snd_soc_dapm_sync takes the dapm_mutex internally, but we currently take
it externally as well. This patch fixes this.
Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 93e6f119c0ce ("ipc/mqueue: cleanup definition names and
locations") added global hardcoded limits to the amount of message
queues that can be created. While these limits are per-namespace,
reality is that it ends up breaking userspace applications.
Historically users have, at least in theory, been able to create up to
INT_MAX queues, and limiting it to just 1024 is way too low and dramatic
for some workloads and use cases. For instance, Madars reports:
"This update imposes bad limits on our multi-process application. As
our app uses approaches that each process opens its own set of queues
(usually something about 3-5 queues per process). In some scenarios
we might run up to 3000 processes or more (which of-course for linux
is not a problem). Thus we might need up to 9000 queues or more. All
processes run under one user."
Other affected users can be found in launchpad bug #1155695:
https://bugs.launchpad.net/ubuntu/+source/manpages/+bug/1155695
Instead of increasing this limit, revert it entirely and fallback to the
original way of dealing queue limits -- where once a user's resource
limit is reached, and all memory is used, new queues cannot be created.
Currently last dqput() can race with dquot_scan_active() causing it to
call callback for an already deactivated dquot. The race is as follows:
CPU1 CPU2
dqput()
spin_lock(&dq_list_lock);
if (atomic_read(&dquot->dq_count) > 1) {
- not taken
if (test_bit(DQ_ACTIVE_B, &dquot->dq_flags)) {
spin_unlock(&dq_list_lock);
->release_dquot(dquot);
if (atomic_read(&dquot->dq_count) > 1)
- not taken
dquot_scan_active()
spin_lock(&dq_list_lock);
if (!test_bit(DQ_ACTIVE_B, &dquot->dq_flags))
- not taken
atomic_inc(&dquot->dq_count);
spin_unlock(&dq_list_lock);
- proceeds to release dquot
ret = fn(dquot, priv);
- called for inactive dquot
Fix the problem by making sure possible ->release_dquot() is finished by
the time we call the callback and new calls to it will notice reference
dquot_scan_active() has taken and bail out.
Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since commit 77873803363c "net_dma: mark broken" we no longer pin dma
engines active for the network-receive-offload use case. As a result
the ->free_chan_resources() that occurs after the driver self test no
longer has a NET_DMA induced ->alloc_chan_resources() to back it up. A
late firing irq can lead to ksoftirqd spinning indefinitely due to the
tasklet_disable() performed by ->free_chan_resources(). Only
->alloc_chan_resources() can clear this condition in affected kernels.
This problem has been present since commit 3e037454bcfa "I/OAT: Add
support for MSI and MSI-X" in 2.6.24, but is now exposed. Given the
NET_DMA use case is deprecated we can revisit moving the driver to use
threaded irqs. For now, just tear down the irq and tasklet properly by:
Cc: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Reported-by: Mike Galbraith <bitbucket@online.de> Reported-by: Stanislav Fomichev <stfomichev@yandex-team.ru> Tested-by: Mike Galbraith <bitbucket@online.de> Tested-by: Stanislav Fomichev <stfomichev@yandex-team.ru> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When writing policy via /sys/fs/selinux/policy I wrote the type and class
of filename trans rules in CPU endian instead of little endian. On
x86_64 this works just fine, but it means that on big endian arch's like
ppc64 and s390 userspace reads the policy and converts it from
le32_to_cpu. So the values are all screwed up. Write the values in le
format like it should have been to start.
Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Stephen Smalley <sds@tycho.nsa.gov> Signed-off-by: Paul Moore <pmoore@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Most in-kernel users want registers spilled on the kernel stack and
don't require PS.EXCM to be set. That means that they don't need fixup
routine and could reuse regular window overflow mechanism for that,
which makes spill routine very simple.
Suggested-by: Chris Zankel <chris@zankel.net> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We need it saved because it contains a3 where we track which register
windows we still need to spill, and fixup handler may call C exception
handlers. Also fix comments.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Enabling SPARSE_IRQ shows up a bug in the irq-orion bridge interrupt
handler. The bridge interrupt is implemented using a single generic
chip. Thus the parameter passed to irq_get_domain_generic_chip()
should always be zero.
Signed-off-by: Andrew Lunn <andrew@lunn.ch> Acked-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Fixes: 9dbd90f17e4f ("irqchip: Add support for Marvell Orion SoCs") Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bridge IRQ_CAUSE bits are asserted regardless of the corresponding bit in
IRQ_MASK register. To avoid interrupt events on stale irqs, we have to clear
them before unmask. This installs an .irq_startup callback to ensure stale
irqs are cleared before initial unmask.
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Tested-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bridge irqs are edge-triggered, i.e. they get asserted on low-to-high
transitions and not on the level of the downstream interrupt line.
This replaces handle_level_irq by the more appropriate handle_edge_irq.
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Tested-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
HP Folio 13 may have a broken BIOS that doesn't set up the mute LED
GPIO properly, and the driver guesses it wrongly, too. Add a new
fixup entry for setting the GPIO pin statically for this laptop.
Drew Richardson reported that he could make the kernel go *boom* when hotplugging
while having perf events active.
It turned out that when you have a group event, the code in
__perf_event_exit_context() fails to remove the group siblings from
the context.
We then proceed with destroying and freeing the event, and when you
re-plug the CPU and try and add another event to that CPU, things go
*boom* because you've still got dead entries there.
Whilst trying to bring-up an SMMUv2 implementation with the table
walker plumbed into a coherent interconnect, I noticed that the memory
transactions targetting the CPU caches from the SMMU were marked as
outer-shareable instead of inner-shareable.
After a bunch of digging, it seems that we actually need to program
CBARn.BPSHCFG for s1-s2-bypass contexts to act as non-shareable in order
for the shareability configured in the corresponding TTBCR not to be
overridden with an outer-shareable attribute.
Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Now that we populate page tables as we traverse them ("iommu/arm-smmu:
fix pud/pmd entry fill sequence"), we need to ensure that we flush out
our zeroed tables after initial allocation, to prevent speculative TLB
fills using bogus data.
This patch adds additional calls to arm_smmu_flush_pgtable during
initial table allocation, and moves the dsb required by coherent table
walkers into the helper.
Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit a44a9791e778 ("iommu/arm-smmu: use mutex instead of spinlock for
locking page tables") replaced the page table spinlock with a mutex, to
allow blocking allocations to satisfy lazy mapping requests.
Unfortunately, it turns out that IOMMU mappings are created from atomic
context (e.g. spinlock held during a dma_map), so this change doesn't
really help us in practice.
This patch is a partial revert of the offending commit, bringing back
the original spinlock but replacing our page table allocations for any
levels below the pgd (which is allocated during domain init) with
GFP_ATOMIC instead of GFP_KERNEL.
Reported-by: Andreas Herrmann <andreas.herrmann@calxeda.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The ARM SMMU driver's population of puds and pmds is broken, since we
iterate over the next level of table repeatedly setting the current
level descriptor to point at the pmd being initialised. This is clearly
wrong when dealing with multiple pmds/puds.
This patch fixes the problem by moving the pud/pmd population out of the
loop and instead performing it when we allocate the next level (like we
correctly do for ptes already). The starting address for the next level
is then calculated prior to entering the loop.
The driver was not able to manage the sensor: during probe function
and wai check, the driver stops and writes: "device name and WhoAmI mismatch."
The correct value of L3GD20H wai is 0xd7 instead of 0xd4.
Dropped support for the sensor.
Releasing the touchscreen lets the internal statemachine left in a wrong state.
Due to this the release coordinate will be reported again by accident when the next
touchscreen event happens. This change sets up the correct state when waiting
for the next touchscreen event.
This has led to reported issues with calibrating the touchscreen.
Bug was introduced somewhere in the series that began with 18da755de59b406ce2371a55fb15ed676eb08ed2
Staging/iio/adc/touchscreen/MXS: add proper clock handling
in which the way this driver worked was substantially changed
to be interrupt driven rather than relying on a busy loop.
This was a regression in the 3.13 kernel.
The change (008fa749e0fe5b2fffd20b7fe4891bb80d072c6a) that moved the
node release code to a separate function broke death notifications in
some cases. When it encountered a reference without a death
notification request, it would skip looking at the remaining
references, and therefore fail to send death notifications for them.
Cc: Colin Cross <ccross@android.com> Cc: Android Kernel Team <kernel-team@android.com> Signed-off-by: Arve Hjønnevåg <arve@android.com> Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1) In current implementation, ecclayout->oobfree->offset is calculated with
respect to ecclayout->eccpos[0] which is incorrect because ECC bytes may not
be stored contiguously in OOB.
So, this patch calculates ecclayout->oobfree->offset with respect to last
ECC byte-position 'eccpos[ecclayout->eccbytes-1]'.
2) ECC layout of some ecc-schemes expects reserved-markers at specific eccpos[]
which should not be over-written by any file-system metadata.
So this patch aligns oobfree->offset taking into account of such markers.
Tested-by: Enric Balletbo i Serra <eballetbo@gmail.com> Tested-by: Stefan Roese <sr@denx.de> Signed-off-by: Pekon Gupta <pekon@ti.com> Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes: commit a919e51161b58ed7e6e663daba99ab7d558808f3
mtd: nand: omap2: clean-up BCHx_HW and BCHx_SW ECC configurations in device_probe
Fixes ecclayout mismatch introduced in above commit for following ecc-schemes:
- OMAP_ECC_BCH4_CODE_HW_DETECTION_SW
- OMAP_ECC_BCH8_CODE_HW_DETECTION_SW
However, this patch also touches other ecc-schemes as the fix required
refactoring common code, into ecc-scheme specific code.
This patch aligns ecc-layout for below ecc-schemes as per reference [1],[2],[3]
[1] ecc-layout expected by ROM code, as specified in SoC TRM under:
Chapter="Initialization"
Section="Device Initialization by ROM code"
Sub-Section="Memory Booting"
Heading="NAND"
Figure="ECC Locations in NAND Spare Areas"
[2] ecc-layout updates in u-boot
http://lists.denx.de/pipermail/u-boot/2013-November/167551.html
[3] u-boot configurations to match above ecc-layout are documented at
https://processors.wiki.ti.com/index.php/Linux_Core_NAND_User%27s_Guide
Reported-by: Enric Balletbo Serra <eballetbo@iseebcn.com> Tested-by: Enric Balletbo i Serra <eballetbo@gmail.com> Tested-by: Stefan Roese <sr@denx.de> Signed-off-by: Pekon Gupta <pekon@ti.com> Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bug fix to allow the setting of maximum voltage for certain LDOs.
What the bug is:
There is a problem caused by an invalid calculation of n_voltages
in the driver. This n_voltages value has the potential to be
different for each regulator.
The value for linear_min_sel is set as DA9063_V##regl_name#
which can be different depending upon the regulator. This is
chosen according to the following definitions in the DA9063
registers.h file:
The calculation for n_voltages is valid for LDOs whose BIAS value
is zero but this is not correct for those LDOs which have a
non-zero value.
What the fix is:
In order to take into account the non-zero linear_min_sel value which
is set for the regulators LDO5, LDO6, LDO7, LDO8, LDO9, LDO10 and
LDO11, the calculation for n_voltages should take into account the
missing term defined by DA9063_V##regl_name#.
This will in turn allow the core constraints calculation to set the
maximum voltage limits correctly and therefore allow users to apply
the maximum expected voltage to all of the LDOs.
Signed-off-by: Steve Twiss <stwiss.opensource@diasemi.com> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a kworker should die, the kworkre is notified through WORKER_DIE
flag instead of kthread_should_stop(). This, IIRC, is primarily to
keep the test synchronized inside worker_pool lock. WORKER_DIE is
first set while holding pool->lock, the lock is dropped and
kthread_stop() is called.
Unfortunately, this means that there's a slight chance that the target
kworker may see WORKER_DIE before kthread_stop() finishes and exits
and frees the target task before or during kthread_stop().
Fix it by pinning the target task before setting WORKER_DIE and
putting it after kthread_stop() is done.
tj: Improved patch description and comment. Moved pinning above
WORKER_DIE for better signify what it's protecting.
In mei_cl_read_start(), if it fails to send flow control request, it
will release "cl->read_cb" but forget to set pointer to NULL, leaving
"cl->read_cb" still pointing to random memory, next time this client is
operated like mei_release(), it has chance to refer to this wrong pointer.
the following patch adds an entry for the PID of a Cressi Leonardo
diving computer interface to kernel 3.13.0.
It is detected as FT232RL.
Works with subsurface.
ehci_irq() and ehci_hrtimer_func() can deadlock on ehci->lock when
threadirqs option is used. To prevent the deadlock use
spin_lock_irqsave() in ehci_irq().
This change can be reverted when hrtimer callbacks become threaded.
High-speed USB connections revert back to full-speed signalling when
the device goes into suspend. This takes several milliseconds, and
during that time it's not possible to tell reliably whether the device
has been disconnected.
On some platforms, the Wake-On-Disconnect circuitry gets confused
during this intermediate state. It generates a false wakeup signal,
which can prevent the controller from going to sleep.
To avoid this problem, this patch adds a 5-ms delay to the
ehci_bus_suspend() routine if any ports have to switch over to
full-speed signalling. (Actually, the delay was already present for
devices using a particular kind of PHY power management; the patch
merely causes the delay to be used more widely.)
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Reviewed-by: Peter Chen <Peter.Chen@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 3dc6475 ("bcm63xx_enet: add support Broadcom BCM6345 Ethernet")
changed the ENETDMA[CS] macros such that they are no longer macros, but
actual register offset definitions. The bcm63xx_udc driver was not
updated, and as a result, causes the following build error to pop up:
CC drivers/usb/gadget/u_ether.o
drivers/usb/gadget/bcm63xx_udc.c: In function 'iudma_write':
drivers/usb/gadget/bcm63xx_udc.c:642:24: error: called object '0' is not
a function
drivers/usb/gadget/bcm63xx_udc.c: In function 'iudma_reset_channel':
drivers/usb/gadget/bcm63xx_udc.c:698:46: error: called object '0' is not
a function
drivers/usb/gadget/bcm63xx_udc.c:700:49: error: called object '0' is not
a function
Fix this by updating usb_dmac_{read,write}l and usb_dmas_{read,write}l to
take an extra channel argument, and use the channel width
(ENETDMA_CHAN_WIDTH) to offset the register we want to access, hence
doing again what the macro implicitely did for us.
Cc: Kevin Cernekee <cernekee@gmail.com> Cc: Jonas Gorski <jogo@openwrt.org> Signed-off-by: Florian Fainelli <florian@openwrt.org> Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ENDPTFLUSH and ENDPTPRIME registers are set by software and clear
by hardware. There is a bit for each endpoint. When we are setting
a bit for an endpoint we should make sure we do not touch other
endpoint bit. There is a race condition if the hardware clear the
bit between the read and the write in hw_write.
Signed-off-by: Peter Chen <peter.chen@freescale.com> Signed-off-by: Matthieu CASTET <matthieu.castet@parrot.com> Tested-by: Michael Grzeschik <mgrzeschik@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LFM (max efficiency ratio) is the max frequency at minimum voltage
supported by the processor. Using LFM as the minimum P state
increases performmance without affecting power. By not using P states
below LFM we avoid using P states that are less power efficient.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
acpi_processor_set_throttling() uses set_cpus_allowed_ptr() to make
sure that the (struct acpi_processor)->acpi_processor_set_throttling()
callback will run on the right CPU. However, the function may be
called from a worker thread already bound to a different CPU in which
case that won't work.
Make acpi_processor_set_throttling() use work_on_cpu() as appropriate
instead of abusing set_cpus_allowed_ptr().
Reported-and-tested-by: Jiri Olsa <jolsa@redhat.com> Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
[rjw: Changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some devices have duplicate entries in there brightness levels table, ie
on my Dell Latitude E6430 the table looks like this:
[ 3.686060] acpi backlight index 0, val 80
[ 3.686095] acpi backlight index 1, val 50
[ 3.686122] acpi backlight index 2, val 5
[ 3.686147] acpi backlight index 3, val 5
[ 3.686172] acpi backlight index 4, val 5
[ 3.686197] acpi backlight index 5, val 5
[ 3.686223] acpi backlight index 6, val 5
[ 3.686248] acpi backlight index 7, val 5
[ 3.686273] acpi backlight index 8, val 6
[ 3.686332] acpi backlight index 9, val 7
[ 3.686356] acpi backlight index 10, val 8
[ 3.686380] acpi backlight index 11, val 9
etc.
Notice that brightness values 0-5 are all mapped to 5. This means that
if userspace writes any value between 0 and 5 to the brightness sysfs attribute
and then reads it, it will always return 0, which is somewhat unexpected.
This is a problem for ie gnome-settings-daemon, which uses read-modify-write
logic when the users presses the brightness up or down keys. This is done
this way to take brightness changes from other sources into account.
On this specific laptop what happens once the brightness has been set to 0,
is that gsd reads 0, adds 5, writes 5, and on the next brightness up key press
again reads 0, so things get stuck at the lowest brightness setting.
Filtering out the duplicate table entries, makes any write to brightness
read back as the written value as one would expect, fixing this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Reviewed-by: Aaron Lu <aaron.lu@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The reference count changes done by pci_get_device can be a little
misleading when the usage diverges from the most common scheme. The
reference count of the device passed as the last parameter is always
decreased, even if the function returns no new device. So if we are
going to try alternative device IDs, we must manually increment the
device reference count before each retry. If we don't, we end up
decreasing the reference count, and after a few modprobe/rmmod cycles
the PCI devices will vanish.
In other words and as Alan put it: without this fix the EDAC code
corrupts the PCI device list.
This fixes kernel bug #50491:
https://bugzilla.kernel.org/show_bug.cgi?id=50491
acpi_pci_link_allocate_irq() can return negative gsi even if
entry != NULL. For that case we have a memory leak, so free
entry before returning from acpi_pci_irq_enable() for gsi < 0.
Signed-off-by: Tomasz Nowicki <tomasz.nowicki@linaro.org>
[rjw: Subject and changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some firmware leaves the Interrupt Disable bit set even if the device uses
INTx interrupts. Clear Interrupt Disable so we get those interrupts.
Based on the report mentioned below, if the user selects the "EHCI only"
option in the Intel Baytrail BIOS, the EHCI device is handed off to the OS
with the PCI_COMMAND_INTX_DISABLE bit set.
Marvell SoCs place the SoC number into the PCIe endpoint device ID. The
SoC stepping is placed into the PCIe revision. The old plat-orion PCIe
driver allowed this information to be seen in user space with a simple
lspci command.
The new driver places a virtual PCI-PCI bridge on top of these endpoints.
It has its own hard coded PCI device ID. Thus it is no longer possible to
see what the SoC is using lspci.
When initializing the PCI-PCI bridge, set its device ID and revision from
the underlying endpoint, thus restoring this functionality. Debian would
like to use this in order to aid installing the correct DTB file.
Fixes: 45361a4fe4464 ("pci: PCIe driver for Marvell Armada 370/XP systems") Signed-off-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This reverts commit c4a391b53a72d2df4ee97f96f78c1d5971b47489. Dave
Chinner <david@fromorbit.com> has reported the commit may cause some
inodes to be left out from sync(2). This is because we can call
redirty_tail() for some inode (which sets i_dirtied_when to current time)
after sync(2) has started or similarly requeue_inode() can set
i_dirtied_when to current time if writeback had to skip some pages. The
real problem is in the functions clobbering i_dirtied_when but fixing
that isn't trivial so revert is a safer choice for now.
Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The powernow-k8 driver maintains a per-cpu data-structure called
powernow_data that is used to perform the frequency transitions.
It initializes this data structure only for the policy->cpu. So,
accesses to this data structure by other CPUs results in various
problems because they would have been uninitialized.
Specifically, if a cpu (!= policy->cpu) invokes the drivers' ->get()
function, it returns 0 as the KHz value, since its per-cpu memory
doesn't point to anything valid. This causes problems during
suspend/resume since cpufreq_update_policy() tries to enforce this
(0 KHz) as the current frequency of the CPU, and this madness gets
propagated to adjust_jiffies() as well. Eventually, lots of things
start breaking down, including the r8169 ethernet card, in one
particularly interesting case reported by Pierre Ossman.
Fix this by initializing the per-cpu data-structures of all the CPUs
in the policy appropriately.
References: https://bugzilla.kernel.org/show_bug.cgi?id=70311 Reported-by: Pierre Ossman <pierre@ossman.eu> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It's a bit odd to see a newer device showing mod15write; however, the
reported behavior is highly consistent and other factors which could
contribute seem to have been verified well enough. Also, both
sata_sil itself and the drive are fairly outdated at this point making
the risk of this change fairly low. It is possible, probably likely,
that other drive models in the same family have the same problem;
however, for now, let's just add the specific model which was tested.
Without the patch the kernel generates the following error.
ata11.15: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
ata11.15: Port Multiplier vendor mismatch '0x197b' != '0x123'
ata11.15: PMP revalidation failed (errno=-19)
ata11.15: failed to recover PMP after 5 tries, giving up
This patch helps to bypass this error and the device becomes
functional.
Vince "Super Tester" Weaver reported a new round of syscall fuzzing (Trinity) failures,
with perf WARN_ON()s triggering. He also provided traces of the failures.
We try and add the {BP,cycles,br_insn} group (fd[3], fd[4], fd[15]).
These events are 0:cycles and 4:br_insn, the BP event isn't x86_pmu so
that's not visible.
group_sched_in()
pmu->start_txn() /* nop - BP pmu */
event_sched_in()
event->pmu->add()
But seeing the below state on x86_pmu_enable(), the must have failed,
because the 0 and 4 events aren't there anymore.
Looking at group_sched_in(), since the BP is the leader, its
event_sched_in() must have succeeded, for otherwise we would not have
seen the sibling adds.
But since neither 0 or 4 are in the below state; their event_sched_in()
must have failed; but I don't see why, the complete state: 0,0,1:p,4
fits perfectly fine on a core2.
However, since we try and schedule 4 it means the 0 event must have
succeeded! Therefore the 4 event must have failed, its failure will
have put group_sched_in() into the fail path, which will call:
event_sched_out()
event->pmu->del()
on 0 and the BP event.
Now x86_pmu_del() will reduce n_events; but it will not reduce n_added;
giving what we see below:
So the problem is that x86_pmu_del(), when called from a
group_sched_in() that fails (for whatever reason), and without x86_pmu
TXN support (because the leader is !x86_pmu), will corrupt the n_added
state.
Reported-and-Tested-by: Vince Weaver <vincent.weaver@maine.edu> Signed-off-by: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Stephane Eranian <eranian@google.com> Cc: Dave Jones <davej@redhat.com> Link: http://lkml.kernel.org/r/20140221150312.GF3104@twins.programming.kicks-ass.net Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Supporting decoding the ioctl 'request' parameter needs more work to
properly support more architectures, the current approach doesn't work
on at least powerpc and sparc, as reported by Ben Hutchings in
http://lkml.kernel.org/r/1391593985.3003.48.camel@deadeye.wl.decadent.org.uk .
Work around that by making it to be ifdefed for the architectures known
to work with the current, limited approach, i386 and x86_64 till better
code is written.
Reported-by: Ben Hutchings <ben@decadent.org.uk> Acked-by: Ben Hutchings <ben@decadent.org.uk> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-ss04k11insqlu329xh5g02q0@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
GFP_ATOMIC is not a single gfp flag, but a macro which expands to the other
flags, where meaningful is the LACK of __GFP_WAIT flag. To check if caller
wants to perform an atomic allocation, the code must test for a lack of the
__GFP_WAIT flag. This patch fixes the issue introduced in v3.5-rc1.
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Samsung's pci-e SSDs with device ID 0x1600 which are found on some
macbooks time out on NCQ commands. Blacklist NCQ on the device so
that the affected machines can at least boot.
The OPAL firmware functions opal_xscom_read and opal_xscom_write
take a 64-bit argument for the XSCOM (PCB) address in order to
support the indirect mode on P8.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In copy_oldmem_page, the current check using max_pfn and min_low_pfn to
decide if the page is backed or not, is not valid when the memory layout is
not continuous.
This happens when running as a QEMU/KVM guest, where RTAS is mapped higher
in the memory. In that case max_pfn points to the end of RTAS, and a hole
between the end of the kdump kernel and RTAS is not backed by PTEs. As a
consequence, the kdump kernel is crashing in copy_oldmem_page when accessing
in a direct way the pages in that hole.
This fix relies on the memblock's service memblock_is_region_memory to
check if the read page is part or not of the directly accessible memory.
Currently we're storing a host endian RTAS token in
rtas_stop_self_args.token. We then pass that directly to rtas. This is
fine on big endian however on little endian the token is not what we
expect.
This will typically result in hitting:
panic("Alas, I survived.\n");
To fix this we always use the stop-self token in host order and always
convert it to be32 before passing this to rtas.
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The new ELFv2 little-endian ABI increases the stack redzone -- the
area below the stack pointer that can be used for storing data --
from 288 bytes to 512 bytes. This means that we need to allow more
space on the user stack when delivering a signal to a 64-bit process.
To make the code a bit clearer, we define new USER_REDZONE_SIZE and
KERNEL_REDZONE_SIZE symbols in ptrace.h. For now, we leave the
kernel redzone size at 288 bytes, since increasing it to 512 bytes
would increase the size of interrupt stack frames correspondingly.
Gcc currently only makes use of 288 bytes of redzone even when
compiling for the new little-endian ABI, and the kernel cannot
currently be compiled with the new ABI anyway.
In the future, hopefully gcc will provide an option to control the
amount of redzone used, and then we could reduce it even more.
This also changes the code in arch_compat_alloc_user_space() to
preserve the expanded redzone. It is not clear why this function would
ever be used on a 64-bit process, though.
Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fix a race in which the RPC client is shutting down while the
gss daemon is processing a downcall. If the RPC client manages to
shut down before the gss daemon is done, then the struct gss_auth
used in gss_release_msg() may have already been freed.
When a send failure occurs due to the socket being out of buffer space,
we call xs_nospace() in order to have the RPC task wait until the
socket has drained enough to make it worth while trying again.
The current patch fixes a race in which the socket is drained before
we get round to setting up the machinery in xs_nospace(), and which
is reported to cause hangs.
Link: http://lkml.kernel.org/r/20140210170315.33dfc621@notabene.brown Fixes: a9a6b52ee1ba (SUNRPC: Don't start the retransmission timer...) Reported-by: Neil Brown <neilb@suse.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a typo in the Limiter2 Release Rate control, a wrong enum for
Limiter1 is assigned. It must point to Limiter2.
Spotted by a compile warning:
In file included from sound/soc/codecs/sta32x.c:34:0:
sound/soc/codecs/sta32x.c:223:29: warning: ‘sta32x_limiter2_release_rate_enum’ defined but not used [-Wunused-variable]
static SOC_ENUM_SINGLE_DECL(sta32x_limiter2_release_rate_enum,
^
include/sound/soc.h:275:18: note: in definition of macro ‘SOC_ENUM_DOUBLE_DECL’
struct soc_enum name = SOC_ENUM_DOUBLE(xreg, xshift_l, xshift_r, \
^
sound/soc/codecs/sta32x.c:223:8: note: in expansion of macro ‘SOC_ENUM_SINGLE_DECL’
static SOC_ENUM_SINGLE_DECL(sta32x_limiter2_release_rate_enum,
^
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The driver reads from the DC offset control registers during callibration
but since the registers are marked as volatile and there is a register
cache the values will not be read from the hardware after the first reading
rendering the callibration ineffective.
It appears that the driver was originally written for the ASoC level
register I/O code but converted to regmap prior to merge and this issue
was missed during the conversion as the framework level volatile register
functionality was not being used.
Signed-off-by: Mark Brown <broonie@linaro.org> Acked-by: Adam Thomson <Adam.Thomson.Opensource@diasemi.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sync regcache when entering STANDBY from OFF. ON isn't entered with
OFF as the current state, so the registers were not being re-synced
after suspend/resume.
The 98088 and 98095 already call regcache_sync from STANDBY.
Signed-off-by: Dylan Reid <dgreid@chromium.org> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The commit 1abe729 (ASoC: fsl: Add missing pm to current machine
drivers) enables pm support for a few IMX machine drivers. But it does
not update dev drvdata to be the pointer to 'card'. This causes the
kernel dump below in system suspend, because snd_soc_suspend() expects
that the dev drvdata points to 'card', while it still points to the
private data of machine driver.
This patch fixes imx-sgtl5000 and imx-wm8962 by attaching 'card' to dev
drvdata and private data to card drvdata. For imx-mc13783, I simply
revert the pm change because it must be broken for the same reason and
I don't have hardware to test pm enabling code.
This patch fixes a crash caused by commit 3bed3344c826
(ASoC: txx9aclc_ac97: Convert to devm_ioremap_resource()).
This is an attempt to assign "drvdata->base" while memory
for "drvdata" is not already allocated.
Fixes: 3bed3344c826 (ASoC: txx9aclc_ac97: Convert to devm_ioremap_resource()) Signed-off-by: Alexander Shiyan <shc_work@mail.ru> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Realtek RT5640 uses ACPI ID "10EC5640" for Intel Baytrail platforms.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the I2C device Ids conflict for the MFD and CODEC so
cannot be both instantiated on one platform. This patch updates
the Ids and names to make them unique from each other.
It should be noted that the I2C addresses for both PMIC and CODEC
are modifiable so instantiation of the two are kept as separate
devices, rather than instantiating the CODEC from the MFD code.
Signed-off-by: Adam Thomson <Adam.Thomson.Opensource@diasemi.com> Acked-by: Mark Brown <broonie@linaro.org> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit e504c9098ed6 (kvm, vmx: Fix lazy FPU on nested guest, 2013-11-13)
highlighted a real problem, but the fix was subtly wrong.
nested_read_cr0 is the CR0 as read by L2, but here we want to look at
the CR0 value reflecting L1's setup. In other words, L2 might think
that TS=0 (so nested_read_cr0 has the bit clear); but if L1 is actually
running it with TS=1, we should inject the fault into L1.
The effective value of CR0 in L2 is contained in vmcs12->guest_cr0, use
it.
The problem occurs when the guest performs a pusha with the stack
address pointing to an mmio address (or an invalid guest physical
address) to start with, but then extending into an ordinary guest
physical address. When doing repeated emulated pushes
emulator_read_write sets mmio_needed to 1 on the first one. On a
later push when the stack points to regular memory,
mmio_nr_fragments is set to 0, but mmio_is_needed is not set to 0.
As a result, KVM exits to userspace, and then returns to
complete_emulated_mmio. In complete_emulated_mmio
vcpu->mmio_cur_fragment is incremented. The termination condition of
vcpu->mmio_cur_fragment == vcpu->mmio_nr_fragments is never achieved.
The code bounces back and fourth to userspace incrementing
mmio_cur_fragment past it's buffer. If the guest does nothing else it
eventually leads to a a crash on a memcpy from invalid memory address.
However if a guest code can cause the vm to be destroyed in another
vcpu with excellent timing, then kvm_clear_async_pf_completion_queue
can be used by the guest to control the data that's pointed to by the
call to cancel_work_item, which can be used to gain execution.
Fixes: f78146b0f9230765c6315b2e14f56112513389ad Signed-off-by: Andrew Honig <ahonig@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
sparse complains about any __ksymtab symbols with the following:
warning: symbol '__ksymtab_...' was not declared. Should it be static?
due to Andi's patch making it non-static.
Mollify sparse by declaring the symbol extern, otherwise we get
drowned in sparse warnings for anything that uses EXPORT_SYMBOL
in the sources, making it easy to miss real warnings.
Fixes: e0f244c63fc9 ("asmlinkage, module: Make ksymtab [...] __visible") Signed-off-by: Johannes Berg <johannes.berg@intel.com> Acked-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently notify_change directly updates i_version for size updates,
which not only is counter to how all other fields are updated through
struct iattr, but also breaks XFS, which need inode updates to happen
under its own lock, and synchronized to the structure that gets written
to the log.
Remove the update in the common code, and it to btrfs and ext4,
XFS already does a proper updaste internally and currently gets a
double update with the existing code.
IMHO this is 3.13 and -stable material and should go in through the XFS
tree.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Acked-by: Jan Kara <jack@suse.cz> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Ben Myers <bpm@sgi.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, there's nothing preventing cgroup_enable_task_cg_lists()
from missing set PF_EXITING and race against cgroup_exit(). Depending
on the timing, cgroup_exit() may finish with the task still linked on
css_set leading to list corruption. Fix it by grabbing siglock in
cgroup_enable_task_cg_lists() so that PF_EXITING is guaranteed to be
visible.
This whole on-demand cg_list optimization is extremely fragile and has
ample possibility to lead to bugs which can cause things like
once-a-year oops during boot. I'm wondering whether the better
approach would be just adding "cgroup_disable=all" handling which
disables the whole cgroup rather than tempting fate with this
on-demand craziness.
cgroup_cfts_commit() walks the cgroup hierarchy that the target
subsystem is attached to and tries to apply the file changes. Due to
the convolution with inode locking, it can't keep cgroup_mutex locked
while iterating. It currently holds only RCU read lock around the
actual iteration and then pins the found cgroup using dget().
Unfortunately, this is incorrect. Although the iteration does check
cgroup_is_dead() before invoking dget(), there's nothing which
prevents the dentry from going away inbetween. Note that this is
different from the usual css iterations where css_tryget() is used to
pin the css - css_tryget() tests whether the css can be pinned and
fails if not.
The problem can be solved by simply holding cgroup_mutex instead of
RCU read lock around the iteration, which actually reduces LOC.
When cgroup_mount() fails to allocate an id for the root, it didn't
set ret before jumping to unlock_drop ending up returning 0 after a
failure. Fix it.
The front headphone and mic jackes on a HP desktop model (Vendor Id:
0x111d76c7 Subsystem Id: 0x103c2b17) can not work, the codec on this
machine has 8 physical ports, 6 of them are routed to rear jackes
and all of them work very well, while the remaining 2 ports are
routed to front headphone and mic jackes, but the corresponding
pin complex node are not defined correctly.
After apply this fix, the front audio jackes can work very well.
[trivial fix of enum definition by tiwai]
BugLink: https://bugs.launchpad.net/bugs/1282369 Cc: David Henningsson <david.henningsson@canonical.com> Tested-by: Gerald Yang <gerald.yang@canonical.com> Signed-off-by: Hui Wang <hui.wang@canonical.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Incorrect ADC is picked in ca0132_capture_pcm_prepare(),
where it assumes multiple streams while there is one stream
per ADC. Note that ca0132_capture_pcm_cleanup() already does
the right thing.
The Chromebook Pixel has a microphone under the keyboard that
is attached to node id 0x8. Before this fix, recording would
always go to the main internal mic (node id 0x7).
When a HDMI stream is opened with the same stream tag
as a following opened stream to ca0132, audio will be
heard from two ports simultaneously.
Fix this issue by change to use snd_hda_codec_setup_stream
and snd_hda_codec_cleanup_stream instead, so that an
inactive stream can be marked as 'dirty' when found
with a conflict stream tag, and then get purified.
When we plug a 3-ring headset on the Dell machines (Vendor ID:
0x10ec0255, Subsystem ID: 0x10280657; Vendor ID: 0x10ec0255,
Subsystem ID: 0x1028065f), the headset mic can't be
detected, after apply this patch, the headset mic can work well.
BugLink: https://bugs.launchpad.net/bugs/1260303 Cc: David Henningsson <david.henningsson@canonical.com> Tested-by: Cyrus Lien <cyrus.lien@canonical.com> Signed-off-by: Hui Wang <hui.wang@canonical.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the driver tries to access Function Unit 10, the KEF X300A
speakers' firmware apparently locks up, making even PCM streaming
impossible. Work around this by ignoring this FU.
batadv_send_skb_prepare_unicast(_4addr) might reallocate the
skb's data. If it does then our ethhdr pointer is not valid
anymore in batadv_send_skb_unicast(), resulting in a kernel
paging error.
Fixing this by refetching the ethhdr pointer after the
potential reallocation.
In the failure path of the orig_node initialization routine
the orig_node->bat_iv.bcast_own field is free'd twice: first
in batadv_iv_ogm_orig_get() and then later in
batadv_orig_node_free_rcu().
Fix it by removing the kfree in batadv_iv_ogm_orig_get().
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com> Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When computing the CRC on a 2byte variable the order of
the bytes obviously alters the final result. This means
that computing the CRC over the same value on two archs
having different endianess leads to different numbers.
The global and local translation table CRC computation
routine makes this mistake while processing the clients
VIDs. The result is a continuous CRC mismatching between
nodes having different endianess.
Fix this by converting the VID to Network Order before
processing it. This guarantees that every node uses the same
byte order.
Since batadv_orig_node_new() sets the refcount to two, assuming that
the calling function will use a reference for putting the orig_node into
a hash or similar, both references must be freed if initialization of
the orig_node fails. Otherwise that object may be leaked in that error
case.
Reported-by: Antonio Quartulli <antonio@meshcoding.com> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch> Signed-off-by: Antonio Quartulli <antonio@meshcoding.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When adding a new neighbour it is important to atomically
perform the following:
- check if the neighbour already exists
- append the neighbour to the proper list
If the two operations are not performed in an atomic context
it is possible that two concurrent insertions add the same
neighbour twice.
Signed-off-by: Antonio Quartulli <antonio@open-mesh.com> Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a refcounter unbalance in the CRC checking routine
invoked on OGM reception. A vlan object is retrieved (thus
its refcounter is increased by one) but it is never properly
released. This leads to a memleak because the vlan object
will never be free'd.
Fix this by releasing the vlan object after having read the
CRC.
Reported-by: Russell Senior <russell@personaltelco.net> Reported-by: Daniel <daniel@makrotopia.org> Reported-by: cmsv <cmsv@wirelesspt.net> Signed-off-by: Antonio Quartulli <antonio@meshcoding.com> Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current MTU computation always returns a value
smaller than 1500bytes even if the real interfaces
have an MTU large enough to compensate the batman-adv
overhead.
Fix the computation by properly returning the highest
admitted value.
sock_alloc_send_pskb() & sk_page_frag_refill()
have a loop trying high order allocations to prepare
skb with low number of fragments as this increases performance.
Problem is that under memory pressure/fragmentation, this can
trigger OOM while the intent was only to try the high order
allocations, then fallback to order-0 allocations.
We had various reports from unexpected regressions.
According to David, setting __GFP_NORETRY should be fine,
as the asynchronous compaction is still enabled, and this
will prevent OOM from kicking as in :
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: David Rientjes <rientjes@google.com> Acked-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Right now the mvneta driver doesn't handle Tx IRQ, and relies on two
mechanisms to flush Tx descriptors : a flush at the end of mvneta_tx()
and a timer. If a burst of packets is emitted faster than the device
can send them, then the queue is stopped until next wake-up of the
timer 10ms later. This causes jerky output traffic with bursts and
pauses, making it difficult to reach line rate with very few streams.
A test on UDP traffic shows that it's not possible to go beyond 134
Mbps / 12 kpps of outgoing traffic with 1500-bytes IP packets. Routed
traffic tends to observe pauses as well if the traffic is bursty,
making it even burstier after the wake-up.
It seems that this feature was inherited from the original driver but
nothing there mentions any reason for not using the interrupt instead,
which the chip supports.
Thus, this patch enables Tx interrupts and removes the timer. It does
the two at once because it's not really possible to make the two
mechanisms coexist, so a split patch doesn't make sense.
First tests performed on a Mirabox (Armada 370) show that less CPU
seems to be used when sending traffic. One reason might be that we now
call the mvneta_tx_done_gbe() with a mask indicating which queues have
been done instead of looping over all of them.
The same UDP test above now happily reaches 987 Mbps / 87.7 kpps.
Single-stream TCP traffic can now more easily reach line rate. HTTP
transfers of 1 MB objects over a single connection went from 730 to
840 Mbps. It is even possible to go significantly higher (>900 Mbps)
by tweaking tcp_tso_win_divisor.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Gregory CLEMENT <gregory.clement@free-electrons.com> Cc: Arnaud Ebalard <arno@natisbad.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Tested-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Marvell has not published the chip's datasheet yet, so it's very hard
to find the relevant bits to manipulate to change the IRQ behaviour.
Fortunately, these bits are described in the proprietary LSP patch set
which is publicly available here :
http://www.plugcomputer.org/downloads/mirabox/
So let's put them back in the driver in order to reduce the burden of
current and future maintenance.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Gregory CLEMENT <gregory.clement@free-electrons.com> Tested-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a queue timeout is reported, we can oops because of some
schedules while the caller is atomic, as shown below :
mvneta d0070000.ethernet eth0: tx timeout
BUG: scheduling while atomic: bash/1528/0x00000100
Modules linked in: slhttp_ethdiv(C) [last unloaded: slhttp_ethdiv]
CPU: 2 PID: 1528 Comm: bash Tainted: G WC 3.13.0-rc4-mvebu-nf #180
[<c0011bd9>] (unwind_backtrace+0x1/0x98) from [<c000f1ab>] (show_stack+0xb/0xc)
[<c000f1ab>] (show_stack+0xb/0xc) from [<c02ad323>] (dump_stack+0x4f/0x64)
[<c02ad323>] (dump_stack+0x4f/0x64) from [<c02abe67>] (__schedule_bug+0x37/0x4c)
[<c02abe67>] (__schedule_bug+0x37/0x4c) from [<c02ae261>] (__schedule+0x325/0x3ec)
[<c02ae261>] (__schedule+0x325/0x3ec) from [<c02adb97>] (schedule_timeout+0xb7/0x118)
[<c02adb97>] (schedule_timeout+0xb7/0x118) from [<c0020a67>] (msleep+0xf/0x14)
[<c0020a67>] (msleep+0xf/0x14) from [<c01dcbe5>] (mvneta_stop_dev+0x21/0x194)
[<c01dcbe5>] (mvneta_stop_dev+0x21/0x194) from [<c01dcfe9>] (mvneta_tx_timeout+0x19/0x24)
[<c01dcfe9>] (mvneta_tx_timeout+0x19/0x24) from [<c024afc7>] (dev_watchdog+0x18b/0x1c4)
[<c024afc7>] (dev_watchdog+0x18b/0x1c4) from [<c0020b53>] (call_timer_fn.isra.27+0x17/0x5c)
[<c0020b53>] (call_timer_fn.isra.27+0x17/0x5c) from [<c0020cad>] (run_timer_softirq+0x115/0x170)
[<c0020cad>] (run_timer_softirq+0x115/0x170) from [<c001ccb9>] (__do_softirq+0xbd/0x1a8)
[<c001ccb9>] (__do_softirq+0xbd/0x1a8) from [<c001cfad>] (irq_exit+0x61/0x98)
[<c001cfad>] (irq_exit+0x61/0x98) from [<c000d4bf>] (handle_IRQ+0x27/0x60)
[<c000d4bf>] (handle_IRQ+0x27/0x60) from [<c000843b>] (armada_370_xp_handle_irq+0x33/0xc8)
[<c000843b>] (armada_370_xp_handle_irq+0x33/0xc8) from [<c000fba9>] (__irq_usr+0x49/0x60)
Ben Hutchings attempted to propose a better fix consisting in using a
scheduled work for this, but while it fixed this panic, it caused other
random freezes and panics proving that the reset sequence in the driver
is unreliable and that additional fixes should be investigated.
When sending multiple streams over a link limited to 100 Mbps, Tx timeouts
happen from time to time, and the driver correctly recovers only when the
function is disabled.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Gregory CLEMENT <gregory.clement@free-electrons.com> Cc: Ben Hutchings <ben@decadent.org.uk> Tested-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stats writers are mvneta_rx() and mvneta_tx(). They don't lock anything
when they update the stats, and as a result, it randomly happens that
the stats freeze on SMP if two updates happen during stats retrieval.
This is very easily reproducible by starting two HTTP servers and binding
each of them to a different CPU, then consulting /proc/net/dev in loops
during transfers, the interface should immediately lock up. This issue
also randomly happens upon link state changes during transfers, because
the stats are collected in this situation, but it takes more attempts to
reproduce it.
The comments in netdevice.h suggest using per_cpu stats instead to get
rid of this issue.
This patch implements this. It merges both rx_stats and tx_stats into
a single "stats" member with a single syncp. Both mvneta_rx() and
mvneta_rx() now only update the a single CPU's counters.
In turn, mvneta_get_stats64() does the summing by iterating over all CPUs
to get their respective stats.
With this change, stats are still correct and no more lockup is encountered.
Note that this bug was present since the first import of the mvneta
driver. It might make sense to backport it to some stable trees. If
so, it depends on "d33dc73 net: mvneta: increase the 64-bit rx/tx stats
out of the hot path".
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Gregory CLEMENT <gregory.clement@free-electrons.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Tested-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Better count packets and bytes in the stack and on 32 bit then
accumulate them at the end for once. This saves two memory writes
and two memory barriers per packet. The incoming packet rate was
increased by 4.7% on the Openblocks AX3 thanks to this.
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Gregory CLEMENT <gregory.clement@free-electrons.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Tested-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Willy Tarreau <w@1wt.eu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Marcelo Ricardo Leitner reported problems when the forwarding link path
has a lower mtu than the incoming one if the inbound interface supports GRO.
Given:
Host <mtu1500> R1 <mtu1200> R2
Host sends tcp stream which is routed via R1 and R2. R1 performs GRO.
In this case, the kernel will fail to send ICMP fragmentation needed
messages (or pkt too big for ipv6), as GSO packets currently bypass dstmtu
checks in forward path. Instead, Linux tries to send out packets exceeding
the mtu.
When locking route MTU on Host (i.e., no ipv4 DF bit set), R1 does
not fragment the packets when forwarding, and again tries to send out
packets exceeding R1-R2 link mtu.
This alters the forwarding dstmtu checks to take the individual gso
segment lengths into account.
For ipv6, we send out pkt too big error for gso if the individual
segments are too big.
For ipv4, we either send icmp fragmentation needed, or, if the DF bit
is not set, perform software segmentation and let the output path
create fragments when the packet is leaving the machine.
It is not 100% correct as the error message will contain the headers of
the GRO skb instead of the original/segmented one, but it seems to
work fine in my (limited) tests.
Eric Dumazet suggested to simply shrink mss via ->gso_size to avoid
sofware segmentation.
However it turns out that skb_segment() assumes skb nr_frags is related
to mss size so we would BUG there. I don't want to mess with it considering
Herbert and Eric disagree on what the correct behavior should be.
Hannes Frederic Sowa notes that when we would shrink gso_size
skb_segment would then also need to deal with the case where
SKB_MAX_FRAGS would be exceeded.
This uses sofware segmentation in the forward path when we hit ipv4
non-DF packets and the outgoing link mtu is too small. Its not perfect,
but given the lack of bug reports wrt. GRO fwd being broken this is a
rare case anyway. Also its not like this could not be improved later
once the dust settles.
Acked-by: Herbert Xu <herbert@gondor.apana.org.au> Reported-by: Marcelo Ricardo Leitner <mleitner@redhat.com> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Will be used by upcoming ipv4 forward path change that needs to
determine feature mask using skb->dst->dev instead of skb->dev.
Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / thirdparty / kernel / stable.git / log
