linux-stable.git/include, branch v4.4.67

mtd: avoid stack overflow in MTD CFI code

2017-05-08T05:46:01+00:00

commit fddcca5107051adf9e4481d2a79ae0616577fd2c upstream.

When map_word gets too large, we use a lot of kernel stack, and for
MTD_MAP_BANK_WIDTH_32, this means we use more than the recommended
1024 bytes in a number of functions:

drivers/mtd/chips/cfi_cmdset_0020.c: In function 'cfi_staa_write_buffers':
drivers/mtd/chips/cfi_cmdset_0020.c:651:1: warning: the frame size of 1336 bytes is larger than 1024 bytes [-Wframe-larger-than=]
drivers/mtd/chips/cfi_cmdset_0020.c: In function 'cfi_staa_erase_varsize':
drivers/mtd/chips/cfi_cmdset_0020.c:972:1: warning: the frame size of 1208 bytes is larger than 1024 bytes [-Wframe-larger-than=]
drivers/mtd/chips/cfi_cmdset_0001.c: In function 'do_write_buffer':
drivers/mtd/chips/cfi_cmdset_0001.c:1835:1: warning: the frame size of 1240 bytes is larger than 1024 bytes [-Wframe-larger-than=]

This can be avoided if all operations on the map word are done
indirectly and the stack gets reused between the calls. We can
mostly achieve this by selecting MTD_COMPLEX_MAPPINGS whenever
MTD_MAP_BANK_WIDTH_32 is set, but for the case that no other
bank width is enabled, we also need to use a non-constant
map_bankwidth() to convince the compiler to use less stack.

Signed-off-by: Arnd Bergmann 
[Brian: this patch mostly achieves its goal by forcing
    MTD_COMPLEX_MAPPINGS (and the accompanying indirection) for 256-bit
    mappings; the rest of the change is mostly a wash, though it helps
    reduce stack size slightly. If we really care about supporting
    256-bit mappings though, we should consider rewriting some of this
    code to avoid keeping and assigning so many 256-bit objects on the
    stack.]
Signed-off-by: Brian Norris 
Signed-off-by: Greg Kroah-Hartman

net: ipv6: RTF_PCPU should not be settable from userspace

2017-05-03T04:19:53+00:00

[ Upstream commit 557c44be917c322860665be3d28376afa84aa936 ]

Andrey reported a fault in the IPv6 route code:

kasan: GPF could be caused by NULL-ptr deref or user memory access
general protection fault: 0000 [#1] SMP KASAN
Modules linked in:
CPU: 1 PID: 4035 Comm: a.out Not tainted 4.11.0-rc7+ #250
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
task: ffff880069809600 task.stack: ffff880062dc8000
RIP: 0010:ip6_rt_cache_alloc+0xa6/0x560 net/ipv6/route.c:975
RSP: 0018:ffff880062dced30 EFLAGS: 00010206
RAX: dffffc0000000000 RBX: ffff8800670561c0 RCX: 0000000000000006
RDX: 0000000000000003 RSI: ffff880062dcfb28 RDI: 0000000000000018
RBP: ffff880062dced68 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
R13: ffff880062dcfb28 R14: dffffc0000000000 R15: 0000000000000000
FS:  00007feebe37e7c0(0000) GS:ffff88006cb00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000205a0fe4 CR3: 000000006b5c9000 CR4: 00000000000006e0
Call Trace:
 ip6_pol_route+0x1512/0x1f20 net/ipv6/route.c:1128
 ip6_pol_route_output+0x4c/0x60 net/ipv6/route.c:1212
...

Andrey's syzkaller program passes rtmsg.rtmsg_flags with the RTF_PCPU bit
set. Flags passed to the kernel are blindly copied to the allocated
rt6_info by ip6_route_info_create making a newly inserted route appear
as though it is a per-cpu route. ip6_rt_cache_alloc sees the flag set
and expects rt->dst.from to be set - which it is not since it is not
really a per-cpu copy. The subsequent call to __ip6_dst_alloc then
generates the fault.

Fix by checking for the flag and failing with EINVAL.

Fixes: d52d3997f843f ("ipv6: Create percpu rt6_info")
Reported-by: Andrey Konovalov 
Signed-off-by: David Ahern 
Acked-by: Martin KaFai Lau 
Tested-by: Andrey Konovalov 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

mnt: Add a per mount namespace limit on the number of mounts

2017-04-30T03:49:28+00:00

commit d29216842a85c7970c536108e093963f02714498 upstream.

CAI Qian  pointed out that the semantics
of shared subtrees make it possible to create an exponentially
increasing number of mounts in a mount namespace.

    mkdir /tmp/1 /tmp/2
    mount --make-rshared /
    for i in $(seq 1 20) ; do mount --bind /tmp/1 /tmp/2 ; done

Will create create 2^20 or 1048576 mounts, which is a practical problem
as some people have managed to hit this by accident.

As such CVE-2016-6213 was assigned.

Ian Kent  described the situation for autofs users
as follows:

> The number of mounts for direct mount maps is usually not very large because of
> the way they are implemented, large direct mount maps can have performance
> problems. There can be anywhere from a few (likely case a few hundred) to less
> than 10000, plus mounts that have been triggered and not yet expired.
>
> Indirect mounts have one autofs mount at the root plus the number of mounts that
> have been triggered and not yet expired.
>
> The number of autofs indirect map entries can range from a few to the common
> case of several thousand and in rare cases up to between 30000 and 50000. I've
> not heard of people with maps larger than 50000 entries.
>
> The larger the number of map entries the greater the possibility for a large
> number of active mounts so it's not hard to expect cases of a 1000 or somewhat
> more active mounts.

So I am setting the default number of mounts allowed per mount
namespace at 100,000.  This is more than enough for any use case I
know of, but small enough to quickly stop an exponential increase
in mounts.  Which should be perfect to catch misconfigurations and
malfunctioning programs.

For anyone who needs a higher limit this can be changed by writing
to the new /proc/sys/fs/mount-max sysctl.

Tested-by: CAI Qian 
Signed-off-by: "Eric W. Biederman" 
[bwh: Backported to 4.4: adjust context]
Signed-off-by: Ben Hutchings 
Signed-off-by: Greg Kroah-Hartman

crypto: ahash - Fix EINPROGRESS notification callback

2017-04-21T07:30:06+00:00

commit ef0579b64e93188710d48667cb5e014926af9f1b upstream.

The ahash API modifies the request's callback function in order
to clean up after itself in some corner cases (unaligned final
and missing finup).

When the request is complete ahash will restore the original
callback and everything is fine.  However, when the request gets
an EBUSY on a full queue, an EINPROGRESS callback is made while
the request is still ongoing.

In this case the ahash API will incorrectly call its own callback.

This patch fixes the problem by creating a temporary request
object on the stack which is used to relay EINPROGRESS back to
the original completion function.

This patch also adds code to preserve the original flags value.

Fixes: ab6bf4e5e5e4 ("crypto: hash - Fix the pointer voodoo in...")
Reported-by: Sabrina Dubroca 
Tested-by: Sabrina Dubroca 
Signed-off-by: Herbert Xu 
Signed-off-by: Greg Kroah-Hartman

cgroup, kthread: close race window where new kthreads can be migrated to non-root cgroups

2017-04-21T07:30:04+00:00

commit 77f88796cee819b9c4562b0b6b44691b3b7755b1 upstream.

Creation of a kthread goes through a couple interlocked stages between
the kthread itself and its creator.  Once the new kthread starts
running, it initializes itself and wakes up the creator.  The creator
then can further configure the kthread and then let it start doing its
job by waking it up.

In this configuration-by-creator stage, the creator is the only one
that can wake it up but the kthread is visible to userland.  When
altering the kthread's attributes from userland is allowed, this is
fine; however, for cases where CPU affinity is critical,
kthread_bind() is used to first disable affinity changes from userland
and then set the affinity.  This also prevents the kthread from being
migrated into non-root cgroups as that can affect the CPU affinity and
many other things.

Unfortunately, the cgroup side of protection is racy.  While the
PF_NO_SETAFFINITY flag prevents further migrations, userland can win
the race before the creator sets the flag with kthread_bind() and put
the kthread in a non-root cgroup, which can lead to all sorts of
problems including incorrect CPU affinity and starvation.

This bug got triggered by userland which periodically tries to migrate
all processes in the root cpuset cgroup to a non-root one.  Per-cpu
workqueue workers got caught while being created and ended up with
incorrected CPU affinity breaking concurrency management and sometimes
stalling workqueue execution.

This patch adds task->no_cgroup_migration which disallows the task to
be migrated by userland.  kthreadd starts with the flag set making
every child kthread start in the root cgroup with migration
disallowed.  The flag is cleared after the kthread finishes
initialization by which time PF_NO_SETAFFINITY is set if the kthread
should stay in the root cgroup.

It'd be better to wait for the initialization instead of failing but I
couldn't think of a way of implementing that without adding either a
new PF flag, or sleeping and retrying from waiting side.  Even if
userland depends on changing cgroup membership of a kthread, it either
has to be synchronized with kthread_create() or periodically repeat,
so it's unlikely that this would break anything.

v2: Switch to a simpler implementation using a new task_struct bit
    field suggested by Oleg.

Signed-off-by: Tejun Heo 
Suggested-by: Oleg Nesterov 
Cc: Linus Torvalds 
Cc: Andrew Morton 
Cc: Peter Zijlstra (Intel) 
Cc: Thomas Gleixner 
Reported-and-debugged-by: Chris Mason 
Signed-off-by: Tejun Heo 
Signed-off-by: Greg Kroah-Hartman

drm/ttm, drm/vmwgfx: Relax permission checking when opening surfaces

2017-04-12T10:38:33+00:00

commit fe25deb7737ce6c0879ccf79c99fa1221d428bf2 upstream.

Previously, when a surface was opened using a legacy (non prime) handle,
it was verified to have been created by a client in the same master realm.
Relax this so that opening is also allowed recursively if the client
already has the surface open.

This works around a regression in svga mesa where opening of a shared
surface is used recursively to obtain surface information.

Signed-off-by: Thomas Hellstrom 
Reviewed-by: Sinclair Yeh 
Signed-off-by: Greg Kroah-Hartman

KVM: kvm_io_bus_unregister_dev() should never fail

2017-04-08T07:53:32+00:00

commit 90db10434b163e46da413d34db8d0e77404cc645 upstream.

No caller currently checks the return value of
kvm_io_bus_unregister_dev(). This is evil, as all callers silently go on
freeing their device. A stale reference will remain in the io_bus,
getting at least used again, when the iobus gets teared down on
kvm_destroy_vm() - leading to use after free errors.

There is nothing the callers could do, except retrying over and over
again.

So let's simply remove the bus altogether, print an error and make
sure no one can access this broken bus again (returning -ENOMEM on any
attempt to access it).

Fixes: e93f8a0f821e ("KVM: convert io_bus to SRCU")
Reported-by: Dmitry Vyukov 
Reviewed-by: Cornelia Huck 
Signed-off-by: David Hildenbrand 
Signed-off-by: Paolo Bonzini 
Signed-off-by: Greg Kroah-Hartman

usb-core: Add LINEAR_FRAME_INTR_BINTERVAL USB quirk

2017-03-30T07:35:16+00:00

commit 3243367b209faed5c320a4e5f9a565ee2a2ba958 upstream.

Some USB 2.0 devices erroneously report millisecond values in
bInterval. The generic config code manages to catch most of them,
but in some cases it's not completely enough.

The case at stake here is a USB 2.0 braille device, which wants to
announce 10ms and thus sets bInterval to 10, but with the USB 2.0
computation that yields to 64ms.  It happens that one can type fast
enough to reach this interval and get the device buffers overflown,
leading to problematic latencies.  The generic config code does not
catch this case because the 64ms is considered a sane enough value.

This change thus adds a USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL quirk
to mark devices which actually report milliseconds in bInterval,
and marks Vario Ultra devices as needing it.

Signed-off-by: Samuel Thibault 
Acked-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

scsi: libiscsi: add lock around task lists to fix list corruption regression

2017-03-26T10:13:19+00:00

commit 6f8830f5bbab16e54f261de187f3df4644a5b977 upstream.

There's a rather long standing regression from the commit "libiscsi:
Reduce locking contention in fast path"

Depending on iSCSI target behavior, it's possible to hit the case in
iscsi_complete_task where the task is still on a pending list
(!list_empty(&task->running)).  When that happens the task is removed
from the list while holding the session back_lock, but other task list
modification occur under the frwd_lock.  That leads to linked list
corruption and eventually a panicked system.

Rather than back out the session lock split entirely, in order to try
and keep some of the performance gains this patch adds another lock to
maintain the task lists integrity.

Major enterprise supported kernels have been backing out the lock split
for while now, thanks to the efforts at IBM where a lab setup has the
most reliable reproducer I've seen on this issue.  This patch has been
tested there successfully.

Signed-off-by: Chris Leech 
Fixes: 659743b02c41 ("[SCSI] libiscsi: Reduce locking contention in fast path")
Reported-by: Prashantha Subbarao 
Reviewed-by: Guilherme G. Piccoli 
Signed-off-by: Martin K. Petersen 
Signed-off-by: Greg Kroah-Hartman

give up on gcc ilog2() constant optimizations

2017-03-26T10:13:18+00:00

commit 474c90156c8dcc2fa815e6716cc9394d7930cb9c upstream.

gcc-7 has an "optimization" pass that completely screws up, and
generates the code expansion for the (impossible) case of calling
ilog2() with a zero constant, even when the code gcc compiles does not
actually have a zero constant.

And we try to generate a compile-time error for anybody doing ilog2() on
a constant where that doesn't make sense (be it zero or negative).  So
now gcc7 will fail the build due to our sanity checking, because it
created that constant-zero case that didn't actually exist in the source
code.

There's a whole long discussion on the kernel mailing about how to work
around this gcc bug.  The gcc people themselevs have discussed their
"feature" in

   https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72785

but it's all water under the bridge, because while it looked at one
point like it would be solved by the time gcc7 was released, that was
not to be.

So now we have to deal with this compiler braindamage.

And the only simple approach seems to be to just delete the code that
tries to warn about bad uses of ilog2().

So now "ilog2()" will just return 0 not just for the value 1, but for
any non-positive value too.

It's not like I can recall anybody having ever actually tried to use
this function on any invalid value, but maybe the sanity check just
meant that such code never made it out in public.

Reported-by: Laura Abbott 
Cc: John Stultz ,
Cc: Thomas Gleixner 
Cc: Ard Biesheuvel 
Signed-off-by: Linus Torvalds 
Cc: Jiri Slaby 
Signed-off-by: Greg Kroah-Hartman