[PATCH 0/2] PCI: meson: Add system suspend/resume support
From: Igor Velkov via B4 Relay
Date: Sat Nov 14 2026 - 17:13:38 EST
On SM1 the PCIe controller loses power in system suspend and the driver
has no PM callbacks, so the first config access after resume hangs the
CPU; on ODROID-HC4 s2idle never returns.
Patch 1 moves the PHY power on and the resets into the host .init and
.deinit callbacks. Patch 2 adds noirq suspend/resume through the DWC
helpers, for SM1 only.
Tested on ODROID-HC4 (SM1, ASM1061 SATA with two disks behind the port)
on an Armbian v7.4-rc2 kernel with this series, which also carries
Hans Zhang's pending MPS series (v9): 20 s2idle and 20 suspend-to-RAM
cycles, the link back at Gen2 x1 and the disk readable after each.
Suspend-to-RAM woke on the VRTC alarm via a patched BL301; the stock
firmware has no timer wake on this board.
Not tested on G12A, G12B or AXG. There patch 1 changes probe and
remove, and patch 2 leaves suspend alone.
After a failed resume, dw_pcie_suspend_noirq() calls .deinit a second
time; pci-imx6 and pcie-rcar-gen4 have no guard against it. Patch 2
guards with a flag; I can fix it in the DWC core instead if preferred.
Signed-off-by: Igor Velkov <iav@xxxxxx>
---
Igor Velkov (2):
PCI: meson: Power the PHY on and reset the controller in host .init
PCI: meson: Add system suspend/resume support
drivers/pci/controller/dwc/pci-meson.c | 108 +++++++++++++++++++++++++++------
1 file changed, 89 insertions(+), 19 deletions(-)
---
base-commit: 35f800669ce835aa77364a2ab5b61be19b778d3c
change-id: 20261114-b4-pci-meson-suspend-20798a7b9956
Best regards,
--
Igor Velkov <iav@xxxxxx>