Should fix LAN speed issues on some devices. This is an updated version
of the previously reverted commit with the same name.
It improves the check for MACs connected to a built-in switch
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Not all mach-* files set all boards correctly in ETH_CFG. They depend on
some preset values by u-boot which were not previously modified by
ath79_setup_qca955x_eth_cfg. Avoiding to modify them in this function keeps
it backward compatible for these boards.
This reverts commit 119b8ab2c2eac237ec4e9c4d0ed53df22b5c6978.
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
SVN-Revision: 49072
The default delays RXD 3. RDV 3, TXD 0, TXE 0 doesn't seem to work for some
boards. These boards depend on the preset values of u-boot which may
differ.
This reverts commit f2d4bb96b62512caa161dcc2867c91692fb16a38.
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
SVN-Revision: 49071
Some u-boot versions for QCA955x change the delays based on the link speed
during boot. This usually breaks the support of other linkspeeds when
OpenWrt is booted. It also conflicts with the
at803x_platform_data::fixup_rgmii_tx_delay. OpenWrt has to set its own
values in QCA955X_GMAC_REG_ETH_CFG.
The default RGMII values from the Atheros u-boot are currently used to
preset the existing mach files. These may have to be adjusted for boards
using different values but which are not currently set them explicitely in
OpenWrt.
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
Cc: Gabor Juhos <juhosg@openwrt.org>
Cc: Imre Kaloz <kaloz@openwrt.org>
Cc: Christian Beier <cb@shoutrlabs.com>
Cc: Chris R Blake <chrisrblake93@gmail.com>
Cc: Benjamin Berg <benjamin@sipsolutions.net>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: Cezary Jackiewicz <cezary.jackiewicz@gmail.com>
Cc: Matthias Schiffer <mschiffer@universe-factory.net>
Cc: Dirk Neukirchen <dirkneukirchen@web.de>
Cc: Christian Mehlis <christian@m3hlis.de>
Cc: Luka Perkov <luka@openwrt.org>
Cc: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 49029
Some u-boot versions for QCA955x set currently not cleared bits depending
on the used link speed. This breaks the rx/tx under OpenWrt. The mach-*.c
file is responsible to select the correct configuration bits and thus the
ath79_setup_qca955x_eth_cfg has to clear the unset.
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
SVN-Revision: 49028
Complete internal switch initialization for QCA956X.
Set default mdio device if the interface mode of GE0 is not SGMII (fix ticket #21520).
Signed-off-by: Weijie Gao <hackpascal@gmail.com>
SVN-Revision: 48937
QCA9563 and QCA9561 are two series of Qualcomm SoC Dragonfly. The only different
is QCA9563 w/o internal switch. It has one GMAC with SGMII interface. But they
have the same device ID(0x1150). So they share the same codes.
Signed-off-by: Miaoqing Pan <miaoqing@codeaurora.org>
SVN-Revision: 46971
The ETH_RXDV_DELAY (17:16) and ETH_RXD_DELAY (15:14) are currently not cleared
by the function ath79_setup_ar934x_eth_cfg. Clearing these in the
ath79_setup_ar934x_eth_cfg may cause problems on some hardware because they
rely on the preset value by the bootloader.
Instead another function is introduced which also works on ETH_CFG on AR934x.
It can be used to safely clear and set ETH_RXDV_DELAY and ETH_RXD_DELAY on
machines which require special settings.
Signed-off-by: Sven Eckelmann <sven@open-mesh.com>
SVN-Revision: 45523
While the AR9331 has a gigabit MAC towards the internal switch, the
integrated PHYs however are only 100-base-tx capable. The existing code
however advertieses gigabit capability in the link status word. If you
attach such a PHY to a gigabit capable switch on the remote end, with
some probability it attempts to negotiate gigabit and fails, falling
baco to the AR9331 assuming a 10mbit half-duplex link. This has been
observed quite frequently with the Carambola2 and gigabit capable
switches.
In ath79_register_eth(), "pdata->has_gbit = 1;" is set unconditionally
for both AR9331 ethernet ports. This is most likely wrong. Despite the
two MAC IP cores being gigabit MACs, the MAC for eth1 is connected to a
100base-T PHY via MII. The has_gbit attribute is used in the ethernet
driver to determine the supported link modes.
So either pdata->has_gbit is not set to 1 anymore, or the ethernet
driver needs to be modified to determine the advertised link code word
on another criteria than pdata->has_gbit. This patch implements the
former solution.
Signed-off-by: Harald Welte <laforge@gnumonks.org>
SVN-Revision: 42432
I don't have access to the specs, so I'm not sure about every detail, but I
haven't seen any problems with my test system, a TL-WR841N v9. It looks pretty
much like a QCA955x without PCI, a little twist in the clock calculation and
a AR9331-compatible switch.
Features not yet supported:
* EHCI (my test system doesn't have USB)
* ? (I have no idea if the QCA953x has any other features I don't know about
that aren't used by the TL-WR841N v9)
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
SVN-Revision: 40399
The hardware supports large ethernet frames. Override
the maximum frame length and packet lenght mask in the
platform data to allow to use large MTU on the ethernet
interfaces.
Limit the feature to AR934x SoCs for now. It should work
on some other SoCs as well, but those has not been tested
yet.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 39149
This enables us to modify the ag71xx_mdio_platform_data from within the
board support files.
Signed-off-by: Felix Kaechele <heffer@fedoraproject.org>
Patchwork: http://patchwork.openwrt.org/patch/4613/
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 39126
Rename the function and extend it in order to make it
usable from board setup code.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 38085
