We assume that the mapping between DMA and virtual addresses is done
on whole pages, so we can find the page offset of an RX buffer using
the lower bits of the DMA address.  However, swiotlb maps in units of
2K, breaking this assumption.

Add an explicit page_offset field to struct efx_rx_buffer.
Signed-off-by: NBen Hutchings <bhutchings@solarflare.com>

We may currently allocate two RX DMA buffers to a page, and only unmap
the page when the second is completed.  We do not sync the first RX
buffer to be completed; this can result in packet loss or corruption
if the last RX buffer completed in a NAPI poll is the first in a page
and is not DMA-coherent.  (In the middle of a NAPI poll, we will
handle the following RX completion and unmap the page *before* looking
at the content of the first buffer.)
Signed-off-by: NBen Hutchings <bhutchings@solarflare.com>

We were using the same variable for iterating two nested loops.
Reported-by: NTijs Van Buggenhout <tvb@able.be>
Signed-off-by: NRafał Miłecki <zajec5@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix missing () & { }
Signed-off-by: NSyam Sidhardhan <s.syam@samsung.com>
Acked-by: NOlof Johansson <olof@lixom.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Sparse warned about several functions that were unnecessarily global.
After making them static, discovered that several functions were actually never used.

Compile tested only.
Signed-off-by: NStephen Hemminger <stephen@networkplumber.org>
Acked-by: NEilon Greenstein <eilong@broadcomo.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit ee873fda ("gianfar: Pack struct
gfar_priv_grp into three cachelines") moved the irq number and names
off into a separate struct and created accessors for them.  However
it was never tested with NET_POLL enabled, and so some conversions
that were simply overlooked went undetected until now.

Make the netpoll ones also use the gfar_irq() accessors.
Reported-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Claudiu Manoil <claudiu.manoil@freescale.com>
Cc: Jianhua Xie <jianhua.xie@freescale.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The mlx4_en driver allocates the number of objects for the CPU affinity
reverse-map based on the number of rx rings of the device. However,
mlx4_assign_eq() calls irq_cpu_rmap_add() as many times as IRQ's are
assigned to EQ's, which can be as large as mlx4_dev->caps.comp_pool. If
caps.comp_pool is larger than rx_ring_num we will eventually hit the
BUG_ON() in cpu_rmap_add().

Fix this problem by allocating space for the maximum number of CPU
affinity reverse-map objects we might want to add.
Signed-off-by: NKleber Sacilotto de Souza <klebers@linux.vnet.ibm.com>
Acked-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory to hold the network device tx_cq is not being allocated with
the correct size in mlx4_en_init_netdev(). It should use MAX_TX_RINGS
instead of MAX_RX_RINGS. This can cause problems if the number of tx
rings being used is greater than MAX_RX_RINGS.
Signed-off-by: NKleber Sacilotto de Souza <klebers@linux.vnet.ibm.com>
Acked-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit 7f7d6c28 (net: fec: Ensure that initialization is done prior to
request_irq()) placed fec_ptp_init() into a point that ptp clock was not
available, which causes a division by zero in fec_ptp_start_cyclecounter():

[   17.895723] Division by zero in kernel.
[   17.899571] Backtrace:
[   17.902094] [<80012564>] (dump_backtrace+0x0/0x10c) from [<8056deec>]
(dump_stack+0x18/0x1c)
[   17.910539]  r6:bfba8500 r5:8075c950 r4:bfba8000 r3:bfbd0000
[   17.916284] [<8056ded4>] (dump_stack+0x0/0x1c) from [<80012688>]
(__div0+0x18/0x20)
[   17.923968] [<80012670>] (__div0+0x0/0x20) from [<802829c4>] (Ldiv0+0x8/0x10)
[   17.931140] [<80398534>] (fec_ptp_start_cyclecounter+0x0/0x110) from
[<80394f64>] (fec_restart+0x6c8/0x754)
[   17.940898] [<8039489c>] (fec_restart+0x0/0x754) from [<803969a0>]
(fec_enet_adjust_link+0xdc/0x108)
[   17.950046] [<803968c4>] (fec_enet_adjust_link+0x0/0x108) from [<80390bc4>]
(phy_state_machine+0x178/0x534)
...

Fix this by rearraging the code so that fec_ptp_init() is called only after
the clocks have been properly acquired.

Tested on both mx53 and mx6 platforms.
Reported-by: NJim Baxter <jim_baxter@mentor.com>
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NManish Chopra <manish.chopra@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Use correct function to clear loopback mode.
Signed-off-by: NShahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently request_irq() is called prior to fec_enet_init() and fec_ptp_init(),
which causes the following crash on a mx53qsb:

Unable to handle kernel NULL pointer dereference at virtual address 00000002
pgd = 80004000
[00000002] *pgd=00000000
Internal error: Oops: 5 [#1] SMP ARM
Modules linked in:
CPU: 0    Not tainted  (3.8.0-rc7-next-20130215+ #346)
PC is at fec_enet_interrupt+0xd0/0x348
LR is at fec_enet_interrupt+0xb8/0x348
pc : [<80372b7c>]    lr : [<80372b64>]    psr: 60000193
sp : df855c20  ip : df855c20  fp : df855c74
r10: 00000516  r9 : 1c000000  r8 : 00000000
r7 : 00000000  r6 : 00000000  r5 : 00000000  r4 : df9b7800
r3 : df9b7df4  r2 : 00000000  r1 : 00000000  r0 : df9b7d34

Ensure that such initialization functions are called prior to requesting the
interrupts, so that all necessary the data structures are in place when the
irqs occur.
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

in the previous commit : f1f220ea1dda078, the BUSY state of buffer is wrongly
deleted. this patch just restore it.
Signed-off-by: Nxiong <xiong@qca.qualcomm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Without this patch b44 always allocates the 2 bytes needed for aligned
access on every platform, now it uses netdev_alloc_skb_ip_align().
Signed-off-by: NHauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Check whether the phy-reset GPIO is valid, prior to requesting it.

In the case a board does not provide a phy-reset GPIO, just returns immediately.

With such gpio validation in place, it is also safe to change from pr_debug to
dev_err in the case the gpio request fails.
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Delete the MAC address of a VM, from the adapter's embedded switch,
after the VM had been migrated out of this adapter/server.
Signed-off-by: NShahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: NRajesh Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Set gso_size to MSS obtained from adapter to avoid incorrect estimation
of receive MSS, which would lead to delayed ACKs in some traffic patterns

Example:
Send two or three packets and wait for ack and only then send
remaining packets.
Signed-off-by: NShahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Cleanly seperate 83xx Legacy interrupt handling code from 82xx
o Update 83xx Legacy interrupt handling code to match with the spec
Signed-off-by: NHimanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Config interrupt is not needed for mailbox interrupts.
Signed-off-by: NSucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Added detailed error messages for FW CDRP command failure
Signed-off-by: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

CPDMA interrupts are not properly acknowledged which leads to interrupt
storm, only cpdma interrupt 0 is acknowledged in Davinci CPDMA driver.
Changed cpdma_ctlr_eoi api to acknowledge 1 and 2 interrupts which are
used for rx and tx respectively.
Reported-by: NPantelis Antoniou <panto@antoniou-consulting.com>
Signed-off-by: NMugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Upper case macros for various chip attributes are slightly
difficult to read and are a bit out of characterto the other
tg3_<foo> attribute functions.

Convert:

GET_ASIC_REV(tp->pci_chip_rev_id)       -> tg3_asic_rev(tp)
GET_CHIP_REV(tp->pci_chip_rev_id)       -> tg3_chip_rev(tp)

Remove:
GET_METAL_REV(tp->pci_chip_rev_id)      -> tg3_metal_rev(tp) (unused)

Add:
tg3_chip_rev_id(tp) for tp->pci_chip_rev_id so access styles
are similar to tg3_asic_rev and tg3_chip_rev.

These macros are not converted to static inline functions
because gcc (tested with 4.7.2) is currently unable to
optimize the object code it produces the same way and code
is otherwise larger.
Signed-off-by: NJoe Perches <joe@perches.com>
Acked-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's the same value as tp->pci_chip_rev_id so use that
instead.  This makes all CHIPREV_ID_<foo> tests the same.
Signed-off-by: NJoe Perches <joe@perches.com>
Acked-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Without this patch I get many unaligned access warnings per packet,
this patches fixes them all. This should improve performance on some
systems like mips.
Signed-off-by: NHauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In ehea.h the minimal entries is 2^7 - 1:
#define EHEA_MIN_ENTRIES_QP  127

Thus change the module param description accordinglly
Signed-off-by: NDave Young <dyoung@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Recent changes have made it so that MAX_SKB_FRAGS is now never less than 16.
As a result we were seeing issues on systems with 64K pages as it would
cause DESC_NEEDED to increase to 68, and we would need over 136 descriptors
free before clean_tx_irq would wake the queue.

This patch makes it so that DESC_NEEDED is always MAX_SKB_FRAGS + 4.  This
should prevent any possible deadlocks on the systems with 64K pages as we will
now only require 42 descriptors to wake.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch makes sure that TXDCTL.WTHRESH is set to 1 when BQL is enabled
and EITR is set to more than 100k interrupts per second to avoid Tx timeouts.
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch adds support for reading data from SFP+ modules over i2c.
Signed-off-by: NAurélien Guillaume <footplus@gmail.com>
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch replaces instances where a return code from i2c operations
were checked against a list of error codes with a much simpler
if ( status != 0 ) check.

Some whitespace cleanups included.
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch makes sure that the SW lock is released after all i2c
operations complete in the retry code path.
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change adds support for the ethtool set_channels operation.

Since the ixgbe driver has to support DCB as well as the other modes the
assumption I made here is that the number of channels in DCB modes refers
to the number of queues per traffic class, not the number of queues total.

CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch adds support for the ethtool get_channels operation.

Since the ixgbe driver has to support DCB as well as the other modes the
assumption I made here is that the number of channels in DCB modes refers
to the number of queues per traffic class, not the number of queues total.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The ixgbe_setup_tc code is essentially the same code we need any time we have
to update the number of queues.  As such I am making it available always and
just stripping the DCB specific bits out when DCB is disabled instead of
stripping the entire function.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change updates the ixgbe driver to use __netdev_pick_tx instead of
the current logic it is using to select a queue.  The main result of this
change is that ixgbe can now fully support XPS, and in the case of non-FCoE
enabled configs it means we don't need to have our own ndo_select_queue.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change adds support for ixgbe to configure the XPS queue mapping on
load.  The result of this change is that on open we will now be resetting
the number of Tx queues, and then setting the default configuration for XPS
based on if ATR is enabled or disabled.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Instead of adjusting the FCoE and Flow director limits based on the number
of CPUs we can define them much sooner.  This allows the user to come
through later and adjust them once we have updated the code to support the
set_channels ethtool operation.

I am still allowing for FCoE and RSS queues to be separated if the number
queues is less than the number of CPUs.  This essentially treats the two
groupings like they are two separate traffic classes.

In addition I am changing the initialization to use the MAX_TX/RX_QUEUES
defines instead of trying to compute the value as it will be possible in
upcoming patches for the user to request the maximum number of queues.

I have also updated things so that the upper limit on queues is exactly 63
instead of allowing it to go up to 64.  The reason for this change is to
address the fact thqt the driver only supports up to 63 queue vectors since
the hardware supports 64 MSI-X vectors, but one must be reserved for "other"
causes.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

On several machines with i350 adapters the ethtool offline self-test sometimes
fails. This happens because link auto negotiation may take longer than the
timeout of 4 seconds. Increasing the timeout by 1 seconds resolves the issue.

Output from a failing i350 offline self-test:
while [ 1 ]; do ethtool -t eth2 offline; done
The test result is PASS
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         0

The test result is FAIL
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         1

The test result is PASS
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         0
Signed-off-by: NStefan Assmann <sassmann@kpanic.de>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change is meant to address several race issues that become possible
because next_to_watch could possibly be set to a value that shows that the
descriptor is done when it is not.  In order to correct that we instead make
next_to_watch a pointer that is set to NULL during cleanup, and set to the
eop_desc after the descriptor rings have been written.

To enforce proper ordering the next_to_watch pointer is not set until after
a wmb writing the values to the last descriptor in a transmit.  In order to
guarantee that the descriptor is not read until after the eop_desc we use the
read_barrier_depends which is only really necessary on the alpha architecture.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Acked-by: NGreg Rose <gregory.v.rose@intel.com>
Tested-by: NSibai Li <sibai.li@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>