b44 chip has some hardware limitations, that need GFP_DMA bounce
buffers in some situations.

In order to not deplete DMA zone, we should keep allocated GFP_DMA skb
only for driver use. At rx time, we copy such skb to newly allocated
skb, reusing existing copybreak infrastructure.

On machines with low amount of memory, all skb meet the hardware limitation,
so no copy is needed. We detect this situation using a new device flag, set
to one if one GFP_DMA skb was ever allocated by b44_alloc_rx_skb().

Previously allocated skb, even outside from DMA zone will then be recycled,
to have minimal impact on DMA zone use.
Signed-off-by: NEric Dumazet <dada1@cosmosbay.com>
Tested-by: NIonut Leonte <ionut.leonte@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As the kernel warning states: "IRQF_DISABLED is not guaranteed on shared
IRQs". Since these IRQs' values are hardcoded and my test system doesn't
show any shared use of IRQs at all, rather make them non-shared than
non-disabled.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Apparently this doesn't make sense. Otherwise the queue gets disabled as
soon as it's getting empty and can only be resurrected by a driver
restart.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Originally this must have been a rewrite error when introducing
'chain_index'. But the original driver did not use the previous chain
item everywhere: when altering the address tx_chain_tail points to, it
should move forward, not backwards.
Also this is not an "index" but rather the penultimate element in the
chain, so rename it accordingly.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Triggering TX before the write to the DMA status mask register leads to
transferring packets with maximum payload no matter what the actual
packet size is.
While here, also trigger RX scheduling after writing the DMA status mask
register, like it was in the original driver before it was sent
upstream.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The called netif_rx_schedule() does all the work for us:
- it checks the return value of netif_rx_schedule_prep() and
- if everything is ok calls __netif_rx_schedule().
Before this change, the driver received absolutely nothing.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This function needs an early exit condition to function properly, or
else caller assumes napi workload wasn't enough to handle all received
packets and korina_rx is called again (and again and again and ...).
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Without this the driver will crash when the NIC is being restarted.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: NFlorian Fainelli <florian@openwrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The new value is the one used in the external patch before and allows at
least a standard MTU of 1500 to be handled correctly. Impact of this
change gets visible when bigger packets are to be received, issuing:
| ping -s 492 <IP>
and bigger payload sized led to 100% packet loss.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Acked-by: Florian Fainelli <florian@openwrt.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Using platform_set_drvdata() here makes no sense, since the driver_data
field has already been filled with valuable data (i.e. the MAC address).
Also having driver_data point to the net_device is rather pointless
since struct korina_device contains an apropriate field for it.
Signed-off-by: NPhil Sutter <n0-1@freewrt.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The "read for interrupts" flag must be set before enabling slow-path
interrupts as well (and not just before fast-path interrupts)
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Too big packets could pass due to wrong filter size
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Wrong initialization of the multi-queue indirection table - it should
be using the function and not the port index
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Calculation bug due to missing brackets
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The size of the doorbell is 4KB, this bug become visible when using
more than 8 queues
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adding missing le_to_cpu and disabling wrong HW endianity flag (the
two complete each other)
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Wrong handling of tagged packet if VLAN offload is disabled caused
packets to get corrupted
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Without this lock, in some race conditions the driver missed link
change indication
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the page size is not 4KB, the FW must be programmed to work with
the right SGE boundaries and fragment list length.

To avoid confusion with the BCM_PAGE_SIZE which is set to 4KB for the
FW sake, another alias for the system page size was added to
explicitly indicate that it is meant for the SGE
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While working on IA64, it became clear that the following memory
barriers are missing
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since slow-path events, including link update, are handled in
work-queue, a race condition was introduced in the self-test that
sometimes caused the link status to fail: the self-test was running
under RTNL lock, and if the link-watch was scheduled it stoped the
shared work-queue (waiting for the RTNL lock) and so the link update
event was not handled until the self-test ended (releasing the RTNL
lock) with failure (since the link status was not updated)
Signed-off-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Due to the loopback functionality in can_send() we can not invoke it
from hardirq context which was done inside the
bcm_tx_timeout_handler() hrtimer callback:

[  700.361154]  [<c012228c>] warn_slowpath+0x80/0xb6
[  700.361163]  [<c013d559>] valid_state+0x125/0x136
[  700.361171]  [<c013d858>] mark_lock+0x18e/0x332
[  700.361180]  [<c013e300>] __lock_acquire+0x12e/0xb1e
[  700.361189]  [<f8ab5915>] bcm_tx_timeout_handler+0x0/0xbc [can_bcm]
[  700.361198]  [<c031e20a>] dev_queue_xmit+0x191/0x479
[  700.361206]  [<c01262a7>] __local_bh_disable+0x2b/0x64
[  700.361213]  [<c031e20a>] dev_queue_xmit+0x191/0x479
[  700.361225]  [<f8aa69a1>] can_send+0xd7/0x11a [can]
[  700.361235]  [<f8ab522b>] bcm_can_tx+0x9d/0xd9 [can_bcm]
[  700.361245]  [<f8ab597f>] bcm_tx_timeout_handler+0x6a/0xbc [can_bcm]
[  700.361255]  [<f8ab5915>] bcm_tx_timeout_handler+0x0/0xbc [can_bcm]
[  700.361263]  [<c0134143>] __run_hrtimer+0x5a/0x86
[  700.361273]  [<f8ab5915>] bcm_tx_timeout_handler+0x0/0xbc [can_bcm]
[  700.361282]  [<c0134a50>] hrtimer_interrupt+0xb9/0x110

This patch moves the rest of the functionality from the hrtimer
callback to the already existing tasklet to fix this slowpath problem.
Signed-off-by: NOliver Hartkopp <oliver@hartkopp.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch hooks up the start_xmit/tx_timeout/get_stats callbacks
in the ax88796 driver since they no longer are installed by the
lib8390 code. Without this patch the function dev_hard_start_xmit()
crashes due to a start_xmit callback with the value NULL.

While at it, update the ax88796 driver to make use of use of struct
net_device_ops.
Signed-off-by: NMagnus Damm <damm@igel.co.jp>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds an init_dummy_netdev() function that gets a network device
structure (allocation and lifetime entirely under caller's control) and
initialize the minimum amount of fields so it can be used to schedule
NAPI polls without registering a full blown interface. This is to be
used by drivers that need to tie several hardware interfaces to a single
NAPI poll scheduler due to HW limitations.

It also updates the ibm_newemac driver to use that, this fixing the
oops on 2.6.29 due to passing NULL as "dev" to netif_napi_add()

Symbol is exported GPL only a I don't think we want binary drivers doing
that sort of acrobatics (if we want them at all).
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Tested-by: NGeert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Taken from http://bugzilla.kernel.org/show_bug.cgi?id=12397

We're doing an sprintf of an 11-char string into an 11-char buffer. 
Whoops.  It breaks firmware uploading.
Reported-by: NJos-Vicente Gilabert <josevteg@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Bail out if pci_map_single() fails while replenishing rx ring.
o Drop packet if pci_map_{single,page}() fail in tx.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some firmware commands like mac address addition/deletion are sent
on the transmit ring. So need to hold the tx lock before touching
tx producer/consumer indices.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This fixes a tiny memory leak when driver is unloaded. The mac
address list maintained in netxen_adapter needs to deleted when
driver is going down.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Fix order or rom register writes.
o Reduce udelays when writing rom registers.

This cuts the firmware init time by 40%.

o Do not reset core/memory clocks when reinitializing driver.
  Firmware willl handle this when initialized.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Read negotiated link speed when link state changes.
o Fix link speed reporting for hybrid nic boards, which have both 1Gbps and
  10Gbps ports.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o fix the ip/tcp hdr offset in tx descriptors for ipv6.
o cleanup xmit function, move the tso checks into separate function,
  this reduces unnecessary endian conversions back and forth.
o optimize macros to initialize tx descriptors.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Set restricted (little endian) data types in firmware command
  requests and responses.
o Remove unnecessary conversion to LE when writing registers.
Signed-off-by: NDhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The attached patch modifies the sis900 driver when the MAC address
read from the hardware is invalid. As suggested, the patch now
generates a random address so that the user can go on and use
the hardware. In any case a message is also shown to warn on the
unexpected condition.
This seems to happen with newer HW implementation of the sis900
chipset, since this never came up before.

Patch is against vanilla 2.6.28 (but the driver doesn't change so often,
so it will probably apply to older/newer versions too).

See bugzilla ID 10201 and 11649 and ignore the previous patch.
Signed-off-by: NDaniele Venzano <venza@brownhat.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If something goes wrong attaching to phy driver, we weren't freeing
the IRQ.
Signed-off-by: NMike Ditto <mditto@consentry.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When we get a GSO packet from an untrusted source, we need to
ensure that it is sufficiently long so that we don't end up
crashing.

Based on discovery and patch by Ian Campbell.
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Tested-by: NIan Campbell <ian.campbell@citrix.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When an skb with page frags is merged into an existing one, we
cannibalise its reference count.  This is OK when the skb is
reused because we set nr_frags to zero in that case.  However,
for the case where the skb is freed through kfree_skb, we didn't
clear nr_frags which causes the page to be freed prematurely.

This is fixed by moving the skb resetting into skb_gro_receive.
Reported-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reported by Jiri Klimes.

Fix suggested by Patrick McHardy.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Minor cleanups, either made possible or obvious after commit d700555b (I4l:
convert to net_device_ops).
Signed-off-by: NPaul Bolle <pebolle@tiscali.nl>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>