Skip to content

Conversation

pamolloy
Copy link
Collaborator

@pamolloy pamolloy commented Aug 28, 2025

git cherry-pick $(git log --reverse --pretty=format:%H v6.12..github/adi/oran/main-6.12.y)

@pamolloy pamolloy requested review from gastmaier and a team August 28, 2025 10:32
@pamolloy
Copy link
Collaborator Author

pamolloy commented Aug 28, 2025

@analogdevicesinc/o-ran-linux Please take a look at the CI/CD results. If you have any questions feel free to ping myself and @gastmaier

@pamolloy pamolloy requested review from a team and removed request for a team August 28, 2025 11:12
woodrowbarlow and others added 25 commits August 28, 2025 14:20
Co-authored-by: Brian Neely <[email protected]>
Co-authored-by: Slawomir Kulig <[email protected]>
Co-authored-by: Daniel Mateu <[email protected]>
Co-authored-by: Kim Holdt <[email protected]>
Co-authored-by: Sheng-X Wang <[email protected]>
Co-authored-by: Hans Schultz <[email protected]>
Co-authored-by: Mike Shi <[email protected]>
Co-authored-by: Eduardo Grande <[email protected]>
Co-authored-by: Landau Zhang <[email protected]>
Co-authored-by: Gerhard DuToit <[email protected]>
Co-authored-by: Jenna Harris <[email protected]>
Co-authored-by: Xin Xu <[email protected]>
Co-authored-by: Caleb Ethridge <[email protected]>
Co-authored-by: Kwangdo Yi <[email protected]>
Co-authored-by: RamGanesh NambiSrinivasan <[email protected]>
Co-authored-by: Devin Wang <[email protected]>
Co-authored-by: Howard Massey <[email protected]>
Co-authored-by: Jie Zhang <[email protected]>
Co-authored-by: Joao Pinto <[email protected]>
Co-authored-by: Anish Venkataraman <[email protected]>
Co-authored-by: Pawel Zielinski <[email protected]>
Co-authored-by: Kim Holdt <[email protected]>
Co-authored-by: Brian Neely <[email protected]>
Co-authored-by: Slawomir Kulig <[email protected]>
Co-authored-by: Xin Xu <[email protected]>
Co-authored-by: Daniel Mateu <[email protected]>
Co-authored-by: Eduardo Grande <[email protected]>
Co-authored-by: Caleb Ethridge <[email protected]>
Co-authored-by: Arash Khabbazibasmenj <[email protected]>
Co-authored-by: Jianhong Xu <[email protected]>
Co-authored-by: Jie Zhang <[email protected]>
Co-authored-by: Johnny Huang <[email protected]>
Co-authored-by: Pawel Zielinski <[email protected]>
Co-authored-by: Sheng Wang <[email protected]>
Co-authored-by: Ajs Stormholt <[email protected]>
Increase size of reserved DDR region to 272MB.

Disable reserved DDR region in device tree to prevent U-Boot warnings when
performing device tree fixup.
CONFIG_HW_RANDOM (optee-rng driver) is enabled.
…urce files

Make relies on dependencies to determine how to build targets. Those dependencies
were incorrectly specified because of name conflicts where the target name
(kernel module) was the same as the intermediate object name (source file). As a result,
the kernel module didn't contain symbols from other intermediate objects.
Updating timeout value to be functional for all Ethernet scenarios regardless of data path (MACSec, switch, 10G/25G).
PCS offers RS-FEC support for 25G link speed. This option
is now enabled by default in PHY device driver and can be
changed from user-space via ethtool.

This commit also includes minor clean-up related to
streamlining macros/function names, and deletion of
unnecessary conditions in the PHY and Ethernet device
drivers.
The evaluation boards has two link status LEDs that can be controlled
by a GPIO pin. These shall indicate the link status of each of the
10G/25G links.

*Enable the GPIO LED driver
*Add PHY triggers to the LED driver to toggle PHY link status leds when
the link state changes
Frames received by NDMA can be fragmented by hardware in multiple
work-units if bigger than RX packet size. Feature added is for enabling
NDMA driver to handle the work-units recieved and returns to Ethernet
Device Driver the frame reconstructed from the fragments content.
The function misses a '~' to remove the previously enabled bit(s).
The run-time print needn't be print when successful unless debug prints
are enabled.
This also includes:
 * SerDes TX sync trigger execution after successful init calibration.
 * Move PCS configuration before starting SerDes init calibration.
 * Define new Generic Netlink command for SerDes App: “Reset 4Pack”.
The PMA's SerDes sends a "Signal OK" when the calibration is done. When
that happens, the PCS's data path should be reset.
Fixed errors in DT schema and DTS files.
Enable HS200, HS400 and HS400ES speed modes

- Enable HSx00 modes
- Enable HS200 tuning
- Fix issue in PHY init configuration
- Ensure card initialization (mmc_rescan) doesn't start before PHY
  initialization is done
- Add workaround to 'HS200 to HS downgrade' issue (CMD6) during HS400 init
  sequence
- fixed the use case when link is unplugged or SERDES is not up
- converted debug message prints to use KERN_DEBUG
…C ADI driver probe) introduced in TPGSWE-18228
The first 2 bytes of the multicast address were consistently
set to 0. This was due to the use of an `unsigned long` variable
type, which can only store 4 bytes.

This commit also includes enable promicious mode by default.
woodrowbarlow and others added 29 commits August 28, 2025 14:20
This reverts commit d7cd5ecf3c52baf8a602f64be25f6813b3ac90b1
This includes:
 - Add FSM for monitoring PLL states
 - Update SerDes FSM with the new states
 - Add functions for Ser/Deser power up/down and reset
 - Update link up procedure
 - Update link down procedure
 - Add PLL relock procedure
 - Add monitoring of health check messages
Move the ToD read before the OP register is cleared in adjust time function
The link down request waits for confirmation from the
serdes-app before powering down the lane. Previously, the
lane power down was executed immediately in the message
handler, which could occur after a link up request from
user space. This fix moves the lane power down execution
to the FMS transition action - now if a link up request
is received before confirmation from the serdes-app, the
lane power down step will be skipped.
We need to apply extra hardware configuration when the carrier
state is changed. The current logic in the callbacks, which
are triggered by link status changes, is insufficient. It only
checks the carrier state, not the state change, causing redundant
configuration applications and affecting kernel logs.
The conditions are now updated to detect the carrier state change.
This improvement includes updating the DMA descriptor list
on the fly.
The page table is not setup for the XCORR memories, which causes
errors when access the page table. Add devicetree nodes for the
XCORR memories to fix this issue.

Add nodes for XCORR memory on the secondary tile.

All reserved memories are now 'no-map'.

Use one memory region for combined input and output xcorr memories.

Use 0x38800000 and 0x3c800000 addresses for xcorr memories
since they are cacheable.
Add RMON (RFC 2819) support for ADRV906x.
Changes include:
* Support for Ethernet driver as a loadable module by merging
  PHY and Ethernet drivers into a single module
* Kernel config cleanup
* Added unregister function for mdio driver
* Fixed issues with stopping pll/serdes threads
Enable PHY RS-FEC by default. Disable it for A0 engineering samples
and boards.
In adi_twi_handle_interrupt(), error detection when the I2C
controller reported the end of transfer, while the TX/RX FIFO
still contained data, was based on checking a snapshot of the
IRQ status register. This approach was incorrect because the
status of the TX/RX FIFO can change during the execution of
adi_twi_handle_interrupt(), as the driver reads or writes data
before this condition is checked. Instead, the status of the
TX/RX FIFO should be checked directly before the error
condition is assessed.
The default configuration of SerDes and PLL in emac_common
is redundant after introducing the new configuration sequence
for PLL and SerDes, as these steps are also performed during
PLL reconfiguration or Link Up/Down procedures.
The ndo_do_ioctl callback has been deprecated in favor of ndo_eth_ioctl
for Ethernet devices.
Avoid the printed error message by detecting whether it's a dual-tile
setup instead of accepting a failure when opening the register map.
By default, all ports in the Ethernet Switch IP are enabled, and
the Ethernet driver did not disable FH ports during the switch init
procedure. This was incorrect because if a switch port is enabled,
packets can be forwarded to this port. If there is no carrier on
this port, the port buffer will eventually fill up, causing
backpressure to be applied to the other port that forwards the
packets, which impacts traffic on that port.
@pamolloy pamolloy self-assigned this Aug 28, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.