[PATCH RFC v3 0/6] PCI: rockchip-host: Handle unavailable downstream config space

Igor Velkov iav at iav.lv
Mon Nov 16 11:10:00 PST 2026


Helios64 (RK3399 + JMB585, x2), v7.4-rc2, 150 unbind/bind cycles,
alternating A53/A72:

- v3 only: JMB585 lost 5/150 - 3 SError panics on A72
  (rockchip_pcie_rd_conf+0xd8), 1 fixup on A53 (no panic, device gone),
  1 gen1 training timeout.
- v3 + my retry patch [1]: 0/150 lost (2 bad bring-ups, both recovered
  by one retry), no fixup hits; s2idle 20/20.

As with v2, the fixup saves A53 from the panic but not the device, and
does not catch the A72 SError. Patches 4/6 and 5/6 were not exercised.

[1] https://lore.kernel.org/r/20261116-rockchip-pcie-resume-v2-1-a633e279be97@iav.lv

--
Igor Velkov



More information about the Linux-rockchip mailing list