The two methods essentially do the same: find the real dentry/inode
belonging to an overlay dentry. The difference is in the usage:
vfs_open() uses ->d_select_inode() and expects the function to perform
copy-up if necessary based on the open flags argument.
file_dentry() uses ->d_real() passing in the overlay dentry as well as the
underlying inode.
vfs_rename() uses ->d_select_inode() but passes zero flags. ->d_real()
with a zero inode would have worked just as well here.
This patch merges the functionality of ->d_select_inode() into ->d_real()
by adding an 'open_flags' argument to the latter.
[Al Viro] Make the signature of d_real() match that of ->d_real() again.
And constify the inode argument, while we are at it.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Introduce a function may_open_dev that tests MNT_NODEV and a new
superblock flab SB_I_NODEV. Use this new function in all of the
places where MNT_NODEV was previously tested.
Add the new SB_I_NODEV s_iflag to proc, sysfs, and mqueuefs as those
filesystems should never support device nodes, and a simple superblock
flags makes that very hard to get wrong. With SB_I_NODEV set if any
device nodes somehow manage to show up on on a filesystem those
device nodes will be unopenable.
Acked-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
We always mixed in the parent pointer into the dentry name hash, but we
did it late at lookup time. It turns out that we can simplify that
lookup-time action by salting the hash with the parent pointer early
instead of late.
A few other users of our string hashes also wanted to mix in their own
pointers into the hash, and those are updated to use the same mechanism.
Hash users that don't have any particular initial salt can just use the
NULL pointer as a no-salt.
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: George Spelvin <linux@sciencehorizons.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs fixes from Al Viro:
"Fixes for crap of assorted ages: EOPENSTALE one is 4.2+, autofs one is
4.6, d_walk - 3.2+.
The atomic_open() and coredump ones are regressions from this window"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
coredump: fix dumping through pipes
fix a regression in atomic_open()
fix d_walk()/non-delayed __d_free() race
autofs braino fix for do_last()
fix EOPENSTALE bug in do_last()
open("/foo/no_such_file", O_RDONLY | O_CREAT) on should fail with
EACCES when /foo is not writable; failing with ENOENT is obviously
wrong. That got broken by a braino introduced when moving the
creat_error logics from atomic_open() to lookup_open(). Easy to
fix, fortunately.
Spotted-by: "Yan, Zheng" <ukernel@gmail.com>
Tested-by: "Yan, Zheng" <ukernel@gmail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The /dev/ptmx device node is changed to lookup the directory entry "pts"
in the same directory as the /dev/ptmx device node was opened in. If
there is a "pts" entry and that entry is a devpts filesystem /dev/ptmx
uses that filesystem. Otherwise the open of /dev/ptmx fails.
The DEVPTS_MULTIPLE_INSTANCES configuration option is removed, so that
userspace can now safely depend on each mount of devpts creating a new
instance of the filesystem.
Each mount of devpts is now a separate and equal filesystem.
Reserved ttys are now available to all instances of devpts where the
mounter is in the initial mount namespace.
A new vfs helper path_pts is introduced that finds a directory entry
named "pts" in the directory of the passed in path, and changes the
passed in path to point to it. The helper path_pts uses a function
path_parent_directory that was factored out of follow_dotdot.
In the implementation of devpts:
- devpts_mnt is killed as it is no longer meaningful if all mounts of
devpts are equal.
- pts_sb_from_inode is replaced by just inode->i_sb as all cached
inodes in the tty layer are now from the devpts filesystem.
- devpts_add_ref is rolled into the new function devpts_ptmx. And the
unnecessary inode hold is removed.
- devpts_del_ref is renamed devpts_release and reduced to just a
deacrivate_super.
- The newinstance mount option continues to be accepted but is now
ignored.
In devpts_fs.h definitions for when !CONFIG_UNIX98_PTYS are removed as
they are never used.
Documentation/filesystems/devices.txt is updated to describe the current
situation.
This has been verified to work properly on openwrt-15.05, centos5,
centos6, centos7, debian-6.0.2, debian-7.9, debian-8.2, ubuntu-14.04.3,
ubuntu-15.10, fedora23, magia-5, mint-17.3, opensuse-42.1,
slackware-14.1, gentoo-20151225 (13.0?), archlinux-2015-12-01. With the
caveat that on centos6 and on slackware-14.1 that there wind up being
two instances of the devpts filesystem mounted on /dev/pts, the lower
copy does not end up getting used.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Greg KH <greg@kroah.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Peter Anvin <hpa@zytor.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Serge Hallyn <serge.hallyn@ubuntu.com>
Cc: Willy Tarreau <w@1wt.eu>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
Cc: Jann Horn <jann@thejh.net>
Cc: Jiri Slaby <jslaby@suse.com>
Cc: Florian Weimer <fw@deneb.enyo.de>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's an analogue of commit 7500c38a (fix the braino in "namei:
massage lookup_slow() to be usable by lookup_one_len_unlocked()").
The same problem (->lookup()-returned unhashed negative dentry
just might be an autofs one with ->d_manage() that would wait
until the daemon makes it positive) applies in do_last() - we
need to do follow_managed() first.
Fortunately, remaining callers of follow_managed() are OK - only
autofs has that weirdness (negative dentry that does not mean
an instant -ENOENT)) and autofs never has its negative dentries
hashed, so we can't pick one from a dcache lookup.
->d_manage() is a bloody mess ;-/
Cc: stable@vger.kernel.org # v4.6
Spotted-by: Ian Kent <raven@themaw.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
EOPENSTALE occuring at the last component of a trailing symlink ends up
with do_last() retrying its lookup. After the symlink body has been
discarded. The thing is, all this retry_lookup logics in there is not
needed at all - the upper layers will do the right thing if we simply
return that -EOPENSTALE as we would with any other error. Trying to
microoptimize in do_last() is a lot of headache for no good reason.
Cc: stable@vger.kernel.org # v4.2+
Tested-by: Oleg Drokin <green@linuxhacker.ru>
Reviewed-and-Tested-by: Jeff Layton <jlayton@poochiereds.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The self-test was updated to cover zero-length strings; the function
needs to be updated, too.
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Fixes: fcfd2fbf22 ("fs/namei.c: Add hashlen_string() function")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The original name was simply hash_string(), but that conflicted with a
function with that name in drivers/base/power/trace.c, and I decided
that calling it "hashlen_" was better anyway.
But you have to do it in two places.
[ This caused build errors for architectures that don't define
CONFIG_DCACHE_WORD_ACCESS - Linus ]
Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Fixes: fcfd2fbf22 ("fs/namei.c: Add hashlen_string() function")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull string hash improvements from George Spelvin:
"This series does several related things:
- Makes the dcache hash (fs/namei.c) useful for general kernel use.
(Thanks to Bruce for noticing the zero-length corner case)
- Converts the string hashes in <linux/sunrpc/svcauth.h> to use the
above.
- Avoids 64-bit multiplies in hash_64() on 32-bit platforms. Two
32-bit multiplies will do well enough.
- Rids the world of the bad hash multipliers in hash_32.
This finishes the job started in commit 689de1d6ca ("Minimal
fix-up of bad hashing behavior of hash_64()")
The vast majority of Linux architectures have hardware support for
32x32-bit multiply and so derive no benefit from "simplified"
multipliers.
The few processors that do not (68000, h8/300 and some models of
Microblaze) have arch-specific implementations added. Those
patches are last in the series.
- Overhauls the dcache hash mixing.
The patch in commit 0fed3ac866 ("namei: Improve hash mixing if
CONFIG_DCACHE_WORD_ACCESS") was an off-the-cuff suggestion.
Replaced with a much more careful design that's simultaneously
faster and better. (My own invention, as there was noting suitable
in the literature I could find. Comments welcome!)
- Modify the hash_name() loop to skip the initial HASH_MIX(). This
would let us salt the hash if we ever wanted to.
- Sort out partial_name_hash().
The hash function is declared as using a long state, even though
it's truncated to 32 bits at the end and the extra internal state
contributes nothing to the result. And some callers do odd things:
- fs/hfs/string.c only allocates 32 bits of state
- fs/hfsplus/unicode.c uses it to hash 16-bit unicode symbols not bytes
- Modify bytemask_from_count to handle inputs of 1..sizeof(long)
rather than 0..sizeof(long)-1. This would simplify users other
than full_name_hash"
Special thanks to Bruce Fields for testing and finding bugs in v1. (I
learned some humbling lessons about "obviously correct" code.)
On the arch-specific front, the m68k assembly has been tested in a
standalone test harness, I've been in contact with the Microblaze
maintainers who mostly don't care, as the hardware multiplier is never
omitted in real-world applications, and I haven't heard anything from
the H8/300 world"
* 'hash' of git://ftp.sciencehorizons.net/linux:
h8300: Add <asm/hash.h>
microblaze: Add <asm/hash.h>
m68k: Add <asm/hash.h>
<linux/hash.h>: Add support for architecture-specific functions
fs/namei.c: Improve dcache hash function
Eliminate bad hash multipliers from hash_32() and hash_64()
Change hash_64() return value to 32 bits
<linux/sunrpc/svcauth.h>: Define hash_str() in terms of hashlen_string()
fs/namei.c: Add hashlen_string() function
Pull out string hash to <linux/stringhash.h>
This is just the infrastructure; there are no users yet.
This is modelled on CONFIG_ARCH_RANDOM; a CONFIG_ symbol declares
the existence of <asm/hash.h>.
That file may define its own versions of various functions, and define
HAVE_* symbols (no CONFIG_ prefix!) to suppress the generic ones.
Included is a self-test (in lib/test_hash.c) that verifies the basics.
It is NOT in general required that the arch-specific functions compute
the same thing as the generic, but if a HAVE_* symbol is defined with
the value 1, then equality is tested.
Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Greg Ungerer <gerg@linux-m68k.org>
Cc: Andreas Schwab <schwab@linux-m68k.org>
Cc: Philippe De Muyter <phdm@macq.eu>
Cc: linux-m68k@lists.linux-m68k.org
Cc: Alistair Francis <alistai@xilinx.com>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: uclinux-h8-devel@lists.sourceforge.jp
Patch 0fed3ac866 improved the hash mixing, but the function is slower
than necessary; there's a 7-instruction dependency chain (10 on x86)
each loop iteration.
Word-at-a-time access is a very tight loop (which is good, because
link_path_walk() is one of the hottest code paths in the entire kernel),
and the hash mixing function must not have a longer latency to avoid
slowing it down.
There do not appear to be any published fast hash functions that:
1) Operate on the input a word at a time, and
2) Don't need to know the length of the input beforehand, and
3) Have a single iterated mixing function, not needing conditional
branches or unrolling to distinguish different loop iterations.
One of the algorithms which comes closest is Yann Collet's xxHash, but
that's two dependent multiplies per word, which is too much.
The key insights in this design are:
1) Barring expensive ops like multiplies, to diffuse one input bit
across 64 bits of hash state takes at least log2(64) = 6 sequentially
dependent instructions. That is more cycles than we'd like.
2) An operation like "hash ^= hash << 13" requires a second temporary
register anyway, and on a 2-operand machine like x86, it's three
instructions.
3) A better use of a second register is to hold a two-word hash state.
With careful design, no temporaries are needed at all, so it doesn't
increase register pressure. And this gets rid of register copying
on 2-operand machines, so the code is smaller and faster.
4) Using two words of state weakens the requirement for one-round mixing;
we now have two rounds of mixing before cancellation is possible.
5) A two-word hash state also allows operations on both halves to be
done in parallel, so on a superscalar processor we get more mixing
in fewer cycles.
I ended up using a mixing function inspired by the ChaCha and Speck
round functions. It is 6 simple instructions and 3 cycles per iteration
(assuming multiply by 9 can be done by an "lea" instruction):
x ^= *input++;
y ^= x; x = ROL(x, K1);
x += y; y = ROL(y, K2);
y *= 9;
Not only is this reversible, two consecutive rounds are reversible:
if you are given the initial and final states, but not the intermediate
state, it is possible to compute both input words. This means that at
least 3 words of input are required to create a collision.
(It also has the property, used by hash_name() to avoid a branch, that
it hashes all-zero to all-zero.)
The rotate constants K1 and K2 were found by experiment. The search took
a sample of random initial states (I used 1023) and considered the effect
of flipping each of the 64 input bits on each of the 128 output bits two
rounds later. Each of the 8192 pairs can be considered a biased coin, and
adding up the Shannon entropy of all of them produces a score.
The best-scoring shifts also did well in other tests (flipping bits in y,
trying 3 or 4 rounds of mixing, flipping all 64*63/2 pairs of input bits),
so the choice was made with the additional constraint that the sum of the
shifts is odd and not too close to the word size.
The final state is then folded into a 32-bit hash value by a less carefully
optimized multiply-based scheme. This also has to be fast, as pathname
components tend to be short (the most common case is one iteration!), but
there's some room for latency, as there is a fair bit of intervening logic
before the hash value is used for anything.
(Performance verified with "bonnie++ -s 0 -n 1536:-2" on tmpfs. I need
a better benchmark; the numbers seem to show a slight dip in performance
between 4.6.0 and this patch, but they're too noisy to quote.)
Special thanks to Bruce fields for diligent testing which uncovered a
nasty fencepost error in an earlier version of this patch.
[checkpatch.pl formatting complaints noted and respectfully disagreed with.]
Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Tested-by: J. Bruce Fields <bfields@redhat.com>
We'd like to make more use of the highly-optimized dcache hash functions
throughout the kernel, rather than have every subsystem create its own,
and a function that hashes basic null-terminated strings is required
for that.
(The name is to emphasize that it returns both hash and length.)
It's actually useful in the dcache itself, specifically d_alloc_name().
Other uses in the next patch.
full_name_hash() is also tweaked to make it more generally useful:
1) Take a "char *" rather than "unsigned char *" argument, to
be consistent with hash_name().
2) Handle zero-length inputs. If we want more callers, we don't want
to make them worry about corner cases.
Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Changes in this update:
o fixes for mount line parsing, sparse warnings, read-only compat
feature remount behaviour
o allow fast path symlink lookups for inline symlinks.
o attribute listing cleanups
o writeback goes direct to bios rather than indirecting through
bufferheads
o transaction allocation cleanup
o optimised kmem_realloc
o added configurable error handling for metadata write errors,
changed default error handling behaviour from "retry forever" to
"retry until unmount then fail"
o fixed several inode cluster writeback lookup vs reclaim race
conditions
o fixed inode cluster writeback checking wrong inode after lookup
o fixed bugs where struct xfs_inode freeing wasn't actually RCU safe
o cleaned up inode reclaim tagging
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=0tYo
-----END PGP SIGNATURE-----
Merge tag 'xfs-for-linus-4.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs
Pull xfs updates from Dave Chinner:
"A pretty average collection of fixes, cleanups and improvements in
this request.
Summary:
- fixes for mount line parsing, sparse warnings, read-only compat
feature remount behaviour
- allow fast path symlink lookups for inline symlinks.
- attribute listing cleanups
- writeback goes direct to bios rather than indirecting through
bufferheads
- transaction allocation cleanup
- optimised kmem_realloc
- added configurable error handling for metadata write errors,
changed default error handling behaviour from "retry forever" to
"retry until unmount then fail"
- fixed several inode cluster writeback lookup vs reclaim race
conditions
- fixed inode cluster writeback checking wrong inode after lookup
- fixed bugs where struct xfs_inode freeing wasn't actually RCU safe
- cleaned up inode reclaim tagging"
* tag 'xfs-for-linus-4.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: (39 commits)
xfs: fix warning in xfs_finish_page_writeback for non-debug builds
xfs: move reclaim tagging functions
xfs: simplify inode reclaim tagging interfaces
xfs: rename variables in xfs_iflush_cluster for clarity
xfs: xfs_iflush_cluster has range issues
xfs: mark reclaimed inodes invalid earlier
xfs: xfs_inode_free() isn't RCU safe
xfs: optimise xfs_iext_destroy
xfs: skip stale inodes in xfs_iflush_cluster
xfs: fix inode validity check in xfs_iflush_cluster
xfs: xfs_iflush_cluster fails to abort on error
xfs: remove xfs_fs_evict_inode()
xfs: add "fail at unmount" error handling configuration
xfs: add configuration handlers for specific errors
xfs: add configuration of error failure speed
xfs: introduce table-based init for error behaviors
xfs: add configurable error support to metadata buffers
xfs: introduce metadata IO error class
xfs: configurable error behavior via sysfs
xfs: buffer ->bi_end_io function requires irq-safe lock
...
Pull security subsystem updates from James Morris:
"Highlights:
- A new LSM, "LoadPin", from Kees Cook is added, which allows forcing
of modules and firmware to be loaded from a specific device (this
is from ChromeOS, where the device as a whole is verified
cryptographically via dm-verity).
This is disabled by default but can be configured to be enabled by
default (don't do this if you don't know what you're doing).
- Keys: allow authentication data to be stored in an asymmetric key.
Lots of general fixes and updates.
- SELinux: add restrictions for loading of kernel modules via
finit_module(). Distinguish non-init user namespace capability
checks. Apply execstack check on thread stacks"
* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security: (48 commits)
LSM: LoadPin: provide enablement CONFIG
Yama: use atomic allocations when reporting
seccomp: Fix comment typo
ima: add support for creating files using the mknodat syscall
ima: fix ima_inode_post_setattr
vfs: forbid write access when reading a file into memory
fs: fix over-zealous use of "const"
selinux: apply execstack check on thread stacks
selinux: distinguish non-init user namespace capability checks
LSM: LoadPin for kernel file loading restrictions
fs: define a string representation of the kernel_read_file_id enumeration
Yama: consolidate error reporting
string_helpers: add kstrdup_quotable_file
string_helpers: add kstrdup_quotable_cmdline
string_helpers: add kstrdup_quotable
selinux: check ss_initialized before revalidating an inode label
selinux: delay inode label lookup as long as possible
selinux: don't revalidate an inode's label when explicitly setting it
selinux: Change bool variable name to index.
KEYS: Add KEYCTL_DH_COMPUTE command
...
Pull 'struct path' constification update from Al Viro:
"'struct path' is passed by reference to a bunch of Linux security
methods; in theory, there's nothing to stop them from modifying the
damn thing and LSM community being what it is, sooner or later some
enterprising soul is going to decide that it's a good idea.
Let's remove the temptation and constify all of those..."
* 'work.const-path' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
constify ima_d_path()
constify security_sb_pivotroot()
constify security_path_chroot()
constify security_path_{link,rename}
apparmor: remove useless checks for NULL ->mnt
constify security_path_{mkdir,mknod,symlink}
constify security_path_{unlink,rmdir}
apparmor: constify common_perm_...()
apparmor: constify aa_path_link()
apparmor: new helper - common_path_perm()
constify chmod_common/security_path_chmod
constify security_sb_mount()
constify chown_common/security_path_chown
tomoyo: constify assorted struct path *
apparmor_path_truncate(): path->mnt is never NULL
constify vfs_truncate()
constify security_path_truncate()
[apparmor] constify struct path * in a bunch of helpers
Pull parallel filesystem directory handling update from Al Viro.
This is the main parallel directory work by Al that makes the vfs layer
able to do lookup and readdir in parallel within a single directory.
That's a big change, since this used to be all protected by the
directory inode mutex.
The inode mutex is replaced by an rwsem, and serialization of lookups of
a single name is done by a "in-progress" dentry marker.
The series begins with xattr cleanups, and then ends with switching
filesystems over to actually doing the readdir in parallel (switching to
the "iterate_shared()" that only takes the read lock).
A more detailed explanation of the process from Al Viro:
"The xattr work starts with some acl fixes, then switches ->getxattr to
passing inode and dentry separately. This is the point where the
things start to get tricky - that got merged into the very beginning
of the -rc3-based #work.lookups, to allow untangling the
security_d_instantiate() mess. The xattr work itself proceeds to
switch a lot of filesystems to generic_...xattr(); no complications
there.
After that initial xattr work, the series then does the following:
- untangle security_d_instantiate()
- convert a bunch of open-coded lookup_one_len_unlocked() to calls of
that thing; one such place (in overlayfs) actually yields a trivial
conflict with overlayfs fixes later in the cycle - overlayfs ended
up switching to a variant of lookup_one_len_unlocked() sans the
permission checks. I would've dropped that commit (it gets
overridden on merge from #ovl-fixes in #for-next; proper resolution
is to use the variant in mainline fs/overlayfs/super.c), but I
didn't want to rebase the damn thing - it was fairly late in the
cycle...
- some filesystems had managed to depend on lookup/lookup exclusion
for *fs-internal* data structures in a way that would break if we
relaxed the VFS exclusion. Fixing hadn't been hard, fortunately.
- core of that series - parallel lookup machinery, replacing
->i_mutex with rwsem, making lookup_slow() take it only shared. At
that point lookups happen in parallel; lookups on the same name
wait for the in-progress one to be done with that dentry.
Surprisingly little code, at that - almost all of it is in
fs/dcache.c, with fs/namei.c changes limited to lookup_slow() -
making it use the new primitive and actually switching to locking
shared.
- parallel readdir stuff - first of all, we provide the exclusion on
per-struct file basis, same as we do for read() vs lseek() for
regular files. That takes care of most of the needed exclusion in
readdir/readdir; however, these guys are trickier than lookups, so
I went for switching them one-by-one. To do that, a new method
'->iterate_shared()' is added and filesystems are switched to it
as they are either confirmed to be OK with shared lock on directory
or fixed to be OK with that. I hope to kill the original method
come next cycle (almost all in-tree filesystems are switched
already), but it's still not quite finished.
- several filesystems get switched to parallel readdir. The
interesting part here is dealing with dcache preseeding by readdir;
that needs minor adjustment to be safe with directory locked only
shared.
Most of the filesystems doing that got switched to in those
commits. Important exception: NFS. Turns out that NFS folks, with
their, er, insistence on VFS getting the fuck out of the way of the
Smart Filesystem Code That Knows How And What To Lock(tm) have
grown the locking of their own. They had their own homegrown
rwsem, with lookup/readdir/atomic_open being *writers* (sillyunlink
is the reader there). Of course, with VFS getting the fuck out of
the way, as requested, the actual smarts of the smart filesystem
code etc. had become exposed...
- do_last/lookup_open/atomic_open cleanups. As the result, open()
without O_CREAT locks the directory only shared. Including the
->atomic_open() case. Backmerge from #for-linus in the middle of
that - atomic_open() fix got brought in.
- then comes NFS switch to saner (VFS-based ;-) locking, killing the
homegrown "lookup and readdir are writers" kinda-sorta rwsem. All
exclusion for sillyunlink/lookup is done by the parallel lookups
mechanism. Exclusion between sillyunlink and rmdir is a real rwsem
now - rmdir being the writer.
Result: NFS lookups/readdirs/O_CREAT-less opens happen in parallel
now.
- the rest of the series consists of switching a lot of filesystems
to parallel readdir; in a lot of cases ->llseek() gets simplified
as well. One backmerge in there (again, #for-linus - rockridge
fix)"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (74 commits)
ext4: switch to ->iterate_shared()
hfs: switch to ->iterate_shared()
hfsplus: switch to ->iterate_shared()
hostfs: switch to ->iterate_shared()
hpfs: switch to ->iterate_shared()
hpfs: handle allocation failures in hpfs_add_pos()
gfs2: switch to ->iterate_shared()
f2fs: switch to ->iterate_shared()
afs: switch to ->iterate_shared()
befs: switch to ->iterate_shared()
befs: constify stuff a bit
isofs: switch to ->iterate_shared()
get_acorn_filename(): deobfuscate a bit
btrfs: switch to ->iterate_shared()
logfs: no need to lock directory in lseek
switch ecryptfs to ->iterate_shared
9p: switch to ->iterate_shared()
fat: switch to ->iterate_shared()
romfs, squashfs: switch to ->iterate_shared()
more trivial ->iterate_shared conversions
...
Backmerge to resolve a conflict in ovl_lookup_real();
"ovl_lookup_real(): use lookup_one_len_unlocked()" instead,
but it was too late in the cycle to rebase.
The hash mixing between adding the next 64 bits of name
was just a bit weak.
Replaced with a still very fast but slightly more effective
mixing function.
Signed-off-by: George Spelvin <linux@horizon.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Overlayfs needs lookup without inode_permission() and already has the name
hash (in form of dentry->d_name on overlayfs dentry). It also doesn't
support filesystems with d_op->d_hash() so basically it only needs
the actual hashed lookup from lookup_one_len_unlocked()
So add a new helper that does unlocked lookup of a hashed name.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
If a file is renamed to a hardlink of itself POSIX specifies that rename(2)
should do nothing and return success.
This condition is checked in vfs_rename(). However it won't detect hard
links on overlayfs where these are given separate inodes on the overlayfs
layer.
Overlayfs itself detects this condition and returns success without doing
anything, but then vfs_rename() will proceed as if this was a successful
rename (detach_mounts(), d_move()).
The correct thing to do is to detect this condition before even calling
into overlayfs. This patch does this by calling vfs_select_inode() to get
the underlying inodes.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: <stable@vger.kernel.org> # v4.2+
It should never return positives; however, with Linux S&M crowd
involved, no bogosity is impossible. Results would be unpleasant...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
make it conditional on *opened & FILE_OPENED; in addition to getting
rid of exit_fput: thing, it simplifies atomic_open() cleanup on
may_open() failure.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Lift IS_DEADDIR handling up into the part common with atomic_open(),
remove it from the latter. Collapse permission checks into the
call of may_o_create(), getting it closer to atomic_open() case.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
do_last() and lookup_open() simpler that way and so does O_PATH
itself. As it bloody well should: we find what the pathname
resolves to, same way as in stat() et.al. and associate it with
FMODE_PATH struct file.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
ta-da!
The main issue is the lack of down_write_killable(), so the places
like readdir.c switched to plain inode_lock(); once killable
variants of rwsem primitives appear, that'll be dealt with.
lockdep side also might need more work
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If we *do* run into an in-lookup match, we need to wait for it to
cease being in-lookup. Fortunately, we do have unused space in
in-lookup dentries - d_lru is never looked at until it stops being
in-lookup.
So we can stash a pointer to wait_queue_head from stack frame of
the caller of ->lookup(). Some precautions are needed while
waiting, but it's not that hard - we do hold a reference to dentry
we are waiting for, so it can't go away. If it's found to be
in-lookup the wait_queue_head is still alive and will remain so
at least while ->d_lock is held. Moreover, the condition we
are waiting for becomes true at the same point where everything
on that wq gets woken up, so we can just add ourselves to the
queue once.
d_alloc_parallel() gets a pointer to wait_queue_head_t from its
caller; lookup_slow() adjusted, d_add_ci() taught to use
d_alloc_parallel() if the dentry passed to it happens to be
in-lookup one (i.e. if it's been called from the parallel lookup).
That's pretty much it - all that remains is to switch ->i_mutex
to rwsem and have lookup_slow() take it shared.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We will need to be able to check if there is an in-lookup
dentry with matching parent/name. Right now it's impossible,
but as soon as start locking directories shared such beasts
will appear.
Add a secondary hash for locating those. Hash chains go through
the same space where d_alias will be once it's not in-lookup anymore.
Search is done under the same bitlock we use for modifications -
with the primary hash we can rely on d_rehash() into the wrong
chain being the worst that could happen, but here the pointers are
buggered once it's removed from the chain. On the other hand,
the chains are not going to be long and normally we'll end up
adding to the chain anyway. That allows us to avoid bothering with
->d_lock when doing the comparisons - everything is stable until
removed from chain.
New helper: d_alloc_parallel(). Right now it allocates, verifies
that no hashed and in-lookup matches exist and adds to in-lookup
hash.
Returns ERR_PTR() for error, hashed match (in the unlikely case it's
been found) or new dentry. In-lookup matches trigger BUG() for
now; that will change in the next commit when we introduce waiting
for ongoing lookup to finish. Note that in-lookup matches won't be
possible until we actually go for shared locking.
lookup_slow() switched to use of d_alloc_parallel().
Again, these commits are separated only for making it easier to
review. All this machinery will start doing something useful only
when we go for shared locking; it's just that the combination is
too large for my taste.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
marked as such when (would be) parallel lookup is about to pass them
to actual ->lookup(); unmarked when
* __d_add() is about to make it hashed, positive or not.
* __d_move() (from d_splice_alias(), directly or via
__d_unalias()) puts a preexisting dentry in its place
* in caller of ->lookup() if it has escaped all of the
above. Bug (WARN_ON, actually) if it reaches the final dput()
or d_instantiate() while still marked such.
As the result, we are guaranteed that for as long as the flag is
set, dentry will
* remain negative unhashed with positive refcount
* never have its ->d_alias looked at
* never have its ->d_lru looked at
* never have its ->d_parent and ->d_name changed
Right now we have at most one such for any given parent directory.
With parallel lookups that restriction will weaken to
* only exist when parent is locked shared
* at most one with given (parent,name) pair (comparison of
names is according to ->d_compare())
* only exist when there's no hashed dentry with the same
(parent,name)
Transition will take the next several commits; unfortunately, we'll
only be able to switch to rwsem at the end of this series. The
reason for not making it a single patch is to simplify review.
New primitives: d_in_lookup() (a predicate checking if dentry is in
the in-lookup state) and d_lookup_done() (tells the system that
we are done with lookup and if it's still marked as in-lookup, it
should cease to be such).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Commit 3034a14 "ima: pass 'opened' flag to identify newly created files"
stopped identifying empty files as new files. However new empty files
can be created using the mknodat syscall. On systems with IMA-appraisal
enabled, these empty files are not labeled with security.ima extended
attributes properly, preventing them from subsequently being opened in
order to write the file data contents. This patch defines a new hook
named ima_post_path_mknod() to mark these empty files, created using
mknodat, as new in order to allow the file data contents to be written.
In addition, files with security.ima xattrs containing a file signature
are considered "immutable" and can not be modified. The file contents
need to be written, before signing the file. This patch relaxes this
requirement for new files, allowing the file signature to be written
before the file contents.
Changelog:
- defer identifying files with signatures stored as security.ima
(based on Dmitry Rozhkov's comments)
- removing tests (eg. dentry, dentry->d_inode, inode->i_size == 0)
(based on Al's review)
Signed-off-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Cc: Al Viro <<viro@zeniv.linux.org.uk>
Tested-by: Dmitry Rozhkov <dmitry.rozhkov@linux.intel.com>
* if we have a hashed negative dentry and either CREAT|EXCL on
r/o filesystem, or CREAT|TRUNC on r/o filesystem, or CREAT|EXCL
with failing may_o_create(), we should fail with EROFS or the
error may_o_create() has returned, but not ENOENT. Which is what
the current code ends up returning.
* if we have CREAT|TRUNC hitting a regular file on a read-only
filesystem, we can't fail with EROFS here. At the very least,
not until we'd done follow_managed() - we might have a writable
file (or a device, for that matter) bound on top of that one.
Moreover, the code downstream will see that O_TRUNC and attempt
to grab the write access (*after* following possible mount), so
if we really should fail with EROFS, it will happen. No need
to do that inside atomic_open().
The real logics is much simpler than what the current code is
trying to do - if we decided to go for simple lookup, ended
up with a negative dentry *and* had create_error set, fail with
create_error. No matter whether we'd got that negative dentry
from lookup_real() or had found it in dcache.
Cc: stable@vger.kernel.org # v3.6+
Acked-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Also drop the now unused readlink_copy export.
[dchinner: use d_inode(dentry) rather than dentry->d_inode]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
When get_acl() is called for an inode whose ACL is not cached yet, the
get_acl inode operation is called to fetch the ACL from the filesystem.
The inode operation is responsible for updating the cached acl with
set_cached_acl(). This is done without locking at the VFS level, so
another task can call set_cached_acl() or forget_cached_acl() before the
get_acl inode operation gets to calling set_cached_acl(), and then
get_acl's call to set_cached_acl() results in caching an outdate ACL.
Prevent this from happening by setting the cached ACL pointer to a
task-specific sentinel value before calling the get_acl inode operation.
Move the responsibility for updating the cached ACL from the get_acl
inode operations to get_acl(). There, only set the cached ACL if the
sentinel value hasn't changed.
The sentinel values are chosen to have odd values. Likewise, the value
of ACL_NOT_CACHED is odd. In contrast, ACL object pointers always have
an even value (ACLs are aligned in memory). This allows to distinguish
uncached ACLs values from ACL objects.
In addition, switch from guarding inode->i_acl and inode->i_default_acl
upates by the inode->i_lock spinlock to using xchg() and cmpxchg().
Filesystems that do not want ACLs returned from their get_acl inode
operations to be cached must call forget_cached_acl() to prevent the VFS
from doing so.
(Patch written by Al Viro and Andreas Gruenbacher.)
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We should try to trigger automount *before* bailing out on negative dentry.
Reported-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Reported-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Reported-by: Arend van Spriel <arend@broadcom.com>
Tested-by: Arend van Spriel <arend@broadcom.com>
Tested-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... and make mountpoint_last() use it. That makes all
candidates for lookup with parent locked shared go
through lookup_slow().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
No need to lock parent just because of ->d_revalidate() on child;
contrary to the stale comment, lookup_dcache() *can* be used without
locking the parent. Result can be moved as soon as we return, of
course, but the same is true for lookup_one_len_unlocked() itself.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Have lookup_fast() return 1 on success and 0 on "need to fall back";
lookup_slow() and follow_managed() return positive (1) on success.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... otherwise d_is_symlink() above might have nothing to do with
the inode value we've got.
Cc: stable@vger.kernel.org # v4.2+
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
both do_last() and walk_component() risk picking a NULL inode out
of dentry about to become positive, *then* checking its flags and
seeing that it's not negative anymore and using (already stale by
then) value they'd fetched earlier. Usually ends up oopsing soon
after that...
Cc: stable@vger.kernel.org # v3.13+
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... into returning a positive to path_openat(), which would interpret that
as "symlink had been encountered" and proceed to corrupt memory, etc.
It can only happen due to a bug in some ->open() instance or in some LSM
hook, etc., so we report any such event *and* make sure it doesn't trick
us into further unpleasantness.
Cc: stable@vger.kernel.org # v3.6+, at least
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
parallel to mutex_{lock,unlock,trylock,is_locked,lock_nested},
inode_foo(inode) being mutex_foo(&inode->i_mutex).
Please, use those for access to ->i_mutex; over the coming cycle
->i_mutex will become rwsem, with ->lookup() done with it held
only shared.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull misc vfs updates from Al Viro:
"All kinds of stuff. That probably should've been 5 or 6 separate
branches, but by the time I'd realized how large and mixed that bag
had become it had been too close to -final to play with rebasing.
Some fs/namei.c cleanups there, memdup_user_nul() introduction and
switching open-coded instances, burying long-dead code, whack-a-mole
of various kinds, several new helpers for ->llseek(), assorted
cleanups and fixes from various people, etc.
One piece probably deserves special mention - Neil's
lookup_one_len_unlocked(). Similar to lookup_one_len(), but gets
called without ->i_mutex and tries to avoid ever taking it. That, of
course, means that it's not useful for any directory modifications,
but things like getting inode attributes in nfds readdirplus are fine
with that. I really should've asked for moratorium on lookup-related
changes this cycle, but since I hadn't done that early enough... I
*am* asking for that for the coming cycle, though - I'm going to try
and get conversion of i_mutex to rwsem with ->lookup() done under lock
taken shared.
There will be a patch closer to the end of the window, along the lines
of the one Linus had posted last May - mechanical conversion of
->i_mutex accesses to inode_lock()/inode_unlock()/inode_trylock()/
inode_is_locked()/inode_lock_nested(). To quote Linus back then:
-----
| This is an automated patch using
|
| sed 's/mutex_lock(&\(.*\)->i_mutex)/inode_lock(\1)/'
| sed 's/mutex_unlock(&\(.*\)->i_mutex)/inode_unlock(\1)/'
| sed 's/mutex_lock_nested(&\(.*\)->i_mutex,[ ]*I_MUTEX_\([A-Z0-9_]*\))/inode_lock_nested(\1, I_MUTEX_\2)/'
| sed 's/mutex_is_locked(&\(.*\)->i_mutex)/inode_is_locked(\1)/'
| sed 's/mutex_trylock(&\(.*\)->i_mutex)/inode_trylock(\1)/'
|
| with a very few manual fixups
-----
I'm going to send that once the ->i_mutex-affecting stuff in -next
gets mostly merged (or when Linus says he's about to stop taking
merges)"
* 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (63 commits)
nfsd: don't hold i_mutex over userspace upcalls
fs:affs:Replace time_t with time64_t
fs/9p: use fscache mutex rather than spinlock
proc: add a reschedule point in proc_readfd_common()
logfs: constify logfs_block_ops structures
fcntl: allow to set O_DIRECT flag on pipe
fs: __generic_file_splice_read retry lookup on AOP_TRUNCATED_PAGE
fs: xattr: Use kvfree()
[s390] page_to_phys() always returns a multiple of PAGE_SIZE
nbd: use ->compat_ioctl()
fs: use block_device name vsprintf helper
lib/vsprintf: add %*pg format specifier
fs: use gendisk->disk_name where possible
poll: plug an unused argument to do_poll
amdkfd: don't open-code memdup_user()
cdrom: don't open-code memdup_user()
rsxx: don't open-code memdup_user()
mtip32xx: don't open-code memdup_user()
[um] mconsole: don't open-code memdup_user_nul()
[um] hostaudio: don't open-code memdup_user()
...
We need information about exports when crossing mountpoints during
lookup or NFSv4 readdir. If we don't already have that information
cached, we may have to ask (and wait for) rpc.mountd.
In both cases we currently hold the i_mutex on the parent of the
directory we're asking rpc.mountd about. We've seen situations where
rpc.mountd performs some operation on that directory that tries to take
the i_mutex again, resulting in deadlock.
With some care, we may be able to avoid that in rpc.mountd. But it
seems better just to avoid holding a mutex while waiting on userspace.
It appears that lookup_one_len is pretty much the only operation that
needs the i_mutex. So we could just drop the i_mutex elsewhere and do
something like
mutex_lock()
lookup_one_len()
mutex_unlock()
In many cases though the lookup would have been cached and not required
the i_mutex, so it's more efficient to create a lookup_one_len() variant
that only takes the i_mutex when necessary.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
new method: ->get_link(); replacement of ->follow_link(). The differences
are:
* inode and dentry are passed separately
* might be called both in RCU and non-RCU mode;
the former is indicated by passing it a NULL dentry.
* when called that way it isn't allowed to block
and should return ERR_PTR(-ECHILD) if it needs to be called
in non-RCU mode.
It's a flagday change - the old method is gone, all in-tree instances
converted. Conversion isn't hard; said that, so far very few instances
do not immediately bail out when called in RCU mode. That'll change
in the next commits.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
kmap() in page_follow_link_light() needed to go - allowing to hold
an arbitrary number of kmaps for long is a great way to deadlocking
the system.
new helper (inode_nohighmem(inode)) needs to be used for pagecache
symlinks inodes; done for all in-tree cases. page_follow_link_light()
instrumented to yell about anything missed.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
that allows to kill the recheck of nd->seq on the way out in
this case, and this check on the way out is left only for
absolute pathnames.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
we already zero it on outermost set_nameidata(), so initialization in
path_init() is pointless and wrong. The same DoS exists on pre-4.2
kernels, but there a slightly different fix will be needed.
Cc: stable@vger.kernel.org # v4.2
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Merge second patch-bomb from Andrew Morton:
- most of the rest of MM
- procfs
- lib/ updates
- printk updates
- bitops infrastructure tweaks
- checkpatch updates
- nilfs2 update
- signals
- various other misc bits: coredump, seqfile, kexec, pidns, zlib, ipc,
dma-debug, dma-mapping, ...
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (102 commits)
ipc,msg: drop dst nil validation in copy_msg
include/linux/zutil.h: fix usage example of zlib_adler32()
panic: release stale console lock to always get the logbuf printed out
dma-debug: check nents in dma_sync_sg*
dma-mapping: tidy up dma_parms default handling
pidns: fix set/getpriority and ioprio_set/get in PRIO_USER mode
kexec: use file name as the output message prefix
fs, seqfile: always allow oom killer
seq_file: reuse string_escape_str()
fs/seq_file: use seq_* helpers in seq_hex_dump()
coredump: change zap_threads() and zap_process() to use for_each_thread()
coredump: ensure all coredumping tasks have SIGNAL_GROUP_COREDUMP
signal: remove jffs2_garbage_collect_thread()->allow_signal(SIGCONT)
signal: introduce kernel_signal_stop() to fix jffs2_garbage_collect_thread()
signal: turn dequeue_signal_lock() into kernel_dequeue_signal()
signals: kill block_all_signals() and unblock_all_signals()
nilfs2: fix gcc uninitialized-variable warnings in powerpc build
nilfs2: fix gcc unused-but-set-variable warnings
MAINTAINERS: nilfs2: add header file for tracing
nilfs2: add tracepoints for analyzing reading and writing metadata files
...
Pull trivial updates from Jiri Kosina:
"Trivial stuff from trivial tree that can be trivially summed up as:
- treewide drop of spurious unlikely() before IS_ERR() from Viresh
Kumar
- cosmetic fixes (that don't really affect basic functionality of the
driver) for pktcdvd and bcache, from Julia Lawall and Petr Mladek
- various comment / printk fixes and updates all over the place"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial:
bcache: Really show state of work pending bit
hwmon: applesmc: fix comment typos
Kconfig: remove comment about scsi_wait_scan module
class_find_device: fix reference to argument "match"
debugfs: document that debugfs_remove*() accepts NULL and error values
net: Drop unlikely before IS_ERR(_OR_NULL)
mm: Drop unlikely before IS_ERR(_OR_NULL)
fs: Drop unlikely before IS_ERR(_OR_NULL)
drivers: net: Drop unlikely before IS_ERR(_OR_NULL)
drivers: misc: Drop unlikely before IS_ERR(_OR_NULL)
UBI: Update comments to reflect UBI_METAONLY flag
pktcdvd: drop null test before destroy functions
There are many places which use mapping_gfp_mask to restrict a more
generic gfp mask which would be used for allocations which are not
directly related to the page cache but they are performed in the same
context.
Let's introduce a helper function which makes the restriction explicit and
easier to track. This patch doesn't introduce any functional changes.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Michal Hocko <mhocko@suse.com>
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull userns hardlink capability check fix from Eric Biederman:
"This round just contains a single patch. There has been a lot of
other work this period but it is not quite ready yet, so I am pushing
it until 4.5.
The remaining change by Dirk Steinmetz wich fixes both Gentoo and
Ubuntu containers allows hardlinks if we have the appropriate
capabilities in the user namespace. Security wise it is really a
gimme as the user namespace root can already call setuid become that
user and create the hardlink"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
namei: permit linking with CAP_FOWNER in userns
Attempting to hardlink to an unsafe file (e.g. a setuid binary) from
within an unprivileged user namespace fails, even if CAP_FOWNER is held
within the namespace. This may cause various failures, such as a gentoo
installation within a lxc container failing to build and install specific
packages.
This change permits hardlinking of files owned by mapped uids, if
CAP_FOWNER is held for that namespace. Furthermore, it improves consistency
by using the existing inode_owner_or_capable(), which is aware of
namespaced capabilities as of 23adbe12ef ("fs,userns: Change
inode_capable to capable_wrt_inode_uidgid").
Signed-off-by: Dirk Steinmetz <public@rsjtdrjgfuzkfg.com>
This is hitting us in Ubuntu during some dpkg upgrades in containers.
When upgrading a file dpkg creates a hard link to the old file to back
it up before overwriting it. When packages upgrade suid files owned by a
non-root user the link isn't permitted, and the package upgrade fails.
This patch fixes our problem.
Tested-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Leandro Awa writes:
"After switching to version 4.1.6, our parallelized and distributed
workflows now fail consistently with errors of the form:
T34: ./regex.c:39:22: error: config.h: No such file or directory
From our 'git bisect' testing, the following commit appears to be the
possible cause of the behavior we've been seeing: commit 766c4cbfacd8"
Al Viro says:
"What happens is that 766c4cbfac got the things subtly wrong.
We used to treat d_is_negative() after lookup_fast() as "fall with
ENOENT". That was wrong - checking ->d_flags outside of ->d_seq
protection is unreliable and failing with hard error on what should've
fallen back to non-RCU pathname resolution is a bug.
Unfortunately, we'd pulled the test too far up and ran afoul of
another kind of staleness. The dentry might have been absolutely
stable from the RCU point of view (and we might be on UP, etc), but
stale from the remote fs point of view. If ->d_revalidate() returns
"it's actually stale", dentry gets thrown away and the original code
wouldn't even have looked at its ->d_flags.
What we need is to check ->d_flags where 766c4cbfac does (prior to
->d_seq validation) but only use the result in cases where we do not
discard this dentry outright"
Reported-by: Leandro Awa <lawa@nvidia.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=104911
Fixes: 766c4cbfac ("namei: d_is_negative() should be checked...")
Tested-by: Leandro Awa <lawa@nvidia.com>
Cc: stable@vger.kernel.org # v4.1+
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
IS_ERR(_OR_NULL) already contain an 'unlikely' compiler flag and there
is no need to do that again from its callers. Drop it.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Jeff Layton <jlayton@poochiereds.net>
Reviewed-by: David Howells <dhowells@redhat.com>
Reviewed-by: Steve French <smfrench@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Fix the following warnings:
Warning(.//fs/namei.c:2422): No description found for parameter 'nd'
Warning(.//fs/namei.c:2422): Excess function parameter 'nameidata'
description in 'path_mountpoint'
Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In rare cases a directory can be renamed out from under a bind mount.
In those cases without special handling it becomes possible to walk up
the directory tree to the root dentry of the filesystem and down
from the root dentry to every other file or directory on the filesystem.
Like division by zero .. from an unconnected path can not be given
a useful semantic as there is no predicting at which path component
the code will realize it is unconnected. We certainly can not match
the current behavior as the current behavior is a security hole.
Therefore when encounting .. when following an unconnected path
return -ENOENT.
- Add a function path_connected to verify path->dentry is reachable
from path->mnt.mnt_root. AKA to validate that rename did not do
something nasty to the bind mount.
To avoid races path_connected must be called after following a path
component to it's next path component.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Now that we can get there in RCU mode, we shouldn't play with
nd->path.dentry->d_inode - it's not guaranteed to be stable.
Use nd->inode instead.
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
In RCU mode we might end up with dentry evicted just we check
that it's a directory. In such case we should return ECHILD
rather than ENOTDIR, so that pathwalk would be retries in non-RCU
mode.
Breakage had been introduced in commit b18825a - prior to that
we were looking at nd->inode, which had been fetched before
verifying that ->d_seq was still valid. That form of check
would only be satisfied if at some point the pathname prefix
would indeed have resolved to a non-directory. The fix consists
of checking ->d_seq after we'd run into a non-directory dentry,
and failing with ECHILD in case of mismatch.
Note that all branches since 3.12 have that problem...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The only caller that cares about its return value can just
as easily pick it from nd->root_seq itself. We used to just
calculate it and return to caller, but these days we are
storing it in nd->root_seq in all cases.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
these guys are always declared next to each other; might as well put
the former (pointer to previous instance) into the latter and simplify
the calling conventions for {set,restore}_nameidata()
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
a) make it reject ERR_PTR() for name
b) make it putname(name) on all other failure exits
c) make it return name on success
again, simplifies the callers
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
a) make it reject ERR_PTR() for name
b) make it putname(name) upon return in all other cases.
seriously simplifies the callers...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Otherwise we are risking a hard error where nonlazy restart would be the right
thing to do; it's a very narrow race with mount --move and most of the time it
ends up being completely harmless, but it's possible to construct a case when
we'll get a bogus hard error instead of falling back to non-lazy walk...
For one thing, when crossing _into_ overmount of parent we need to check for
mount_lock bumps when we get NULL from __lookup_mnt() as well.
For another, and less exotically, we need to make sure that the data fetched
in follow_up_rcu() had been consistent. ->mnt_mountpoint is pinned for as
long as it is a mountpoint, but we need to check mount_lock after fetching
to verify that.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
touch_atime is not RCU-safe, and so cannot be called on an RCU walk.
However, in situations where RCU-walk makes a difference, the symlink
will likely to accessed much more often than it is useful to update
the atime.
So split out the test of "Does the atime actually need to be updated"
into atime_needs_update(), and have get_link() unlazy if it finds that
it will need to do that update.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We are almost done - primitives for leaving RCU mode are aware of nd->stack
now, a new primitive for going to non-RCU mode when we have a symlink on hands
added.
The thing we are heavily relying upon is that *any* unlazy failure will be
shortly followed by terminate_walk(), with no access to nameidata in between.
So it's enough to leave the things in a state terminate_walk() would cope with.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We *can't* call that audit garbage in RCU mode - it's doing a weird
mix of allocations (GFP_NOFS, immediately followed by GFP_KERNEL)
and I'm not touching that... thing again.
So if this security sclero^Whardening feature gets triggered when
we are in RCU mode, tough - we'll fail with -ECHILD and have
everything restarted in non-RCU mode. Only to hit the same test
and fail, this time with EACCES and with (oh, rapture) an audit spew
produced.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
very simple - just make path_put() conditional on !RCU.
Note that right now it doesn't get called in RCU mode -
we leave it before getting anything into stack.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
inode_follow_link now takes an inode and rcu flag as well as the
dentry.
inode is used in preference to d_backing_inode(dentry), particularly
in RCU-walk mode.
selinux_inode_follow_link() gets dentry_has_perm() and
inode_has_perm() open-coded into it so that it can call
avc_has_perm_flags() in way that is safe if LOOKUP_RCU is set.
Calling avc_has_perm_flags() with rcu_read_lock() held means
that when avc_has_perm_noaudit calls avc_compute_av(), the attempt
to rcu_read_unlock() before calling security_compute_av() will not
actually drop the RCU read-lock.
However as security_compute_av() is completely in a read_lock()ed
region, it should be safe with the RCU read-lock held.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Make use of d_backing_inode() in pathwalk to gain access to an
inode or dentry that's on a lower layer.
Signed-off-by: David Howells <dhowells@redhat.com>
Lift it from link_path_walk(), trailing_symlink(), lookup_last(),
mountpoint_last(), complete_walk() and do_last(). A _lot_ of
those suckers merge.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Make trailing_symlink() return the pathname to traverse or ERR_PTR(-E...).
A subtle point is that for "magic" symlinks it returns "" now - that
leads to link_path_walk("", nd), which is immediately returning 0 and
we are back to the treatment of the last component, at whereever the
damn thing has left us.
Reduces the stack footprint - link_path_walk() called on more shallow
stack now.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* lift link_path_walk() into callers; moving it down into path_init()
had been a mistake. Stack footprint, among other things...
* do _not_ call path_cleanup() after path_init() failure; on all failure
exits out of it we have nothing for path_cleanup() to do
* have path_init() return pathname or ERR_PTR(-E...)
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
we can do fdput() under rcu_read_lock() just fine; all we need to take
care of is fetching nd->inode value first.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Makes the situation much more regular - we avoid a strange state
when the element just after the top of stack is used to store
struct path of symlink, but isn't counted in nd->depth. This
is much more regular, so the normal failure exits, etc., work
fine.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Just store it in nd->stack[nd->depth].link right in pick_link().
Now that we make sure of stack expansion in pick_link(), we can
do so...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... and don't open-code unlazy_walk() in there - the only reason
for that is to avoid verfication of cached nd->root, which is
trivially avoided by discarding said cached nd->root first.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
rather than letting the callers handle the jump-to-root part of
semantics, do it right in get_link() and return the rest of the
body for the caller to deal with - at that point it's treated
the same way as relative symlinks would be. And return NULL
when there's no "rest of the body" - those are treated the same
as pure jump symlink would be.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Instead of saving name and branching to OK:, where we'll immediately restore
it, and call walk_component() with WALK_PUT|WALK_GET and nd->last_type being
LAST_BIND, which is equivalent to put_link(nd), err = 0, we can just treat
that the same way we'd treat procfs-style "jump" symlinks - do put_link(nd)
and move on.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
when cookie is NULL, put_link() is equivalent to path_put(), so
as soon as we'd set last->cookie to NULL, we can bump nd->depth and
let the normal logics in terminate_walk() to take care of cleanups.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
now that it gets nameidata, no reason to have setting LOOKUP_JUMPED on
mountpoint crossing and calling path_put_conditional() on failures
done in every caller.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
task_struct currently contains two ad-hoc members for use by the VFS:
link_count and total_link_count. These are only interesting to fs/namei.c,
so exposing them explicitly is poor layering. Incidentally, link_count
isn't used anymore, so it can just die.
This patches replaces those with a single pointer to 'struct nameidata'.
This structure represents the current filename lookup of which
there can only be one per process, and is a natural place to
store total_link_count.
This will allow the current "nameidata" argument to all
follow_link operations to be removed as current->nameidata
can be used instead in the _very_ few instances that care about
it at all.
As there are occasional circumstances where pathname lookup can
recurse, such as through kern_path_locked, we always save and old
current->nameidata (if there is one) when setting a new value, and
make sure any active link_counts are preserved.
follow_mount and follow_automount now get a 'struct nameidata *'
rather than 'int flags' so that they can directly access
total_link_count, rather than going through 'current'.
Suggested-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
instead of a single flag (!= 0 => we want to follow symlinks) pass
two bits - WALK_GET (want to follow symlinks) and WALK_PUT (put_link()
once we are done looking at the name). The latter matters only for
success exits - on failure the caller will discard everything anyway.
Suggestions for better variant are welcome; what this thing aims for
is making sure that pending put_link() is done *before* walk_component()
decides to pick a symlink up, rather than between picking it up and
acting upon it. See the next commit for payoff.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
All callers of terminate_walk() are followed by more or less
open-coded eqiuvalent of "do put_link() on everything left
in nd->stack". Better done in terminate_walk() itself, and
when we go for RCU symlink traversal we'll have to do it
there anyway.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
rationale: we'll need to have terminate_walk() do put_link() on
everything, which will mean that in some cases ..._last() will do
put_link() anyway. Easier to have them do it in all cases.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
follow_dotdot_rcu() does an equivalent of terminate_walk() on failure;
shifting it into callers makes for simpler rules and those callers
already have terminate_walk() on other failure exits.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The only reason why we needed one more was that purely nested
MAXSYMLINKS symlinks could lead to path_init() using that many
entries in addition to nd->stack[0] which it left unused.
That can't happen now - path_init() starts with entry 0 (and
trailing_symlink() is called only when we'd already encountered
one symlink, so no more than MAXSYMLINKS-1 are left).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
get rid of orig_depth - we only use it on error exit to tell whether
to stop doing put_link() when depth reaches 0 (call from path_init())
or when it reaches 1 (call from trailing_symlink()). However, in
the latter case the caller would immediately follow with one more
put_link(). Just keep doing it until the depth reaches zero (and
simplify trailing_symlink() as the result).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Get rid of orig_depth checks in OK: logics. If nd->depth is
zero, we had been called from path_init() and we are done.
If it is greater than 1, we are not done, whether we'd been
called from path_init() or trailing_symlink(). And in
case when it's 1, we might have been called from path_init()
and reached the end of nested symlink (in which case
nd->stack[0].name will point to the rest of pathname and
we are not done) or from trailing_symlink(), in which case
we are done.
Just have trailing_symlink() leave NULL in nd->stack[0].name
and use that to discriminate between those cases.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Make link_path_walk() work with any value of nd->depth on entry -
memorize it and use it in tests instead of comparing with 1.
Don't bother with increment/decrement in path_init().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
move increment of ->depth to the point where we'd discovered
that get_link() has not returned an error, adjust exits
accordingly.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
nd->stack[0] is unused until the handling of trailing symlinks and
we want to get rid of that. Having fucked that transformation up
several times, I went for bloody pedantic series of provably equivalent
transformations. Sorry.
Step 1: keep nd->depth higher by one in link_path_walk() - increment upon
entry, decrement on exits, adjust the arithmetics inside and surround the
calls of functions that care about nd->depth value (nd_alloc_stack(),
get_link(), put_link()) with decrement/increment pairs.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The only restriction is that on the total amount of symlinks
crossed; how they are nested does not matter
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Array of MAX_NESTED_LINKS + 1 elements put into nameidata;
what used to be a local array in link_path_walk() occupies
entries 1 .. MAX_NESTED_LINKS in it, link and cookie from
the trailing symlink handling loops - entry 0.
This is _not_ the final arrangement; just an easily verified
incremental step.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Deal with skipping leading slashes before what used to be the
recursive call. That way we can get rid of that goto completely.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
absolutely straightforward now - the only variables we need to preserve
across the recursive call are name, link and cookie, and recursion depth
is limited (and can is equal to nd->depth). So arrange an array of
triples to hold instances of those and be done with that.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
reduce the number of returns in there - turn all places
where it returns zero into goto OK and places where it
returns non-zero into goto Err. The only non-trivial
detail is that all breaks in the loop are guaranteed
to be with non-zero err.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
What we do after the second walk_component() + put_link() + depth
decrement in there is exactly equivalent to what's done right
after the first walk_component(). Easy to verify and not at all
surprising, seeing that there we have just walked the last
component of nested symlink.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull the block after the if-else in the end of what used to be do-while
body into all branches there. We are almost done with the massage...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If we get ERR_PTR() from get_link(), we are guaranteed to get err != 0
when we break out of do-while, so we are going to hit if (err) return err;
shortly after it. Pull that into the if (IS_ERR(s)) body.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
... and strip __always_inline from follow_link() - remaining callers
don't need that.
Now link_path_walk() recursion is a direct one.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
shares space with nameidata->next, walk_component() et.al. store
the struct path of symlink instead of returning it into a variable
passed by caller.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Split a piece of fs/namei.c:follow_link() that does obtaining the link
body into a separate function. follow_link() itself is converted to
calling get_link() and then doing the body traversal (if any).
The next step will expand follow_link() call in link_path_walk()
and this helps to keep the size down...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
a) instead of storing the symlink body (via nd_set_link()) and returning
an opaque pointer later passed to ->put_link(), ->follow_link() _stores_
that opaque pointer (into void * passed by address by caller) and returns
the symlink body. Returning ERR_PTR() on error, NULL on jump (procfs magic
symlinks) and pointer to symlink body for normal symlinks. Stored pointer
is ignored in all cases except the last one.
Storing NULL for opaque pointer (or not storing it at all) means no call
of ->put_link().
b) the body used to be passed to ->put_link() implicitly (via nameidata).
Now only the opaque pointer is. In the cases when we used the symlink body
to free stuff, ->follow_link() now should store it as opaque pointer in addition
to returning it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
when we go for on-demand allocation of saved state in
link_path_walk(), we'll want nameidata to stay around
for all 3 calls of path_mountpoint().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
that avoids having nameidata on stack during the calls of
->rmdir()/->unlink() and *two* of those during the calls
of ->rename().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
it's a convenient helper, but we'll want to shift nameidata
down the call chain, so it won't be available there...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
With LOOKUP_FOLLOW we unlazy and return 1; without it we either
fail with ELOOP or, for O_PATH opens, succeed. No need to mix
those cases...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
When O_PATH is present, O_CREAT isn't, so symlink_ok is always equal to
(open_flags & O_PATH) && !(nd->flags & LOOKUP_FOLLOW).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
No ->inode_follow_link() methods use the nameidata arg, and
it is about to become private to namei.c.
So remove from all inode_follow_link() functions.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
path_openat() jumps to the wrong place after do_tmpfile() - it has
already done path_cleanup() (as part of path_lookupat() called by
do_tmpfile()), so doing that again can lead to double fput().
Cc: stable@vger.kernel.org # v3.11+
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Fetching ->d_inode, verifying ->d_seq and finding d_is_negative() to
be true does *not* mean that inode we'd fetched had been NULL - that
holds only while ->d_seq is still unchanged.
Shift d_is_negative() checks into lookup_fast() prior to ->d_seq
verification.
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Tested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Calling unlazy_walk() in walk_component() and do_last() when we find
a symlink that needs to be followed doesn't acquire a reference to vfsmount.
That's fine when the symlink is on the same vfsmount as the parent directory
(which is almost always the case), but it's not always true - one _can_
manage to bind a symlink on top of something. And in such cases we end up
with excessive mntput().
Cc: stable@vger.kernel.org # since 2.6.39
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Make pathwalk use d_is_reg() rather than S_ISREG() to determine whether to
honour O_TRUNC. Since this occurs after complete_walk(), the dentry type
field cannot change and the inode pointer cannot change as we hold a ref on
the dentry, so this should be safe.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Where we have:
if (!dentry->d_inode || d_is_negative(dentry)) {
type constructions in pathwalk we should be able to eliminate the check of
d_inode and rely solely on the result of d_is_negative() or d_is_positive().
What we do have to take care to do is to read d_inode after calling a
d_is_xxx() typecheck function to get the barriering right.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Miklos Szeredi