linux.git/drivers/net, branch v5.3-rc8

virtio-net: lower min ring num_free for efficiency

2019-09-04T10:21:17+00:00

This change lowers ring buffer reclaim threshold from 1/2*queue to budget
for better performance. According to our test with qemu + dpdk, packet
dropping happens when the guest is not able to provide free buffer in
avail ring timely with default 1/2*queue. The value in the patch has been
tested and does show better performance.

Test setup: iperf3 to generate packets to guest (total 30mins, pps 400k, UDP)
avg packets drop before: 2842
avg packets drop after: 360(-87.3%)

Further, current code suffers from a starvation problem: the amount of
work done by try_fill_recv is not bounded by the budget parameter, thus
(with large queues) once in a while userspace gets blocked for a long
time while queue is being refilled. Trigger refills earlier to make sure
the amount of work to do is limited.

Signed-off-by: jiangkidd 
Acked-by: Jason Wang 
Acked-by: Michael S. Tsirkin 
Signed-off-by: Michael S. Tsirkin

net: seeq: Fix the function used to release some memory in an error handling path

2019-09-01T19:10:11+00:00

In commit 99cd149efe82 ("sgiseeq: replace use of dma_cache_wback_inv"),
a call to 'get_zeroed_page()' has been turned into a call to
'dma_alloc_coherent()'. Only the remove function has been updated to turn
the corresponding 'free_page()' into 'dma_free_attrs()'.
The error hndling path of the probe function has not been updated.

Fix it now.

Rename the corresponding label to something more in line.

Fixes: 99cd149efe82 ("sgiseeq: replace use of dma_cache_wback_inv")
Signed-off-by: Christophe JAILLET 
Reviewed-by: Thomas Bogendoerfer 
Signed-off-by: David S. Miller

enetc: Add missing call to 'pci_free_irq_vectors()' in probe and remove functions

2019-09-01T06:53:17+00:00

Call to 'pci_free_irq_vectors()' are missing both in the error handling
path of the probe function, and in the remove function.
Add them.

Fixes: 19971f5ea0ab ("enetc: add PTP clock driver")
Signed-off-by: Christophe JAILLET 
Signed-off-by: David S. Miller

net: bcmgenet: use ethtool_op_get_ts_info()

2019-09-01T06:51:19+00:00

This change enables the use of SW timestamping on the Raspberry Pi 4.

bcmgenet's transmit function bcmgenet_xmit() implements software
timestamping. However the SOF_TIMESTAMPING_TX_SOFTWARE capability was
missing and only SOF_TIMESTAMPING_RX_SOFTWARE was announced. By using
ethtool_ops bcmgenet_ethtool_ops() as get_ts_info(), the
SOF_TIMESTAMPING_TX_SOFTWARE capability is announced.

Similar to commit a8f5cb9e7991 ("smsc95xx: use ethtool_op_get_ts_info()")

Signed-off-by: Ryan M. Collins 
Acked-by: Florian Fainelli 
Acked-by: Doug Berger 
Signed-off-by: David S. Miller

net: dsa: microchip: add KSZ8563 compatibility string

2019-09-01T06:36:37+00:00

It is a 3-Port 10/100 Ethernet Switch with 1588v2 PTP.

Signed-off-by: Razvan Stefanescu 
Signed-off-by: David S. Miller

net: aquantia: fix out of memory condition on rx side

2019-09-01T02:07:11+00:00

On embedded environments with hard memory limits it is a normal although
rare case when skb can't be allocated on rx part under high traffic.

In such OOM cases napi_complete_done() was not called.
So the napi object became in an invalid state like it is "scheduled".
Kernel do not re-schedules the poll of that napi object.

Consequently, kernel can not remove that object the system hangs on
`ifconfig down` waiting for a poll.

We are fixing this by gracefully closing napi poll routine with correct
invocation of napi_complete_done.

This was reproduced with artificially failing the allocation of skb to
simulate an "out of memory" error case and check that traffic does
not get stuck.

Fixes: 970a2e9864b0 ("net: ethernet: aquantia: Vector operations")
Signed-off-by: Igor Russkikh 
Signed-off-by: Dmitry Bogdanov 
Signed-off-by: David S. Miller

net: aquantia: linkstate irq should be oneshot

2019-09-01T02:07:11+00:00

Declaring threaded irq handler should also indicate the irq is
oneshot. It is oneshot indeed, because HW implements irq automasking
on trigger.

Not declaring this causes some kernel configurations to fail
on interface up, because request_threaded_irq returned an err code.

The issue was originally hidden on normal x86_64 configuration with
latest kernel, because depending on interrupt controller, irq driver
added ONESHOT flag on its own.

Issue was observed on older kernels (4.14) where no such logic exists.

Fixes: 4c83f170b3ac ("net: aquantia: link status irq handling")
Signed-off-by: Igor Russkikh 
Reported-by: Michael Symolkin 
Signed-off-by: David S. Miller

net: aquantia: reapply vlan filters on up

2019-09-01T02:07:11+00:00

In case of device reconfiguration the driver may reset the device invisible
for other modules, vlan module in particular. So vlans will not be
removed&created and vlan filters will not be configured in the device.
The patch reapplies the vlan filters at device start.

Fixes: 7975d2aff5afb ("net: aquantia: add support of rx-vlan-filter offload")
Signed-off-by: Dmitry Bogdanov 
Signed-off-by: Igor Russkikh 
Signed-off-by: David S. Miller

net: aquantia: fix limit of vlan filters

2019-09-01T02:07:11+00:00

Fix a limit condition of vlans on the interface before setting vlan
promiscuous mode

Fixes: 48dd73d08d4dd ("net: aquantia: fix vlans not working over bridged network")
Signed-off-by: Dmitry Bogdanov 
Signed-off-by: Igor Russkikh 
Signed-off-by: David S. Miller

net: aquantia: fix removal of vlan 0

2019-09-01T02:07:10+00:00

Due to absence of checking against the rx flow rule when vlan 0 is being
removed, the other rule could be removed instead of the rule with vlan 0

Fixes: 7975d2aff5afb ("net: aquantia: add support of rx-vlan-filter offload")
Signed-off-by: Dmitry Bogdanov 
Signed-off-by: Igor Russkikh 
Signed-off-by: David S. Miller