RTL8221B unfortunately suffers from a issue where reading registers from
MMD 30 can break the PHY. Once it is in the broken state, recovery is
only possible by hardware reset.
Add a hack which prevents the kernel from reading MMD 30 while probing
for PHYs when a RTL8221B is detected, based on the ID from MMD 1.
Fixes: https://github.com/openwrt/openwrt/issues/22140
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/25017
Signed-off-by: Jonas Jelonek <jelonek.jonas@gmail.com>
The Rollball command byte needs to be written last. Otherwise the
controller might access the wrong register or write the wrong value.
Fixes: 1fc19bc06e ("realtek: rtl93xx: mdio-smbus support for clause 45 and Rollball SFPs")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/23049
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
At least the XikeStor SKT-2.5G-100M SFP module seems to internally use
MDIO address 0 to access the PHY. This module allows accessing PHY
registers using Rollball protocol on address 0x51, and also provides
read-only C22 access on address 0x56. However, after disabling the
PHYAD0 configuration bit, only 0xffff can be read via both methods
(except for MMD device 30 which can still be accessed).
Since having MDIO address 0 enabled shouldn't do any harm on SFP modules
just leave the configuration bit alone in that case.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/23065
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The RTL8226 PHYs in Zyxel XGS1010-10 and XGS1210-10 rev A1 have swapped
MDI lanes. Specify this in the device tree, so the driver can configure
it. With this change, the PHYs no longer require initialization by the
bootloader.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21261
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The PHY supports swapping the MDI pairs (ABCD->DCBA) to simplify board
layout. On devices making use of this, it needs to be configured in the
driver, otherwise the PHY won't work properly.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21261
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Use NVMEM in device tree to set the label and eth0 MAC address based on
the U-Boot environment.
Invididual port MAC addresses and bridge MAC are still handled in the
02_network script to maintain the current assignment.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/22055
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
These devices contain a single MAC address in the U-Boot environment.
Set it as eth0 and label MAC in device tree.
To maintain the current state, the 02_network script still sets
individual port MAC addresses and the bridge MAC address.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/22024
Signed-off-by: Robert Marko <robimarko@gmail.com>
The MAC addresses for eth0 and the individual LAN ports are now
configured via device tree. The assignment itself stays the same as
before, matching factory firmware.
The 02_network script still sets the bridge MAC address, as it is
different from the lowest port MAC address.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21976
Signed-off-by: Robert Marko <robimarko@gmail.com>
There is now an implementation of .set_autoneg and .restart_autoneg for
all variants. Remove the helper function which checks for it, and just
call the operation directly.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21934
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
RTL83xx uses the BMCR and ADVERTISE registers like RTL93xx to configure
in-band auto-negotiation. Split out the common parts as a new generic
implementation and use it for RTL83xx. RTL93xx retains its own variant
of set_autoneg to support XSGMII, but calls into the generic version for
all other modes.
Tested 1000Base-X auto-negotiation on HPE 1920-8G (RTL8380). Also tested
HPE 1920-24G (RTL8382) and HPE-1920-48G (RTL8393) to make sure this does
not affect PHY ports.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21934
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Only configure the eth0 MAC address when it is not already done in the
device tree. To do this, create a new variable "eth0_mac".
Also avoid setting "label_mac" for devices already having it defined in
the device tree.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21644
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Explicitly specify all devices where the MAC address is configured based
on the U-Boot environment.
This change makes it clearer which devices use this method. Also makes
things simpler for any future devices which handle MAC address
configuration entirely via device tree.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21644
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Currently, the 02_network script always configures MAC addresses for
each individual LAN port unless "lan_mac_start" is set to "skip". This
behaviour can be unexpected, and is also somewhat broken, as it even
continues to do so when "lan_mac_start" is empty.
Change it to only do the configuration if "lan_mac_start" is non-empty,
and also remove the fallback to "lan_mac", making this more obvious and
less error-prone.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21644
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Enabling and disabling in-band auto-negotation is already supported on
RTL93xx. However, so far the advertisement is left unchanged at the
default of 0x1a0 (full duplex + pause + asymmetric pause).
Instead, set the advertisement to reflect the current configuration for
1000Base-X and 2500Base-X. Nothing needs to be done for SGMII, as the
advertisement register is ignored in that case.
Testing shows that negotiation of flow control works for 1000Base-X and
2500Base-X (tested with RTL930x on both ends of the link).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21869
Signed-off-by: Robert Marko <robimarko@gmail.com>
Add a new SerDes operation for restarting in-band negotiation including
an implemenation for RTL93xx, and call it from .pcs_an_restart.
This is a prerequisite for configuration of the in-band advertisement,
as changing it requires triggering a restart of auto-negotiation.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21869
Signed-off-by: Robert Marko <robimarko@gmail.com>
SGMII only works correctly on this device if inband auto-negotiation is
enabled. Configure the PHY for SGMII and in-band mode in the device tree
to make this happen.
For 2.5G link speeds the PHY will still switch to 2500Base-X without AN.
The same configuration also works on RTL8226, so it is fine to apply
this change to the A1 revision of XGS1010-12/XGS1210-12 as well.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/21605
Signed-off-by: Robert Marko <robimarko@gmail.com>
The initramfs.trx image can be flashed from the web interface of factory
firmware.
Unfortunately, the default boot command of the bootloader does not load
the ramdisk in the FIT image. This means that the image can only be
built when the option TARGET_ROOTFS_INITRAMFS_SEPARATE is disabled.
Tested with firmware 3.0.0.4.388_33965 (U-Boot 2022.10 / 2.0.0.5).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/20841
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
On RTL930x, each SerDes pair shares a set of PLLs with different
capabilities (LC PLL: 1G/2.5G/10G, ring PLL: 1G/2.5G). In principle,
this allows any combination of speeds on a SerDes pair. However, it
creates a special case when trying to configure a SerDes for 10G while
the LC PLL is already in use at a slower speed for the neighbor SerDes.
The current implementation just gives up in that case. Instead, free up
the LC PLL by reconfiguring the neighbor SerDes to the ring PLL.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/20568
Signed-off-by: Robert Marko <robimarko@gmail.com>
This device is very similar to the already supported XGS1210-12 A1. For
now, only revision A1 is supported (not marked on the label).
Hardware:
- RTL9302B SoC
- 16 MiB NOR flash
- 128 MiB DDR3 SDRAM
- 8x 1G RJ45 (RTL8218D)
- 2x 2.5G RJ45 (2x RTL8226)
- 2x SFP+ (supporting 1G/2.5G/10G)
- 3.3V UART serial (115200 baud 8N1) on the right side of the case
(from bottom to top: GND, RX, TX, VCC)
It is originally an unmanaged switch, so there are a few differences:
- No reset button
- Different partition layout: There is some reserved space in the middle
of the flash which might be used by the bootloader for flash testing.
The remaining space in between is used for OpenWrt using mtd-concat.
The largest contiguous area is at the beginning, allowing a maximum
kernel size of 7 MiB.
- No individual MAC address: This device ships with an empty U-Boot
environment. When an OpenWrt squashfs image is booted for the first
time, a random MAC address will be written to the environment (but
only if the environment has been initialized from the bootloader
before and contains the default MAC address).
Steps to boot initramfs image via network:
- Configure a TFTP server to provide the OpenWrt initramfs image
- Connect to device using serial (see hardware information above)
- Power on the device and enter U-Boot using Esc when prompted
- Run the following commands (adjust as necessary):
# rtk network on
# tftpboot 0x84f00000 192.168.1.100:openwrt-xgs1010-initramfs.bin
# bootm
Installation on flash:
- Boot initramfs image as described above
- Now is a good time to create a backup of all flash partitions! You'll
need this if you want to revert to the unmanaged factory firmware at
some point.
- Use sysupgrade to install OpenWrt
- After restart enter U-Boot again and set the boot command:
# setenv bootcmd 'rtk network on; bootm 0xb4900000'
# saveenv
# run bootcmd
Note: The command "rtk network on" is only needed because the drivers
currently rely on some setup by the bootloader (without this the RJ45
ports don't work). If the drivers improve in the future, it should be
removed (i.e. change the boot command to "bootm 0xb4900000").
Reverting to factory firmware:
- Write back your backup of the firmware partition (or write just the
fwconcat1 partition, and erase the other two fwconcat partitions)
- Change the boot command back to "boota" (or just erase the u-boot-env
partition so the default gets used)
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/20469
Signed-off-by: Robert Marko <robimarko@gmail.com>
This is a preparation for adding support for XGS1010-12, which is almost
identical to XGS1210-12, with some small differences (partition layout,
missing reset key).
In addition to moving the common parts to a new file, also simplify the
definition of the 2.5G PHYs to reduce duplication. With this change, the
revision-specific files only have to specify the SMI addresses.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/20469
Signed-off-by: Robert Marko <robimarko@gmail.com>
There are SFP modules which only work if auto-negotiation is disabled,
like some "OEM SFP-2.5G-T" modules. This also seems to be necessary for
RTL8226/RTL8221B PHYs when using 2500Base-X.
However, currently, it is always enabled, so add support for configuring
it to make these SFP modules and PHYs work.
This also adds locking which should be useful for future extension of
the PCS driver.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19518
Signed-off-by: Robert Marko <robimarko@gmail.com>
The bootloader on these devices uses 0x81000000 as load address for the
compressed image. Since the kernel uses a load address 0x80100000, this
only leaves a space of 15 MiB for the uncompressed image. For larger
images, the compressed data starts to get overwritten, and at some point
the boot will fail:
## Booting image from partition ... 0
## Booting kernel from Legacy Image at 81000000 ...
Version: 9.9.9.9
Created: 2025-08-07 14:56:09 UTC
Data Size: 6756645 Bytes = 6.4 MB
Checksum ... OK
Uncompressing ... LZMA: uncompress or overwrite error 1 - must RESET board to recover
Currently, initramfs images with default config are already over the
limit. And while they still happen to work regardless, adding additional
packages easily pushes the size so much that the boot fails.
Fix this by switching to rt-loader (which relocates the data to the
upper end of the RAM before decompression). The switch includes regular
kernel images to avoid this becoming an issue again in the future.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19734
Signed-off-by: Robert Marko <robimarko@gmail.com>
Add SoC revision, CPU part number, and a flag for engineering samples to
the rtl83xx_soc_info structure.
Also extend the system type string to include this information.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19653
Signed-off-by: Robert Marko <robimarko@gmail.com>
Move the definitions to mach-rtl83xx.h, so they can be used during init
to read more detailed SoC information. Also rename the RTL931X register,
as it has the same address on all RTL93xx.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19653
Signed-off-by: Robert Marko <robimarko@gmail.com>
Read model name from the register instead of using hard-coded values.
Also remove detection of the unsupported Realtek ESW/SSW SoCs. The Fast
Ethernet variants of the Maple and Cypress series stay for now, but are
moved to the RTL8380/RTL8390 families.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19653
Signed-off-by: Robert Marko <robimarko@gmail.com>
Since Linux 6.7, introduced with commit 5c2f7727d437 ("mtd: mtdpart:
check for subpartitions parsing result"), errors during subpartition
parsing cause all MTD partitions to be torn down.
Since the current mtdsplit driver for devices using H3C VFS returns
-EINVAL if it does not find a file system containing an OpenWrt image,
this makes initial installation of OpenWrt impossible.
Work around this by returning 0 when the file system contains unexpected
data. Also print a message in this case to show what is going on.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19475
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Maintain 64 bit counters by polling the hardware counters and adding up
the differences. Polling needs to happen just often enough to catch
every single overflow.
As we now have non-overflowing counters now, we can safely calculate
composite counters without getting weird results on overflow. Use this
to follow RFC 3635 more accurately by mapping the hardware counters to
the proper counters, while taking into account hardware quirks as best
as possible.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/18415
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
By default, the network interface stats are based on software counters,
which only consider traffic from and to the CPU. Implementing the
get_stats64 method allows to report the full hardware counters instead.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/18415
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The kernel offers several alternatives to get_ethtool_stats which allow
to report some stats in a more structured way. Use them where possible.
Ideally, we should follow RFC 3635 to translate the hardware counters to
the supported frame and octet counters. However, this is not feasible,
as some of the counters are 32-bit only (so it would produce incorrect
results as soon as one of them overflows).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/18415
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The MIB registers contain different stats depending on the SoC, and for
RTL930x some stats are in an additional register.
Create separate MIB descs for each SoC to implement this. Also make
reading 64-bit counters more robust, by protecting against an overflow
of the lower 32 bits during the read.
RTL931x remains unsupported, because it uses a table and thus requires
a separate implementation.
While we are at it, rename structs/functions to use the rtldsa prefix.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/18415
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The interrupt handler reads from the mailbox if no other reason for the
interrupt is known. If a spurious interrupt is received just after a
mailbox message has been sent, this means that the response to the
previous message is read again and returned by DSL_BSP_SendCMV instead
of the actual response.
To fix this, check the status register before reading from the mailbox
in the interrupt handler.
Tested on Fritzbox 7320. Without this change, there is occasionally a
kernel panic due to an out-of-bounds memory access in the ltq-adsl
driver (in DSL_DRV_DEV_G997_SnrAllocationNscGet), as a result of an
incorrect value returned by DSL_DRV_DANUBE_CmvRead. This is reproducible
by calling "dsl_cpe_pipe.sh g997dsnrg 1 1" multiple times.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/19385
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
On exit, the app tries to do an orderly shutdown of the DSL connection
before it is stopped forcibly. Since the driver does 3 attempts with a
timeout of 2 seconds each, this might take about 6 seconds in the worst
case.
This is problematic on sysupgrade, because any process that doesn't exit
within 4 seconds is killed. This means that the DSL connection might not
be stopped at all before the actual system upgrade begins.
To avoid this, use the newly added option in the driver to not retry the
L3 request on failure.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://patchwork.ozlabs.org/project/openwrt/patch/20250130102108.1606919-3-jan@3e8.eu/
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Commit 7bc487c12e ("kernel: ltq-vdsl-vr9-mei: fix warning about
field-spanning write") patched the driver to use unsafe_memcpy in
MEI_IoctlCmdMsgWrite.
However, this is not actually necessary. The assignment of the variable
"pDestPtr" can be modified so that the compiler knows about the correct
size. This way, the check in the fortified memcpy works correctly.
While at it, also adjust all places where similar code is used to copy
from a CMV_STD_MESSAGE_T struct.
Also mark all related structs as packed, because the code (and the
driver in general) seems to rely on that anyway.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://patchwork.ozlabs.org/project/openwrt/patch/20250206225444.2521817-2-jan@3e8.eu/
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Since the update to kernel 6.1, a warning like this appears in the
kernel log:
[ 49.773953] ------------[ cut here ]------------
[ 49.773998] WARNING: CPU: 3 PID: 2349 at target-arm_cortex-a7+neon-vfpv4_musl_eabi/linux-ipq40xx_generic/dsl_cpe_mei-ugw_8.5.2.10/src/drv_mei_cpe_msg_process.c:3570 MEI_IoctlCmdMsgWrite+0x290/0x2c8 [drv_mei_cpe]
[ 49.777670] memcpy: detected field-spanning write (size 4) of single field "pDestPtr" at target-arm_cortex-a7+neon-vfpv4_musl_eabi/linux-ipq40xx_generic/dsl_cpe_mei-ugw_8.5.2.10/src/drv_mei_cpe_msg_process.c:3570 (size 2)
...
[ 50.087078] ---[ end trace 0000000000000000 ]---
The variable "pDestPtr" points to the field "header.index" in a
CMV_STD_MESSAGE_T struct (header is a CMV_STD_MESSAGE_HEADER_T struct).
The offending code intentionally copies data beyond this field, which is
followed by "header.length" and "payload".
To fix this, change the assignment of "pDestPtr" to use the pointer to
the message plus the offset of the "header.index" field. This way, the
compiler knows about the size and thus the false positive warning
disappears.
While at it, also adjust all places where similar code is used to copy
from a CMV_STD_MESSAGE_T struct.
Also mark all related structs as packed, because the code (and the
driver in general) seems to rely on that anyway.
Fixes: https://github.com/openwrt/openwrt/issues/17142
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://patchwork.ozlabs.org/project/openwrt/patch/20250206225444.2521817-1-jan@3e8.eu/
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
With the current LED configuration using "mxl,led-config", the WAN LED
stops working after the interface is brought down and up again.
Since the driver also properly supports PHY LEDs now, switch to that
instead. This makes the LED work properly, but requires configuration
from userspace.
Fixes: #17782
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://github.com/openwrt/openwrt/pull/17785
Signed-off-by: Robert Marko <robimarko@gmail.com>
This is a preparation for adding support for dsl_cpe_pipe.sh with a
similar set of commands compared to the VDSL variant.
The configuration is simplified by using the "--enable-model" option.
Other options are chosen to match the VDSL variant, while also making
sure that previously enabled options stay like that. However, ReTx
options stay disabled, because of incompatibility with the ubus code.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://patchwork.ozlabs.org/project/openwrt/patch/20241019174041.1281093-3-jan@3e8.eu/
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This is a preparation for adding support for dsl_cpe_pipe.sh with a
similar set of commands compared to the VDSL variant.
The configuration is simplified by using the "--enable-model" option.
Other options are chosen to match the VDSL variant, while also making
sure that previously enabled options stay like that. However, ReTx
options stay disabled, because of incompatibility with the ubus code.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Link: https://patchwork.ozlabs.org/project/openwrt/patch/20241019174041.1281093-2-jan@3e8.eu/
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Try to do a clean disconnection via L3 request before the connection is
stopped.
Because this might take up to 6 seconds (the driver does 3 attempts with
a timeout of 2 seconds each), a termination timeout needs to be defined
in the init script.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Move the code for disconnection on exit to a separate function, and also
call it in the code paths for SIGINT and the "quit" CLI command.
While at it, make the patch description a bit clearer.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Move the code for disconnection on exit to a separate function, and also
call it in the code path for the "quit" CLI command.
While at it, make the patch description a bit clearer.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Use the correct return value in error message.
Fixes: 6e4c9738be ("ltq-vdsl-vr11-app: add version 4.23.1 for vr11 targets")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Use the correct return value in error message.
Fixes: 1daaef31b3 ("ltq-vdsl-app: disconnect when service is stopped")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The target was marked source-only due do the broken Ethernet port on
some devices. With that fixed, it can be enabled again.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The new cpsw-switch driver reserves VLAN 1 for internal use, which
conflicts with the default network configuration of OpenWrt.
Switch back to the older cpsw driver to make the network connection on
the affected devices (BeagleBone Black and AM335x EVM) usable again.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
All callers of the rtl83xx_mc_group_* functions already do the same
check, so these aren't needed.
For rtl83xx_mc_group_alloc, this branch also incorrectly returned 0
instead of a negative value. If the branch wasn't effectively dead code
anyway, this could potentially have caused bugs, as 0 is a valid
multicast group entry index.
Fixes: cde31976e3 ("realtek: Add support for Layer 2 Multicast")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The current implementation only works when store and load are called for
the same port without any other calls in between. This is because the
store function only saves a single port number instead of a portmask for
each group. It also doesn't take into account that the allocation of
multicast group entries might change between store/load calls.
As a result, the multicast port mask table gets corrupted. This also
includes the reserved entry for unknown multicast, which gets corrupted
even when no other mdb entries have been added.
Remove the code for storing/loading multicast groups entirely, as the
original commit message doesn't offer a convincing reason why this would
be necessary in the first place.
Fixes: 724e4af530 ("realtek: Store and Restore MC memberships for port enable/disable")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
There shouldn't be any reason to forward all multicast to the CPU. The
original commit message also doesn't provide a reason for this seemingly
unrelated change.
The current implementation of the delete method is also broken, as it
entirely removes any entry when the portmask contains only the CPU port,
even if it was explicitly created.
Fixes: 724e4af530 ("realtek: Store and Restore MC memberships for port enable/disable")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
There doesn't appear to be a reason to do this, as only the last entry
is actually reserved for unknown multicast.
This also fixes two issues:
- As the increment happened after the bounds check, the value of the
actually reserved last entry could be overwritten.
- On deletion of entries, a corresponding decrement was missing,
causing the wrong entry to be marked as free.
Fixes: cde31976e3 ("realtek: Add support for Layer 2 Multicast")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Actually use the index returned by rtl83xx_find_l2_cam_entry.
Fixes: cde31976e3 ("realtek: Add support for Layer 2 Multicast")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The port_vlan_add method may be called while a port is already a member
of that VLAN, so it needs to be able to handle changed flags. Fix it to
properly handle when the PVID or UNTAGGED flag was previously set, but
now no longer is.
To reduce duplication, move PVID configuration to a separate function.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The registers L2_PORT_STATIC_MV_ACT seem to specify the action to take
when the source address of a packet exists as a static fdb entry on
another port. By default the configured action is to drop such packets.
For standalone ports, this behaviour is undesired, as all traffic should
be forwarded to the CPU. So change the action to forward on standalone
ports.
A situation where this issue can occur is when a non-offloaded bond
interface is part of a bridge. In that case, the CPU port will have fdb
entries for devices connected to the bond interface, which are managed
by the assisted learning feature.
For now, this is only implemented for RTL838x/RTL839x, as the available
set of registers differs for the other devices.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
All ports are disabled by default, so configure the port isolation masks
and the pm field accordingly in the setup function. When port_enable is
called for a port, the isolation masks will be set up so that traffic
can flow between the port and the CPU.
While at it, change the code to also use the traffic_set method in
rtl83xx_setup, instead of writing to the RTL838x_PORT_ISO_CTRL(i)
registers directly.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Correctly update the isolation mask of the port being configured. The
port_bitmap variable should contain all other bridge members and needs
to be actually removed from the isolation mask instead of added to it.
Also actually remove the port being configured from the pm field of the
other ports, so that any other ports that are currently disabled will be
configured correctly when they are enabled.
Fixes: df8e6be59a ("rtl838x: add new architecture")
[fixed updating pm field of other ports]
Fixes: 2b88563ee5 ("realtek: update the tree to the latest refactored version")
[reintroduced incorrect pm field update]
Fixes: 27029277f9 ("realtek: add switch driver support for the RTL93XX based switches")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The switch driver actually expects every port to have a PHY handle, and
several branches in the code determine if a port is valid by checking
for a non-zero phy field.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The net_event_work struct is allocated, but only freed in a single case.
Move the allocation to the branch where it is actually needed, and free
it after the work has been done.
Fixes: 03e1d93e07 ("realtek: add driver support for routing offload")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
These devices have a partition table stored in flash, which compensates
for any pre-existing bad blocks by enlarging the respective partition.
This means that the current static partition table is only correct for
devices without any bad blocks.
Typical results of this mismatch are degraded wireless performance and
wrong MAC addresses, when the factory partition is shifted due to a bad
block somewhere before it. If there is a bad block already before the
ubi partition, then OpenWrt may not run at all because the kernel can't
find the rootfs.
Use the on-flash partition table to fix these issues. Replace the two
reserved partitions by the full partition list, as the driver does not
allow merging them.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The upstream value read from the device seems to already be in bits per
second, so there is no need to multiply by 1000 again (which for typical
values causes an overflow of the 32-bit unsigned integer).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
This is no longer needed now that the kernel is built with a load
address that matches the one hard-coded in the bootloader.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
A full loop accessing all FDB entries can take several milliseconds
(on RTL839x about 20 ms), so give other kernel tasks a chance to run.
This is especially important for rtl83xx_port_fdb_dump which is itself
called in a loop for all ports by the kernel.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
These two functions are identical apart from writing different values to
the read/write bit. Create a new function rtl_table_exec to reduce code
duplication.
Also replace the unbounded busy-waiting loop. The new implementation may
sleep, but as the hardware typically responds before the first poll, any
callers doing many table accesses still need to make sure not to block
other kernel tasks themselves.
So far, polling timeout errors are only handled by logging an error, but
a return value is added to allow proper handling in the future.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
This function currently prints three messages for every switch port at
KERN_INFO level. This takes a considerable amount of time during bootup
and can even trigger an external watchdog.
Replace these log messages by a single one at KERN_DEBUG level.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
As learning for the CPU port is now disabled globally, the bit in the
TX header doesn't have any effect anymore. Remove it to make the header
consistent with the global configuration.
Originally, this change was intended to be applied before commit
eb456aedfe ("realtek: use assisted learning on CPU port"), which is
why the commit message incorrectly mentions that the TX header already
disables learning.
The reason for disabling learning on the CPU port in the first place is
that it doesn't work correctly when packets are trapped to the CPU and
then forwarded by the CPU to other ports. In that case, the switch would
incorrectly learn the CPU port as source. An example that triggered this
issue are Multicast Listener Reports and IGMP membership reports.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
L2 learning on the CPU port is currently not consistently configured and
relies on the default configuration of the device. On RTL83xx, it is
disabled for packets transmitted with a TX header, as hardware learning
corrupts the forwarding table otherwise. As a result, unneeded flooding
of traffic for the CPU port can already happen on some devices now. It
is also likely that similar issues exist on RTL93xx, which doesn't have
a field to disable learning in the TX header.
To address this, disable hardware learning for the CPU port globally on
all devices. Instead, enable assisted learning to let DSA write FDB
entries to the switch.
For now, this does not sync local/bridge entries to the switch. However,
support for that was added in Linux 5.14, so the next switch to a newer
kernel version is going to fix this.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Initialize the data structure using memset to avoid the possibility of
writing garbage values to the hardware.
Always set a valid entry type, which should fix writing unicast entries
on RTL930x.
For unicast entries, set the is_static flag to prevent the switch from
aging them out.
Also set the rvid field for unicast entries. This is not strictly
necessary, as the switch fills it in automatically from a non-zero vid.
However, this makes the code consistent with multicast entry setup.
While at it, reorder the statements and fix some style issues (double
space, comma instead of semicolon at end of statement). Also remove the
unneeded priv parameter and debug print for the multicast entry setup
function.
Fixes: cde31976e3 ("realtek: Add support for Layer 2 Multicast")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Currently several messages at KERN_INFO level are printed for every FDB
del/dump operation. This can cause a significant slowdown for example
while using "bridge fdb", and may even trigger a watchdog.
Remove most of these log messages, as the new L2 table debugfs node
should be a good replacement. Change the remaining messages to
KERN_DEBUG level.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Switch to a polling implementation similar to the one for RTL838x, to
allow other kernel tasks to run while waiting.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Don't overwrite AS_DPM and L2LEARNING flags when dest_port is >= 32.
Fixes: 1773264a0c ("realtek: correct egress frame port verification")
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Hardware information:
---------------------
- HPE 1920-8G:
- RTL8380 SoC
- 8 Gigabit RJ45 ports (built-in RTL8218B)
- 2 SFP ports (built-in SerDes)
- HPE 1920-16G / HPE 1920-24G (same board):
- RTL8382 SoC
- 16/24 Gigabit RJ45 ports (built-in RTL8218B, 1/2 external RTL8218D)
- 4 SFP ports (external RTL8214FC)
- Common:
- RJ45 RS232 port on front panel
- 32 MiB NOR Flash
- 128 MiB DDR3 DRAM
- PT7A7514 watchdog
Booting initramfs image:
------------------------
- Prepare a FTP or TFTP server serving the OpenWrt initramfs image and
connect the server to a switch port.
- Connect to the console port of the device and enter the extended
boot menu by typing Ctrl+B when prompted.
- Choose the menu option "<3> Enter Ethernet SubMenu".
- Set network parameters via the option "<5> Modify Ethernet Parameter".
Enter the FTP/TFTP filename as "Load File Name" ("Target File Name"
can be left blank, it is not required for booting from RAM). Note that
the configuration is saved on flash, so it only needs to be done once.
- Select "<1> Download Application Program To SDRAM And Run".
Initial installation:
---------------------
- Boot an initramfs image as described above, then use sysupgrade to
install OpenWrt permanently. After initial installation, the
bootloader needs to be configured to load the correct image file
- Enter the extended boot menu again and choose "<4> File Control",
then select "<2> Set Application File type".
- Enter the number of the file "openwrt-kernel.bin" (should be 1), and
use the option "<1> +Main" to select it as boot image.
- Choose "<0> Exit To Main Menu" and then "<1> Boot System".
NOTE: The bootloader on these devices can only boot from the VFS
filesystem which normally spans most of the flash. With OpenWrt, only
the first part of the firmware partition contains a valid filesystem,
the rest is used for rootfs. As the bootloader does not know about this,
you must not do any file operations in the bootloader, as this may
corrupt the OpenWrt installation (selecting the boot image is an
exception, as it only stores a flag in the bootloader data, but doesn't
write to the filesystem).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Add the 7zr command line tool, which is a version of the 7z application
that only supports 7z archives.
7z is one of the two compression formats supported in H3C firmware
images (the alternative would be ARJ).
(Alternatively, the 7zr command line tool could also be built from a
current version of the public-domain LZMA SDK. That would require
repackaging the source package, as it is only provided in 7z format.)
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The bootloader on some H3C devices (for example HPE 1920 switches) only
supports booting from flash by reading an image from an "VFS" filesystem
which spans most of the available flash. The filesystem size is hard-
coded in the bootloader. However, as long as no write operations are
performed in the bootloader menu, it is sufficient if the start of the
partition contains a valid filesystem with the kernel image.
This mtdsplit parser reads the size and location of the kernel image and
finds the location of the rootfs stored after it. It assumes that the
filesystem image matches the layout of one generated by mkh3cvfs, with
a filename of "openwrt-kernel.bin" for the kernel image.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Don't use udelay to allow other kernel tasks to execute if the kernel
has been built without preemption. Also determine the timeout based on
jiffies instead of loop iterations.
This is especially important on devices containing a watchdog with a
short timeout. Without this change, the watchdog is not serviced during
PHY patching which can take multiple seconds.
Tested-by: Birger Koblitz <mail@birger-koblitz.de>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Probe the SFP module during PHY initialization and implement
insertion/removal handlers to automatically configure the media type
of the respective port.
Suggested-by: Birger Koblitz <git@birger-koblitz.de>
Tested-by: Birger Koblitz <mail@birger-koblitz.de>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Move RTL8214FC power configuration to newly created suspend and resume
methods. A media change now only results in power configuration if the
PHY is not suspended, to avoid powering up a port when the interface is
currently not up.
While at it, remove the rtl8380 prefix from function names, as this is
actually not SoC-specific.
Tested-by: Birger Koblitz <mail@birger-koblitz.de>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Toggle power on the individual PHY instead of the package. Otherwise
a media change always toggles power on the first port, and not the one
that is being configured.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Stop the connection when the control daemon is terminated. The code is
a modified version of the termination routine in version 4.23.1 of the
daemon (which doesn't support VR9 modems anymore).
This could also be implemented by calling the acos and acs commands via
dsl_cpe_pipe.sh in the init script. However, doing it in the daemon
itself has the advantage of also working if it is terminated in another
way (for example during sysupgrade).
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
The driver maintains elapsed times by repeatedly accumulating the time
since the previous update in a loop. For the elapsed showtime time, the
time difference is truncated to seconds before adding it, leading to a
sizable error over time.
Move the truncation to before calculation of the time difference in
order to remove this error. Also maintain the total elapsed time in the
same way in full seconds, to prevent the unsigned 32-bit counter from
wrapping around after about 50 days.
Testing on a VR9 device shows that the reported line uptime now matches
the actual elapsed wall time. The ADSL variant is only compile-tested,
but it should also work as the relevant code is identical.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Right now, both ltq-adsl-mei and ltq-vdsl-mei are always built, even
when they aren't necessary for the selected variant. This can cause the
build to fail, for example ltq-vdsl-mei doesn't build successfully here
on xway target due to the vectoring callback.
Make these dependencies conditional on the specific package variants,
so they are only built when actually needed.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Some users noticed repeated resyncs at random intervals, which go away
when the MEI driver is configured to use polling instead of interrupts.
Debugging shows that this seems to be caused by concurrent calls to
MEI_ReadMailbox (in the interrupt handler) and MEI_WriteMailbox. This
appears to be mostly triggered when there is an interrupt for vectoring
error reports.
In polling mode, calls to MEI_ReadMailbox are protected by the same
semaphore as is used in MEI_WriteMailbox. When interrupts are used,
MEI_WriteMailbox appears to rely on MEI_DisableDeviceInt and
MEI_EnableDeviceInt to provide mutual exclusion with the interrupt
handler. These functions mask/unmask interrupts, and there is an
additional check of the mask in the interrupt handler itself. However,
this is not sufficient on systems with SMP, as the interrupt handler
may be running in parallel, and could already be past the interrupt
mask check at this point.
This adds a lock to the interrupt handler, and also acquires this lock
in MEI_DisableDeviceInt. This should make sure that after a call to
MEI_DisableDeviceInt the interrupt is masked, and the interrupt handler
is either not running, has alread finished its work, or is still before
the interrupt mask check, and is thus going to detect the change.
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
This tells the modem about the WAN MAC address, which is used as source
address for vectoring error reports that are generated by the firmware.
It needs to be set early, as the MEI driver only actually writes the
value to the modem when is in reset state (i.e. the firmware has been
loaded, but connection has not started yet).
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
This re-enables the vectoring error sample callback and adds a
dependency to the corresponding driver.
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
In order to calculate the required pre-distortion for downstream
vectoring, the vectoring control entity (VCE) at the carrier office
needs error samples from the modem. On Lantiq VR9 modems, error reports
are generated by the firmware, but need to be multiplexed into the data
stream by the driver on the main processor when L2 encapsulation is
selected by the VCE.
This driver provides the necessary callback function, which is called by
the MEI driver after receiving an error report from the firmware.
Originally, it is part of the Lantiq PPA driver, but after a few changes
it also works with the PTM driver used in OpenWrt. The direct call to
ndo_start_xmit needs to be replaced, as the PTM driver relies on locks
from the kernel. Instead dev_queue_xmit is used, which is called from a
work queue, as it is not safe to call from an interrupt handler.
Additional changes include fixes to support recent kernel versions and
a change of the used interface from ptm0 to dsl0.
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
Without a partition subnode ofpart_core still parses direct subnodes as
partitions, but it ignores nodes with a compatible property. Due to
this, the switch to nvmem-cells made the urlader partition inaccessible.
As a result, the wireless network was broken, as the calibration data
is read from that partition by a script.
Fixes: #8983
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
The userspace application now uses the model=full option to match the
configuration of the kernel module. The source no longer contains SOAP
support, which was the primary reason to build only typical instead
of full before.
This makes several CLI commands, which were already supported in the
kernel module, available in the userspace application. For example, this
includes bbsg which allows to get information about VDSL2 bands.
Some previously applied build options were redundant. Disabling ADSL MIB
support is unnecessary, as it only applies to Danube. ADSL LED support
is no longer included in the source. ReTx counters are already included
with model type full.
This increases the size of the userspace application by approximately
15 kB (uncompressed). The kernel module does not change at all.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
This can be useful when a DFS channel is configured, as the ACS channel
list is taken into account when switching channels after a radar event.
For example, this allows to prevent the SRD channels from being used in
that case.
Signed-off-by: Jan Hoffmann <jan@3e8.eu>
[reorder structure]
Signed-off-by: David Bauer <mail@david-bauer.net>
Jan Hoffmann