o Consider number of Tx queues while calculating the length of
  Tx statistics as part of ethtool stats.
o Calculate statistics lenght properly for 82xx and 83xx adapter
Signed-off-by: NShahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Driver was not updating TX stats so it was not populating
  statistics in `ifconfig` command output.
Signed-off-by: NManish Chopra <manish.chopra@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, the tx queue were selected implicitly in ndo_dfwd_start_xmit(). The
will cause several issues:

- NETIF_F_LLTX were removed for macvlan, so txq lock were done for macvlan
  instead of lower device which misses the necessary txq synchronization for
  lower device such as txq stopping or frozen required by dev watchdog or
  control path.
- dev_hard_start_xmit() was called with NULL txq which bypasses the net device
  watchdog.
- dev_hard_start_xmit() does not check txq everywhere which will lead a crash
  when tso is disabled for lower device.

Fix this by explicitly introducing a new param for .ndo_select_queue() for just
selecting queues in the case of l2 forwarding offload. netdev_pick_tx() was also
extended to accept this parameter and dev_queue_xmit_accel() was used to do l2
forwarding transmission.

With this fixes, NETIF_F_LLTX could be preserved for macvlan and there's no need
to check txq against NULL in dev_hard_start_xmit(). Also there's no need to keep
a dedicated ndo_dfwd_start_xmit() and we can just reuse the code of
dev_queue_xmit() to do the transmission.

In the future, it was also required for macvtap l2 forwarding support since it
provides a necessary synchronization method.

Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: e1000-devel@lists.sourceforge.net
Signed-off-by: NJason Wang <jasowang@redhat.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Acked-by: NJohn Fastabend <john.r.fastabend@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bnx2x triggers warnings with CONFIG_DMA_API_DEBUG=y:

  WARNING: CPU: 0 PID: 2253 at lib/dma-debug.c:887 check_unmap+0xf8/0x920()
  bnx2x 0000:28:00.0: DMA-API: device driver frees DMA memory with
  different size [device address=0x00000000da2b389e] [map size=1490 bytes]
  [unmap size=66 bytes]

The reason is that bnx2x splits a TSO BD into two BDs (headers + data)
using one DMA mapping for both, but it uses only the length of the first
BD when unmapping.

This patch fixes the bug by unmapping the whole length of the two BDs.
Signed-off-by: NMichal Schmidt <mschmidt@redhat.com>
Reviewed-by: NEric Dumazet <edumazet@google.com>
Acked-by: NDmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Starting with commit 80c33ddd "net: add might_sleep() call to napi_disable"
bnx2x fails the might_sleep tests causing a stack trace to appear whenever
the driver is unloaded, as local_bh_disable() is being called before
napi_disable().

This changes the locking schematics related to CONFIG_NET_RX_BUSY_POLL,
preventing the need for calling local_bh_disable() and thus eliminating
the issue.
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NDmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The driver wrongly assumes 16 EQs/vectors are available for each BE3 PF.
When SR-IOV is enabled, a BE3 PF can support only a max of 8 EQs.
Signed-off-by: NSuresh Reddy <suresh.reddy@emulex.com>
Signed-off-by: NSathya Perla <sathya.perla@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The loopback test FW cmd may need upto 15 seconds to complete on
certain PHYs. This patch also fixes the name of the completion variable
used to synchronize FW cmd completions as it not used by the flashing
cmd alone anymore.
Signed-off-by: NSuresh Reddy <suresh.reddy@emulex.com>
Signed-off-by: NSathya Perla <sathya.perla@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When *only* the default RXQ is used, the RSS policy must be disabled so
that all IP and no-IP traffic is placed into the default RXQ. If not,
IP traffic is dropped.

Also, issue the RSS_CONFIG cmd only if FW advertises RSS capability for
the interface.
Signed-off-by: NVasundhara Volam <vasundhara.volam@emulex.com>
Signed-off-by: NSathya Perla <sathya.perla@emulex.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In order to keep DT compatibility we need to revert this, otherwise the original
dts files will no longer work with this driver change.

This reverts commit 7a399e3a.
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If the hypervisor configures a vlan for the VF via the PF, the expected
result is that only packets tagged by said vlan will be received by the VF
(and that vlan will be silently removed).
Due to an incorrect manipulation of vlan filters in the driver, the
VF can receive untagged traffic even if the hypervisor configured
some vlan for it.

This patch corrects the behaviour.
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are 2 different (related) flows in the slowpath configuration
that utilize the same pointer and cast it to different structs;
This is obviously incorrect as the intended allocated memory is that
of the smaller struct, possibly causing the flow utilizing the larger
struct to corrupt other slowpath configuration.

Since both flows are exclusive, set the allocated memory to be a union
of both structs.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a PF receives a VF message indicating a change in RSS properties
it should clean the flags' bit-fields; Otherwise, it's possible that
some random values will be considered as flags by the lower layers configuring
the RSS in FW.
Signed-off-by: NMichal Kalderon <michals@broadcom.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Number of VFs in PCIe configuration space is zero-based. Driver incorrectly
sets the number of VFs to be larger by one than what actually is feasible by
HW, which might cause later VFs to fail to allocate their MSI-X interrupts.
Signed-off-by: NMichal Kalderon <michals@broadcom.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Original straightforward division may lead to zeroing number of SB and
null-pointer dereference when device is short of MSIX vectors or lacks
MSIX capabilities.
Reported-by: NVladislav Zolotarov <vladz@cloudius-systems.com>
Signed-off-by: NDmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Driver is using common tx_clean_lock for all Tx queues. This patch
  adds per queue tx_clean_lock.
o Driver is not updating sw_consumer while processing Tx completion
  when interface is going down. Fixed in this patch.
Signed-off-by: NShahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: NManish Chopra <manish.chopra@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since commit 52367a76
("cxgb4/cxgb4vf: Code cleanup to enable T4 Configuration File support"),
we have failures like this during cxgb4 probe:

cxgb4 0000:01:00.4: bad SGE FL page buffer sizes [65536, 65536]
cxgb4: probe of 0000:01:00.4 failed with error -22

This happens whenever software parameters are used, without a
configuration file. That happens when the hardware was already
initialized (after kexec, or after csiostor is loaded).

It happens that these values are acceptable, rendering fl_pg_order equal
to 0, which is the case of a hard init when the page size is equal or
larger than 65536.

Accepting fl_large_pg equal to fl_small_pg solves the issue, and
shouldn't cause any trouble besides a possible performance reduction
when smaller pages are used. And that can be fixed by a configuration
file.
Signed-off-by: NThadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o TX queues allocation was getting distributed equally among all the
  functions of the port including VFs and PF. Which was leading to failure
  in PF's multiple TX queues creation.

o Instead of dividing queues equally allocate one TX queue for each VF as VF
  doesn't support multiple TX queues.
Signed-off-by: NManish Chopra <manish.chopra@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Adapter requires that if the port is in loopback mode no traffic should
  be flowing through that port, so on arrival of Link up AEN, do not advertise
  Link up to the stack until port is out of loopback mode
Signed-off-by: NManish Chopra <manish.chopra@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

KR2 work-around is based on detecting non-KR2 devices which may not link up
in this mode. One such link-partner is the BCM8073 which has specific
advertisement characteristics in specific mode, and this condition was not set
correctly.
Signed-off-by: NYaniv Rosner <yanivr@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix a problem where link is reported to be up when SFP+ module is plugged in
without cable. This occurs with specific module types which may generate
temporary TX_FAULT indication. Solution is to avoid changing any link parameters
when checking TX_FAULT indication while physical link is down.
Signed-off-by: NYaniv Rosner <yanivr@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

BCM54618SE is used to advertise half-duplex even if HD was not requested by the
user. This change makes the legacy speed/duplex advertisement for this PHY
exactly according to the requested speed and duplex.
Signed-off-by: NYaniv Rosner <yanivr@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix Passive DAC detection for specific cables, such that even in case
SFP_CABLE_TECHNOLOGY option is not set in the EEPROM (offset 8), treat it as a
passive DAC cable, since some cables don't have this indication.
Signed-off-by: NYaniv Rosner <yanivr@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix a problem where 578xx-KR is unable to get link when connected to 1G link
partner. Two fixes were required:
One was to force CL37 sync_status low to prevent Warpcore from getting stuck in
CL73 parallel detect loop while link partner is sending.
Second fix was to enable auto-detect mode, thus allowing the Warpcore to select
the higher speed protocol between 10G-KR (over CL73), or go down to 1G over CL73
when there's indication for it.
Signed-off-by: NYaniv Rosner <yanivr@broadcom.com>
Signed-off-by: NAriel Elior <ariele@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch corrects a problem in stmmac_ptp.c, functions
stmmac_adjust_time and stmmac_adjust_freq where the incorrect spinlocks
were released. This patch also addresses a problem in stmmac_main,
function stmmac_init_ptp where the capability detection for
advanced timestamping was masked by message masking.

This patch was touch tested using linuxptp, and runs without the previously
observed instabilities. More extensive testing is ongoing.

Vince
Signed-off-by: NVince Bridgers <vbridgers2013@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We should not assume that the PHY reset is always active low.

Retrieve this information from the device tree instead, so that the PHY reset
can work on both cases.
Reported-by: NPhilipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

netxen_process_lro() contains two bounds checks.  One for the ring number
against the number of rings, and one for the Rx buffer ID against the
array of receive buffers.

Both of these have off-by-one errors, using > instead of >=. The correct
versions are used in netxen_process_rcv(), they're just wrong in
netxen_process_lro().
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NEric Dumazet <edumazet@google.com>

skb_tx_timestamp(skb) should be called _before_ TX completion
has a chance to trigger, otherwise it is too late and we access
freed memory.

Fixes: e4f2379d ("ethernet/arc/arc_emac - Add new driver")
From: Eric Dumazet <edumazet@google.com>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The current driver assumes that an skb fragment can only be upto jumbo
size. Presumably this was a fast-path optimization. This assumption is
no longer true as fragments can be upto 32k.

v2: Remove unnecessary parantheses per Eric Dumazet.

Cc: stable@vger.kernel.org
Signed-off-by: NNithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

During checking the interrupts with "cat /proc/interrupts", it is showing
device name as (null), this change was done with commit id aa1a15e2 where
request_irq is changed to devm_request_irq also changing the irq name from
platform device name to net device name, but the net device is not
registered at this point with the network frame work, so devm_request_irq
is called with device name as NULL, by which it is showed as "(null)" in
"cat /proc/interrupts". So this patch changes back irq name to platform
device name itself in devm_request_irq so that the device name shows as
below.

Previous to this patch
root@am335x-evm:~# cat /proc/interrupts
           CPU0
 28:       2265      INTC  12  edma
 30:         80      INTC  14  edma_error
 56:          0      INTC  40  (null)
 57:       1794      INTC  41  (null)
 58:          7      INTC  42  (null)
 59:          0      INTC  43  (null)

With this patch
root@am335x-evm:~# cat /proc/interrupts
           CPU0
 28:        213      INTC  12  edma
 30:          9      INTC  14  edma_error
 56:          0      INTC  40  4a100000.ethernet
 57:      16097      INTC  41  4a100000.ethernet
 58:      11964      INTC  42  4a100000.ethernet
 59:          0      INTC  43  4a100000.ethernet
Signed-off-by: NMugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adds API cxgb4_select_ntuple so as to enable Upper Level Drivers to correctly
calculate the tuple fields.

Adds constant definitions for TP_VLAN_PRI_MAP for the Compressed
Filter Tuple field widths and structures and uses them.

Also, the CPL Parameters field for T5 is 40 bits so we need to prototype
cxgb4_select_ntuple() to calculate and return u64 values.

Based on original work by Casey Leedom <leedom@chelsio.com>
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

IPv6 uses 2 TIDs with CLIP enabled and 4 TIDs without CLIP.
Currently we are incrementing STIDs in use by 1 for both IPv4 and IPv6 which
is wrong.

Further, driver currently does not have interface to query if CLIP is programmed
for particular IPv6 address. So, in this patch we increment/decrement TIDs in use
by 4 for IPv6 assuming absence of CLIP. Such assumption keeps us on safe side and
we don't end up allocating more stids for IPv6 than actually supported.

Based on original work by Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The LE workaround code is incorrectly reusing the TCAM TIDs
(meant for allocation by firmware in case of hash collisions) for filter
servers. This patch assigns the filter server TIDs properly starting from
sftid_base index.

Based on original work by Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We were creating LE Workaround Server Filters without specifying
IPPROTO_TCP (6) in the filters (when F_PROTOCOL is set in TP_VLAN_PRI_MAP).
This meant that UDP packets with matching IP Addresses/Ports would get
caught up in the filter and be delivered to ULDs like iw_cxgb4.
So, include the protocol information in the server filter properly.

Based on original work by Casey Leedom <leedom@chelsio.com>
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When creating offload server entries, an IPv6 passive connection request
can trigger a reply with a null STID, whereas the driver would expect
the reply 'STID to match the value used for the request.
This happens due to h/w limitation on T4 and T5.

This patch ensures that STID 0 is never used if the stid range starts
from zero.

Based on original work by Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This version corrects the whitespace issue.

orion_mdio_wait_ready uses wait_event_timeout to wait for the
SMI interrupt to fire.  wait_event_timeout waits for between
"timeout - 1" and "timeout" jiffies.  In this case a 1ms timeout
when HZ is 1000 results in a wait of 0 to 1 jiffies, causing
premature timeouts.

This fix ensures a minimum timeout of 2 jiffies, ensuring
wait_event_timeout will always wait at least 1 jiffie.

Issue reported by Nicolas Schichan.
Tested-by: NNicolas Schichan <nschichan@freebox.fr>
Signed-off-by: NLeigh Brown <leigh@solinno.co.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function atl1c_reset_pcie() does not check the return from
pci_find_ext_cabability() where it is getting the postion of the
PCI_EXT_CAP_ID_ERR. It is possible for the return to be 0.
Signed-off-by: NBetty Dall <betty.dall@hp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

skb_tx_timestamp(skb) should be called _before_ TX completion
has a chance to trigger, otherwise it is too late and we access
freed memory.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Fixes: de5fb0a0 ("net: fec: put tx to napi poll function to fix dead lock")
Cc: Frank Li <Frank.Li@freescale.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Acked-by: NRichard Cochran <richardcochran@gmail.com>
Acked-by: NFrank Li <Frank.Li@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

"valid ME register value" is not an error. It should be logged for
debugging only.
Signed-off-by: NMichal Schmidt <mschmidt@redhat.com>
Acked-by: NYuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for enabling In Band mode in 10 mbps speed.
RGMII supports 1 Gig and 100 mbps mode for Forced mode of operation.
For 10mbps mode it should be configured to in band mode so that link
status, duplexity and speed are determined from the RGMII input data
stream
Signed-off-by: NMugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If CONFIG_PCI_IOV isn't defined we get an "unused variable" warining so
now wrap the variable declaration like it's usage already was.
Signed-off-by: NDon Skidmore <donald.c.skidmore@intel.com>
Acked-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>