Jann Horn points out that the vmacache_flush_all() function is not only
potentially expensive, it's buggy too. It also happens to be entirely
unnecessary, because the sequence number overflow case can be avoided by
simply making the sequence number be 64-bit. That doesn't even grow the
data structures in question, because the other adjacent fields are
already 64-bit.
So simplify the whole thing by just making the sequence number overflow
case go away entirely, which gets rid of all the complications and makes
the code faster too. Win-win.
[ Oleg Nesterov points out that the VMACACHE_FULL_FLUSHES statistics
also just goes away entirely with this ]
Reported-by: Jann Horn <jannh@google.com> Suggested-by: Will Deacon <will.deacon@arm.com> Acked-by: Davidlohr Bueso <dave@stgolabs.net> Cc: Oleg Nesterov <oleg@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: drop changes to mm debug code] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We will want to filter away MSR_IA32_SMBASE from the emulated_msrs if
the host CPU does not support SMM virtualization. Introduce the
logic to do that, and also move paravirt MSRs to emulated_msrs for
simplicity and to get rid of KVM_SAVE_MSRS_BEGIN.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
CPUID Fn8000_0007_EDX[CPB] is wrongly 0 on models up to B1. But they do
support CPB (AMD's Core Performance Boosting cpufreq CPU feature), so fix that.
Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sherry Hurwitz <sherry.hurwitz@amd.com> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20170907170821.16021-1-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: David Woodhouse <dwmw@amazon.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16:
- Change the added case into an if-statement
- s/x86_stepping/x86_mask/] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The debug control MSR is "highly magical" as the blockstep bit can be
cleared by hardware under not well documented circumstances.
So a task switch relying on the bit set by the previous task (according to
the previous tasks thread flags) can trip over this and not update the flag
for the next task.
To fix this its required to handle DEBUGCTLMSR_BTF when either the previous
or the next or both tasks have the TIF_BLOCKSTEP flag set.
While at it avoid branching within the TIF_BLOCKSTEP case and evaluating
boot_cpu_data twice in kernels without CONFIG_X86_DEBUGCTLMSR.
x86_64: arch/x86/kernel/process.o
text data bss dec hex
3024 8577 16 11617 2d61 Before
3008 8577 16 11601 2d51 After
i386: No change
[ tglx: Made the shift value explicit, use a local variable to make the
code readable and massaged changelog]
Originally-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Kyle Huey <khuey@kylehuey.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Andy Lutomirski <luto@kernel.org> Link: http://lkml.kernel.org/r/20170214081104.9244-3-khuey@kylehuey.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: David Woodhouse <dwmw@amazon.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Help the compiler to avoid reevaluating the thread flags for each checked
bit by reordering the bit checks and providing an explicit xor for
evaluation.
With default defconfigs for each arch,
x86_64: arch/x86/kernel/process.o
text data bss dec hex
3056 8577 16 11649 2d81 Before
3024 8577 16 11617 2d61 After
i386: arch/x86/kernel/process.o
text data bss dec hex
2957 8673 8 11638 2d76 Before
2925 8673 8 11606 2d56 After
Originally-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Kyle Huey <khuey@kylehuey.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Andy Lutomirski <luto@kernel.org> Link: http://lkml.kernel.org/r/20170214081104.9244-2-khuey@kylehuey.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[bwh: Backported to 3.16:
- We don't do refresh_tr_limit() here
- Use ACCESS_ONCE() instead of READ_ONCE()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This adds the new "seccomp" syscall with both an "operation" and "flags"
parameter for future expansion. The third argument is a pointer value,
used with the SECCOMP_SET_MODE_FILTER operation. Currently, flags must
be 0. This is functionally equivalent to prctl(PR_SET_SECCOMP, ...).
In addition to the TSYNC flag later in this patch series, there is a
non-zero chance that this syscall could be used for configuring a fixed
argument area for seccomp-tracer-aware processes to pass syscall arguments
in the future. Hence, the use of "seccomp" not simply "seccomp_add_filter"
for this syscall. Additionally, this syscall uses operation, flags,
and user pointer for arguments because strictly passing arguments via
a user pointer would mean seccomp itself would be unable to trivially
filter the seccomp syscall itself.
Signed-off-by: Kees Cook <keescook@chromium.org> Reviewed-by: Oleg Nesterov <oleg@redhat.com> Reviewed-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
In preparation for having other callers of the seccomp mode setting
logic, split the prctl entry point away from the core logic that performs
seccomp mode setting.
Signed-off-by: Kees Cook <keescook@chromium.org> Reviewed-by: Oleg Nesterov <oleg@redhat.com> Reviewed-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
xfs_attr3_leaf_create may have errored out before instantiating a buffer,
for example if the blkno is out of range. In that case there is no work
to do to remove it, and in fact xfs_da_shrink_inode will lead to an oops
if we try.
This also seems to fix a flaw where the original error from
xfs_attr3_leaf_create gets overwritten in the cleanup case, and it
removes a pointless assignment to bp which isn't used after this.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=199969 Reported-by: Xu, Wen <wen.xu@gatech.edu> Tested-by: Xu, Wen <wen.xu@gatech.edu> Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[bwh: Backported to 3.16: adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
A recent fuzzed filesystem image cached random dcache corruption
when the reproducer was run. This often showed up as panics in
lookup_slow() on a null inode->i_ops pointer when doing pathwalks.
but had many different failure modes including deadlocks trying to
lock the inode that was just allocated or KASAN reports of
use-after-free violations.
The cause of the problem was a corrupt INOBT on a v4 fs where the
root inode was marked as free in the inobt record. Hence when we
allocated an inode, it chose the root inode to allocate, found it in
the cache and re-initialised it.
We recently fixed a similar inode allocation issue caused by inobt
record corruption problem in xfs_iget_cache_miss() in commit ee457001ed6c ("xfs: catch inode allocation state mismatch
corruption"). This change adds similar checks to the cache-hit path
to catch it, and turns the reproducer into a corruption shutdown
situation.
Reported-by: Wen Xu <wen.xu@gatech.edu> Signed-Off-By: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
[darrick: fix typos in comment] Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[bwh: Backported to 3.16:
- Look up mode in XFS inode, not VFS inode
- Use positive error codes] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We recently came across a V4 filesystem causing memory corruption
due to a newly allocated inode being setup twice and being added to
the superblock inode list twice. From code inspection, the only way
this could happen is if a newly allocated inode was not marked as
free on disk (i.e. di_mode wasn't zero).
Running the metadump on an upstream debug kernel fails during inode
allocation like so:
Confirms that it is not a free inode on disk. xfs_repair
also trips over this inode:
.....
zero length extent (off = 0, fsbno = 0) in ino 184452204
correcting nextents for inode 184452204
bad attribute fork in inode 184452204, would clear attr fork
bad nblocks 1 for inode 184452204, would reset to 0
bad anextents 1 for inode 184452204, would reset to 0
imap claims in-use inode 184452204 is free, would correct imap
would have cleared inode 184452204
.....
disconnected inode 184452204, would move to lost+found
And so we have a situation where the directory structure and the
inobt thinks the inode is free, but the inode on disk thinks it is
still in use. Where this corruption came from is not possible to
diagnose, but we can detect it and prevent the kernel from oopsing
on lookup. The reproducer now results in:
[ 54.843517] XFS (loop0): Corruption detected! Free inode 0xafe846c not=
marked free on disk
[ 54.845885] XFS (loop0): Internal error xfs_trans_cancel at line 1023 =
of file fs/xfs/xfs_trans.c. Caller xfs_create+0x425/0x670
[ 54.848994] CPU: 10 PID: 3541 Comm: mkdir Not tainted 4.16.0-rc5-dgc #=
443
[ 54.850753] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIO=
S 1.10.2-1 04/01/2014
[ 54.852859] Call Trace:
[ 54.853531] dump_stack+0x85/0xc5
[ 54.854385] xfs_trans_cancel+0x197/0x1c0
[ 54.855421] xfs_create+0x425/0x670
[ 54.856314] xfs_generic_create+0x1f7/0x2c0
[ 54.857390] ? capable_wrt_inode_uidgid+0x3f/0x50
[ 54.858586] vfs_mkdir+0xfb/0x1b0
[ 54.859458] SyS_mkdir+0xcf/0xf0
[ 54.860254] do_syscall_64+0x73/0x1a0
[ 54.861193] entry_SYSCALL_64_after_hwframe+0x42/0xb7
[ 54.862492] RIP: 0033:0x7fb73bddf547
[ 54.863358] RSP: 002b:00007ffdaa553338 EFLAGS: 00000246 ORIG_RAX: 0000= 000000000053
[ 54.865133] RAX: ffffffffffffffda RBX: 00007ffdaa55449a RCX: 00007fb73= bddf547
[ 54.866766] RDX: 0000000000000001 RSI: 00000000000001ff RDI: 00007ffda= a55449a
[ 54.868432] RBP: 00007ffdaa55449a R08: 00000000000001ff R09: 00005623a= 8670dd0
[ 54.870110] R10: 00007fb73be72d5b R11: 0000000000000246 R12: 000000000= 00001ff
[ 54.871752] R13: 00007ffdaa5534b0 R14: 0000000000000000 R15: 00007ffda= a553500
[ 54.873429] XFS (loop0): xfs_do_force_shutdown(0x8) called from line 1=
024 of file fs/xfs/xfs_trans.c. Return address = ffffffff814cd050
[ 54.882790] XFS (loop0): Corruption of in-memory data detected. Shutt=
ing down filesystem
[ 54.884597] XFS (loop0): Please umount the filesystem and rectify the =
problem(s)
Note that this crash is only possible on v4 filesystemsi or v5
filesystems mounted with the ikeep mount option. For all other V5
filesystems, this problem cannot occur because we don't read inodes
we are allocating from disk - we simply overwrite them with the new
inode information.
Signed-Off-By: Dave Chinner <dchinner@redhat.com> Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com> Tested-by: Carlos Maiolino <cmaiolino@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[bwh: Backported to 3.16:
- Look up mode in XFS inode, not VFS inode
- Use positive error codes] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
An HFS+ filesystem can be mounted read-only without having a metadata
directory, which is needed to support hardlinks. But if the catalog
data is corrupted, a directory lookup may still find dentries claiming
to be hardlinks.
hfsplus_lookup() does check that ->hidden_dir is not NULL in such a
situation, but mistakenly does so after dereferencing it for the first
time. Reorder this check to prevent a crash.
This happens when looking up corrupted catalog data (dentry) on a
filesystem with no metadata directory (this could only ever happen on a
read-only mount). Wen Xu sent the replication steps in detail to the
fsdevel list: https://bugzilla.kernel.org/show_bug.cgi?id=200297
Link: http://lkml.kernel.org/r/20180712215344.q44dyrhymm4ajkao@eaf Signed-off-by: Ernesto A. Fernández <ernesto.mnd.fernandez@gmail.com> Reported-by: Wen Xu <wen.xu@gatech.edu> Cc: Viacheslav Dubeyko <slava@dubeyko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Invalid reloc tree can cause kernel NULL pointer dereference when btrfs
does some cleanup of the reloc roots.
It turns out that fs_info::reloc_ctl can be NULL in
btrfs_recover_relocation() as we allocate relocation control after all
reloc roots have been verified.
So when we hit: note, we haven't called set_reloc_control() thus
fs_info::reloc_ctl is still NULL.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=199833 Reported-by: Xu Wen <wen.xu@gatech.edu> Signed-off-by: Qu Wenruo <wqu@suse.com> Tested-by: Gu Jinxiang <gujx@cn.fujitsu.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
on some architecture spin_is_locked() always return false in
uniprocessor configuration and therefore it would be advise
to replace with lockdep_assert_held().
Signed-off-by: Sanjeev Sharma <Sanjeev_Sharma@mentor.com> Acked-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Like d88b6d04: "cdrom: information leak in cdrom_ioctl_media_changed()"
There is another cast from unsigned long to int which causes
a bounds check to fail with specially crafted input. The value is
then used as an index in the slot array in cdrom_slot_status().
Signed-off-by: Scott Bauer <scott.bauer@intel.com> Signed-off-by: Scott Bauer <sbauer@plzdonthack.me> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Nadav reported that on guests we're failing to rewrite the indirect
calls to CALLEE_SAVE paravirt functions. In particular the
pv_queued_spin_unlock() call is left unpatched and that is all over the
place. This obviously wrecks Spectre-v2 mitigation (for paravirt
guests) which relies on not actually having indirect calls around.
The reason is an incorrect clobber test in paravirt_patch_call(); this
function rewrites an indirect call with a direct call to the _SAME_
function, there is no possible way the clobbers can be different
because of this.
Therefore remove this clobber check. Also put WARNs on the other patch
failure case (not enough room for the instruction) which I've not seen
trigger in my (limited) testing.
Three live kernel image disassemblies for lock_sock_nested (as a small
function that illustrates the problem nicely). PRE is the current
situation for guests, POST is with this patch applied and NATIVE is with
or without the patch for !guests.
The article "Spectre Returns! Speculation Attacks using the Return Stack
Buffer" [1] describes two new (sub-)variants of spectrev2-like attacks,
making use solely of the RSB contents even on CPUs that don't fallback to
BTB on RSB underflow (Skylake+).
Mitigate userspace-userspace attacks by always unconditionally filling RSB on
context switch when the generic spectrev2 mitigation has been enabled.
[1] https://arxiv.org/pdf/1807.07940.pdf
Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Tim Chen <tim.c.chen@linux.intel.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Borislav Petkov <bp@suse.de> Cc: David Woodhouse <dwmw@amazon.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: https://lkml.kernel.org/r/nycvar.YFH.7.76.1807261308190.997@cbobk.fhfr.pm Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The SNDRV_RAWMIDI_IOCTL_PARAMS ioctl may resize the buffers and the
current code is racy. For example, the sequencer client may write to
buffer while it being resized.
As a simple workaround, let's switch to the resized buffer inside the
stream runtime lock.
Reported-by: syzbot+52f83f0ea8df16932f7f@syzkaller.appspotmail.com Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
In general, accessing userspace memory beyond the length of the supplied
buffer in VFS read/write handlers can lead to both kernel memory corruption
(via kernel_read()/kernel_write(), which can e.g. be triggered via
sys_splice()) and privilege escalation inside userspace.
Fix it by using simple_read_from_buffer() instead of custom logic.
error_entry and error_exit communicate the user vs. kernel status of
the frame using %ebx. This is unnecessary -- the information is in
regs->cs. Just use regs->cs.
This makes error_entry simpler and makes error_exit more robust.
It also fixes a nasty bug. Before all the Spectre nonsense, the
xen_failsafe_callback entry point returned like this:
And it did not go through error_entry. This was bogus: RBX
contained garbage, and error_exit expected a flag in RBX.
Fortunately, it generally contained *nonzero* garbage, so the
correct code path was used. As part of the Spectre fixes, code was
added to clear RBX to mitigate certain speculation attacks. Now,
depending on kernel configuration, RBX got zeroed and, when running
some Wine workloads, the kernel crashes. This was introduced by:
commit 3ac6d8c787b8 ("x86/entry/64: Clear registers for exceptions/interrupts, to reduce speculation attack surface")
With this patch applied, RBX is no longer needed as a flag, and the
problem goes away.
I suspect that malicious userspace could use this bug to crash the
kernel even without the offending patch applied, though.
[ Historical note: I wrote this patch as a cleanup before I was aware
of the bug it fixed. ]
[ Note to stable maintainers: this should probably get applied to all
kernels. If you're nervous about that, a more conservative fix to
add xorl %ebx,%ebx; incl %ebx before the jump to error_exit should
also fix the problem. ]
Reported-and-tested-by: M. Vefa Bicakci <m.v.b@runbox.com> Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Dominik Brodowski <linux@dominikbrodowski.net> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: xen-devel@lists.xenproject.org Fixes: 3ac6d8c787b8 ("x86/entry/64: Clear registers for exceptions/interrupts, to reduce speculation attack surface") Link: http://lkml.kernel.org/r/b5010a090d3586b2d6e06c7ad3ec5542d1241c45.1532282627.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.16:
- error_exit moved EBX to EAX before testing it, so delete both instructions
- error_exit does RESTORE_REST earlier, so adjust the offset to saved CS
accordingly
- Drop inapplicable comment changes
- Adjust filename, context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
sgid directories have special semantics, making newly created files in
the directory belong to the group of the directory, and newly created
subdirectories will also become sgid. This is historically used for
group-shared directories.
But group directories writable by non-group members should not imply
that such non-group members can magically join the group, so make sure
to clear the sgid bit on non-directories for non-members (but remember
that sgid without group execute means "mandatory locking", just to
confuse things even more).
Reported-by: Jann Horn <jannh@google.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Use a separate journal transaction if it turns out that we need to
convert an inline file to use an data block. Otherwise we could end
up failing due to not having journal credits.
Do not set the b_modified flag in block's journal head should not
until after we're sure that jbd2_journal_dirty_metadat() will not
abort with an error due to there not being enough space reserved in
the jbd2 handle.
Otherwise, future attempts to modify the buffer may lead a large
number of spurious errors and warnings.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
[bwh: Backported to 3.16: Drop the added logging statement, as it's on
a code path that doesn't exist here] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When converting from an inode from storing the data in-line to a data
block, ext4_destroy_inline_data_nolock() was only clearing the on-disk
copy of the i_blocks[] array. It was not clearing copy of the
i_blocks[] in ext4_inode_info, in i_data[], which is the copy actually
used by ext4_map_blocks().
This didn't matter much if we are using extents, since the extents
header would be invalid and thus the extents could would re-initialize
the extents tree. But if we are using indirect blocks, the previous
contents of the i_blocks array will be treated as block numbers, with
potentially catastrophic results to the file system integrity and/or
user data.
This gets worse if the file system is using a 1k block size and
s_first_data is zero, but even without this, the file system can get
quite badly corrupted.
When expanding the extra isize space, we must never move the
system.data xattr out of the inode body. For performance reasons, it
doesn't make any sense, and the inline data implementation assumes
that system.data xattr is never in the external xattr block.
If there an inode points to a block which is also some other type of
metadata block (such as a block allocation bitmap), the
buffer_verified flag can be set when it was validated as that other
metadata block type; however, it would make a really terrible external
attribute block. The reason why we use the verified flag is to avoid
constantly reverifying the block. However, it doesn't take much
overhead to make sure the magic number of the xattr block is correct,
and this will avoid potential crashes.
In theory this should have been caught earlier when the xattr list was
verified, but in case it got missed, it's simple enough to add check
to make sure we don't overrun the xattr buffer.
Ext4_check_descriptors() was getting called before s_gdb_count was
initialized. So for file systems w/o the meta_bg feature, allocation
bitmaps could overlap the block group descriptors and ext4 wouldn't
notice.
For file systems with the meta_bg feature enabled, there was a
fencepost error which would cause the ext4_check_descriptors() to
incorrectly believe that the block allocation bitmap overlaps with the
block group descriptor blocks, and it would reject the mount.
Fix both of these problems.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
It's really bad when the allocation bitmaps and the inode table
overlap with the block group descriptors, since it causes random
corruption of the bg descriptors. So we really want to head those off
at the pass.
If some metadata block, such as an allocation bitmap, overlaps the
superblock, it's very likely that if the file system is mounted
read/write, the results will not be pretty. So disallow r/w mounts
for file systems corrupted in this particular way.
Backport notes:
3.18.y is missing bc98a42c1f7d ("VFS: Convert sb->s_flags & MS_RDONLY
to sb_rdonly(sb)") and e462ec50cb5f ("VFS: Differentiate mount flags
(MS_*) from internal superblock flags") so we simply use the sb
MS_RDONLY check from pre bc98a42c1f7d in place of the sb_rdonly
function used in the upstream variant of the patch.
Regardless of whether the flex_bg feature is set, we should always
check to make sure the bits we are setting in the block bitmap are
within the block group bounds.
Commit 8844618d8aa7: "ext4: only look at the bg_flags field if it is
valid" will complain if block group zero does not have the
EXT4_BG_INODE_ZEROED flag set. Unfortunately, this is not correct,
since a freshly created file system has this flag cleared. It gets
almost immediately after the file system is mounted read-write --- but
the following somewhat unlikely sequence will end up triggering a
false positive report of a corrupted file system:
mkfs.ext4 /dev/vdc
mount -o ro /dev/vdc /vdc
mount -o remount,rw /dev/vdc
Instead, when initializing the inode table for block group zero, test
to make sure that itable_unused count is not too large, since that is
the case that will result in some or all of the reserved inodes
getting cleared.
This fixes the failures reported by Eric Whiteney when running
generic/230 and generic/231 in the the nojournal test case.
Fixes: 8844618d8aa7 ("ext4: only look at the bg_flags field if it is valid") Reported-by: Eric Whitney <enwlinux@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The bg_flags field in the block group descripts is only valid if the
uninit_bg or metadata_csum feature is enabled. We were not
consistently looking at this field; fix this.
Also block group #0 must never have uninitialized allocation bitmaps,
or need to be zeroed, since that's where the root inode, and other
special inodes are set up. Check for these conditions and mark the
file system as corrupted if they are detected.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
[bwh: Backported to 3.16:
- ext4_read_block_bitmap_nowait() and ext4_read_inode_bitmap() return
a pointer (NULL on error) instead of an error code
- Open-code sb_rdonly()
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If xfs_bmap_extents_to_btree fails in a mode where we call
xfs_iroot_realloc(-1) to de-allocate the root, set the
format back to extents.
Otherwise we can assume we can dereference ifp->if_broot
based on the XFS_DINODE_FMT_BTREE format, and crash.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=199423 Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[bwh: Backported to 3.16:
- Only one failure path needs to be patched
- Adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When ata device doing EH, some commands still attached with tasks are
not passed to libata when abort failed or recover failed, so libata did
not handle these commands. After these commands done, sas task is freed,
but ata qc is not freed. This will cause ata qc leak and trigger a
warning like below:
If ata qc leaked too many, ata tag allocation will fail and io blocked
for ever.
As suggested by Dan Williams, defer ata device commands to libata and
merge sas_eh_finish_cmd() with sas_eh_defer_cmd(). libata will handle
ata qcs correctly after this.
Signed-off-by: Jason Yan <yanaijie@huawei.com> CC: Xiaofei Tan <tanxiaofei@huawei.com> CC: John Garry <john.garry@huawei.com> CC: Dan Williams <dan.j.williams@intel.com> Reviewed-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The buffer length is unsigned at all layers, but gets cast to int and
checked in hidp_process_report and can lead to a buffer overflow.
Switch len parameter to unsigned int to resolve issue.
This affects 3.18 and newer kernels.
Signed-off-by: Mark Salyzyn <salyzyn@android.com> Fixes: a4b1b5877b514b276f0f31efe02388a9c2836728 ("HID: Bluetooth: hidp: make sure input buffers are big enough") Cc: Marcel Holtmann <marcel@holtmann.org> Cc: Johan Hedberg <johan.hedberg@gmail.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Kees Cook <keescook@chromium.org> Cc: Benjamin Tissoires <benjamin.tissoires@redhat.com> Cc: linux-bluetooth@vger.kernel.org Cc: netdev@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: security@kernel.org Cc: kernel-team@android.com Acked-by: Kees Cook <keescook@chromium.org> Signed-off-by: Marcel Holtmann <marcel@holtmann.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This shall help avoid copying uninitialized memory to the userspace when
calling ioctl(fd, SG_IO) with an empty command.
Reported-by: syzbot+7d26fc1eea198488deab@syzkaller.appspotmail.com Signed-off-by: Alexander Potapenko <glider@google.com> Acked-by: Douglas Gilbert <dgilbert@interlog.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We're casting the CDROM layer request_sense to the SCSI sense
buffer, but the former is 64 bytes and the latter is 96 bytes.
As we generally allocate these on the stack, we end up blowing
up the stack.
Fix this by wrapping the scsi_execute() call with a properly
sized sense buffer, and copying back the bits for the CDROM
layer.
Reported-by: Piotr Gabriel Kosinski <pg.kosinski@gmail.com> Reported-by: Daniel Shapira <daniel@twistlock.com> Tested-by: Kees Cook <keescook@chromium.org> Fixes: 82ed4db499b8 ("block: split scsi_request out of struct request") Signed-off-by: Jens Axboe <axboe@kernel.dk>
[bwh: Despite what the "Fixes" field says, a buffer overrun was already
possible if the sense data was really > 64 bytes long.
Backported to 3.16:
- We always need to allocate a sense buffer in order to call
scsi_normalize_sense()
- Remove the existing conditional heap-allocation of the sense buffer] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The functions that were used in the emulation of fxrstor, fxsave, sgdt and
sidt were originally meant for task switching, and as such they did not
check privilege levels. This is very bad when the same functions are used
in the emulation of unprivileged instructions. This is CVE-2018-10853.
The obvious fix is to add a new argument to ops->read_std and ops->write_std,
which decides whether the access is a "system" access or should use the
processor's CPL.
Fixes: 129a72a0d3c8 ("KVM: x86: Introduce segmented_write_std", 2017-01-12) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[bwh: Backported to 3.16: Drop change in handle_ud()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Int the next patch the emulator's .read_std and .write_std callbacks will
grow another argument, which is not needed in kvm_read_guest_virt and
kvm_write_guest_virt_system's callers. Since we have to make separate
functions, let's give the currently existing names a nicer interface, too.
Fixes: 129a72a0d3c8 ("KVM: x86: Introduce segmented_write_std", 2017-01-12) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[bwh: Backported to 3.16:
- Drop change to handle_invvpid()
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The current implementation ignores the LDTR/TR base high 32-bits on long-mode.
As a result the loaded segment descriptor may be incorrect.
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Commit 65d8fc777f6d ("futex: Remove requirement for lock_page() in
get_futex_key()") removed an unnecessary lock_page() with the
side-effect that page->mapping needed to be treated very carefully.
Two defensive warnings were added in case any assumption was missed and
the first warning assumed a correct application would not alter a
mapping backing a futex key. Since merging, it has not triggered for
any unexpected case but Mark Rutland reported the following bug
triggering due to the first warning.
kernel BUG at kernel/futex.c:679!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 3695 Comm: syz-executor1 Not tainted 4.13.0-rc3-00020-g307fec773ba3 #3
Hardware name: linux,dummy-virt (DT)
task: ffff80001e271780 task.stack: ffff000010908000
PC is at get_futex_key+0x6a4/0xcf0 kernel/futex.c:679
LR is at get_futex_key+0x6a4/0xcf0 kernel/futex.c:679
pc : [<ffff00000821ac14>] lr : [<ffff00000821ac14>] pstate: 80000145
The fact that it's a bug instead of a warning was due to an unrelated
arm64 problem, but the warning itself triggered because the underlying
mapping changed.
This is an application issue but from a kernel perspective it's a
recoverable situation and the warning is unnecessary so this patch
removes the warning. The warning may potentially be triggered with the
following test program from Mark although it may be necessary to adjust
NR_FUTEX_THREADS to be a value smaller than the number of CPUs in the
system.
static int futex_wrapper(int *uaddr, int op, int val,
const struct timespec *timeout,
int *uaddr2, int val3)
{
syscall(SYS_futex, uaddr, op, val, timeout, uaddr2, val3);
}
void *poll_futex(void *unused)
{
for (;;) {
futex_wrapper(mem, FUTEX_CMP_REQUEUE_PI, 1, NULL, mem + 4, 1);
}
}
int main(int argc, char *argv[])
{
int i;
mem = mmap(NULL, MEM_SIZE, MEM_PROT,
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
printf("Mapping @ %p\n", mem);
printf("Creating futex threads...\n");
for (i = 0; i < NR_FUTEX_THREADS; i++)
pthread_create(&threads[i], NULL, poll_futex, NULL);
When dealing with key handling for shared futexes, we can drastically reduce
the usage/need of the page lock. 1) For anonymous pages, the associated futex
object is the mm_struct which does not require the page lock. 2) For inode
based, keys, we can check under RCU read lock if the page mapping is still
valid and take reference to the inode. This just leaves one rare race that
requires the page lock in the slow path when examining the swapcache.
Additionally realtime users currently have a problem with the page lock being
contended for unbounded periods of time during futex operations.
Task A
get_futex_key()
lock_page()
---> preempted
Now any other task trying to lock that page will have to wait until
task A gets scheduled back in, which is an unbound time.
With this patch, we pretty much have a lockless futex_get_key().
Experiments show that this patch can boost/speedup the hashing of shared
futexes with the perf futex benchmarks (which is good for measuring such
change) by up to 45% when there are high (> 100) thread counts on a 60 core
Westmere. Lower counts are pretty much in the noise range or less than 10%,
but mid range can be seen at over 30% overall throughput (hash ops/sec).
This makes anon-mem shared futexes much closer to its private counterpart.
Signed-off-by: Mel Gorman <mgorman@suse.de>
[ Ported on top of thp refcount rework, changelog, comments, fixes. ] Signed-off-by: Davidlohr Bueso <dbueso@suse.de> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Chris Mason <clm@fb.com> Cc: Darren Hart <dvhart@linux.intel.com> Cc: Hugh Dickins <hughd@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: dave@stgolabs.net Link: http://lkml.kernel.org/r/1455045314-8305-3-git-send-email-dave@stgolabs.net Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Chenbo Feng <fengc@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: s/READ_ONCE/ACCESS_ONCE/] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
stub_probe() calls put_busid_priv() in an error path when device isn't
found in the busid_table. Fix it by making put_busid_priv() safe to be
called with null struct bus_id_priv pointer.
This problem happens when "usbip bind" is run without loading usbip_host
driver and then running modprobe. The first failed bind attempt unbinds
the device from the original driver and when usbip_host is modprobed,
stub_probe() runs and doesn't find the device in its busid table and calls
put_busid_priv(0 with null bus_id_priv pointer.
usbip-host 3-10.2: 3-10.2 is not in match_busid table... skip!
[ 367.359679] =====================================
[ 367.359681] WARNING: bad unlock balance detected!
[ 367.359683] 4.17.0-rc4+ #5 Not tainted
[ 367.359685] -------------------------------------
[ 367.359688] modprobe/2768 is trying to release lock (
[ 367.359689]
==================================================================
[ 367.359696] BUG: KASAN: null-ptr-deref in print_unlock_imbalance_bug+0x99/0x110
[ 367.359699] Read of size 8 at addr 0000000000000058 by task modprobe/2768
usbip_host updates device status without holding lock from stub probe,
disconnect and rebind code paths. When multiple requests to import a
device are received, these unprotected code paths step all over each
other and drive fails with NULL-ptr deref and use-after-free errors.
The driver uses a table lock to protect the busid array for adding and
deleting busids to the table. However, the probe, disconnect and rebind
paths get the busid table entry and update the status without holding
the busid table lock. Add a new finer grain lock to protect the busid
entry. This new lock will be held to search and update the busid entry
fields from get_busid_idx(), add_match_busid() and del_match_busid().
match_busid_show() does the same to access the busid entry fields.
get_busid_priv() changed to return the pointer to the busid entry holding
the busid lock. stub_probe(), stub_disconnect() and stub_device_rebind()
call put_busid_priv() to release the busid lock before returning. This
changes fixes the unprotected code paths eliminating the race conditions
in updating the busid entries.
Reported-by: Jakub Jirasek Signed-off-by: Shuah Khan (Samsung OSG) <shuah@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: adjust filenames, context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
After removing usbip_host module, devices it releases are left without
a driver. For example, when a keyboard or a mass storage device are
bound to usbip_host when it is removed, these devices are no longer
bound to any driver.
Fix it to run device_attach() from the module exit routine to restore
the devices to their original drivers. This includes cleanup changes
and moving device_attach() code to a common routine to be called from
rebind_store() and usbip_host_exit().
Signed-off-by: Shuah Khan (Samsung OSG) <shuah@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: adjust filenames] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Device is left in the busid_table after unbind and rebind. Rebind
initiates usb bus scan and the original driver claims the device.
After rescan the device should be deleted from the busid_table as
it no longer belongs to usbip_host.
Fix it to delete the device after device_attach() succeeds.
Signed-off-by: Shuah Khan (Samsung OSG) <shuah@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.16: adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
sk->sk_prot and sk->sk_prot_creator can differ when the app uses
IPV6_ADDRFORM (transforming an IPv6-socket to an IPv4-one).
Which is why sk_prot_creator is there to make sure that sk_prot_free()
does the kmem_cache_free() on the right kmem_cache slab.
Now, if such a socket gets transformed back to a listening socket (using
connect() with AF_UNSPEC) we will allocate an IPv4 tcp_sock through
sk_clone_lock() when a new connection comes in. But sk_prot_creator will
still point to the IPv6 kmem_cache (as everything got copied in
sk_clone_lock()). When freeing, we will thus put this
memory back into the IPv6 kmem_cache although it was allocated in the
IPv4 cache. I have seen memory corruption happening because of this.
With slub-debugging and MEMCG_KMEM enabled this gives the warning
"cache_from_obj: Wrong slab cache. TCPv6 but object is from TCP"
A C-program to trigger this:
void main(void)
{
int fd = socket(AF_INET6, SOCK_STREAM, IPPROTO_TCP);
int new_fd, newest_fd, client_fd;
struct sockaddr_in6 bind_addr;
struct sockaddr_in bind_addr4, client_addr1, client_addr2;
struct sockaddr unsp;
int val;
As far as I can see, this bug has been there since the beginning of the
git-days.
Signed-off-by: Christoph Paasch <cpaasch@apple.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Cc: Thomas King <thomaskingnew@gmail.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Ben Hutchings [Sun, 24 Jun 2018 23:25:48 +0000 (00:25 +0100)]
x86/fpu: Default eagerfpu if FPU and FXSR are enabled
This is a limited version of commit 58122bf1d856 "x86/fpu: Default
eagerfpu=on on all CPUs". That commit revealed bugs in the use of
eagerfpu together with math emulation or without the FXSR feature.
Although those bugs have been fixed upstream, the fixes do not seem to
be practical to backport to 3.16.
The security issue that motivates using eagerfpu (CVE-2018-3665) is an
information leak through speculative execution, and most CPUs lacking
the FXSR feature also don't implement speculative execution. The
exceptions I am aware of are the Intel Pentium Pro and AMD K6 family,
which will remain vulnerable to this issue.
Move the eagerfpu variable and associated initialisation into
fpu_init(), since xstate_enable_boot_cpu() won't be called at all if
XSAVE is disabled.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Cc: Andy Lutomirski <luto@kernel.org> Cc: x86@kernel.org
In some situation vlan packets do not have ethernet headers. One example
is packets from tun devices. Users can specify vlan protocol in tun_pi
field instead of IP protocol, and skb_vlan_untag() attempts to untag such
packets.
skb_vlan_untag() (more precisely, skb_reorder_vlan_header() called by it)
however did not expect packets without ethernet headers, so in such a case
size argument for memmove() underflowed and triggered crash.
We don't need to copy headers for packets which do not have preceding
headers of vlan headers, so skip memmove() in that case.
Fixes: 4bbb3e0e8239 ("net: Fix vlan untag for bridge and vlan_dev with reorder_hdr off") Reported-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
It turns out that the loop where we read manufacturer
jedec_read_mfd() can under some circumstances get a
CFI_MFR_CONTINUATION repeatedly, making the loop go
over all banks and eventually hit the end of the
map and crash because of an access violation:
Fix this by breaking the loop with a return 0 if
the offset exceeds the map size.
Fixes: 5c9c11e1c47c ("[MTD] [NOR] Add support for flash chips with ID in bank other than 0") Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Ensure that device exists prior to accessing its properties.
Reported-by: <syzbot+71655d44855ac3e76366@syzkaller.appspotmail.com> Fixes: 75216638572f ("RDMA/cma: Export rdma cm interface to userspace") Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Fixes: 75216638572f ("RDMA/cma: Export rdma cm interface to userspace") Fixes: d541e45500bd ("IB/core: Convert ah_attr from OPA to IB when copying to user") Reported-by: <syzbot+7b62c837c2516f8f38c8@syzkaller.appspotmail.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Global pause and PFC configuration should be mutually exclusive (i.e. only
one of them at most can be set). However, once PFC was turned off,
driver automatically turned Global pause on. This is a bug.
Fix the driver behaviour to turn off PFC/Global once the user turned the
other on.
This also fixed a weird behaviour that at a current time, the profile
had both PFC and global pause configuration turned on, which is
Hardware-wise impossible and caused returning false positive indication
to query tools.
In addition, fix error code when setting global pause or PFC to change
metadata only upon successful change.
Also, removed useless debug print.
Fixes: af7d51852631 ("net/mlx4_en: Add DCB PFC support through CEE netlink commands") Fixes: c27a02cd94d6 ("mlx4_en: Add driver for Mellanox ConnectX 10GbE NIC") Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com> Signed-off-by: Tariq Toukan <tariqt@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16:
- Drop changes to mlx4_en_dcbnl_set_all()
- Don't call mlx4_en_update_pfc_stats_bitmap()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The driver does not support pause autonegotiation so it should return
-EINVAL when the function is called with non-zero autoneg.
Cc: Amir Vadai <amirv@mellanox.com> Signed-off-by: Ivan Vecera <ivecera@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
skb mac header is not necessarily set at the time skb_network_protocol()
is called. Use skb->data instead.
BUG: KASAN: slab-out-of-bounds in skb_network_protocol+0x46b/0x4b0 net/core/dev.c:2739
Read of size 2 at addr ffff8801b3097a0b by task syz-executor5/14242
Fixes: 19acc327258a ("gso: Handle Trans-Ether-Bridging protocol in skb_network_protocol()") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Pravin B Shelar <pshelar@ovn.org> Reported-by: Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When dev_set_promiscuity(1) succeeds but dev_set_allmulti(1) fails,
dev_set_promiscuity(-1) should be done before going to the err path.
Otherwise, dev->promiscuity will leak.
Fixes: 7e1a1ac1fbaa ("bonding: Check return of dev_set_promiscuity/allmulti") Signed-off-by: Xin Long <lucien.xin@gmail.com> Acked-by: Andy Gospodarek <andy@greyhouse.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This is actually a dead lock caused by sync slave hwaddr from master when
the master is the slave's 'slave'. This dead loop check is actually done
by netdev_master_upper_dev_link. However, Commit 1f718f0f4f97 ("bonding:
populate neighbour's private on enslave") moved it after dev_mc_sync.
This patch is to fix it by moving dev_mc_sync after master_upper_dev_link,
so that this loop check would be earlier than dev_mc_sync. It also moves
if (mode == BOND_MODE_8023AD) into if (!bond_uses_primary) clause as an
improvement.
Note team driver also has this issue, I will fix it in another patch.
Fixes: 1f718f0f4f97 ("bonding: populate neighbour's private on enslave") Reported-by: Beniamino Galvani <bgalvani@redhat.com> Signed-off-by: Xin Long <lucien.xin@gmail.com> Acked-by: Andy Gospodarek <andy@greyhouse.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
vlan_vids_add_by_dev is called right after dev hwaddr sync, so on
the err path it should unsync dev hwaddr. Otherwise, the slave
dev's hwaddr will never be unsync when this err happens.
Fixes: 1ff412ad7714 ("bonding: change the bond's vlan syncing functions with the standard ones") Signed-off-by: Xin Long <lucien.xin@gmail.com> Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com> Acked-by: Andy Gospodarek <andy@greyhouse.net> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When trying to use the driver (e.g. aplay *.wav), the 4MiB DMA buffer
will get mmapp'ed in 16KiB chunks. But this fails with the 2nd 16KiB
area, as the page offset is outside of the VMA range (size), which is
currently used as size parameter in snd_pcm_lib_default_mmap(). By
using the DMA buffer size (dma_bytes) instead, the complete DMA buffer
can be mmapp'ed and the issue is fixed.
This issue was detected on an ARM platform (TI AM57xx) using the RME
HDSP MADI PCIe soundcard.
Fixes: 657b1989dacf ("ALSA: pcm - Use dma_mmap_coherent() if available") Signed-off-by: Stefan Roese <sr@denx.de> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
While building ipv6 datagram we currently allow arbitrary large
extheaders, even beyond pmtu size. The syzbot has found a way
to exploit the above to trigger the following splat:
When a host fragments an IPv6 datagram, it MUST include the entire
IPv6 Header Chain in the First Fragment.
So this patch addresses the issue dropping datagrams with excessive
extheader length. It also updates the error path to report to the
calling socket nonnegative pmtu values.
The issue apparently predates git history.
v1 -> v2: cleanup error path, as per Eric's suggestion
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: syzbot+91e6f9932ff122fa4410@syzkaller.appspotmail.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16: Adjust context, indentation] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
KMSAN reports use of uninitialized memory in the case when |alen| is
smaller than sizeof(struct sockaddr_nl), and therefore |nladdr| isn't
fully copied from the userspace.
Signed-off-by: Alexander Potapenko <glider@google.com> Fixes: 1da177e4c3f41524 ("Linux-2.6.12-rc2") Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
DHCP connectivity issues can currently occur if the following conditions
are met:
1) A DHCP packet from a client to a server
2) This packet has a multicast destination
3) This destination has a matching entry in the translation table
(FF:FF:FF:FF:FF:FF for IPv4, 33:33:00:01:00:02/33:33:00:01:00:03
for IPv6)
4) The orig-node determined by TT for the multicast destination
does not match the orig-node determined by best-gateway-selection
In this case the DHCP packet will be dropped.
The "gateway-out-of-range" check is supposed to only be applied to
unicasted DHCP packets to a specific DHCP server.
In that case dropping the the unicasted frame forces the client to
retry via a broadcasted one, but now directed to the new best
gateway.
A DHCP packet with broadcast/multicast destination is already ensured to
always be delivered to the best gateway. Dropping a multicasted
DHCP packet here will only prevent completing DHCP as there is no
other fallback.
So far, it seems the unicast check was implicitly performed by
expecting the batadv_transtable_search() to return NULL for multicast
destinations. However, a multicast address could have always ended up in
the translation table and in fact is now common.
To fix this potential loss of a DHCP client-to-server packet to a
multicast address this patch adds an explicit multicast destination
check to reliably bail out of the gateway-out-of-range check for such
destinations.
The issue and fix were tested in the following three node setup:
- Line topology, A-B-C
- A: gateway client, DHCP client
- B: gateway server, hop-penalty increased: 30->60, DHCP server
- C: gateway server, code modifications to announce FF:FF:FF:FF:FF:FF
Without this patch, A would never transmit its DHCP Discover packet
due to an always "out-of-range" condition. With this patch,
a full DHCP handshake between A and B was possible again.
Fixes: be7af5cf9cae ("batman-adv: refactoring gateway handling code") Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue> Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
[bwh: Backported to 3.16: Drop redundant change to initialisation] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
For multicast frames AP isolation is only supposed to be checked on
the receiving nodes and never on the originating one.
Furthermore, the isolation or wifi flag bits should only be intepreted
as such for unicast and never multicast TT entries.
By injecting flags to the multicast TT entry claimed by a single
target node it was verified in tests that this multicast address
becomes unreachable, leading to packet loss.
Omitting the "src" parameter to the batadv_transtable_search() call
successfully skipped the AP isolation check and made the target
reachable again.
Fixes: 1d8ab8d3c176 ("batman-adv: Modified forwarding behaviour for multicast packets") Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue> Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
For tunnels created with IFLA_MTU, MTU of the netdevice is set by
rtnl_create_link() (called from rtnl_newlink()) before the device is
registered. However without IFLA_MTU that's not done.
rtnl_newlink() proceeds by calling struct rtnl_link_ops.newlink, which
via ip_tunnel_newlink() calls register_netdevice(), and that emits
NETDEV_REGISTER. Thus any listeners that inspect the netdevice get the
MTU of 0.
After ip_tunnel_newlink() corrects the MTU after registering the
netdevice, but since there's no event, the listeners don't get to know
about the MTU until something else happens--such as a NETDEV_UP event.
That's not ideal.
So instead of setting the MTU directly, go through dev_set_mtu(), which
takes care of distributing the necessary NETDEV_PRECHANGEMTU and
NETDEV_CHANGEMTU events.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Petr Machata <petrm@mellanox.com> Signed-off-by: Ido Schimmel <idosch@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16: Resolve conflict with commit 24fc79798b8d
"ip_tunnel: Clamp MTU to bounds on new link", referring to commit 5568cdc368c3 "ip_tunnel: Resolve ipsec merge conflict properly."] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
@SYM[+|-offs] : Fetch memory at SYM +|- offs (SYM should be a data symbol)
However, the parser doesn't parse minus offset correctly, since
commit 2fba0c8867af ("tracing/kprobes: Fix probe offset to be
unsigned") drops minus ("-") offset support for kprobe probe
address usage.
This fixes the traceprobe_split_symbol_offset() to parse minus
offset again with checking the offset range, and add a minus
offset check in kprobe probe address usage.
Link: http://lkml.kernel.org/r/152129028983.31874.13419301530285775521.stgit@devbox Cc: Ingo Molnar <mingo@redhat.com> Cc: Tom Zanussi <tom.zanussi@linux.intel.com> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Fixes: 2fba0c8867af ("tracing/kprobes: Fix probe offset to be unsigned") Acked-by: Namhyung Kim <namhyung@kernel.org> Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Alexander reported a use of uninitialized memory in __mpol_equal(),
which is caused by incorrect use of preferred_node.
When mempolicy in mode MPOL_PREFERRED with flags MPOL_F_LOCAL, it uses
numa_node_id() instead of preferred_node, however, __mpol_equal() uses
preferred_node without checking whether it is MPOL_F_LOCAL or not.
[akpm@linux-foundation.org: slight comment tweak] Link: http://lkml.kernel.org/r/4ebee1c2-57f6-bcb8-0e2d-1833d1ee0bb7@huawei.com Fixes: fc36b8d3d819 ("mempolicy: use MPOL_F_LOCAL to Indicate Preferred Local Policy") Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com> Reported-by: Alexander Potapenko <glider@google.com> Tested-by: Alexander Potapenko <glider@google.com> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Fix it by free'ing the netdev straight after unregistering. This also
fixes the sysfs-driven layer switch case (qeth_dev_layer2_store()),
where the need to free the current netdevice was not considered at all.
Note that free_netdev() takes care of the netif_napi_del() for us too.
Fixes: 4a71df50047f ("qeth: new qeth device driver") Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com> Reviewed-by: Ursula Braun <ubraun@linux.vnet.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The clockid argument of clockid_to_kclock() comes straight from user space
via various syscalls and is used as index into the posix_clocks array.
Protect it against spectre v1 array out of bounds speculation. Remove the
redundant check for !posix_clock[id] as this is another source for
speculation and does not provide any advantage over the return
posix_clock[id] path which returns NULL in that case anyway.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Dan Williams <dan.j.williams@intel.com> Cc: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: David Woodhouse <dwmw@amazon.co.uk> Link: https://lkml.kernel.org/r/alpine.DEB.2.21.1802151718320.1296@nanos.tec.linutronix.de
[bwh: Backported to 3.16:
- Move the test of the clock_getres field below the lookup using
array_index_nospec()
- Adjust filename, context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
In loopback_open() and loopback_close(), we assign and release the
substream object to the corresponding cable in a racy way. It's
neither locked nor done in the right position. The open callback
assigns the substream before its preparation finishes, hence the other
side of the cable may pick it up, which may lead to the invalid memory
access.
This patch addresses these: move the assignment to the end of the open
callback, and wrap with cable->lock for avoiding concurrent accesses.
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The aloop driver tries to stop the pending timer via timer_del() in
the trigger callback and in the close callback. The former is
correct, as it's an atomic operation, while the latter expects that
the timer gets really removed and proceeds the resource releases after
that. But timer_del() doesn't synchronize, hence the running timer
may still access the released resources.
A similar situation can be also seen in the prepare callback after
trigger(STOP) where the prepare tries to re-initialize the things
while a timer is still running.
The problems like the above are seen indirectly in some syzkaller
reports (although it's not 100% clear whether this is the only cause,
as the race condition is quite narrow and not always easy to
trigger).
For addressing these issues, this patch adds the explicit alls of
timer_del_sync() in some places, so that the pending timer is properly
killed / synced.
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
ralink_halt() does nothing that machine_halt() doesn't already do, so it
adds no value.
It actually causes incorrect behaviour due to the "unreachable()" at the
end. This tells the compiler that the end of the function will never be
reached, which isn't true. The compiler responds by not adding a
'return' instruction, so control simply moves on to whatever bytes come
afterwards in memory. In my tested, that was the ralink_restart()
function. This means that an attempt to 'halt' the machine would
actually cause a reboot.
So remove ralink_halt() so that a 'halt' really does halt.
Fixes: c06e836ada59 ("MIPS: ralink: adds reset code") Signed-off-by: NeilBrown <neil@brown.name> Cc: John Crispin <john@phrozen.org> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/18851/ Signed-off-by: James Hogan <jhogan@kernel.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Prior to access UCMA commands, the context should be initialized
and connected to CM_ID with ucma_create_id(). In case user skips
this step, he can provide non-valid ctx without CM_ID and cause
to multiple NULL dereferences.
Also there are situations where the create_id can be raced with
other user access, ensure that the context is only shared to
other threads once it is fully initialized to avoid the races.
Fixes: 75216638572f ("RDMA/cma: Export rdma cm interface to userspace") Reported-by: <syzbot+36712f50b0552615bf59@syzkaller.appspotmail.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The error in ucma_create_id() left ctx in the list of contexts belong
to ucma file descriptor. The attempt to close this file descriptor causes
to use-after-free accesses while iterating over such list.
Fixes: 75216638572f ("RDMA/cma: Export rdma cm interface to userspace") Reported-by: <syzbot+dcfd344365a56fbebd0f@syzkaller.appspotmail.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Reviewed-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently, the offsets in the UAC2 processing unit descriptor are
calculated incorrectly. It causes an issue when connecting the device which
provides such a feature:
~~~~
[84126.724420] usb 1-1.3.1: invalid Processing Unit descriptor (id 18)
~~~~
After this patch is applied, the UAC2 processing unit inits w/o this error.
Fixes: 23caaf19b11e ("ALSA: usb-mixer: Add support for Audio Class v2.0") Signed-off-by: Kirill Marinushkin <k.marinushkin@gmail.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When commit 9c7be59fc519af ("libata: Apply NOLPM quirk to Crucial MX100
512GB SSDs") was added it inherited the ATA_HORKAGE_NO_NCQ_TRIM quirk
from the existing "Crucial_CT*MX100*" entry, but that entry sets model_rev
to "MU01", where as the entry adding the NOLPM quirk sets it to NULL.
This means that after this commit we no apply the NO_NCQ_TRIM quirk to
all "Crucial_CT512MX100*" SSDs even if they have the fixed "MU02"
firmware. This commit splits the "Crucial_CT512MX100*" quirk into 2
quirks, one for the "MU01" firmware and one for all other firmware
versions, so that we once again only apply the NO_NCQ_TRIM quirk to the
"MU01" firmware version.
Fixes: 9c7be59fc519af ("libata: Apply NOLPM quirk to ... MX100 512GB SSDs") Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Tejun Heo <tj@kernel.org>
[bwh: Backported to 3.16: There's no ATA_HORKAGE_ZERO_AFTER_TRIM flag] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Commit b17e5729a630 ("libata: disable LPM for Crucial BX100 SSD 500GB
drive"), introduced a ATA_HORKAGE_NOLPM quirk for Crucial BX100 500GB SSDs
but limited this to the MU02 firmware version, according to:
http://www.crucial.com/usa/en/support-ssd-firmware
MU02 is the last version, so there are no newer possibly fixed versions
and if the MU02 version has broken LPM then the MU01 almost certainly
also has broken LPM, so this commit changes the quirk to apply to all
firmware versions.
Fixes: b17e5729a630 ("libata: disable LPM for Crucial BX100 SSD 500GB...") Cc: Kai-Heng Feng <kai.heng.feng@canonical.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
There have been reports of the Crucial M500 480GB model not working
with LPM set to min_power / med_power_with_dipm level.
It has not been tested with medium_power, but that typically has no
measurable power-savings.
Note the reporters Crucial_CT480M500SSD3 has a firmware version of MU03
and there is a MU05 update available, but that update does not mention any
LPM fixes in its changelog, so the quirk matches all firmware versions.
In my experience the LPM problems with (older) Crucial SSDs seem to be
limited to higher capacity versions of the SSDs (different firmware?),
so this commit adds a NOLPM quirk for the 480 and 960GB versions of the
M500, to avoid LPM causing issues with these SSDs.
Reported-and-tested-by: Martin Steigerwald <martin@lichtvoll.de> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Tejun Heo <tj@kernel.org>
[bwh: Backported to 3.16: There's no ATA_HORKAGE_ZERO_AFTER_TRIM flag] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Don't hardcode a MTU value on vti tunnel initialization,
ip_tunnel_newlink() is able to deal with this already. See also
commit ffc2b6ee4174 ("ip_gre: fix IFLA_MTU ignored on NEWLINK").
Fixes: 1181412c1a67 ("net/ipv4: VTI support new module for ip_vti.") Signed-off-by: Stefano Brivio <sbrivio@redhat.com> Acked-by: Sabrina Dubroca <sd@queasysnail.net> Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>