Enhance the AM65 CPSW NUSS driver to perform a MDIO reset using a GPIO
line. Logic is also added to perform a pre and post delay around reset
using the optional 'reset-delay-us' and 'reset-post-delay-us' properties.
This is similar to the reset being performed in the Linux kernel. The
reset is done once when the CPSW MDIO bus is being initialized.
Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
This patch adds support for MediaTek MT7988.
MT7988 features MediaTek NETSYS v3, including three GMACs, and two
of them supports 10Gbps USXGMII.
MT7988 embeds a MT7531 switch (not MCM) which supports accessing
internal registers through MMIO instead of MDIO.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
MT7981 has its GMAC2 PHY shared with USB3. To enable GMAC2, mux
register must be set to connect the SGMII phy to GMAC2.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
Existing SGMII support of mtk-eth is actually a MediaTek-specific
2.5Gbps high-speed SGMII (HSGMII) which does not support
auto-negotiation mode.
This patch adds SGMII 1Gbps auto-negotiation mode and rename the
existing HSGMII to 2500basex.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
mt7531_mmd_ind_read and mt753x_switch_init are defined without static.
Since they're not used outside this file, we should add them back.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
fixup to add static qualifier
The original direct MDIO clause 45 access via SoC is missing the
data output. This patch adds it back to ensure MDIO clause 45 can
work properly for external PHYs.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
Not all switches requires 1 second delay after deasserting reset.
MT7531 requires only maximum 200ms.
This patch defines dedicated reset wait time for each switch chip, and will
significantly improve the boot time for boards using MT7531.
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
So far the switch is initialized in probe stage and is connected to PSE
unconditionally. This will cause all packets being flooded to PSE and may
cause PSE hang before entering linux.
This patch changes the connection between switch and PSE:
- Still initialize switch in probe stage, but disconnect it with PSE
- Connect switch with PSE on eth start
- Disconnect on eth stop
Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
Add rtl8169 NIC device ID and reorder the device ID.
Signed-off-by: Minda Chen <minda.chen@starfivetech.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
For RISC-V architeture, hardware maintain the dcache coherency.
Software do not flush the cache. So even cache-line size larger
than descriptor size, driver can work.
Signed-off-by: Minda Chen <minda.chen@starfivetech.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
While compiling rtl8169.c, There are many "make pointer from
integer without a cast" compile warnings. fix them with
adding cast.
Signed-off-by: Minda Chen <minda.chen@starfivetech.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
Approved DT binding has the port mode register in the
"phys" property. Get it from there instead of the custom
"cpsw-phy-sel" property.
This will allow us to keep DT in sync with Linux.
Signed-off-by: Roger Quadros <rogerq@kernel.org>
Acked-by: Nishanth Menon <nm@ti.com>
The approved DT property for MAC efuse (ROM) address is
"ti,syscon-efuse".
Use that and drop custom property "mac_efuse".
Signed-off-by: Roger Quadros <rogerq@kernel.org>
Acked-by: Nishanth Menon <nm@ti.com>
The binding represents the MDIO controller as a child device tree
node of the MAC device tree node.
The U-Boot driver mostly ignores that child device tree node and just
hardcodes the resources it uses to support both the MAC and MDIO in a
single driver.
However, some resources like pinctrl muxing states are thus ignored.
This has been a problem with some device trees that will put some
pinctrl states on the MDIO device tree node, like the SK-AM62 Device
Tree does.
Let's rework the driver a bit to create a dummy MDIO driver that we will
then get during our initialization to force the core to select the right
muxing.
Signed-off-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Acked-by: Roger Quadros <rogerq@kernel.org>
Acked-by: Nishanth Menon <nm@ti.com>
The adin phy has extended registers that can be accessed using
adin_ext_read and adin_ext_write. These registers can be read directly
using the mdio command using readext and writext. For example:
=> mdio rx ethernet@428a0000 0xff23
Reading from bus ethernet@428a0000
PHY at address 0:
65315 - 0xe01
Signed-off-by: Nate Drude <nate.d@variscite.com>
Reviewed-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Copy and tweak the required code from the linux kernel.
Only the KSZ9893 has been tested.
Signed-off-by: Karsten Wiese <karsten.wiese@protechna.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
If there is no incoming package than axiemac_recv will return -1 which
in turn leads to printing `eth_rx: recv() returned error -1` error
message in eth_rx function. But missing a package is not an fatal error,
so return -EAGAIN in that case would be more suitable.
Signed-off-by: Maksim Kiselev <bigunclemax@gmail.com>
Link: https://lore.kernel.org/r/20230719065337.69280-1-bigunclemax@gmail.com
Signed-off-by: Michal Simek <michal.simek@amd.com>
As per spec MDC must not exceed 2.5MHz, read the pclk clock
from the device tree and update the MDC clock divisor.
GEM devices support larger clock divisors and have a different
range of divisors. Program the MDIO clock divisors based on
the clock rate of the pclk clock.
Signed-off-by: Venkatesh Yadav Abbarapu <venkatesh.abbarapu@amd.com>
Link: https://lore.kernel.org/r/20230619034922.24019-1-venkatesh.abbarapu@amd.com
Signed-off-by: Michal Simek <michal.simek@amd.com>
The different CPSW sub-system Ethernet ports have different PHY mode
control registers. In order to allow the modes to get configured
independently only the register for the port in question must be
accessed, otherwise we would just be re-configuring the mode for port 1,
while leaving all others at their power-on defaults. Fix this issue by
adding a port-number based offset to the mode control base register
address based on the fact that the control registers for the different
ports are spaced exactly 0x4 bytes apart.
Fixes: 9d0dca1199 ("net: ethernet: ti: Introduce am654 gigabit eth switch subsystem driver")
Signed-off-by: Andreas Dannenberg <dannenberg@ti.com>
Reviewed-by: Siddharth Vadapalli <s-vadapalli@ti.com>
The StarFive ETHQOS hardware has its own clock and reset,so add a
corresponding glue driver to configure them.
Signed-off-by: Yanhong Wang <yanhong.wang@starfivetech.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Add a driver for the motorcomm yt8531 gigabit ethernet phy. We have
verified the driver on StarFive VisionFive2 board.
Signed-off-by: Yanhong Wang <yanhong.wang@starfivetech.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
This patch targets the last remaining commands left to sync to their
latest form - mainly the mc_get_version() API.
Besides this, remove any macro which is now of no help.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Path IO APIs to their latest form, this means the layout
of each command is created based on structures which clearly describe
the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Path Soft Parser APIs to their latest form, this
means the layout of each command is created based on structures which
clearly describe the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Path Network Interface APIs to their latest form, this
means the layout of each command is created based on structures which
clearly describe the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Resource Container APIs to their latest form, this means
the layout of each command is created based on structures which clearly
describe the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Path MAC APIs to their latest form, this means the
layout of each command is created based on structures which clearly
describe the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Sync the Data Path Buffer Pool APIs to their latest form, this means the
layout of each command is created based on structures which clearly
describe the endianness of each field rather than some macros.
The command version is kept in place, meaning that the minimum MC
version accepted is not changed in any way.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
There are multiple MC APIs which were added years ago but they are not
used at all in the u-boot source code. Remove all these APIs.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
The rtl8169 driver uses calls to dm_pci_bus_to_phys,
which are compiled under CONFIG_PCI.
Without CONFIG_PCI, this happens:
drivers/net/rtl8169.o: in function `rtl_recv_common':
drivers/net/rtl8169.c:555: undefined reference to `dm_pci_bus_to_phys'
It is only natural that this driver depends on CONFIG_PCI then.
The device does not work connected in another way anyway, and the driver
does not assume anything else at this moment.
Signed-off-by: Eugen Hristev <eugen.hristev@collabora.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Export the already existing DPNI and DPMAC counters through the newly
added callbacks.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
The ldpaa_eth driver already had a DPMAC statistics dump, this patch
extends the list of stats and adds a bit more structure to the code.
For a bit more context, the DPAA2 u-boot software architecture uses a
default network interface object - a DPNI - which, at runtime, will get
connected to the currently used DPMAC object.
Each time the .stop() eth callback is called, the DPMAC is destroyed
thus any previous counters will get lost.
As a preparation for the next patches, we add a software kept set of
DPMAC counters which will get updated before each destroy operation
takes place.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
The ldpaa_eth driver already had a DPNI statistics dump, this patch
extends the list of stats and adds a bit more structure to the code.
For a bit more context, the DPAA2 u-boot software architecture uses a
default network interface object - a DPNI - which, at runtime, will get
connected to the currently used DPMAC object.
Each time the .stop() eth callback is called, the DPNI is reset to its
original state, including its counters.
As a preparation for the next patches, we add a software kept set of
DPNI counters which will get updated before each reset operation takes
place.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
In order to simplify code, dpni_statistics can be written as a union.
Using the raw accessors we can just loop through all the statistics from
a page without trying to access each an every one independently.
Make this change to a union.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
In case RX error occurs, and the RD_RFE bit is set, the descriptor
is never returned back to the queue. Make sh_eth_recv_start return
zero length in this case so that the descriptor can be released
and pushed back to the list. Also return the more appropriate
-EAGAIN instead of -EINVAL if the descriptor is not ready yet.
Signed-off-by: Valentine Barshak <valentine.barshak@cogentembedded.com>
Reviewed-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
U-Boot writes to RX packets when constructing replies.
This can cause stale cached data to be written to RX
buffer while we're receiving a packet. This causes RX
packet corruption because we invalidate the cache right
before processing the packet. Invalidate packet buffer
cache when preparing RX descriptor as well. This seems
to fix RX packet drops with high RX traffic.
While at it flush the descriptors right before enabling
RX/TX in sh_eth_tx_desc_init/sh_eth_rx_desc_init callbacks
when they are ready instead of flushing after allocation.
Signed-off-by: Valentine Barshak <valentine.barshak@cogentembedded.com>
Reviewed-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
The common code is now capable of handling reset GPIO associated
with PHY. Drop the local ad-hoc code in favor of common code.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
The common code is now capable of handling reset GPIO associated
with PHY. Drop the local ad-hoc code in favor of common code.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
These two properties are used by various DTs in place of
current reset-assert-us/reset-deassert-us , handle both .
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
In case a new PHY is created and DM_ETH_PHY is enabled, bind a
generic PHY driver from ETH_PHY uclass to the PHY to have a
matching DM representation of that PHY.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
The phy_connect_dev() is legacy API, now that there are no users,
make it internal to phy.c and unpublish it from headers.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Use new U-Boot phy_connect() API which also supports fixed PHYs.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
