https://bugzilla.kernel.org/show_bug.cgi?id=220705
--- Comment #17 from Lukas Wunner ([email protected]) --- Created attachment 308934 --> https://bugzilla.kernel.org/attachment.cgi?id=308934&action=edit Patch to intercept writes to LnkCtl + L1SSCtl registers So there are at least three problems here: (1) We're doing something that causes the card to fall of the bus. (2) If a device is attached to a hotplug port on resume from system sleep, the BIOS may have configured ASPM on the hotplug port and the hotplugged device. But the kernel overwrites the ASPM settings on the Root Port with the settings it had on suspend through pci_restore_state(). If aspm_policy is POLICY_DEFAULT, the kernel will leave the BIOS settings of the hotplugged device as is, even though they may not match with what pci_restore_state() has set on the Root Port. (3) If BIOS screwed up and configured inconsistent settings (not the case here), we're not rectifying those settings. So I've provided a patch for (3) and I can come up with a patch for (2), but I'm still stumped as to the cause of (1), even after staring at the code and logs for hours. In the lspci output you've provided here: https://lore.kernel.org/all/lpntymy3w6ryvyo2trpqkl7i3aibofzqcp7p5jhxjlkse645iq@fepikfj4tcyk/ ... ASPM is disabled on the Root Port in all four variants. The Wifi card always has L1 enabled but with varying L1 Substate settings. coreboot seems to enable L1 and all L1 substates both on the Root Port as well as on the Wii card, so how do we get to these weird inconsistent settings? I'm attaching another debug patch which dumps not only all writes to the Link Control register, but also to the L1SS PM Control 1 and 2 registers and emits some additional printk()'s for good measure. If you could apply this on top of a stock v6.18-rc1 or newer kernel and share the full dmesg output (with POLICY_DEFAULT), I'd be grateful. Thanks! -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug. _______________________________________________ acpi-bugzilla mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/acpi-bugzilla
