In the xrx200_close() function we call napi_disable(), that could
sleep, with priv->hw->chan[i].lock held. This could lead to deadlock
and causes the kernel to complain.
Look at the code I couldn't convince myself about why we
need to protect that specific code part with the lock. IMHO there
seems no reason to protect the refcount variables, because AFAIK
ndo_close() and ndo_open() callbacks are already called with a
semaphore held. Neither I could figure out why napi_disable() have to
be called with that lock held. The only remaining code part for
which I could guess the lock is useful for is ltq_dma_close()
function call.
This patch reduces the lock to the said function call, avoiding the
sleep-with-spinlock-held situation
Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
[fold into 0025-NET-MIPS-lantiq-adds-xrx200-net.patch, backport to
kernel 4.4]
Signed-off-by: Mathias Kresin <dev@kresin.me>
The two phy operation mode where one phy is assigned to an interface
without lantiq,* device tree property and the other phy is assigned to
an interface with the lantiq,wan device property was broken with the
multicast package leaks between vlans fixes.
Move the multicast packages relevant portmap settings to the condition
which handles multicast packages for better readability.
Replace the priv->port_map based port_map only for the interface which
has the lantiq,switch device tree property set, to allow tagged
multicast packages in two phy mode where the lantiq,switch device tree
property isn't used.
Signed-off-by: Mathias Kresin <dev@kresin.me>
The ethernet driver uses a port map override via special tag to control
the ports on which multicast packets are sent. This was added to work
around an issue in the switch that was occasionally leaking packets onto
the wrong vlan.
Unfortunately the change had some leftover lines that were overwrting
the port map with a list of all ports, thus always leaking packets onto
the wront vlan.
Fix this by only enabling the override with the VLAN port map and only
if a matching VLAN port map was actually found
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Use priv->wan instead of priv->id as indicator if packets should go to the
Ethernet WAN group (DPID=1) or not (DPID=0). This way, it's independant of
interface names or indexes.
Signed-off-by: Martin Schiller <mschiller@tdt.de>
With an own DMA TX channel for each network device (eth0 + eth1) there
won't be any "tx ring full" errors any more.
This patch also move the spinlocks to the channel level instead of locking
the whole xrx200_hw structure.
Signed-off-by: Martin Schiller <mschiller@tdt.de>
The following patches were dropped because they are already applied
upstream:
- 0038-MIPS-lantiq-fpi-on-ar9.patch
- 0039-MIPS-lantiq-initialize-usb-on-boot.patch
- 0042-USB-DWC2-big-endian-support.patch
- 0043-gpio-stp-xway-fix-phy-mask.patch
All other patches were simply refreshed, except the following:
- 0001-MIPS-lantiq-add-pcie-driver.patch
Changes to arch/mips/lantiq/xway/sysctrl.c (these changes disabled
some PMU gates for the vrx200 / VR9 SoCs) were removed since the
upstream kernel disables unused PMU gates automatically (since
95135bfa7ead1becc2879230f72583dde2b71a0c
"MIPS: Lantiq: Deactivate most of the devices by default").
- 0025-NET-MIPS-lantiq-adds-xrx200-net.patch
Since OpenWrt commit 55ba20afcc2fe785146316e5be2c2473cb329885 drivers
should use of_get_mac_address(). of_get_mac_address_mtd is not
available for drivers anymore since it's called automatically within
of_get_mac_address().
- 0028-NET-lantiq-various-etop-fixes.patch
Same changes as in 0025-NET-MIPS-lantiq-adds-xrx200-net.patch
While refreshing the kernel configuration SPI support had to be moved to
config-4.4 because otherwise M25P80 was disabled.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
SVN-Revision: 48307
