Occasionally, the PHY can be initially inaccessible when the first read of
a PHY register, e.g. PHY_ID1, happens (signified by the returned value
0xFFFF) but subsequent accesses of the PHY work as expected.  Add a retry
counter similar to how it is done in the generic e1000_get_phy_id().

Also, when the PHY is completely inaccessible (i.e. when subsequent reads
of the PHY_IDx registers returns all F's) and the MDIO access mode must be
set to slow before attempting to read the PHY ID again, the functions that
do these latter two actions expect the SW/FW/HW semaphore is not already
set so the semaphore must be released before and re-acquired after calling
them otherwise there is an unnecessarily inordinate amount of delay during
device initialization.
Reported-by: NHeikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NHeikki Krogerus <heikki.krogerus@linux.intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

SYNCH bit and IV bit of RXCW register are sticky. Before examining these bits,
RXCW should be read twice to filter out one-time false events and have correct
values for these bits. Incorrect values of these bits in link check logic can
cause weird link stability issues if auto-negotiation fails.

CC: stable <stable@vger.kernel.org> [2.6.38+]
Reported-by: NDean Nelson <dnelson@redhat.com>
Signed-off-by: NTushar Dave <tushar.n.dave@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NJeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

In rare cases, bnx2x_free_tx_skbs() can unmap the wrong DMA address
when it gets to the last entry of the tx ring.  We were not using
the proper macro to skip the last entry when advancing the tx index.
Reported-by: NZongyun Lai <zlai@vmware.com>
Reviewed-by: NJeffrey Huang <huangjw@broadcom.com>
Signed-off-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit db83d136 (gianfar: Fix missing sock reference when
processing TX time stamps) added a potential sk_wmem_alloc imbalance

If the new skb has a different truesize than old one, we can get a
negative sk_wmem_alloc once new skb is orphaned at TX completion.

Now we no longer early orphan skbs in dev_hard_start_xmit(), this
probably can lead to fatal bugs.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Tested-by: NPaul Gortmaker <paul.gortmaker@windriver.com>
Cc: Manfred Rudigier <manfred.rudigier@omicron.at>
Cc: Claudiu Manoil <claudiu.manoil@freescale.com>
Cc: Jiajun Wu <b06378@freescale.com>
Cc: Andy Fleming <afleming@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If list_for_each_entry, etc complete a traversal of the list, the iterator
variable ends up pointing to an address at an offset from the list head,
and not a meaningful structure.  Thus this value should not be used after
the end of the iterator.  There does not seem to be a meaningful value to
provide to netdev_warn.  Replace with pr_warn, since pr_err is used
elsewhere.

This problem was found using Coccinelle (http://coccinelle.lip6.fr/).
Signed-off-by: NJulia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For the higher mtu sizes requiring the buffer size greater than 8192,
the buffers are sent or received using multiple dma descriptors/ same
descriptor with option of multi buffer handling.
It was observed during tests that the driver was missing on data
packets during the normal ping operations if the data buffers being used
catered to jumbo frame handling.

The memory barrriers are added in between preparation of dma descriptors
in the jumbo frame handling path to ensure all instructions before
enabling the dma are complete.
Signed-off-by: NDeepak Sikri <deepak.sikri@st.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It was observed that during multiple reboots nfs hangs. The status of
receive descriptors shows that all the descriptors were in control of
CPU, and none were assigned to DMA.
Also the DMA status register confirmed that the Rx buffer is
unavailable.

This patch adds the fix for the same by adding the memory barriers to
ascertain that the all instructions before enabling the Rx or Tx DMA are
completed which involves the proper setting of the ownership bit in DMA
descriptors.
Signed-off-by: NDeepak Sikri <deepak.sikri@st.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit c0357e97
    bnx2: stop using net_device.{base_addr, irq}.

removed netdev->base_addr so we need to update cnic to get the MMIO
base address from pci_resource_start().  Otherwise, mmap of the uio
device will fail.
Signed-off-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

DCB and SR-IOV cannot currently be enabled at the same time as the queueing
schemes are incompatible.  If they are both enabled it will result in Tx
hangs since only the first Tx queue will be able to transmit any traffic.

This simple fix for this is to block us from enabling TCs in ixgbe_setup_tc
if SR-IOV is enabled.  This change will be reverted once we can support
SR-IOV and DCB coexistence.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Acked-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: NRoss Brattain <ross.b.brattain@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

some people report atl1c could cause system hang with following
kernel trace info:
---------------------------------------
WARNING: at.../net/sched/sch_generic.c:258 dev_watchdog+0x1db/0x1d0()
...
NETDEV WATCHDOG: eth0 (atl1c): transmit queue 0 timed out
...
---------------------------------------
This is caused by netif_stop_queue calling when cable Link is down.
So remove netif_stop_queue, because link_watch will take it over.
Signed-off-by: Nxiong <xiong@qca.qualcomm.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: NCloud Ren <cjren@qca.qualcomm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit a1c7fff7 (net: netdev_alloc_skb() use build_skb()) broke b44 on
some 64bit machines.

It appears b44 and b43 use __netdev_alloc_skb() instead of alloc_skb()
for their bounce buffers.

There is no need to add an extra NET_SKB_PAD reservation for bounce
buffers :

- In TX path, NET_SKB_PAD is useless

- In RX path in b44, we force a copy of incoming frames if
  GFP_DMA allocations were needed.
Reported-and-bisected-by: NStefan Bader <stefan.bader@canonical.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently only used when packet split mode is enabled with jumbo frames,
IP payload checksum (for fragmented UDP packets) is mutually exclusive with
receive hashing offload since the hardware uses the same space in the
receive descriptor for the hardware-provided packet checksum and the RSS
hash, respectively.  Users currently must disable jumbos when receive
hashing offload is enabled, or vice versa, because of this incompatibility.
Since testing has shown that IP payload checksum does not provide any real
benefit, just remove it so that there is no longer a choice between jumbos
or receive hashing offload but not both as done in other Intel GbE drivers
(e.g. e1000, igb).

Also, add a missing check for IP checksum error reported by the hardware;
let the stack verify the checksum when this happens.

CC: stable <stable@vger.kernel.org> [3.4]
Signed-off-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Using ethtool -C ethX rx-usecs 0 crashes with a divide by zero.
Refactor this function to fix this issue and make it more clear
what the intent of each conditional is. Add comment regarding
using a setting of zero.

CC: stable <stable@vger.kernel.org> [3.3+]
CC: David Ahern <daahern@cisco.com>
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This fixes a number of warnings such as:

  CC      drivers/net/ethernet/ti/davinci_cpdma.o
drivers/net/ethernet/ti/davinci_cpdma.c:279:1: warning: data definition
has no type or storage class
drivers/net/ethernet/ti/davinci_cpdma.c:279:1: warning: type defaults to
‘int’ in declaration of ‘EXPORT_SYMBOL_GPL’
drivers/net/ethernet/ti/davinci_cpdma.c:279:1: warning: parameter names
(without types) in function declaration
Signed-off-by: NDaniel Mack <zonque@gmail.com>
Cc: Vaibhav Hiremath <hvaibhav@ti.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Christian Riesch <christian.riesch@omicron.at>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The correct behavior is to program the interrupt coalescing regs
(RXICr/TXICr) in accordance with the Rx/Tx Q's "rx/txcoalescing"
flag. That is, if the coalescing flag is 0 for a given Rx/Tx queue
then the corresponding coalescing register should be cleared.
This behavior is correctly implemented for the single-queue mode
(SQ_SG_MODE), but not for the multi-queue mode (MQ_MG_MODE).
This fixes the later case.
Signed-off-by: NClaudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

FCoE target mode was experiencing issues due to the fact that we were
sending up data frames that were padded to 60 bytes after the DDP logic had
already stripped the frame down to 52 or 56 depending on the use of VLANs.
This was resulting in the FCoE DDP logic having issues since it thought the
frame still had data in it due to the padding.

To resolve this, adding code so that we do not pad FCoE frames prior to
handling them to the stack.

CC: <stable@vger.kernel.org>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: NRoss Brattain <ross.b.brattain@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add a missing resource release in ring cleanup.
Not doing this leaves a range of QPs that are being reserved,
and no one can use them.
Signed-off-by: NYevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix a crash at the error flow of NOP command which caused the driver to try and use
a completion vector which wasn't allocated.
Signed-off-by: NYevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Set valid port parameters: MTU and flow control configuration when
configuring the port during HW device initialization,
prior to the net device open() being called.
Using  invalid parameters (such as all zeros)
could lead to bad firmware behavior.
Signed-off-by: NYevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The following commit couldn't work if the RMCR is not set to 1.

"net: sh_eth: fix the rxdesc pointer when rx descriptor empty happens"
commit id 79fba9f5

If RMCR is not set, the controller will clear the EDRRR after it received
a frame. In this case, the driver doesn't need to fix the value of
cur_rx/dirty_rx. The driver only needs it when the controll detects
receive descriptors are empty.
Signed-off-by: NYoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The 8168evl (RTL_GIGA_MAC_VER_34) based Gigabyte GA-990FXA motherboards
are very prone to NETDEV watchdog problems without this change. See
https://bugzilla.kernel.org/show_bug.cgi?id=42899 for instance.

I don't know why it *works*. It's depressingly effective though.

For the record:
- the problem may go along IOMMU (AMD-Vi) errors but it really looks
  like a red herring.
- the patch sets the RX_MULTI_EN bit. If the 8168c doc is any guide,
  the chipset now fetches several Rx descriptors at a time.
- long ago the driver ignored the RX_MULTI_EN bit.
  e542a226 changed the RxConfig
  settings. Whatever the problem it's now labeled a regression.
- Realtek's own driver can identify two different 8168evl devices
  (CFG_METHOD_16 and CFG_METHOD_17) where the r8169 driver only
  sees one. It sucks.
Signed-off-by: NFrancois Romieu <romieu@fr.zoreil.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes a memory leak that was introduced in the 3.4 kernel.  The
leak occurred when FCoE was enabled and traffic was passed over the FCoE
rings reserved for FCoE.  The memory leak was due to us not populating the
compound page information on the order 1 pages needed for FCoE.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: NRoss Brattain <ross.b.brattain@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Fix Kconfig file to make sure that PTP and IGB/IXGBE are both either
in-kernel or modules, not mixed.  Having the build status mixed causes
compile errors.
Signed-off-by: NCarolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: NJeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

i210/i211 device has only 16 RAR address filters like 82575, instead of
32 like i350.  This patch removes the entries for i210/i211 in the
get_invariants function which was setting them for 32. This ensures that
they will get the default value which is the correct one.
Signed-off-by: NCarolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: NJeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Increasing the hardware statistics structure to accomodate statistics for skyhawk.
Signed-off-by: NVasundhara Volam <vasundhara.volam@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Modify IOCTL error message to print subsystem also.
Signed-off-by: NVasundhara Volam <vasundhara.volam@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The maximum size of packet that can be handled by controller including ethernet
header is 65535. Reducing gso_max_size accordingly.
Signed-off-by: NSarveshwar Bandi <sarveshwar.bandi@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes a potential hole when configuring the cycle counter used to
generate the nanosecond time clock. This clock is based off of the SYSTIME
registers along with the TIMINCA registers. The TIMINCA register determines
the increment to be added to the SYSTIME registers every DMA clock tick. This
register needs to be reconfigured whenever the link-speed changes. However,
the value calculated stays the same when link is down and when link is up.
Misconfiguration can occur if the link status changes due to a reset, which
causes the TIMINCA register to be reset. This reset puts the device in an
unstable state where the SYSTIME registers stop incrementing and the PTP
protocol does not function.

The solution is to double check the TIMINCA value and always reset the value
if the register is zero. This prevents a misconfiguration bug that halts the
PHC.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Acked-by: NDon Skidmore <donald.c.skidmore@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NYaniv Rosner <yaniv.rosner@broadcom.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When I2C is not responding it's usually due to a previous
unexpected reset during I2C operation. We release it by
powering down and up the SFP+ module.
Signed-off-by: NYaniv Rosner <yaniv.rosner@broadcom.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The adapter->npars[] array has QLCNIC_MAX_PCI_FUNC elements.  We
allocate it that way a few lines earlier in the function.  So this test
is off by one.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NAnirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is a off by one error in the minimal number of BD in
bnx2x_start_xmit() and bnx2x_tx_int() before stopping/resuming tx queue.

A full size GSO packet, with data included in skb->head really needs
(MAX_SKB_FRAGS + 4) BDs, because of bnx2x_tx_split()

This error triggers if BQL is disabled and heavy TCP transmit traffic
occurs.

bnx2x_tx_split() definitely can be called, remove a wrong comment.
Reported-by: NTomas Hruby <thruby@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Cc: Yaniv Rosner <yanivr@broadcom.com>
Cc: Merav Sicron <meravs@broadcom.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Robert Evans <evansr@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bnx2x driver incorrectly sets ip_summed to CHECKSUM_UNNECESSARY on
encapsulated segments. TCP stack happily accepts frames with bad
checksums, if they are inside a GRE or IPIP encapsulation.

Our understanding is that if no IP or L4 csum validation was done by the
hardware, we should leave ip_summed as is (CHECKSUM_NONE), since
hardware doesn't provide CHECKSUM_COMPLETE support in its cqe.

Then, if IP/L4 checksumming was done by the hardware, set
CHECKSUM_UNNECESSARY if no error was flagged.

Patch based on findings and analysis from Robert Evans
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Cc: Yaniv Rosner <yanivr@broadcom.com>
Cc: Merav Sicron <meravs@broadcom.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Robert Evans <evansr@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Acked-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This change adds support for the tilegx network driver based on the
GXIO IORPC support in the tilegx software stack, using the on-chip
mPIPE packet processing engine.
Signed-off-by: NChris Metcalf <cmetcalf@tilera.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

5906 devices also need the short DMA fragment workaround.  This patch
makes the necessary change.
Signed-off-by: NMatt Carlson <mcarlson@broadcom.com>
Tested-by: NChristian Kujau <lists@nerdbynature.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

clk_{un}prepare is mandatory for platforms using common clock framework.
Since these drivers are used by SPEAr platform, which supports common
clock framework, add clk_{un}prepare() support for them. Otherwise
the clocks are not correctly en-/disabled and ethernet support doesn't
work.
Signed-off-by: NStefan Roese <sr@denx.de>
Cc: Viresh Kumar <viresh.linux@gmail.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

__lpc_handle_xmit() has two bugs :

1) It can leak skbs in case TXSTATUS_ERROR is set

2) It can wake up txqueue while no slot was freed.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NRoland Stigge <stigge@antcom.de>
Tested-by: NRoland Stigge <stigge@antcom.de>
Cc: Kevin Wells <kevin.wells@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

lpc_eth does a copy of transmitted skbs to DMA area, without checking
skb lengths, so can trigger buffer overflows :

memcpy(pldat->tx_buff_v + txidx * ENET_MAXF_SIZE, skb->data, len);

One way to get bigger skbs is to allow MTU changes above the 1500 limit.

Calling eth_change_mtu() in ndo_change_mtu() makes sure this cannot
happen.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Roland Stigge <stigge@antcom.de>
Cc: Kevin Wells <kevin.wells@nxp.com>
Acked-by: NRoland Stigge <stigge@antcom.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While reworking the r8169 driver a few months ago to perform the
smallest amount of work in the irq handler, I took care of avoiding
any irq mask register operation in the slow work dedicated user
context thread. The slow work thread scheduled an extra round of NAPI
work which would ultimately set the irq mask register as required,
thus keeping such irq mask operations in the NAPI handler.
It would eventually race with the irq handler and delay NAPI execution
for - assuming no further irq - a whole ksoftirqd period. Mildly a
problem for rare link changes or corner case PCI events.

The race was always lost after the last bh disabling lock had been
removed from the work thread and people started wondering where those
pesky "NOHZ: local_softirq_pending 08" messages came from.

Actually the irq mask register _can_ be set up directly in the slow
work thread.
Signed-off-by: NFrancois Romieu <romieu@fr.zoreil.com>
Reported-by: NDave Jones <davej@redhat.com>
Tested-by: NMarc Dionne <marc.c.dionne@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As soon as hardware is notified of a transmit, we no longer can assume
skb can be dereferenced, as TX completion might have freed the packet.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>