The _nxp_enet_dma_buffer_section is configured to __nocache area.
That makes the RxPkt performance very low (<= 50Mbps). By defining
it to the cacheable area, the RxPkt performance is >10x better.
Example with Zperf:
DUT command - zperf udp download 5001
PC command - iperf -u -c 192.0.2.1 -p 5001 -b 800M
Signed-off-by: Shrek Wang <shrek.wang@nxp.com>
Previously, ipv6 being disabled was a dependency of the hw acceleration
of the checksums for the nxp enet driver, because this ethernet has an
errata causing icmpv6 checksum to not be supported. Now, there is a new
config type in ethernet api for checksum types, so we can re-enable
hardware acceleration for ipv6 by implementing this type in the
get_config api in this driver.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Still mark the iface as down after ethernet_init, but then actually
check the link state and initialize carrier appropriately
This fixes the case where, the phy driver doesn't give a callback after
iface init due to the link already being up, there was no change from
the phy driver perspective, so callback wouldn't happen, and therefore
the interface could remain marked as down after boot even if carrier is up.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
This commit moves one function before another, to make the diff of the
next commit clearer of what it's doing.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Use net_if_carrier_off during iface init instead of net_eth_carrier_off,
to immediately mark net if as down
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
The Drivers using Pinctrl should be turning Pinctrl on
this should not be the responsibility of the board. This
commit removes CONFIG_PINCTRL from the boards side for nxp boards.
Signed-off-by: Emilio Benavente <emilio.benavente@nxp.com>
The code path for error handling and timestamping in tx path
was meant to be identical in function to old eth_mcux driver,
but there is actually a discrepancy causing no error handling
for timestamped frames, fix it.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Add missing braces to comply with MISRA C:2012 Rule 15.6 and
also following Zephyr's style guideline.
Signed-off-by: Pisit Sawangvonganan <pisit@ndrsolution.com>
Add required initialisation of OCOTP. The IMXRT10XX
variants don't support fuseWords to be greater than 1.
MAC0 fuse map address is 0x22 instead of 0x620.
Fill in mac_addr buffer correctly.
Signed-off-by: Matthias Alleman <matthias.alleman@basalte.be>
Fix the busy wait in the MDIO driver that was causing timing
problems in systems with real time requirements performing tasks
more frequently than about a millisecond.
Restructure the code to be less redundant and change the busy wait
kconfig to microseconds instead of millliseconds. Also actually signal
to the mdio driver that it can use the interrupt instead of busy
waiting, this seems to have been forgotten.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
This addresses the coverity issue ID 363724.
This issue seems to have been a false positive, but the root complaint
from coverity was that within the callstack of ENET_Up, some functions
use the buffer config pointer as an array, whereas in the zephyr driver
it is declared as a singleton. Address it by redeclaring this as an
array of size 1 instead of a singleton.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
This commit fixes coverity issue 392512.
The issue was that the return value of clock_control_on was not being
checked. Fix by checking the return value.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Fixed#74370, the issue is caused by the update to introduce MMIO
mapping, after MMIO mapping, should use data->base for device
base memory address.
Signed-off-by: Jiafei Pan <Jiafei.Pan@nxp.com>
Adding MMIO mapping support in ENET module driver before MDIO, MAC
and PTP drivers initialization, so that they can reuse this mapping.
Signed-off-by: Jiafei Pan <Jiafei.Pan@nxp.com>
nxp,unique-mac actually is not meant to be universally
unique, the LAA bit should therefore be set, and fix the
description of the property in the binding to clarify
the intended usage of this property.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
- Added nxp,enet1g compatible to distinguish between ENET (nxp,enet)
and ENET_1G (nxp,enet1g) peripherals within the same driver.
- Added config ETH_NXP_ENET_1G to enable 1G mode of operation on ENET_1G.
- Support RGMII mode of connection between MDIO and PHY to be
able to work with ENET_1G peripheral and support 1000M speed.
- Removed performing of PHY reset before configuring link - it is
not desirable for RTL8211F PHY connected to ENET_1G on RT1170.
Reset of other PHYs can be performed by PHY driver itself if required.
Fixes: #66348
Signed-off-by: Stanislav Poboril <stanislav.poboril@nxp.com>
As far as I can tell it appears that the hardware
does not support acceleration of ICMPV6 checksums.
For now, the easiest way to fix the runtime failure of
IPV6 is just to disable the hardware acceleration if
IPV6 is expected to be used.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Make DSA dependent on ETH_DSA_SUPPORT which is selected by driver
kconfigs, rather than having a list of dependencies.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Deprecate ETH_MCUX, by:
- Marking it as DEPRECATED in Kconfig, obviously.
- Unmarking the new driver as experimental.
- Putting the new and old drivers in the same folder.
- Reworking the menu appearance of the driver selection.
Note that technically now it is possible to choose the wrong
driver than what is enabled in DT, this is intentional, but
the correct one will obviously be enabled by default.
- Convert all sample overlays to the new Kconfigs. This was
part of the motivation for the shared overlays, as it was
causing twister/CI logs to look ugly and misleading due
to Kconfig warnings.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
