[PATCH net-next v2 0/2] net: ethernet: mtk_eth_soc: 9K jumbo frame support

[Sandeep Haemoon <sandeeph@loljews.com>]

This series makes the MediaTek SoC ethernet driver support 9K jumbo
frames on the 10G XGMII MACs (MT7988).  The driver advertises a 2030-byte
max_mtu and programs the MAC max-rx field for at most 2048 bytes, so
MTU > 2K is rejected and 9K frames are not usable.

Patch 1 gates page_pool on the ring buffer size.  The pool hands out
order-0 pages and is created with MTK_PP_MAX_BUF_SIZE, so rings that need
larger buffers - the HW-LRO ring already, and the rings used for jumbo
after patch 2 - cannot be backed by it.

Patch 2 adds the MTK_NETSYS_RX_9K capability, advertises 9K max_mtu on
the XGMII MACs (from the DT phy_mode at probe, so DSA can raise the
conduit MTU before the netdev is opened), programs XMAC_RX_CFG2, derives
the shared RX buffer length from the largest MTU, sizes the RX rings and
programs the PDMA RX SDL max frame from it.  A runtime MTU change that
would need a different ring buffer size is rejected with -EBUSY while DMA
is running; the rings are sized at the next open.

This work is a forward-port of the MediaTek downstream jumbo series:
Mason Chang's "change default rx buffer length", "add 9k jumbo frame
support" and "add dynamic rx buffer adjustment support" (MediaTek
OpenWrt feed 999-eth-30/31/33, carried in
frank-w/BPI-Router-Linux), plus Frank Wunderlich's "try to fix crash
the right way".  The per-ring page_pool rule is added here to make the
9K path safe without disabling page_pool for every ring.

The equivalent change was tested on a BPI-R4 Pro 8X (MT7988, 10G lan6 /
MxL86252) on the 6.18.60 backport: with both ends at MTU 9000, 8K ICMP
passes and iperf3 with MSS 8900 reaches ~9.87 Gbit/s (vs ~2.57 Gbit/s at
MSS 1460); the DSA conduit is raised to 9004 at boot.  This v7.3 series
is compile-tested with CONFIG_NET_MEDIATEK_SOC=y and COMPILE_TEST=y; no
hardware run of this exact tree yet.

Changes in v2:
- put the page_pool change first, so the series builds up to jumbo
  instead of adding it and then fixing it, as suggested by Simon Horman;
- drop the forward declarations by moving mtk_set_mcr_max_rx() above
  mtk_mac_config() and mtk_max_gmac_mtu() above mtk_dma_init(), as
  requested by Jakub Kicinski;
- wrap three lines that exceeded 80 columns;
- send patch 2 as a reply to the cover letter instead of chaining it off
  patch 1;
- Cc the maintainers listed by get_maintainer.pl.

v1: https://lore.kernel.org/all/patch-4f863a941889f0bceca1c082f2d79624@loljews.com/

The runtime-reallocation patch is not included: it reallocates the shared
RX rings when the MTU changes while an interface is up, removing the
-EBUSY restriction. It is a larger change and will be sent separately.

Sandeep Haemoon (2):
  net: ethernet: mtk_eth_soc: skip page_pool for large RX buffers
  net: ethernet: mtk_eth_soc: add 9K jumbo frame support

 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 185 ++++++++++++++------
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |   9 +-
 2 files changed, 144 insertions(+), 50 deletions(-)

base-commit: c03c1a14507dfaf8b856c2608f9ac216021fc222
-- 
2.53.0