20 Aug, 2010
35 commits
-
Introduce proto_ports_offset() for getting the position of the ports or
SPI in the message of a protocol.Signed-off-by: Changli Gao
Signed-off-by: David S. Miller -
Fragmented IP packets may have no transfer header, so when computing
rxhash, we should skip them.Signed-off-by: Changli Gao
Acked-by: Eric Dumazet
Signed-off-by: David S. Miller -
skb_get_rxhash() assumes the network header pointer of the skb is set
properly after the commit:commit bfb564e7391340638afe4ad67744a8f3858e7566
Author: Krishna Kumar
Date: Wed Aug 4 06:15:52 2010 +0000core: Factor out flow calculation from get_rps_cpu
Signed-off-by: Changli Gao
Signed-off-by: David S. Miller -
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Fw can get stuck while holding pci semaphore. Driver will not
be able to perform fw initialization, without this lock.
Release semaphore forcefully in that case.Signed-off-by: Sucheta Chakraborty
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Cleanup legacy code which is not valid for Qlogic
CNA adapters.Signed-off-by: Sony Chacko
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Mark device state failed in error path.
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Privilege functions should wait for npar state to be operational.
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Administrator can configure to drop packet in transmit,
if it doesn't match interface mac address, in case of virtual function.Signed-off-by: Sony Chacko
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Device is not capable of enabling/disabling offload setting per
port in case of Nic Partition.So offload settings needs to be
enabled/disabled per eswitch and it will affect all the function
on that eswitch.Signed-off-by: Rajesh Borundia
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
o Nic partition capable devices has embedded switch, this needs to support
various features like external switch.Signed-off-by: Rajesh Borundia
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Signed-off-by: Sony Chacko
Signed-off-by: David S. Miller -
Serial number references are not used in driver.
Signed-off-by: Sony Chacko
Signed-off-by: David S. Miller -
Spin_lock(rds_ring->lock) is not required while posting buffers
from qlcnic_open and freeing buffers from qlcnic_down.Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
This change makes it so that the ethtool loopback test uses the standard
ring configuration and allocation functions. As a result the loopback test
will be much more effective at testing core driver functionality.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
All of the DESC_ADV macros are currently needing the pointers to be
de-referenced before accessing the ring. Instead of having to add all of
the asterisks it is easier to just update the macro to expect a pointer to
the ring.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
The Rx init is currently split over ixgbe_configure, ixgbe_configure_rx,
and ixgbe_up_complete. Instead of leaving it split over 3 function it is
easier to consolidate them all into ixgbe_configure_rx.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
The Tx init was spread out over ixgbe_configure, ixgbe_configure_tx, and
ixgbe_up_complete. This change combines all of that into the
ixgbe_configure_tx function in order to simplify the Tx init path.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This change moves all GPIE register configuration into a single function.
The advantage of this is that we can avoid a number of unnecessary
read/modify/write cycles on the register.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This change moves the configuration that was done in configure_rx into a
separate virtualization configuration function.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This change moves all of the Rx DMA control register writes to one central
location. This should help to avoid accidentally overwriting existing
settings.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This change consolidates all of the Rx max frame size and Rx buffer length
configuration into a single function.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
The vmolr is configured already in ixgbe_set_rx_mode for the PF so there is
no need to set it again in ixgbe_configure_rx.Instead of using the variable name reg, it is easier to just rename it to
gcr_ext to reflect the register contents that the variable holds.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
Bump the header size for packet split to 512 bytes since this makes the
best use of the 1k buffer that is allocated for any skb 512 bytes or
smaller.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
We are accessing the FCTRL register in multiple spots in the init path and
we can simplify things by combining the configuration all into
ixgbe_set_rx_mode.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
The configuration of PSRTYPE was being done conditionally on if packet
split is enabled or not. It can be configured always since it will not
have any effect when packet split is not enabled.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
These changes add ixgbe_configure_rx_ring which is used to setup the base
function pointers for the ring.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This change simplifies the configuration of MRQC by consolidating the
setting of it into one function. As such the register is no longer set in
multiple places which should make any future changes easier to work with.
In addition we can combine RSS related register writes into the call since
enabling all of those bits without enabling RSS itself in MRQC should have
no effect.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This patch moves the Tx ring configuration into a separate function. In
addition the function drops the setting of the head writeback RO bit since
head writeback is no longer used within ixgbe.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This patch moves the configuration of the MTQC register into it's own
function call similar to ixgbe_setup_mrqc.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
In ixgbe_up_complete we were doing a read-modify-write of TXDCTL followed
by another one just a few lines further down. Instead of performing two
separate read-modify-writes it would make more sense to combine the two
into one.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
We are unnecessarily modifying the GSO size for all HW when we don't need
to. The code can be simplified by moving the check for DCB and the
adjustment of the GSO size for 82598 into ixgbe_configure_dcb.Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
This patch removes the redundant DMA alignment code from the Rx buffer
allocation path. This code is no longer necessary since all x86 buffers
are now DMA aligned due to recent changes to NET_IP_ALIGN and NET_SKB_PAD.It also moves the setting of the Rx queue value into the allocation path
since it is more likely that the queue mapping will still be in the cache
at the time of allocation.
Signed-off-by: Alexander Duyck
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller -
Make the code in ixgbe_get_drvinfo more protected against buffer overflow
boundaries by using snprintf and sizeof with snprintf.Signed-off-by: Don Skidmore
Signed-off-by: Jeff Kirsher
Signed-off-by: David S. Miller
19 Aug, 2010
5 commits
-
Netconsole requires poll support.
Signed-off-by: Michal Simek
Signed-off-by: David S. Miller -
Use integrated net_device_stats instead of a static one, and make sure
no transient values are feeded. ndo_get_stats() can be called by
concurrent cpus.Signed-off-by: Eric Dumazet
Signed-off-by: David S. Miller -
vxge_get_stats() is racy, since it clears a block of memory (net_stats)
possibly still used by other cpus.We can update this driver to full 64bit stats, since ndo_get_stats64()
provides a private block to store results, and driver maintains 64bit
counters already.We also remove net_stats field from struct vxge_sw_stats
Signed-off-by: Eric Dumazet
Acked-by: Jon Mason
Signed-off-by: David S. Miller -
No need to clear device stats in lec_open()
Signed-off-by: Eric Dumazet
Signed-off-by: David S. Miller -
struct net_device has its own struct net_device_stats member, so use
this one instead of a private copy in the ether1_priv struct. As the new
ndo_get_stats function would just return dev->stats we can omit it. This
patch also removes an incorrect memset of the stats on open.Signed-off-by: Tobias Klauser
Signed-off-by: David S. Miller
