[PATCH net-next 0/3] net: ethernet: cortina: Use page_pool for Gemini RX buffers

[Linus Walleij <linusw@kernel.org>]

The Gemini ethernet has this really convoluted queue mechanism using
a custom queue.

Move the Gemini RX buffer management to a parent-owned page_pool.

This series fixes DMA mapping and synchronization, fragment recycling,
and teardown while retaining the hardware free queue shared by both ports.

We need to track fragments by the exact DMA address returned by
hardware, since RX software IDs originate in the classification hash
table rather than the free queue.

Scale the default RX rings and shared free queue on Gemini systems
with 32 or 64 MiB of RAM, as the RAM usage may be a bit heavy on
those.

Assisted-by: LLM
Signed-off-by: Linus Walleij <linusw@kernel.org>
---
Linus Walleij (3):
      net: ethernet: cortina: Keep shared free queue parent-owned
      net: ethernet: cortina: Manage RX buffers with page_pool
      net: ethernet: cortina: Scale Gemini RX queues to system memory

 drivers/net/ethernet/cortina/Kconfig  |   1 +
 drivers/net/ethernet/cortina/gemini.c | 389 ++++++++++++++++++++++------------
 2 files changed, 252 insertions(+), 138 deletions(-)
---
base-commit: 2ff4ca33660706f4092866b220eb0c94940f7e0c
change-id: 20260919-gemini-ethernet-fixes-3-f0403653f23a

Best regards,
--  
Linus Walleij <linusw@kernel.org>