dax_pmem_percpu_exit() waits for dax_pmem_percpu_release() to invoke the
dax_pmem->cmp completion.  Unfortunately this approach to cleaning up
the percpu_ref only works after devm_memremap_pages() was successful.

If devm_add_action_or_reset() or devm_memremap_pages() fails,
dax_pmem_percpu_release() is not invoked.  Therefore
dax_pmem_percpu_exit() hangs waiting for the completion:

  rc = devm_add_action_or_reset(dev, dax_pmem_percpu_exit,
  				&dax_pmem->ref);
  if (rc)
  	return rc;

  dax_pmem->pgmap.ref = &dax_pmem->ref;
  addr = devm_memremap_pages(dev, &dax_pmem->pgmap);

Avoid the hang by calling percpu_ref_exit() in the error paths instead
of going through dax_pmem_percpu_exit().
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

Function persistent_memory_claim doesn't need to get local pointer
dummy_addr from direct_access. Using NULL instead of having to pass
in a useless local pointer that caller then just throw away.
Suggested-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NHuaisheng Ye <yehs1@lenovo.com>
Reviewed-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Acked-by: NMike Snitzer <snitzer@redhat.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

Function __bdev_dax_supported doesn't need to get local pointer kaddr
from direct_access. Using NULL instead of having to pass in a useless
local pointer that caller then just throw away.
Signed-off-by: NHuaisheng Ye <yehs1@lenovo.com>
Reviewed-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

dcssblk_direct_access() needs to check the validity of pointers kaddr
and pfn for NULL assignment. If anyone equals to NULL, it doesn't need
to calculate the value.

If either of them is equal to NULL, that is to say callers may
have no need for kaddr or pfn, so this patch is prepared for allowing
them to pass in NULL instead of having to pass in a pointer or local
variable that they then just throw away.
Signed-off-by: NHuaisheng Ye <yehs1@lenovo.com>
Reviewed-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Acked-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: NDave Jiang <dave.jianig@intel.com>

pmem_direct_access() needs to check the validity of pointers kaddr
and pfn for NULL assignment. If anyone equals to NULL, it doesn't need
to calculate the value.

If pointer equals to NULL, that is to say callers may have no need for
kaddr or pfn, so this patch is prepared for allowing them to pass in
NULL instead of having to pass in a pointer or local variable that
they then just throw away.
Signed-off-by: NHuaisheng Ye <yehs1@lenovo.com>
Reviewed-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

When the ACPI UC error notifier gets called and ARS_REQ bit is set
with the passed in flag, we can receive -EBUSY if ARS_REQ bit is already
set for the nfit_spa->ars_state. When that happens, the ARS request is
dropped. That can potentially cause us to miss the unreported errors that
the on going ARS request does not receive. Add an ARS_REQ_REDO state that
will request short ARS upon ARS completion to grab any errors we missed.
Signed-off-by: NDave Jiang <dave.jiang@intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>

The 'available_size' attribute showing the combined total of all
unallocated space isn't always useful to know how large of a namespace
a user may be able to allocate if the region is fragmented. This patch
will export the largest extent of unallocated space that may be allocated
to create a new namespace.
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

This patch will find the max contiguous area to determine the largest
pmem namespace size that can be created. If the requested size exceeds
the largest available, ENOSPC error will be returned.

This fixes the allocation underrun error and wrong error return code
that have otherwise been observed as the following kernel warning:

  WARNING: CPU: <CPU> PID: <PID> at drivers/nvdimm/namespace_devs.c:913 size_store

Fixes: a1f3e4d6 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support")
Cc: <stable@vger.kernel.org>
Signed-off-by: NKeith Busch <keith.busch@intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDave Jiang <dave.jiang@intel.com>

The _LSR method indicates locked status via error-code-3 returned in the
_LSR payload. When any error is returned the payload of _LSR is
truncated to a zero-length buffer.

The _DSM path in comparison allows system software to retrieve the
locked status *and* namespace label area contents.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

When a DIMM is locked its namespace label area may not be. Introduce the
distinction of locked namespaces to allow namespace enumeration while
the capacity is locked.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

cmd_rc is passed in by reference to the acpi_nfit_ctl() function and the
caller expects a value returned. However, when the package is pass through
via the ND_CMD_CALL command, cmd_rc is not touched. Make sure cmd_rc is
always set.

Fixes: aef25338 ("libnvdimm, nfit: centralize command status translation")
Signed-off-by: NDave Jiang <dave.jiang@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

This is easily triggered from userspace, so let's ratelimit the
messages.
Signed-off-by: NJeff Moyer <jmoyer@redhat.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Commit 60622d68 "x86/asm/memcpy_mcsafe: Return bytes remaining"
converted callers of memcpy_mcsafe() to expect a positive 'bytes
remaining' value rather than a negative error code. The nsio_rw_bytes()
conversion failed to return success. The failure is benign in that
nsio_rw_bytes() will end up writing back what it just read.

Fixes: 60622d68 ("x86/asm/memcpy_mcsafe: Return bytes remaining")
Cc: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Remove dependencies on HAS_DMA where a Kconfig symbol depends on another
symbol that implies HAS_DMA, and, optionally, on "|| COMPILE_TEST".
In most cases this other symbol is an architecture or platform specific
symbol, or PCI.

Generic symbols and drivers without platform dependencies keep their
dependencies on HAS_DMA, to prevent compiling subsystems or drivers that
cannot work anyway.

This simplifies the dependencies, and allows to improve compile-testing.
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: NMark Brown <broonie@kernel.org>
Acked-by: NRobin Murphy <robin.murphy@arm.com>
Reviewed-by: NMatias Bjørling <mb@lightnvm.io>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Enabling LPIs was made a lot stricter recently, by checking that they are
disabled before enabling them. By doing so, the CPU hotplug case was missed
altogether, which leaves LPIs enabled on hotplug off (expecting the CPU to
eventually come back), and won't write a different value anyway on hotplug
on.

So skip that check if that particular case is detected

Fixes: 6eb486b6 ("irqchip/gic-v3: Ensure GICR_CTLR.EnableLPI=0 is observed before enabling")
Reported-by: NSumit Garg <sumit.garg@linaro.org>
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Tested-by: NSumit Garg <sumit.garg@linaro.org>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lkml.kernel.org/r/20180622095254.5906-8-marc.zyngier@arm.com

Similarily to the SYNC operation, it must be verified that the VPE
targetted by a VLPI is backed by a valid collection in the GIC driver data
structures.
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Yang Yingliang <yangyingliang@huawei.com>
Cc: Sumit Garg <sumit.garg@linaro.org>
Link: https://lkml.kernel.org/r/20180622095254.5906-7-marc.zyngier@arm.com

It is possible, under obscure circumstances, to convince the ITS driver to
emit a SYNC operation that targets a collection that is not bound to any
redistributor (and the target_address field is zero) because the
corresponding CPU has not been seen yet (the system has been booted with
max_cpus="something small").

If the ITS is using the linear CPU number as the target, this is not a big
deal, as we just end-up issuing a SYNC to CPU0. But if the ITS requires the
physical address of the redistributor (with GITS_TYPER.PTA==1), we end-up
asking the ITS to write to the physical address zero, which is not exactly
a good idea (there has been report of the ITS locking up). This should of
course never happen, but hey, this is SW...

In order to avoid the above disaster, let's track which collections have
been actually initialized, and let's not generate a SYNC if the collection
hasn't been properly bound to a redistributor.  Take this opportunity to
spit our a warning, in the hope that someone may report the issue if it
arrises again.
Reported-by: NYang Yingliang <yangyingliang@huawei.com>
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Sumit Garg <sumit.garg@linaro.org>
Link: https://lkml.kernel.org/r/20180622095254.5906-6-marc.zyngier@arm.com

On a NUMA system, if an ITS is local to an offline node, the ITS driver may
pick an offline CPU to bind the LPI.  In this case, pick an online CPU (and
the first one will do).

But on some systems, binding an LPI to non-local node CPU may cause
deadlock (see Cavium erratum 23144).  In this case, just fail the activate
and return an error code.
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Sumit Garg <sumit.garg@linaro.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20180622095254.5906-5-marc.zyngier@arm.com

On failing to allocate the required SPIs, the actual number of interrupts
should be freed and not its log2 value.

Fixes: de337ee3 ("irqchip/gic-v2m: Add PCI Multi-MSI support")
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Yang Yingliang <yangyingliang@huawei.com>
Cc: Sumit Garg <sumit.garg@linaro.org>
Link: https://lkml.kernel.org/r/20180622095254.5906-4-marc.zyngier@arm.com

The ls-scfs-msi driver is not dealing with the effective affinity
as it should. Let's fix that, and make it clear that the effective
affinity is restricted to a single CPU. Also prevent the driver from
messing with the internals of the affinity setting infrastructure.
Reported-by: NAlexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: NMarc Zyngier <marc.zyngier@arm.com>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Tested-by: NAlexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Yang Yingliang <yangyingliang@huawei.com>
Cc: Sumit Garg <sumit.garg@linaro.org>
Link: https://lkml.kernel.org/r/20180622095254.5906-3-marc.zyngier@arm.com

Commit:

  79832f0b ("efi/libstub/tpm: Initialize pointer variables to zero for mixed mode")

fixes a problem with the tpm code on mixed mode (64-bit kernel on 32-bit UEFI),
where 64-bit pointer variables are not fully initialized by the 32-bit EFI code.

A similar problem applies to the efi_physical_addr_t variables which
are written by the ->get_event_log() EFI call. Even though efi_physical_addr_t
is 64-bit everywhere, it seems that some 32-bit UEFI implementations only
fill in the lower 32 bits when passed a pointer to an efi_physical_addr_t
to fill.

This commit initializes these to 0 to, to ensure the upper 32 bits are
0 in mixed mode. This fixes recent kernels sometimes hanging during
early boot on mixed mode UEFI systems.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Cc: <stable@vger.kernel.org> # v4.16+
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-efi@vger.kernel.org
Link: http://lkml.kernel.org/r/20180622064222.11633-2-ard.biesheuvel@linaro.orgSigned-off-by: NIngo Molnar <mingo@kernel.org>

Commit 910f8bef ("xen/pirq: fix error path cleanup when binding
MSIs") fixed a couple of errors in error cleanup path of
xen_bind_pirq_msi_to_irq(). This cleanup allowed a call to
__unbind_from_irq() with an unbound irq, which would result in
triggering the BUG_ON there.

Since there is really no reason for the BUG_ON (xen_free_irq() can
operate on unbound irqs) we can remove it.
Reported-by: NBen Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: NBoris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: stable@vger.kernel.org
Reviewed-by: NJuergen Gross <jgross@suse.com>
Signed-off-by: NJuergen Gross <jgross@suse.com>

For passing arbitrary data from user land to the Xen hypervisor the
Xen tools today are using mlock()ed buffers. Unfortunately the kernel
might change access rights of such buffers for brief periods of time
e.g. for page migration or compaction, leading to access faults in the
hypervisor, as the hypervisor can't use the locks of the kernel.

In order to solve this problem add a new device node to the Xen privcmd
driver to easily allocate hypercall buffers via mmap(). The memory is
allocated in the kernel and just mapped into user space. Marked as
VM_IO the user mapping will not be subject to page migration et al.
Signed-off-by: NJuergen Gross <jgross@suse.com>
Reviewed-by: NBoris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: NJuergen Gross <jgross@suse.com>

nvme requires an sg table allocation for each request. If the request
is large, then the allocation can become quite large. For instance,
with our default software settings of 1280KB IO size, we'll need
10248 bytes of sg table. That turns into a 2nd order allocation,
which we can't always guarantee. If we fail the allocation, blk-mq
will retry it later. But there's no guarantee that we'll EVER be
able to allocate that much contigious memory.

Limit the IO size such that we never need more than a single page
of memory. That's a lot faster and more reliable. Then back that
allocation with a mempool, so that we know we'll always be able
to succeed the allocation at some point.
Signed-off-by: NJens Axboe <axboe@kernel.dk>
Acked-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

There is race between nvme_remove and nvme_reset_work that can
lead to io hang.

nvme_remove                    nvme_reset_work
                               -> nvme_remove_dead_ctrl
                                 -> nvme_dev_disable
                                   -> quiesce request_queue
                                 -> queue remove_work
-> cancel_work_sync reset_work
-> nvme_remove_namespaces
  -> splice ctrl->namespaces
                               nvme_remove_dead_ctrl_work
                               -> nvme_kill_queues
  -> nvme_ns_remove               do nothing
    -> blk_cleanup_queue
      -> blk_freeze_queue

Finally, the request_queue is quiesced state when wait freeze,
we will get io hang here. To fix it, move the nvme_kill_queues
from nvme_remove_dead_ctrl_work to nvme_remove_dead_ctrl.
Suggested-by: NKeith Busch <keith.busch@linux.intel.com>
Signed-off-by: NJianchao Wang <jianchao.w.wang@oracle.com>
Reviewed-by: NKeith Busch <keith.busch@intel.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Currently packed pixel modes in MHL2 can't be displayed. The device
automatically recognizes output format, so setting format other than
RGB causes failure. Fix it by writing proper values to registers.

Tested on MHL1 and MHL2 using various vendors' dongles both in
DVI and HDMI mode.
Signed-off-by: NMaciej Purski <m.purski@samsung.com>
Signed-off-by: NAndrzej Hajda <a.hajda@samsung.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1516706239-9104-1-git-send-email-m.purski@samsung.com

Rather than leaving io queues quiesced after tearing down an association,
restart them. This allows ios to be replayed, with fastfail ios terminating
and non-fastfail getting into loops of retry.

This follows rdma's lead.
Signed-off-by: NJames Smart <james.smart@broadcom.com>
Reviewed-by: NSagi Grimberg <sagi@grimber.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

If NBD_DISCONNECT_ON_CLOSE is set on a device, then the driver will
issue a disconnect from nbd_release if the device has no remaining
bdev->bd_openers.

Fix ret val so reconfigure with only setting the flag succeeds.
Reviewed-by: NJosef Bacik <josef@toxicpanda.com>
Signed-off-by: NDoron Roberts-Kedes <doronrk@fb.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Controllers that are not yet enabled should not really enforce keep alive
timeouts, but we still want to track a timeout and cleanup in case a host
died before it enabled the controller.  Hence, simply reset the keep
alive timer when the controller is enabled.
Suggested-by: NMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

That is user argument, and theoretically controller limits can change
over time (over reconnects/resets).  Instead, use the sqsize controller
attribute to check queue depth boundaries and use it to the tagset
allocation.
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The race is between completing the request at error recovery work and
rdma completions.  If we cancel the request before getting the good
rdma completion we get a NULL deref of the request MR at
nvme_rdma_process_nvme_rsp().

When Canceling the request we return its mr to the mr pool (set mr to
NULL) and also unmap its data.  Canceling the requests while the rdma
queues are active is not safe.  Because rdma queues are active and we
get good rdma completions that can use the mr pointer which may be NULL.
Completing the request too soon may lead also to performing DMA to/from
user buffers which might have been already unmapped.

The commit fixes the race by draining the QP before starting the abort
commands mechanism.
Signed-off-by: NIsrael Rukshin <israelr@mellanox.com>
Reviewed-by: NMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

If nvme_rdma_configure_admin_queue fails before we allocated
the async event buffer, we will falsly free it because
nvme_rdma_free_queue is freeing it. Fix it by allocating the buffer right
after nvme_rdma_alloc_queue and free it right before nvme_rdma_queue_free
to maintain orderly reverse cleanup sequence.
Reported-by: NIsrael Rukshin <israelr@mellanox.com>
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Reviewed-by: NMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Failures after nvme_init_ctrl will defer resource cleanups to .free_ctrl
when the reference is released, hence we should not free the controller
queues for these failures.

Fix that by moving controller queues allocation before controller
initialization and correctly freeing them for failures before
initialization and skip them for failures after initialization.
Signed-off-by: NSagi Grimberg <sagi@grimberg.me>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The HID descriptor for the 2nd-gen Intuos Pro large (PTH-860) contains
a typo which defines an incorrect logical maximum Y value. This causes
a small portion of the bottom of the tablet to become unusable (both
because the area is below the "bottom" of the tablet and because
'wacom_wac_event' ignores out-of-range values). It also results in a
skewed aspect ratio.

To fix this, we add a quirk to 'wacom_usage_mapping' which overwrites
the data with the correct value.
Signed-off-by: NJason Gerecke <jason.gerecke@wacom.com>
CC: stable@vger.kernel.org # v4.10+
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

Current ISH driver only registers suspend/resume PM callbacks which don't
support hibernation (suspend to disk). Basically after hiberation, the ISH
can't resume properly and user may not see sensor events (for example: screen
		rotation may not work).

User will not see a crash or panic or anything except the following message
in log:

	hid-sensor-hub 001F:8086:22D8.0001: timeout waiting for response from ISHTP device

So this patch adds support for S4/hiberbation to ISH by using the
SIMPLE_DEV_PM_OPS() MACRO instead of struct dev_pm_ops directly. The suspend
and resume functions will now be used for both suspend to RAM and hibernation.

If power management is disabled, SIMPLE_DEV_PM_OPS will do nothing, the suspend
and resume related functions won't be used, so mark them as __maybe_unused to
clarify that this is the intended behavior, and remove #ifdefs for power
management.

Cc: stable@vger.kernel.org
Signed-off-by: NEven Xu <even.xu@intel.com>
Acked-by: NSrinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

When creating the low-level hidraw device, the reference to steam_device
was stored using hid_set_drvdata(). But this value is not guaranteed to
be kept when set before calling probe. If this pointer is reset, it
crashes when opening the emulated hidraw device.

It looks like hid_set_drvdata() is for users "avobe" this hid_device,
while hid_device.driver_data it for users "below" this one.

In this case, we are creating a virtual hidraw device, so we must use
hid_device.driver_data.
Signed-off-by: NRodrigo Rivas Costa <rodrigorivascosta@gmail.com>
Tested-by: NMariusz Ceier <mceier+kernel@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

HW does not support Half-duplex mode in multi-queue
scenario. Fix it by not advertising the Half-Duplex
mode if multi-queue enabled.
Signed-off-by: NBhadram Varka <vbhadram@nvidia.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The Stratix10 platform has an additional reset line, OCP(Open Core Protocol),
that also needs to get deasserted for the stmmac ethernet controller to work.
Thus we need to update the Kconfig to include ARCH_STRATIX10 in order to build
dwmac-socfpga.

Also, remove the redundant check for the reset controller pointer. The
reset driver already checks for the pointer and returns 0 if the pointer
is NULL.
Signed-off-by: NDinh Nguyen <dinguyen@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After commit 88078d98 ("net: pskb_trim_rcsum() and CHECKSUM_COMPLETE
are friends"), sungem owners reported the infamous "eth0: hw csum failure"
message.

CHECKSUM_COMPLETE has in fact never worked for this driver, but this
was masked by the fact that upper stacks had to strip the FCS, and
therefore skb->ip_summed was set back to CHECKSUM_NONE before
my recent change.

Driver configures a number of bytes to skip when the chip computes
the checksum, and for some reason only half of the Ethernet header
was skipped.

Then a second problem is that we should strip the FCS by default,
unless the driver is updated to eventually support NETIF_F_RXFCS in
the future.

Finally, a driver should check if NETIF_F_RXCSUM feature is enabled
or not, so that the admin can turn off rx checksum if wanted.

Many thanks to Andreas Schwab and Mathieu Malaterre for their
help in debugging this issue.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NMeelis Roos <mroos@linux.ee>
Reported-by: NMathieu Malaterre <malat@debian.org>
Reported-by: NAndreas Schwab <schwab@linux-m68k.org>
Tested-by: NAndreas Schwab <schwab@linux-m68k.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The code does:

 if (hrtimer_active(&t))
    hrtimer_cancel(&t);

However, hrtimer_cancel() checks if the timer is active, so the
test above is pointless.
Signed-off-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>