferencing the array. In
addition, we fix all iterators in the iSeries code to always skip CPU
whose number is above 63 which is the maximum size of that array and
the maximum number of supported CPUs on these machines.

Additionally, we make sure the boot_paca is properly initialized
in our early startup code.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agoMerge branch 'for-2.6.38' of git://linux-nfs.org/~bfields/linux
Linus Torvalds [Wed, 9 Mar 2011 22:52:09 +0000 (14:52 -0800)]
Merge branch 'for-2.6.38' of git://linux-nfs.org/~bfields/linux

* 'for-2.6.38' of git://linux-nfs.org/~bfields/linux:
  nfsd: wrong index used in inner loop
  nfsd4: fix bad pointer on failure to find delegation
  NFSD: fix decode_cb_sequence4resok

13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Wed, 9 Mar 2011 22:04:40 +0000 (14:04 -0800)]
Merge git://git./linux/kernel/git/wim/linux-2.6-watchdog

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  watchdog: sbc_fitpc2_wdt, fix crash on systems without DMI_BOARD_NAME

13 years agoMerge branch 'for-2639-rc7/i2c-fixes' of git://git.fluff.org/bjdooks/linux
Linus Torvalds [Wed, 9 Mar 2011 22:03:59 +0000 (14:03 -0800)]
Merge branch 'for-2639-rc7/i2c-fixes' of git://git.fluff.org/bjdooks/linux

* 'for-2639-rc7/i2c-fixes' of git://git.fluff.org/bjdooks/linux:
  i2c-eg20t: include slab.h for memory allocations
  i2c-ocores: Fix pointer type mismatch error
  i2c-omap: Program I2C_WE on OMAP4 to enable i2c wakeup

13 years agoalpha: fix compile error from IRQ clean up
Matt Turner [Wed, 9 Mar 2011 16:15:13 +0000 (11:15 -0500)]
alpha: fix compile error from IRQ clean up

Signed-off-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Wed, 9 Mar 2011 22:01:42 +0000 (14:01 -0800)]
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] pcc-cpufreq: don't load driver if get_freq fails during init.

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc
Linus Torvalds [Wed, 9 Mar 2011 22:00:44 +0000 (14:00 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/cjb/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc:
  mmc: fix CONFIG_MMC_UNSAFE_RESUME regression

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Wed, 9 Mar 2011 21:55:51 +0000 (13:55 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  nd->inode is not set on the second attempt in path_walk()
  unfuck proc_sysctl ->d_compare()
  minimal fix for do_filp_open() race

13 years agowatchdog: sbc_fitpc2_wdt, fix crash on systems without DMI_BOARD_NAME
Jiri Slaby [Mon, 28 Feb 2011 09:16:29 +0000 (10:16 +0100)]
watchdog: sbc_fitpc2_wdt, fix crash on systems without DMI_BOARD_NAME

Some systems don't provide DMI_BOARD_NAME in their DMI tables. Avoid
crash in such situations in fitpc2_wdt_init.

The fix is to check if the dmi_get_system_info return value is NULL.

The oops:
BUG: unable to handle kernel NULL pointer dereference at           (null)
IP: [<ffffffff81253ae6>] strstr+0x26/0xa0
PGD 3966e067 PUD 39605067 PMD 0
Oops: 0000 [#1] SMP
last sysfs file: /sys/devices/system/cpu/cpu1/cache/index2/shared_cpu_map
CPU 1
Modules linked in: ...
Pid: 1748, comm: modprobe Not tainted 2.6.37-22-default #1 /Bochs
RIP: 0010:[<ffffffff81253ae6>]  [<ffffffff81253ae6>] strstr+0x26/0xa0
RSP: 0018:ffff88003ad73f18  EFLAGS: 00010206
RAX: 0000000000000000 RBX: 00000000ffffffed RCX: 00000000ffffffff
RDX: ffffffffa003f4cc RSI: ffffffffa003f4c2 RDI: 0000000000000000
...
CR2: 0000000000000000 CR3: 000000003b7ac000 CR4: 00000000000006e0
...
Process modprobe (pid: 1748, threadinfo ffff88003ad72000, task ffff88002e6365c0)
Stack: ...
Call Trace:
 [<ffffffffa004201f>] fitpc2_wdt_init+0x1f/0x13c [sbc_fitpc2_wdt]
 [<ffffffff810002da>] do_one_initcall+0x3a/0x170
...
Code: f3 c3 0f 1f 00 80 3e 00 53 48 89 f8 74 1b 48 89 f2 0f 1f 40 00 48 83 c2 01 80 3a 00 75 f7 49 89 d0 48 89 f8 49 29 f0 75 02 5b c3 <80> 3f 00 74 0e 0f 1f 44 00 00 48 83 c0 01 80 38 00 75 f7 49 89

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
13 years ago[CPUFREQ] pcc-cpufreq: don't load driver if get_freq fails during init.
Naga Chumbalkar [Wed, 9 Mar 2011 14:02:49 +0000 (14:02 +0000)]
[CPUFREQ] pcc-cpufreq: don't load driver if get_freq fails during init.

Return 0 on failure. This will cause the initialization of the driver
to fail and prevent the driver from loading if the BIOS cannot handle
the PCC interface command to "get frequency". Otherwise, the driver
will load and display a very high value like "4294967274" (which is
actually -EINVAL) for frequency:

# cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
4294967274

Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
CC: stable@kernel.org
Signed-off-by: Dave Jones <davej@redhat.com>
13 years agoperf symbols: Avoid resolving [kernel.kallsyms] to real path for buildid cache
Lin Ming [Thu, 3 Mar 2011 15:23:57 +0000 (23:23 +0800)]
perf symbols: Avoid resolving [kernel.kallsyms] to real path for buildid cache

kallsyms has a virtual file name [kernel.kallsyms].  Currently, it can't
be added to buildid cache successfully because the code
(build_id_cache__add_s) tries to resolve [kernel.kallsyms] to a real
absolute pathname and that fails.

Fixes it by not resolving it and just use the name [kernel.kallsyms].
So dir ~/.debug/[kernel.kallsyms] is created.

Original bug report at:
https://lkml.org/lkml/2011/3/1/524

Tested-by: Han Pingtian <phan@redhat.com>
Cc: Han Pingtian <phan@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1299165837-27817-1-git-send-email-ming.m.lin@intel.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agox86, UV: Initialize the broadcast assist unit base destination node id properly
Cliff Wickman [Wed, 9 Mar 2011 14:15:57 +0000 (08:15 -0600)]
x86, UV: Initialize the broadcast assist unit base destination node id properly

The BAU's initialization of the broadcast description header is
lacking the coherence domain (high bits) in the nasid.  This
causes a catastrophic system failure when running on a system
with multiple coherence domains.

Signed-off-by: Cliff Wickman <cpw@sgi.com>
LKML-Reference: <E1PxKBB-0005F0-3U@eag09.americas.sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agond->inode is not set on the second attempt in path_walk()
Al Viro [Wed, 9 Mar 2011 02:16:28 +0000 (21:16 -0500)]
nd->inode is not set on the second attempt in path_walk()

We leave it at whatever it had been pointing to after the
first link_path_walk() had failed with -ESTALE.  Things
do not work well after that...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agonfsd: wrong index used in inner loop
roel [Tue, 8 Mar 2011 21:32:26 +0000 (22:32 +0100)]
nfsd: wrong index used in inner loop

Index i was already used in the outer loop

Cc: stable@kernel.org
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
13 years agoi2c-eg20t: include slab.h for memory allocations
Wolfram Sang [Wed, 23 Feb 2011 10:11:35 +0000 (11:11 +0100)]
i2c-eg20t: include slab.h for memory allocations

Fixes (with v2.6.38-rc3/parisc/parisc-allmodconfig):
  src/drivers/i2c/busses/i2c-eg20t.c:720: error: implicit declaration of function 'kzalloc'
  src/drivers/i2c/busses/i2c-eg20t.c:790: error: implicit declaration of function 'kfree'

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Cc: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
13 years agoi2c-ocores: Fix pointer type mismatch error
Grant Likely [Mon, 28 Feb 2011 20:52:32 +0000 (13:52 -0700)]
i2c-ocores: Fix pointer type mismatch error

ocores_i2c_of_probe needs to use a const __be32 type for handing
device tree property values.  This patch fixed the following build
warning:

 CC      drivers/i2c/busses/i2c-ocores.o
  drivers/i2c/busses/i2c-ocores.c: In function 'ocores_i2c_of_probe':
  drivers/i2c/busses/i2c-ocores.c:254: warning: assignment discards qualifiers from pointer target type
  drivers/i2c/busses/i2c-ocores.c:261: warning: assignment discards qualifiers from pointer target type

Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Cc: Peter Korsgaard <jacmet@sunsite.dk>
Cc: Ben Dooks <ben-linux@fluff.org>
Cc: linux-i2c@vger.kernel.org
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
13 years agommc: fix CONFIG_MMC_UNSAFE_RESUME regression
Ohad Ben-Cohen [Tue, 8 Mar 2011 21:32:02 +0000 (23:32 +0200)]
mmc: fix CONFIG_MMC_UNSAFE_RESUME regression

30201e7f3 ("mmc: skip detection of nonremovable cards on rescan")
allowed skipping detection of nonremovable cards on mmc_rescan().
The intention was to only skip detection of hardwired cards that
cannot be removed, so make sure this is indeed the case by directly
checking for (lack of) MMC_CAP_NONREMOVABLE, instead of using
mmc_card_is_removable(), which is overloaded with
CONFIG_MMC_UNSAFE_RESUME semantics.

The user-visible symptom of the bug this patch fixes is that no
"mmc: card XXXX removed" message appears in dmesg when a card is
removed and CONFIG_MMC_UNSAFE_RESUME=y.

Reported-and-tested-by: Dmitry Shmidt <dimitrysh@google.com>
Reported-and-tested-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
13 years agounfuck proc_sysctl ->d_compare()
Al Viro [Tue, 8 Mar 2011 06:25:28 +0000 (01:25 -0500)]
unfuck proc_sysctl ->d_compare()

a) struct inode is not going to be freed under ->d_compare();
however, the thing PROC_I(inode)->sysctl points to just might.
Fortunately, it's enough to make freeing that sucker delayed,
provided that we don't step on its ->unregistering, clear
the pointer to it in PROC_I(inode) before dropping the reference
and check if it's NULL in ->d_compare().

b) I'm not sure that we *can* walk into NULL inode here (we recheck
dentry->seq between verifying that it's still hashed / fetching
dentry->d_inode and passing it to ->d_compare() and there's no
negative hashed dentries in /proc/sys/*), but if we can walk into
that, we really should not have ->d_compare() return 0 on it!
Said that, I really suspect that this check can be simply killed.
Nick?

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
13 years agoLinux 2.6.38-rc8
Linus Torvalds [Tue, 8 Mar 2011 05:09:37 +0000 (21:09 -0800)]
Linux 2.6.38-rc8

13 years agoMerge branch 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 8 Mar 2011 04:46:39 +0000 (20:46 -0800)]
Merge branch 's5p-fixes-for-linus' of git://git./linux/kernel/git/kgene/linux-samsung

* 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
  ARM: S3C64XX: Update regulator names for debugfs compatiblity on SMDK6410
  ARM: S3C64XX: Fix build with WM1190 disabled and WM1192 enabled on SMDK6410
  ARM: S3C64XX: Reduce output of s3c64xx_dma_init1()
  ARM: S3C64XX: Tone down SDHCI debugging
  ARM: S3C64XX: Add clock for i2c1
  ARM: S3C64XX: Staticise non-exported GPIO to interrupt functions
  ARM: SAMSUNG: Include devs.h in dev-uart.c to prototype devices
  ARM: S3C64XX: Fix keypad setup to configure correct number of rows
  ARM: S3C2440: Fix usage gpio bank j pin definitions on GTA02
  ARM: S5P64X0: Fix number of GPIO lines in Bank F
  ARM: S3C2440: Select missing S3C_DEV_USB_HOST on GTA02