Re: [PPC] Boot problems after the pci-v6.18-changes
From: Bjorn Helgaas <helgaas@kernel.org>
Date: 2025-11-13 00:44:55
Also in:
linux-pci, regressions
On Wed, Nov 12, 2025 at 04:40:18AM +0100, Christian Zigotzky wrote:
On 11/11/2025 01:20 PM, Bjorn Helgaas wrote:quoted
On Tue, Nov 11, 2025 at 06:15:20AM +0100, Christian Zigotzky wrote:quoted
On 11/07/2025 06:06 AM, Christian Zigotzky wrote:quoted
On 11/05/2025 11:09 PM, Bjorn Helgaas wrote:quoted
quoted
I tested your patch with the RC4 of kernel 6.18 today. Unfortunatelyitquoted
quoted
doesn't solve the boot issue.Thanks for testing that. I see now why that approach doesn't work: quirk_disable_aspm_l0s_l1() calls pci_disable_link_state(), which updates the permissible ASPM link states, but pci_disable_link_state() only works for devices at the downstream end of a link. It doesn't work at all for Root Ports, which are at the upstream end of a link. Christian, you originally reported that both X5000 and X1000 were broken. I suspect X1000 may have been fixed in v6.18-rc3 by df5192d9bb0e ("PCI/ASPM: Enable only L0s and L1 for devicetree platforms"), but I would love to have confirmation of that.Hello Bjorn, I will enable CONFIG_PCIEASPM and CONFIG_PCIEASPM_DEFAULT for the RC5 of kernel 6.18 and test it with the X1000.I tested the RC5 of kernel 6.18 with CONFIG_PCIEASPM and CONFIG_PCIEASPM_DEFAULT enabled on my X1000 today. Unfortunately the boot problems are still present.Thanks. Can you post a dmesg somewhere so I can see what the relevant device IDs are? Can be with any kernel, doesn't have to be v6.18. We need the Vendor and Device IDs to add a quirk.X1000 kernel 6.18.0-rc5 dmesg: https://github.com/user-attachments/files/23491291/dmesg_x1000.txt
Thanks! This shows all the Root Ports are [1959:a002]: pci 0000:00:10.0: [1959:a002] type 01 class 0x060400 PCIe Root Port pci 0000:00:10.0: PCI bridge to [bus 01] pci 0000:01:00.0: [1002:6898] type 00 class 0x030000 PCIe Legacy Endpoint I'm confused because Hypexed's 6.18.0-a7-dmesg.log from https://github.com/chzigotzky/kernels/issues/17#issuecomment-3400419966 shows the same Root Ports, and apparently it booted fine even though we enabled *everything*: pci 0000:00:10.0: [1959:a002] type 01 class 0x060400 PCIe Root Port pci 0000:00:10.0: PCI bridge to [bus 01] pci 0000:01:00.0: [1002:6610] type 00 class 0x030000 PCIe Legacy Endpoint pci 0000:01:00.0: ASPM: DT platform, enabling L0s-up L0s-dw L1 ASPM-L1.1 ASPM-L1.2 PCI-PM-L1.1 PCI-PM-L1.2 It's *possible* that ASPM doesn't work on your endpoint (AMD Cypress XT [Radeon HD 5870]) but does work on Hypexed's endpoint (AMD Oland XT [Radeon HD 8670 / R5 340X OEM / R7 250/350/350X OEM]), but that seems unlikely to me. Maybe we just give up on ASPM on the [1959:a002] device. A web search doesn't show many users of it, so I don't know how many people would care. Bjorn