Reduces CPU utilization significantly with sendfile for example.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch replaces dev->mc_count in all drivers (hopefully I didn't miss
anything). Used spatch and did small tweaks and conding style changes when
it was suitable.
Jirka
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes two potential races in the velocity driver:
* Move the ACK and error handler to the interrupt handler. This fixes a
potential race with shared interrupts when the other device interrupts
before the NAPI poll handler has finished. As the velocity driver hasn't
acked it's own interrupt, it will then steal the interrupt from the
other device.
* Use spin_lock_irqsave in velocity_poll. In the current code, the
interrupt handler will deadlock if e.g., the NAPI poll handler is
executing when an interrupt (for another device) comes in since it
tries to take the already held lock.
Also unlock the spinlock only after enabling the interrupt in
velocity_poll.
The error path is moved to the interrupt handler since this is where the
ISR is checked now.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: Anders Grafstrom <anders.grafstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
velocity_set_coalesce touches ISR and some other sensitive registers not
covered by the rtnl lock, so take the velocity spinlock.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DEFINE_PCI_DEVICE_TABLE() so we get place PCI ids table into correct section
in every case.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
velocity_open() calls velocity_give_many_rx_descs(), which gives RX
descriptors to the NIC, before installing an interrupt handler or
calling velocity_init_registers(). I think this is very unsafe and it
appears to explain the bug report <http://bugs.debian.org/508527>.
On MTU change, velocity_give_many_rx_descs() is again called before
velocity_init_registers(). I'm not sure whether this is unsafe but
it does look wrong.
Therefore, move the calls to velocity_give_many_rx_descs() after
request_irq() and velocity_init_registers().
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Tested-by: Jan Ceuleers <jan.ceuleers@computer.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the %pM kernel extension to display the MAC address.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only files where David Miller is the primary git-signer.
wireless, wimax, ixgbe, etc are not modified.
Compile tested x86 allyesconfig only
Not all files compiled (not x86 compatible)
Added a few > 80 column lines, which I ignored.
Existing checkpatch complaints ignored.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The velocity hardware can handle up to 7 memory segments. This can be
turned on and off via ethtool. The support was removed in commit
83c98a8cd0
but is re-enabled and cleaned up here. It's off by default.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The VIA driver has changed the default for the DMA_LENGTH_DEF parameter.
Together with adaptive interrupt supression and NAPI support, this
improves performance quite a bit
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds NAPI support for VIA velocity. The new velocity_poll
function also pairs tx/rx handling twice which improves perforamance on
some workloads (e.g., netperf UDP_STREAM) significantly (that part is
from the VIA driver).
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
(Partially from the upstream VIA driver). Tweaking the number of
frames-per-interrupt and timer-until-interrupt can reduce the amount of
CPU work quite a lot.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
(From the VIA driver). The current code does not guarantee 64-byte
alignment since it simply does
int add = skb->data & 63;
skb->data += add;
(via skb_reserve). So for example, if the skb->data address would be
0x10, this would result in 32-byte alignment (0x10 + 0x10).
Correct by adding
64 - (skb->data & 63)
instead.
Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
In this file, function names are otherwise used as pointers without &.
A simplified version of the semantic patch that makes this change is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r@
identifier f;
@@
f(...) { ... }
@@
identifier r.f;
@@
- &f
+ f
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
via-velocity maintains a list of its devices in order to determine
whether a netdev notification applies to one of them. That can be
determined simply by checking the netdev_ops pointer, so the list can
be removed.
Compile-tested only.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
dev_ioctl() already checks capable(CAP_NET_ADMIN) before calling the
driver's implementation of MDIO ioctls.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a couple of cases collapse some extra code like:
int retval = NETDEV_TX_OK;
...
return retval;
into
return NETDEV_TX_OK;
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Test whether VELOCITY_DUPLEX_FULL bit is set in mii_status.
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By moving functions to before their first call, we eliminate
the need to define forward references.
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is the result of an automatic spatch transformation to convert
all ndo_start_xmit() return values of 0 to NETDEV_TX_OK.
Some occurences are missed by the automatic conversion, those will be
handled in a seperate patch.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a packet is greater than ETH_ZLEN, we end up assigning the
boolean result of a comparison to the size we unmap.
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use struct net_device_stats provided in struct net_device instead of
private ones.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix this warning:
drivers/net/via-velocity.c:1924: warning: passing argument 2 of 'request_irq' from incompatible pointer type
Signed-off-by: Seguier Regis <rseguier@e-teleport.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Dave Jones <davej@redhat.com>
The dma-debug changes caught that this driver uses the
wrong DMA mapping length when skb_padto() does something.
With suggestions from Eric Dumazet.
Signed-off-by: David S. Miller <davem@davemloft.net>
while(--j >= 0) keeps spinning when j is unsigned:
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Many drivers lost the ability to set ethernet address accidently
during the net_device_ops conversion.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch moves neigh_setup and hard_start_xmit into the network device ops
structure. For bisection, fix all the previously converted drivers as well.
Bonding driver took the biggest hit on this.
Added a prefetch of the hard_start_xmit in the fast path to try and reduce
any impact this would have.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert this driver to network device ops. Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The generic packet receive code takes care of setting
netdev->last_rx when necessary, for the sake of the
bonding ARP monitor.
Drivers need not do it any more.
Some cases had to be skipped over because the drivers
were making use of the ->last_rx value themselves.
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch corrects a message bug in the via-velocity driver which
bothered me for some time.
The messages printed during device init look like the following:
[ 8.486422] eth%d: set value of parameter Wake On Lan options to 0
^^!
[ 8.487340] eth0: VIA Networking Velocity Family Gigabit Ethernet Adapter
Note the unresolved format string.
dev->name is unavailable before register_netdev, so use
dev_driver_string(&pdev->dev), which is also consistent with other
drivers.
"char *devname" parameters had to be converted to "const char *devname" to
be consistent with dev_driver_string return value.
Signed-off-by: Sven Hartge <sven@svenhartge.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Clean up the various different email addresses of mine listed in the code
to a single current and valid address. As Dave says his network merges
for 2.6.28 are now done this seems a good point to send them in where
they won't risk disrupting real changes.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allocate and free everyting outside of the locked section.
Spotted-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Fixed-by: Seguier Regis <rseguier@e-teleport.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
It should help people fix the bugs in my code :o)
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
It removes a dependancy from velocity_init_rd_ring to dev->mtu.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Updates of the RBRDU have two different meanings depending on their
context:
1. the receiving process has not started - the value which is written
into the RBRDU register is supposed to be the free rx descriptors
count (rounded to a multiple of 4)
2. the receiving process is running - the value increments the count
above (sic)
The update is currently issued deep inside the rx replenish chain (see
velocity_give_many_rx_descs).
Let's propagate enough information to the caller so that the rx
replenish functions do not depend on hardware any more.
It is needed to perform the Rx/Tx buffers housekeeping when MTU changes.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
- PCI consistent areas need no memset
- use dev_err instead of plain printk
- avoid a few casts
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Executive summary: the bounce buffers are in my way
- they use something like a 64 * 1500 bytes area of PCI
consistent area
- they are not resized when the MTU changes
- they are used
- to hand-pad undersized packets. skb_pad anyone ?
- to linearize fragmented skbs whose fragment count
goes beyond the 7 fragments hardware limit in order
to claim scatter-gather support
Actually the SG code is commented out and I wonder if it
could not be implemented (ab-)using the large send feature
of the chipset since the latter should support some
multi-descriptor packet transmitting.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Fixed-by: Séguier Régis <rseguier@e-teleport.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
