It is simpler to have the TPA mode as one three-state variable.
Signed-off-by: NMichal Schmidt <mschmidt@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If disable_tpa is set, remove NETIF_F_LRO from hw_features, so ethtool sees
it as "off [fixed]".

Note that setting the NETIF_F_LRO bit in dev->features in the 'else'
branch is not needed, because the bit was already set by
bnx2x_init_dev().

Then the check for disable_tpa in in bnx2x_fix_features() becomes unnecessary.
Signed-off-by: NMichal Schmidt <mschmidt@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bnx2x's 'disable_tpa=1' module option is not respected properly and TPA
(transparent packet aggregation) remains enabled. Even though the
module option causes LRO to be disabled, TPA is enabled in GRO mode.

Additionally, disabling GRO via ethtool then has no effect. One can
still observe tpa_* statistics increase and large packets being received
in tcpdump.

The bug was an unintended consequence of commit aebf6244 "bnx2x: Be
more forgiving toward SW GRO".

Fix it by following the bp->disable_tpa flag when initializing fp's.
Signed-off-by: NMichal Schmidt <mschmidt@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When having permanent EEH error, the PCI device will be removed
from the system. For this case, we shouldn't set pcierr_recovery
to true wrongly, which blocks the driver to release the allocated
interrupts and their handlers. Eventually, we can't disable MSI
or MSIx successfully because of the MSI or MSIx interrupts still
have associated interrupt actions, which is turned into following
stack dump.

Oops: Exception in kernel mode, sig: 5 [#1]
        :
[c0000000003b76a8] .free_msi_irqs+0x80/0x1a0 (unreliable)
[c00000000039f388] .pci_remove_bus_device+0x98/0x110
[c0000000000790f4] .pcibios_remove_pci_devices+0x9c/0x128
[c000000000077b98] .handle_eeh_events+0x2d8/0x4b0
[c0000000000782d0] .eeh_event_handler+0x130/0x1c0
[c000000000022bd4] .kernel_thread+0x54/0x70
Signed-off-by: NGavin Shan <gwshan@linux.vnet.ibm.com>
Acked-by: NPrashant Sreedharan <prashant@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After d75b1ade ("net: less interrupt masking in NAPI") polling
function has to return whole budget when it wants NAPI to call it again.
Signed-off-by: NRafał Miłecki <zajec5@gmail.com>
Cc: Felix Fietkau <nbd@openwrt.org>
Fixes: eb64e292 ("bgmac: leave interrupts disabled as long as there is work to do")
Acked-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On some feature changes, driver employes an inner-reload flow where it
resets the function and re-configures it with the new required set of
parameters.

Such a flow proves fatal to any VF since those were not intended to be used
while HW is being reset underneath, causing them [at best] to lose all
connectivity.

This changes driver behavior to fail all configuration changes [e.g., mtu
change] requested of the driver in case VFs are active.
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 9a2620c8 ("bnx2x: prevent WARN during driver unload")
switched the napi/busy_lock locking mechanism from spin_lock() into
spin_lock_bh(), breaking inter-operability with netconsole, as netpoll
disables interrupts prior to calling our napi mechanism.

This switches the driver into using atomic assignments instead of the
spinlock mechanisms previously employed.

Based on initial patch from Yuval Mintz & Ariel Elior

I basically added softirq starvation avoidance, and mixture
of atomic operations, plain writes and barriers.

Note this slightly reduces the overhead for this driver when no
busy_poll sockets are in use.

Fixes: 9a2620c8 ("bnx2x: prevent WARN during driver unload")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The ring size is always known at compile time, so make the code a bit
more efficient
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The driver needs to inform the hardware about the first invalid (not yet
filled) rx slot, by writing its DMA descriptor pointer offset to the
BGMAC_DMA_RX_INDEX register.

This register was set to a value exceeding the rx ring size, effectively
allowing the hardware constant access to the full ring, regardless of
which slots are initialized.

To fix this issue, always mark the last filled rx slot as invalid.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of allocating buffers at device init time and initializing
descriptors at device open, do both at the same time (during open).
Free all buffers when closing the device.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Acked-by: NRafał Miłecki <zajec5@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Limiting it to 511 looks like a failed attempt at leaving one descriptor
empty to allow the hardware to stop processing a buffer that has not
been prepared yet. However, this doesn't work because this affects the
total ring size as well
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In very rare cases, the MAC can catch an internal buffer that is bigger
than it's supposed to be. Instead of crashing the kernel, simply pass
the buffer back to the hardware
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Allocate a new buffer before processing the completed one. If allocation
fails, reuse the old buffer.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Acked-by: NRafał Miłecki <zajec5@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A packet buffer offset of 30 bytes is inefficient, because the first 2
bytes end up in a different cacheline.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Always poll rx and tx during NAPI poll instead of relying on the status
of the first interrupt. This prevents bgmac_poll from leaving unfinished
work around until the next IRQ.
In my tests this makes bridging/routing throughput under heavy load more
stable and ensures that no new IRQs arrive as long as bgmac_poll uses up
the entire budget.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Keep incrementing ring->start and ring->end instead of pointing it to
the actual ring slot entry. This simplifies the calculation of the
number of free slots.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Acked-by: NRafał Miłecki <zajec5@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, MoCA fixed PHYs are always in link-up state, regardless of
whether the link is actually up or not.

Add code to properly detect MoCA link state changes and to reflect the
new state in MoCA fixed PHY. Only GENET V3 and V4 MACs are capable of
detecting MoCA link state changes.

The code works as follows:
1. GENET MAC detects MoCA link state change and issues UMAC_IRQ_LINK_UP
   or UMAC_IRQ_LINK_DOWN interrupt.
2. Link up/down interrupt is processed in bcmgenet_irq_task(), which
   calls phy_mac_interrupt().
3. phy_mac_interrupt() updates the fixed PHY phydev->link and kicks
   the PHY state machine.
4. PHY state machine proceeds to read the fixed PHY link status
   register.
5. When the fixed PHY link status register is being read, the new
   function bcmgenet_fixed_phy_link_update() gets called. It copies
   the fixed PHY phydev->link value to the fixed PHY status->link.
6. PHY state machine receives the new link state of the fixed PHY.
7. MoCA fixed PHY link state now correctly reflects the real MoCA
   hardware link state.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When IOMM-vtd is active, once main kernel crashes unfinished DMAE transactions
will be blocked, putting the HW in an error state which will cause further
transactions to timeout.

Current employed logic uses wrong macros, causing the first function to be the
only function that cleanups that error state during its probe/load.

This patch allows all the functions to successfully re-load in kdump kernel.
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When running in a kdump kernel, it's very likely that due to sync. loss with
management firmware the first PCI function to probe and reach the previous
unload flow would decide it can reset the chip and continue onward. While doing
so, it will only close its own Rx port.

On a 4-port device where 2nd port on engine is a 1g-port, the 2nd port would
allow ingress traffic after the chip is reset [assuming it was active on the
first kernel]. This would later cause a HW attention.

This changes driver flow to close both ports' 1g capabilities during the
previous driver unload flow prior to the chip reset.
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are two reasons for having this header in the common place:
1) Simplifying drivers that read NVRAM entries. We will be able to
   safely call bcm47xx_nvram_* functions without #ifdef-s.
2) Getting NVRAM driver out of MIPS arch code. This is needed to support
   BCM5301X arch which also requires this NVRAM driver. Patch for that
   will follow once we get is reviewed.
Signed-off-by: NRafał Miłecki <zajec5@gmail.com>
Acked-by: NHauke Mehrtens <hauke@hauke-m.de>
Cc: linux-mips@linux-mips.org
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Paul Walmsley <paul@pwsan.com>
Cc: linux-soc@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/8619/Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

This patch changes the driver to use ns_to_timespec64() instead of
open coding the same logic.

Compile tested only.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch changes the driver to use ns_to_timespec64() and
timespec64_to_ns() instead of open coding the same logic.

Compile tested only.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On phy_mac_interrupt() call, the new_link parameter should be 0 or 1.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add #define UMAC_IRQ_LINK_EVENT to simplify link event handling code.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If bcmgenet_init_dma() fails, it cleans up after itself. Rx and Tx
DMAs are off, and NAPI instances haven't been netif_napi_add()'ed.
Therefore, we need to skip calling bcmgenet_fini_dma() on the error
handling path. bcmgenet_resume() already does this correctly.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bcmgenet_fini_dma() is the only caller of __bcmgenet_fini_dma().
Move __bcmgenet_fini_dma() code inside bcmgenet_fini_dma().
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add #define for UMAC_IRQ_RXDMA_DONE and UMAC_IRQ_TXDMA_DONE in order
to simplify the code that handles Rx and Tx default queue interrupts.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use bool constants as the return values instead of 1 and 0.
Signed-off-by: NJoe Perches <joe@perches.com>
Acked-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's possible that due to errors [either on PCI or on device itself]
registers reads would fail, returning all-Fs.

This adds a check as early as possible so that driver will not read junk
values and make incorrect probe decisions according to them; instead,
gracefully fail the probe.
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Number of link changes are now being stored in shared memory [by all possible
link owners], for management use [as well as possible debug information for
dumps].
Signed-off-by: NYaniv Rosner <Yaniv.Rosner@qlogic.com>
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Enable controlling Post2, coeff, IPreDriver and IFir according to NVRAM setup.
Signed-off-by: NYaniv Rosner <Yaniv.Rosner@qlogic.com>
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The device appears to use a 64 bit nanoseconds register, and so with
this patch the driver should be ready for the year 2038.

Compile tested only.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This driver's clock is implemented using a timecounter, and so with
this patch the driver is ready for the year 2038.

Compile tested only.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Acked-by: NSony Chacko <sony.chacko@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To allow drivers to handle the features check for multiple tags,
move the check to ndo_features_check().
As no drivers currently handle multiple tagged TSO, introduce
dflt_features_check() and call it if the driver does not have
ndo_features_check().
Signed-off-by: NToshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add support for multiple Rx queues:
1. Add NAPI context per Rx queue
2. Modify Rx interrupt and Rx NAPI code to handle multiple Rx queues
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Introduce new bcmgenet functions to handle the NAPI calls to:
netif_napi_add()
napi_enable()
napi_disable()
netif_napi_del()
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Introduce new bcmgenet functions to handle the NAPI calls to:
netif_napi_add()
napi_enable()
napi_disable()
netif_napi_del()
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use more meaningful variable names int0_enable and int1_enable when
enabling bcmgenet interrupts.

For Rx default queue interrupts, use:
UMAC_IRQ_RXDMA_BDONE | UMAC_IRQ_RXDMA_PDONE

For Tx default queue interrupts, use:
UMAC_IRQ_TXDMA_BDONE | UMAC_IRQ_TXDMA_PDONE
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Do the two kcalloc() calls first, before proceeding into Rx/Tx DMA init.
Makes the error case handling much simpler.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Reviewed-by: NJaedon Shin <jaedon.shin@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove unnecessary function parameter priv. Use ring->priv instead.
Signed-off-by: NPetri Gynther <pgynther@google.com>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>