Special care should be taken when slow path is hit in ip_fragment() :
When walking through frags, we transfert truesize ownership from skb to
frags. Then if we hit a slow_path condition, we must undo this or risk
uncharging frags->truesize twice, and in the end, having negative socket
sk_wmem_alloc counter, or even freeing socket sooner than expected.
Many thanks to Nick Bowler, who provided a very clean bug report and
test program.
Thanks to Jarek for reviewing my first patch and providing a V2
While Nick bisection pointed to commit 2b85a34e911 (net: No more
expensive sock_hold()/sock_put() on each tx), underlying bug is older
(2.6.12-rc5)
A side effect is to extend work done in commit b2722b1c3a893e
(ip_fragment: also adjust skb->truesize for packets not owned by a
socket) to ipv6 as well.
Reported-and-bisected-by: Nick Bowler <nbowler@elliptictech.com> Tested-by: Nick Bowler <nbowler@elliptictech.com> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> CC: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Don't try to "optimize" rds_page_copy_user() by using kmap_atomic() and
the unsafe atomic user mode accessor functions. It's actually slower
than the straightforward code on any reasonable modern CPU.
Back when the code was written (although probably not by the time it was
actually merged, though), 32-bit x86 may have been the dominant
architecture. And there kmap_atomic() can be a lot faster than kmap()
(unless you have very good locality, in which case the virtual address
caching by kmap() can overcome all the downsides).
But these days, x86-64 may not be more populous, but it's getting there
(and if you care about performance, it's definitely already there -
you'd have upgraded your CPU's already in the last few years). And on
x86-64, the non-kmap_atomic() version is faster, simply because the code
is simpler and doesn't have the "re-try page fault" case.
People with old hardware are not likely to care about RDS anyway, and
the optimization for the 32-bit case is simply buggy, since it doesn't
verify the user addresses properly.
Reported-by: Dan Rosenberg <drosenberg@vsecurity.com> Acked-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This fixes possible cases of not collecting valid error info in
the MCE error thresholding groups on F10h hardware.
The current code contains a subtle problem of checking only the
Valid bit of MSR0000_0413 (which is MC4_MISC0 - DRAM
thresholding group) in its first iteration and breaking out if
the bit is cleared.
But (!), this MSR contains an offset value, BlkPtr[31:24], which
points to the remaining MSRs in this thresholding group which
might contain valid information too. But if we bail out only
after we checked the valid bit in the first MSR and not the
block pointer too, we miss that other information.
The thing is, MC4_MISC0[BlkPtr] is not predicated on
MCi_STATUS[MiscV] or MC4_MISC0[Valid] and should be checked
prior to iterating over the MCI_MISCj thresholding group,
irrespective of the MC4_MISC0[Valid] setting.
Before that commit, register_active_regions() is called for every SRAT memory
entry right away.
Use nodememblk_range[] instead of nodes[] in order to make sure we
capture the actual memory blocks registered with each node. nodes[]
contains an extended range which spans all memory regions associated
with a node, but that does not mean that all the memory in between are
included.
Reported-by: Russ Anderson <rja@sgi.com> Tested-by: Russ Anderson <rja@sgi.com> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <4CB27BDF.5000800@kernel.org> Acked-by: David Rientjes <rientjes@google.com> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
adapter->cmb.cmb is initialized when the device is opened and freed when
it's closed. Accessing it unconditionally during resume results either
in a crash (NULL pointer dereference, when the interface has not been
opened yet) or data corruption (when the interface has been used and
brought down adapter->cmb.cmb points to a deallocated memory area).
Signed-off-by: Luca Tettamanti <kronos.it@gmail.com> Acked-by: Chris Snook <chris.snook@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a driver doesn't fill the entire buffer, old
heap contents may remain, and if it also doesn't
update the length properly, this old heap content
will be copied back to userspace.
It is very unlikely that this happens in any of
the drivers using private ioctls since it would
show up as junk being reported by iwpriv, but it
seems better to be safe here, so use kzalloc.
Reported-by: Jeff Mahoney <jeffm@suse.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ocfs2 fast symlinks are NUL terminated strings stored inline in the
inode data area. However, disk corruption or a local attacker could, in
theory, remove that NUL. Because we're using strlen() (my fault,
introduced in a731d1 when removing vfs_follow_link()), we could walk off
the end of that string.
Signed-off-by: Joel Becker <joel.becker@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Prevent from recursively locking the reiserfs lock in reiserfs_unpack()
because we may call journal_begin() that requires the lock to be taken
only once, otherwise it won't be able to release the lock while taking
other mutexes, ending up in inverted dependencies between the journal
mutex and the reiserfs lock for example.
This fixes:
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.35.4.4a #3
-------------------------------------------------------
lilo/1620 is trying to acquire lock:
(&journal->j_mutex){+.+...}, at: [<d0325bff>] do_journal_begin_r+0x7f/0x340 [reiserfs]
but task is already holding lock:
(&REISERFS_SB(s)->lock){+.+.+.}, at: [<d032a278>] reiserfs_write_lock+0x28/0x40 [reiserfs]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
The reiserfs mutex already depends on the inode mutex, so we can't lock
the inode mutex in reiserfs_unpack() without using the safe locking API,
because reiserfs_unpack() is always called with the reiserfs mutex locked.
This fixes:
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.35c #13
-------------------------------------------------------
lilo/1606 is trying to acquire lock:
(&sb->s_type->i_mutex_key#8){+.+.+.}, at: [<d0329450>] reiserfs_unpack+0x60/0x110 [reiserfs]
but task is already holding lock:
(&REISERFS_SB(s)->lock){+.+.+.}, at: [<d032a268>] reiserfs_write_lock+0x28/0x40 [reiserfs]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
CPU X CPU Y
remove_hrtimer
// state & QUEUED == 0
timer->state = CALLBACK
unlock timer base
timer->f(n) //very long
hrtimer_start
lock timer base
remove_hrtimer // no effect
hrtimer_enqueue
timer->state = CALLBACK |
QUEUED
unlock timer base
hrtimer_start
lock timer base
remove_hrtimer
mode = INACTIVE
// CALLBACK bit lost!
switch_hrtimer_base
CALLBACK bit not set:
timer->base
changes to a
different CPU.
lock this CPU's timer base
The bug was introduced with commit ca109491f (hrtimer: removing all ur
callback modes) in 2.6.29
[ tglx: Feed new state via local variable and add a comment. ]
Signed-off-by: Salman Qazi <sqazi@google.com> Cc: akpm@linux-foundation.org Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <20101012142351.8485.21823.stgit@dungbeetle.mtv.corp.google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When using simultaneously the two DMA channels on a same engine, some
transfers are never completed. For example, an endless lock can occur
while writing heavily on a RAID5 array (with async-tx offload support
enabled).
Note that this issue can also be reproduced by using the DMA test
client.
On a same engine, the interrupt cause register is shared between two
DMA channels. This patch make sure that the cause bit is only cleared
for the requested channel.
Signed-off-by: Simon Guinot <sguinot@lacie.com> Tested-by: Luc Saillard <luc@saillard.org> Acked-by: saeed bishara <saeed.bishara@gmail.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Time stamps for the ring buffer are created by the difference between
two events. Each page of the ring buffer holds a full 64 bit timestamp.
Each event has a 27 bit delta stamp from the last event. The unit of time
is nanoseconds, so 27 bits can hold ~134 milliseconds. If two events
happen more than 134 milliseconds apart, a time extend is inserted
to add more bits for the delta. The time extend has 59 bits, which
is good for ~18 years.
Currently the time extend is committed separately from the event.
If an event is discarded before it is committed, due to filtering,
the time extend still exists. If all events are being filtered, then
after ~134 milliseconds a new time extend will be added to the buffer.
This can only happen till the end of the page. Since each page holds
a full timestamp, there is no reason to add a time extend to the
beginning of a page. Time extends can only fill a page that has actual
data at the beginning, so there is no fear that time extends will fill
more than a page without any data.
When reading an event, a loop is made to skip over time extends
since they are only used to maintain the time stamp and are never
given to the caller. As a paranoid check to prevent the loop running
forever, with the knowledge that time extends may only fill a page,
a check is made that tests the iteration of the loop, and if the
iteration is more than the number of time extends that can fit in a page
a warning is printed and the ring buffer is disabled (all of ftrace
is also disabled with it).
There is another event type that is called a TIMESTAMP which can
hold 64 bits of data in the theoretical case that two events happen
18 years apart. This code has not been implemented, but the name
of this event exists, as well as the structure for it. The
size of a TIMESTAMP is 16 bytes, where as a time extend is only
8 bytes. The macro used to calculate how many time extends can fit on
a page used the TIMESTAMP size instead of the time extend size
cutting the amount in half.
The following test case can easily trigger the warning since we only
need to have half the page filled with time extends to trigger the
warning:
Enabling the function tracer and then setting the filter to only trace
functions where the process id is negative (no events), then clearing
the trace buffer to ensure that we have nothing in the buffer,
then write to trace_marker to add an event to the beginning of a page,
sleep for 2 minutes (only 35 seconds is probably needed, but this
guarantees the bug), and then finally reading the trace which will
trigger the bug.
This patch fixes the typo and prevents the false positive of that warning.
Reported-by: Hans J. Koch <hjk@linutronix.de> Tested-by: Hans J. Koch <hjk@linutronix.de> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit f81f2f7c (ubd: drop unnecessary rq->sector manipulation)
dropped request->sector manipulation in preparation for global request
handling cleanup; unfortunately, it incorrectly assumed that the
updated sector wasn't being used.
ubd tries to issue as many requests as possible to io_thread. When
issuing fails due to memory pressure or other reasons, the device is
put on the restart list and issuing stops. On IO completion, devices
on the restart list are scanned and IO issuing is restarted.
ubd issues IOs sg-by-sg and issuing can be stopped in the middle of a
request, so each device on the restart queue needs to remember where
to restart in its current request. ubd needs to keep track of the
issue position itself because,
* blk_rq_pos(req) is now updated by the block layer to keep track of
_completion_ position.
* Multiple io_req's for the current request may be in flight, so it's
difficult to tell where blk_rq_pos(req) currently is.
Add ubd->rq_pos to keep track of the issue position and use it to
correctly restart io_req issue.
Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Richard Weinberger <richard@nod.at> Tested-by: Richard Weinberger <richard@nod.at> Tested-by: Chris Frey <cdfrey@foursquare.net> Signed-off-by: Jens Axboe <jaxboe@fusionio.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
free_irq_cfg() is not freeing the cpumask_vars in irq_cfg. Fixing this
triggers a use after free caused by the fact that copying struct
irq_cfg is done with memcpy, which copies the pointer not the cpumask.
Fix both places.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Yinghai Lu <yhlu.kernel@gmail.com>
LKML-Reference: <alpine.LFD.2.00.1009282052570.2416@localhost6.localdomain6> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
moved the CMTR flag into the skb's status, and
in doing so introduced a use-after-free -- when
the skb has been handed to cooked monitors the
status setting will touch now invalid memory.
Additionally, moving it there has effectively
discarded the optimisation -- since the bit is
only ever set on freed SKBs, and those were a
copy, it could never be checked.
For the current release, fixing this properly
is a bit too involved, so let's just remove the
problematic code and leave userspace with one
copy of each frame for each virtual interface.
Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fixed JSIOCSAXMAP ioctl to update absmap, the map from hardware axis to
event axis in addition to abspam. This fixes a regression introduced
by 999b874f.
The compat code for the VIDIOCSMICROCODE ioctl is totally buggered.
It's only used by the VIDEO_STRADIS driver, and that one is scheduled to
staging and eventually removed unless somebody steps up to maintain it
(at which point it should use request_firmware() rather than some magic
ioctl). So we'll get rid of it eventually.
But in the meantime, the compatibility ioctl code is broken, and this
tries to get it to at least limp along (even if Mauro suggested just
deleting it entirely, which may be the right thing to do - I don't think
the compatibility translation code has ever worked unless you were very
lucky).
The guest can use the paravirt clock in kvmclock.c which is used
by sched_clock(), which in turn is used by the tracing mechanism
for timestamps, which leads to infinite recursion.
Disable mcount/tracing for kvmclock.o.
Cc: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When using a paravirt clock, pvclock.c can be used by sched_clock(),
which in turn is used by the tracing mechanism for timestamps,
which leads to infinite recursion.
Disable mcount/tracing for pvclock.o.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
LKML-Reference: <4C9A9A3F.4040201@goop.org> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch adds a workaround for an IOMMU BIOS problem to
the AMD IOMMU driver. The result of the bug is that the
IOMMU does not execute commands anymore when the system
comes out of the S3 state resulting in system failure. The
bug in the BIOS is that is does not restore certain hardware
specific registers correctly. This workaround reads out the
contents of these registers at boot time and restores them
on resume from S3. The workaround is limited to the specific
IOMMU chipset where this problem occurs.
In the __unmap_single function the dma_addr is rounded down
to a page boundary before the dma pages are unmapped. The
address is later also used to flush the TLB entries for that
mapping. But without the offset into the dma page the amount
of pages to flush might be miscalculated in the TLB flushing
path. This patch fixes this bug by using the original
address to flush the TLB.
This patch moves the setting of the configuration and
feature flags out out the acpi table parsing path and moves
it into the iommu-enable path. This is needed to reliably
fix resume-from-s3.
If not all clocks have been defined in platform data, the driver will
cause a null pointer dereference when it is removed. This patch fixes
this issue.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Chris Ball <cjb@laptop.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The HW by default has RX coalescing on. For iWARP connections, this
causes a 100ms delay in connection establishement due to the ingress
MPA Start message being stalled in HW. So explicitly turn RX
coalescing off when setting up iWARP connections.
This was causing very bad performance for NP64 gather operations using
Open MPI, due to the way it sets up connections on larger jobs.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Recent changes in the usbhid layer exposed a bug in usbcore. If
CONFIG_USB_DYNAMIC_MINORS is enabled then an interface may be assigned
a minor number of 0. However interfaces that aren't registered as USB
class devices also have their minor number set to 0, during
initialization. As a result usb_find_interface() may return the
wrong interface, leading to a crash.
This patch (as1418) fixes the problem by initializing every
interface's minor number to -1. It also cleans up the
usb_register_dev() function, which besides being somewhat awkwardly
written, does not unwind completely on all its error paths.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Tested-by: Philip J. Turmel <philip@turmel.org> Tested-by: Gabriel Craciunescu <nix.or.die@googlemail.com> Tested-by: Alex Riesen <raa.lkml@gmail.com> Tested-by: Matthias Bayer <jackdachef@gmail.com> CC: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a driver module is unloaded and the last still open file is a raw
MIDI device, the card and its devices will be actually freed in the
snd_card_file_remove() call when that file is closed. Afterwards, rmidi
and rmidi->card point into freed memory, so the module pointer is likely
to be garbage.
(This was introduced by commit 9a1b64caac82aa02cb74587ffc798e6f42c6170a.)
Signed-off-by: Clemens Ladisch <clemens@ladisch.de> Reported-by: Krzysztof Foltman <wdev@foltman.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The snd_ctl_new() function in sound/core/control.c allocates space for a
snd_kcontrol struct by performing arithmetic operations on a
user-provided size without checking for integer overflow. If a user
provides a large enough size, an overflow will occur, the allocated
chunk will be too small, and a second user-influenced value will be
written repeatedly past the bounds of this chunk. This code is
reachable by unprivileged users who have permission to open
a /dev/snd/controlC* device (on many distros, this is group "audio") via
the SNDRV_CTL_IOCTL_ELEM_ADD and SNDRV_CTL_IOCTL_ELEM_REPLACE ioctls.
On the HT-Omega Claro halo card, the ADC data must be captured from the
second I2S input. Using the default first input, which isn't connected
to anything, would result in silence.
Signed-off-by: Erik J. Staab <ejs@insightbb.com> Signed-off-by: Clemens Ladisch <clemens@ladisch.de> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The SNDRV_HDSP_IOCTL_GET_CONFIG_INFO and
SNDRV_HDSP_IOCTL_GET_CONFIG_INFO ioctls in hdspm.c and hdsp.c allow
unprivileged users to read uninitialized kernel stack memory, because
several fields of the hdsp{m}_config_info structs declared on the stack
are not altered or zeroed before being copied back to the user. This
patch takes care of it.
For some unknown reason, on a MacBookPro5,3 the iSight sometimes report
a different video format GUID. This patch add the other (wrong) GUID to
the format table, making the iSight work always w/o other problems.
ath5k assumes ah_current_channel is always a valid pointer in
several places, but a newly created interface may not have a
channel. To avoid null pointer dereferences, set it up to point
to the first available channel until later reconfigured.
This fixes the following oops:
$ rmmod ath5k
$ insmod ath5k
$ iw phy0 set distance 11000
Reported-by: Steve Brown <sbrown@cortland.com> Signed-off-by: Bob Copeland <me@bobcopeland.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The function ecryptfs_uid_hash wrongly assumes that the
second parameter to hash_long() is the number of hash
buckets instead of the number of hash bits.
This patch fixes that and renames the variable
ecryptfs_hash_buckets to ecryptfs_hash_bits to make it
clearer.
Typo in down_spin() meant it only read the low 32 bits of the
"serve" value, instead of the full 64 bits. This results in the
system hanging when the values in ticket/serve get larger than
32-bits. A big enough system running the right test can hit this
in a just a few hours.
> program: /home/alan/GitTrees/linux-2.6-mid-ref/scripts/mod/modpost -o
> Module.symvers -S vmlinux.o
>
> Program received signal SIGSEGV, Segmentation fault.
It just hit me.
It's the offset calculation in reloc_location() which overflows:
return (void *)elf->hdr + sechdrs[section].sh_offset +
(r->r_offset - sechdrs[section].sh_addr);
E.g. for the first rodata r entry:
r->r_offset < sechdrs[section].sh_addr
and the expression in the parenthesis produces 0xFFFFFFE0 or something
equally wise.
Reported-by: Alan <alan@clueserver.org> Signed-off-by: Krzysztof Hałasa <khc@pm.waw.pl> Tested-by: Alan <alan@clueserver.org> Signed-off-by: Michal Marek <mmarek@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The flow id (scd_flow) in a compressed BA packet should match the txq_id
of the queue from which the aggregated packets were sent. However, in
some hardware like the 1000 series, sometimes the flow id is 0 for the
txq_id (10 to 19). This can cause the annoying message:
[ 2213.306191] iwlagn 0000:01:00.0: Received BA when not expected
[ 2213.310178] iwlagn 0000:01:00.0: Read index for DMA queue txq id (0),
index 5, is out of range [0-256] 7 7.
And even worse, if agg->wait_for_ba is true when the bad BA is arriving,
this can cause system hang due to NULL pointer dereference because the
code is operating in a wrong tx queue!
I encountered an issue that not to link up on cxgb3 fabric.
I bisected and found that this regression was introduced by 0f07c4ee8c800923ae7918c231532a9256233eed.
Correct to pass phy_addr to cphy_init() at t3_xaui_direct_phy_prep().
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com> Acked-by: Divy Le Ray <divy@chelsio.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The driver attempts to select an IRQ for the NIC automatically by
testing which of the supported IRQs are available and then probing
each available IRQ with probe_irq_{on,off}(). There are obvious race
conditions here, besides which:
1. The test for availability is done by passing a NULL handler, which
now always returns -EINVAL, thus the device cannot be opened:
<http://bugs.debian.org/566522>
2. probe_irq_off() will report only the first ISA IRQ handled,
potentially leading to a false negative.
There was another bug that meant it ignored all error codes from
request_irq() except -EBUSY, so it would 'succeed' despite this
(possibly causing conflicts with other ISA devices). This was fixed
by ab08999d6029bb2c79c16be5405d63d2bedbdfea 'WARNING: some
request_irq() failures ignored in el2_open()', which exposed bug 1.
This patch:
1. Replaces the use of probe_irq_{on,off}() with a real interrupt handler
2. Adds a delay before checking the interrupt-seen flag
3. Disables interrupts on all failure paths
4. Distinguishes error codes from the second request_irq() call,
consistently with the first
Compile-tested only.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In the FPU emulator code of the MIPS, the Cause bits of the FCSR register
are not currently writeable by the ctc1 instruction. In odd corner cases,
this can cause problems. For example, a case existed where a divide-by-zero
exception was generated by the FPU, and the signal handler attempted to
restore the FPU registers to their state before the exception occurred. In
this particular setup, writing the old value to the FCSR register would
cause another divide-by-zero exception to occur immediately. The solution
is to change the ctc1 instruction emulator code to allow the Cause bits of
the FCSR register to be writeable. This is the behaviour of the hardware
that the code is emulating.
This problem was found by Shane McDonald, but the credit for the fix goes
to Kevin Kissell. In Kevin's words:
I submit that the bug is indeed in that ctc_op: case of the emulator. The
Cause bits (17:12) are supposed to be writable by that instruction, but the
CTC1 emulation won't let them be updated by the instruction. I think that
actually if you just completely removed lines 387-388 [...] things would
work a good deal better. At least, it would be a more accurate emulation of
the architecturally defined FPU. If I wanted to be really, really pedantic
(which I sometimes do), I'd also protect the reserved bits that aren't
necessarily writable.
Signed-off-by: Shane McDonald <mcdonald.shane@gmail.com>
To: anemo@mba.ocn.ne.jp
To: kevink@paralogos.com
To: sshtylyov@mvista.com
Patchwork: http://patchwork.linux-mips.org/patch/1205/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Please find attached a patch which adds the device ID for the Belkin F5D8053 v6 to the rtl8192su driver. I've tested this in 2.6.34-rc3
(Ubuntu 9.10 amd64) and the network adapter is working flawlessly.
Signed-off-by: Richard Airlie <richard@backtrace.co.uk> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch fixes the TX_LIMIT feature flag. The previous logic check
for TX_LIMIT2 also took into account a device that only had TX_LIMIT
set.
Reported-by: Stephen Mulcahu <stephen.mulcahy@deri.org> Reported-by: Ben Huchings <ben@decadent.org.uk> Signed-off-by: Ayaz Abdulla <aabdulla@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
IR support on FusionHDTV cards is broken since kernel 2.6.31. One side
effect of the switch to the standard binding model for IR I2C devices
was to let i2c-core do the probing instead of the ir-kbd-i2c driver.
There is a slight difference between the two probe methods: i2c-core
uses 0-byte writes, while the ir-kbd-i2c was using 0-byte reads. As
some IR I2C devices only support reads, the new probe method fails to
detect them.
For now, revert to letting the driver do the probe, using 0-byte
reads. In the future, i2c-core will be extended to let callers of
i2c_new_probed_device() provide a custom probing function.
Signed-off-by: Jean Delvare <khali@linux-fr.org> Tested-by: "Timothy D. Lenz" <tlenz@vorgon.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix ULE decapsulation bug when less than 4 bytes of ULE SNDU is packed
into the remaining bytes of a MPEG2-TS frame
ULE (Unidirectional Lightweight Encapsulation RFC 4326) decapsulation
code has a bug that incorrectly treats ULE SNDU packed into the
remaining 2 or 3 bytes of a MPEG2-TS frame as having invalid pointer
field on the subsequent MPEG2-TS frame.
Signed-off-by: Ang Way Chuang <wcang@nav6.org> Acked-by: Jarod Wilson <jarod@redhat.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch fixes a division by zero error in the irq handler.
There is a small window between the hw_params() callback and when
runtime->frame_bits is set by ALSA middle layer. When another substream is
already running, if an interrupt is delivered during that window the irq
handler calls pcm_pointer() which does a division by zero. The patch below
makes the irq handler skip substreams that are initialized but not started
yet. Cc to Clemens Ladisch because he proposed an alternate fix.
For more information, please read the original thread in the linux-kernel
mailing list: http://lkml.org/lkml/2010/2/2/187
395913d0b1db37092ea3d9d69b832183b1dd84c5 ("[CPUFREQ] remove rwsem lock
from CPUFREQ_GOV_STOP call (second call site)") is not needed, because
there is no rwsem lock in cpufreq_ondemand and cpufreq_conservative
anymore. Lock should not be released until the work done.
fix memory leak introduced by the patch 6e03a201bbe:
firmware: speed up request_firmware()
1. vfree won't release pages there were allocated explicitly and mapped
using vmap. The memory has to be vunmap-ed and the pages needs
to be freed explicitly
2. page array is moved into the 'struct
firmware' so that we can free it from release_firmware()
and not only in fw_dev_release()
The fix doesn't break the firmware load speed.
Cc: Johannes Berg <johannes@sipsolutions.net> Cc: Ming Lei <tom.leiming@gmail.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Singed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: David Woodhouse <David.Woodhouse@intel.com> Signed-off-by: Tomas Winkler <tomas.winkler@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
dm9000_set_rx_csum and dm9000_hash_table are called from atomic context (in
dm9000_init_dm9000), and from non-atomic context (via ethtool_ops and
net_device_ops respectively). This causes a spinlock recursion BUG. Fix this by
renaming these functions to *_unlocked for the atomic context, and make the
original functions locking wrappers for use in the non-atomic context.
Signed-off-by: Baruch Siach <baruch@tkos.co.il> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When operating in 1-bit mode, SDAT1 is used as dedicated interrupt line.
However, the 8686 will only drive this line when the ECSI bit is set in
the CCCR_IF register.
Thanks to Alagu Sankar for pointing me in the right direction.
Signed-off-by: Daniel Mack <daniel@caiaq.de> Cc: Alagu Sankar <alagusankar@embwise.com> Cc: Volker Ernst <volker.ernst@txtr.com> Cc: Dan Williams <dcbw@redhat.com> Cc: John W. Linville <linville@tuxdriver.com> Cc: Holger Schurig <hs4233@mail.mn-solutions.de> Cc: Bing Zhao <bzhao@marvell.com> Cc: libertas-dev@lists.infradead.org Cc: linux-wireless@vger.kernel.org Cc: linux-mmc@vger.kernel.org Acked-by: Dan Williams <dcbw@redhat.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The ACPI spec tells us that the firmware will reenable SCI_EN on resume.
Reality disagrees in some cases. The ACPI spec tells us that the only way
to set SCI_EN is via an SMM call.
https://bugzilla.kernel.org/show_bug.cgi?id=13745 shows us that doing so
may break machines. Tracing the ACPI calls made by Windows shows that it
unconditionally sets SCI_EN on resume with a direct register write, and
therefore the overwhelming probability is that everything is fine with
this behaviour.
Signed-off-by: Matthew Garrett <mjg@redhat.com> Tested-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com> Cc: Kamal Mostafa <kamal@canonical.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It turns out that there is a bit in the _CST for Intel FFH C3
that tells the OS if we should be checking BM_STS or not.
Linux has been unconditionally checking BM_STS.
If the chip-set is configured to enable BM_STS,
it can retard or completely prevent entry into
deep C-states -- as illustrated by turbostat:
ref: Intel Processor Vendor-Specific ACPI Interface Specification
table 4 "_CST FFH GAS Field Encoding"
Bit 1: Set to 1 if OSPM should use Bus Master avoidance for this C-state
https://bugzilla.kernel.org/show_bug.cgi?id=15886
Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Save/restore MISC_ENABLE register on suspend/resume.
This fixes OOPS (invalid opcode) on resume from STR on Asus P4P800-VM,
which wakes up with MWAIT disabled.
Signed-off-by: Ondrej Zary <linux@rainbow-software.org> Tested-by: Alan Stern <stern@rowland.harvard.edu> Acked-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
futex_find_get_task is currently used (through lookup_pi_state) from two
contexts, futex_requeue and futex_lock_pi_atomic. None of the paths
looks it needs the credentials check, though. Different (e)uids
shouldn't matter at all because the only thing that is important for
shared futex is the accessibility of the shared memory.
The credentail check results in glibc assert failure or process hang (if
glibc is compiled without assert support) for shared robust pthread
mutex with priority inheritance if a process tries to lock already held
lock owned by a process with a different euid:
The problem is that futex_lock_pi_atomic which is called when we try to
lock already held lock checks the current holder (tid is stored in the
futex value) to get the PI state. It uses lookup_pi_state which in turn
gets task struct from futex_find_get_task. ESRCH is returned either
when the task is not found or if credentials check fails.
futex_lock_pi_atomic simply returns if it gets ESRCH. glibc code,
however, doesn't expect that robust lock returns with ESRCH because it
should get either success or owner died.
Signed-off-by: Michal Hocko <mhocko@suse.cz> Acked-by: Darren Hart <dvhltc@us.ibm.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Nick Piggin <npiggin@suse.de> Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Marcelo Tosatti [Fri, 28 May 2010 12:44:59 +0000 (09:44 -0300)]
KVM: MMU: invalidate and flush on spte small->large page size change
Always invalidate spte and flush TLBs when changing page size, to make
sure different sized translations for the same address are never cached
in a CPU's TLB.
Currently the only case where this occurs is when a non-leaf spte pointer is
overwritten by a leaf, large spte entry. This can happen after dirty
logging is disabled on a memslot, for example.
Noticed by Andrea.
KVM-Stable-Tag Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
(cherry picked from commit 3be2264be3c00865116f997dc53ebcc90fe7fc4b)
Joerg Roedel [Mon, 17 May 2010 12:43:35 +0000 (14:43 +0200)]
KVM: SVM: Implement workaround for Erratum 383
This patch implements a workaround for AMD erratum 383 into
KVM. Without this erratum fix it is possible for a guest to
kill the host machine. This patch implements the suggested
workaround for hypervisors which will be published by the
next revision guide update.
Joerg Roedel [Mon, 17 May 2010 12:43:34 +0000 (14:43 +0200)]
KVM: SVM: Handle MCEs early in the vmexit process
This patch moves handling of the MC vmexits to an earlier
point in the vmexit. The handle_exit function is too late
because the vcpu might alreadry have changed its physical
cpu.
Cc: stable@kernel.org Signed-off-by: Joerg Roedel <joerg.roedel@amd.com> Signed-off-by: Avi Kivity <avi@redhat.com>
(cherry picked from commit fe5913e4e1700cbfc337f4b1da9ddb26f6a55586)
Avi Kivity [Thu, 27 May 2010 11:35:58 +0000 (14:35 +0300)]
KVM: MMU: Remove user access when allowing kernel access to gpte.w=0 page
If cr0.wp=0, we have to allow the guest kernel access to a page with pte.w=0.
We do that by setting spte.w=1, since the host cr0.wp must remain set so the
host can write protect pages. Once we allow write access, we must remove
user access otherwise we mistakenly allow the user to write the page.
Reviewed-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Avi Kivity <avi@redhat.com>
(cherry picked from commit 69325a122580d3a7b26589e8efdd6663001c3297)
struct ethtool_rxnfc was originally defined in 2.6.27 for the
ETHTOOL_{G,S}RXFH command with only the cmd, flow_type and data
fields. It was then extended in 2.6.30 to support various additional
commands. These commands should have been defined to use a new
structure, but it is too late to change that now.
Since user-space may still be using the old structure definition
for the ETHTOOL_{G,S}RXFH commands, and since they do not need the
additional fields, only copy the originally defined fields to and
from user-space.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an attempt is made to read the interface strings of the Artisman
Watchdog USB dongle (idVendor:idProduct 04b4:0526) an error is written
to the dmesg log (uhci_result_common: failed with status 440000) and the
dongle resets itself, resulting in a disconnect/reconnect loop.
Adding the dongle to the list of devices in quirks.c, with the same
quirk Alan Stern's previous patch for the Saitek Cyborg Gold 3D
joystick, stops the device from resetting and allows it to be used with
no problems.
Signed-off-by: Paul Mortier <mortier@btinternet.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1403) is a partial reversion of an earlier change
(commit 5f677f1d45b2bf08085bbba7394392dfa586fa8e "USB: fix remote
wakeup settings during system sleep"). After hearing from a user, I
realized that remote wakeup should be enabled during system sleep
whenever userspace allows it, and not only if a driver requests it
too.
Indeed, there could be a device with no driver, that does nothing but
generate a wakeup request when the user presses a button. Such a
device should be allowed to do its job.
The problem fixed by the earlier patch -- device generating a wakeup
request for no reason, causing system suspend to abort -- was also
addressed by a later patch ("USB: don't enable remote wakeup by
default", accepted but not yet merged into mainline). The device
won't be able to generate the bogus wakeup requests because it will be
disabled for remote wakeup by default. Hence this reversion will not
re-introduce any old problems.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
hpet_disable is called unconditionally on machine reboot if hpet support
is compiled in the kernel.
hpet_disable only checks if the machine is hpet capable but doesn't make
sure that hpet has been initialized.
[ tglx: Made it a one liner and removed the redundant hpet_address check ]
Signed-off-by: Bin Yang <bin.yang@marvell.com> Signed-off-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
RealView boards with certain revisions of the L220 cache controller (ARM11*
processors only) may have issues (hardware deadlock) with the recent changes to
the mb() barrier implementation (DSB followed by an L2 cache sync). The patch
redefines the RealView ARM11MPCore mandatory barriers without the outer_sync()
call.
Tested-by: Linus Walleij <linus.walleij@stericsson.com> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Nokia RX51 board code (arch/arm/mach-omap2/board-rx51-peripherals.c)
defines a key map for the matrix keypad keyboard. The hardware seems to
use all of the 8 rows and 8 columns of the keypad, although not all
possible locations are used.
The TWL4030 supports keypads with at most 8 rows and 8 columns. Most keys
are defined with a row and column number between 0 and 7, except
which represent keycodes that should be emitted when entire row is
connected to the ground. since the driver handles this case as if we
had an extra column in the key matrix. Unfortunately we do not allocate
enough space and end up owerwriting some random memory.
Gigabyte "Spring Peak" notebook indicates wrong chassis-type, tripping up
i8042 and breaking the touchpad. Add this model to i8042_dmi_noloop_table[]
to resolve.
Sumeet Lahorani <sumeet.lahorani@oracle.com> reported that the IPoIB
child entries are world-writable; however we don't want ordinary users
to be able to create and destroy child interfaces, so fix them to be
writable only by root.
Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Found one x2apic system kexec loop test failed
when CONFIG_NMI_WATCHDOG=y (old) or CONFIG_LOCKUP_DETECTOR=y (current tip)
first kernel can kexec second kernel, but second kernel can not kexec third one.
it can be duplicated on another system with BIOS preenabled x2apic.
First kernel can not kexec second kernel.
It turns out, when kernel boot with pre-enabled x2apic, it will not execute
disable_local_APIC on shutdown path.
when init_apic_mappings() is called in setup_arch, it will skip setting of
apic_phys when x2apic_mode is set. ( x2apic_mode is much early check_x2apic())
Then later, disable_local_APIC() will bail out early because !apic_phys.
So check !x2apic_mode in x2apic_mode in disable_local_APIC with !apic_phys.
another solution could be updating init_apic_mappings() to set apic_phys even
for preenabled x2apic system. Actually even for x2apic system, that lapic
address is mapped already in early stage.
BTW: is there any x2apic preenabled system with apicid of boot cpu > 255?
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <4C3EB22B.3000701@kernel.org> Acked-by: Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since commit 5753c082f66eca5be81f6bda85c1718c5eea6ada ("powerpc/85xx:
Kconfig cleanup"), there is no MPC85xx Kconfig symbol anymore, so the
driver became non-selectable.
This patch fixes the issue by switching to PPC_85xx symbol.
Signed-off-by: Anton Vorontsov <avorontsov@mvista.com> Cc: Doug Thompson <dougthompson@xmission.com> Cc: Peter Tyser <ptyser@xes-inc.com> Cc: Dave Jiang <djiang@mvista.com> Cc: Kumar Gala <galak@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
System will crash sooner or later once the memory with the code of the
s3c-sdhci.ko module is reused for something else. I really have no idea
how the lack of remove function went unnoticed into the mainline code.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Warnings are treated as errors for arch/powerpc code, so build fails
with CONFIG_I2C_SPI_UCODE_PATCH=y:
CC arch/powerpc/sysdev/micropatch.o
cc1: warnings being treated as errors
arch/powerpc/sysdev/micropatch.c: In function 'cpm_load_patch':
arch/powerpc/sysdev/micropatch.c:630: warning: unused variable 'smp'
make[1]: *** [arch/powerpc/sysdev/micropatch.o] Error 1
And with CONFIG_USB_SOF_UCODE_PATCH=y:
CC arch/powerpc/sysdev/micropatch.o
cc1: warnings being treated as errors
arch/powerpc/sysdev/micropatch.c: In function 'cpm_load_patch':
arch/powerpc/sysdev/micropatch.c:629: warning: unused variable 'spp'
arch/powerpc/sysdev/micropatch.c:628: warning: unused variable 'iip'
make[1]: *** [arch/powerpc/sysdev/micropatch.o] Error 1
This patch fixes these issues by introducing proper #ifdefs.
Signed-off-by: Anton Vorontsov <avorontsov@mvista.com> Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
spi_t was removed in commit 644b2a680ccc51a9ec4d6beb12e9d47d2dee98e2
("powerpc/cpm: Remove SPI defines and spi structs"), the commit assumed
that spi_t isn't used anywhere outside of the spi_mpc8xxx driver. But
it appears that the struct is needed for micropatch code. So, let's
reintroduce the struct.
Fixes the following build issue:
CC arch/powerpc/sysdev/micropatch.o
micropatch.c: In function 'cpm_load_patch':
micropatch.c:629: error: expected '=', ',', ';', 'asm' or '__attribute__' before '*' token
micropatch.c:629: error: 'spp' undeclared (first use in this function)
micropatch.c:629: error: (Each undeclared identifier is reported only once
micropatch.c:629: error: for each function it appears in.)
Reported-by: LEROY Christophe <christophe.leroy@c-s.fr> Reported-by: Tony Breeds <tony@bakeyournoodle.com> Signed-off-by: Anton Vorontsov <avorontsov@mvista.com> Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When SPARSE_IRQ is set, irq_to_desc() can
return NULL. While the code here has a
check for NULL, it's not really correct.
Fix it by separating the check for it.
This fixes CPU hot unplug for me.
Reported-by: Alastair Bridgewater <alastair.bridgewater@gmail.com> Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
On a 32-bit machine, info.rule_cnt >= 0x40000000 leads to integer
overflow and the buffer may be smaller than needed. Since
ETHTOOL_GRXCLSRLALL is unprivileged, this can presumably be used for at
least denial of service.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
For yet unknown reason, MCP89 on MBP 7,1 doesn't work w/ ahci under
linux but the controller doesn't require explicit mode setting and
works fine with ata_generic. Make ahci ignore the controller on MBP
7,1 and let ata_generic take it for now.
Reported in bko#15923.
https://bugzilla.kernel.org/show_bug.cgi?id=15923
NVIDIA is investigating why ahci mode doesn't work.
The ds1307 driver misreads the ds1388 registers when checking for 12 or 24
hour mode. Instead of checking the hour register it reads the minute
register. Therefore the driver thinks minutes >= 40 has the 12HR bit set
and resets the minute register by zeroing the high bits. This results in
minutes are reset to 0-9, jumping back in time 40 or 50 minutes. The time
jump is also written back to the RTC.
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se> Cc: Wan ZongShun <mcuos.com@gmail.com> Cc: Alessandro Zummo <a.zummo@towertech.it> Cc: Paul Gortmaker <p_gortmaker@yahoo.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A user reported a kernel bug when running a particular program that did
the following:
created 32 threads
- each thread took a mutex, grabbed a global offset, added a buffer size
to that offset, released the lock
- read from the given offset in the file
- created a new thread to do the same
- exited
The result is that cfq's close cooperator logic would trigger, as the
threads were issuing I/O within the mean seek distance of one another.
This workload managed to routinely trigger a use after free bug when
walking the list of merge candidates for a particular cfqq
(cfqq->new_cfqq). The logic used for merging queues looks like this:
/* Avoid a circular list and skip interim queue merges */
while ((__cfqq = new_cfqq->new_cfqq)) {
if (__cfqq == cfqq)
return;
new_cfqq = __cfqq;
}
process_refs = cfqq_process_refs(cfqq);
/*
* If the process for the cfqq has gone away, there is no
* sense in merging the queues.
*/
if (process_refs == 0)
return;
/*
* Merge in the direction of the lesser amount of work.
*/
new_process_refs = cfqq_process_refs(new_cfqq);
if (new_process_refs >= process_refs) {
cfqq->new_cfqq = new_cfqq;
atomic_add(process_refs, &new_cfqq->ref);
} else {
new_cfqq->new_cfqq = cfqq;
atomic_add(new_process_refs, &cfqq->ref);
}
}
When a merge candidate is found, we add the process references for the
queue with less references to the queue with more. The actual merging
of queues happens when a new request is issued for a given cfqq. In the
case of the test program, it only does a single pread call to read in
1MB, so the actual merge never happens.
Normally, this is fine, as when the queue exits, we simply drop the
references we took on the other cfqqs in the merge chain:
/*
* If this queue was scheduled to merge with another queue, be
* sure to drop the reference taken on that queue (and others in
* the merge chain). See cfq_setup_merge and cfq_merge_cfqqs.
*/
__cfqq = cfqq->new_cfqq;
while (__cfqq) {
if (__cfqq == cfqq) {
WARN(1, "cfqq->new_cfqq loop detected\n");
break;
}
next = __cfqq->new_cfqq;
cfq_put_queue(__cfqq);
__cfqq = next;
}
However, there is a hole in this logic. Consider the following (and
keep in mind that each I/O keeps a reference to the cfqq):
q1->new_cfqq = q2 // q2 now has 2 process references
q3->new_cfqq = q2 // q2 now has 3 process references
// the process associated with q2 exits
// q2 now has 2 process references
// queue 1 exits, drops its reference on q2
// q2 now has 1 process reference
// q3 exits, so has 0 process references, and hence drops its references
// to q2, which leaves q2 also with 0 process references
q4 comes along and wants to merge with q3
q3->new_cfqq still points at q2! We follow that link and end up at an
already freed cfqq.
So, the fix is to not follow a merge chain if the top-most queue does
not have a process reference, otherwise any queue in the chain could be
already freed. I also changed the logic to disallow merging with a
queue that does not have any process references. Previously, we did
this check for one of the merge candidates, but not the other. That
doesn't really make sense.
Without the attached patch, my system would BUG within a couple of
seconds of running the reproducer program. With the patch applied, my
system ran the program for over an hour without issues.
This addresses the following bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=16217
Thanks a ton to Phil Carns for providing the bug report and an excellent
reproducer.
[ Note for stable: this applies to 2.6.32/33/34 ].
Signed-off-by: Jeff Moyer <jmoyer@redhat.com> Reported-by: Phil Carns <carns@mcs.anl.gov> Signed-off-by: Jens Axboe <jaxboe@fusionio.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The set_type() function can change the chip implementation when the
trigger mode changes. That might result in using an non-initialized
irq chip when called from __setup_irq() or when called via
set_irq_type() on an already enabled irq.
The set_irq_type() function should not be called on an enabled irq,
but because we forgot to put a check into it, we have a bunch of users
which grew the habit of doing that and it never blew up as the
function is serialized via desc->lock against all users of desc->chip
and they never hit the non-initialized irq chip issue.
The easy fix for the __setup_irq() issue would be to move the
irq_chip_set_defaults(desc->chip) call after the trigger setting to
make sure that a chip change is covered.
But as we have already users, which do the type setting after
request_irq(), the safe fix for now is to call irq_chip_set_defaults()
from __irq_set_trigger() when desc->set_type() changed the irq chip.
It needs a deeper analysis whether we should refuse to change the chip
on an already enabled irq, but that'd be a large scale change to fix
all the existing users. So that's neither stable nor 2.6.35 material.
Reported-by: Esben Haabendal <eha@doredevelopment.dk> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: linuxppc-dev <linuxppc-dev@ozlabs.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
projects / karo-tx-linux.git / log