Re: [PATCH 01/17] PCI: Add concurrency safe clear_and_set variants for LNKCTL{,2}

[Bjorn Helgaas]

[+cc Emmanuel, Rafael, Heiner]

On Thu, May 11, 2023 at 11:28:05PM +0300, Ilpo Järvinen wrote:
> ...
> One question before I trying to come up something is when PCIEASPM is =n, 
> should I provide some simple function that just does the LNKCTL write to 
> disable it?

The current pci_disable_link_state() stub when CONFIG_PCIEASPM is
unset seems clearly wrong.  In fact, it returns *success* when it
actually did nothing.

I think it should probably clear ASPM Control, at least when the OS
has ownership via _OSC.  I kind of think it should do that independent
of _OSC, but that depends on the conversation at [1].

Bjorn

[1] https://lore.kernel.org/r/ZF1dsvJYYnl8Wv0v@bhelgaas