Re: [PATCH V4 00/12] PCIe TPH and cache direct injection support

[Wei Huang <wei.huang2@amd.com>]

On 9/4/24 15:20, Bjorn Helgaas wrote:
> To me, this series would make more sense if we squashed these
> together:
> 
>   PCI: Introduce PCIe TPH support framework
>   PCI: Add TPH related register definition
>   PCI/TPH: Add pcie_enable_tph() to enable TPH
>   PCI/TPH: Add pcie_disable_tph() to disable TPH
>   PCI/TPH: Add save/restore support for TPH
> 
> These would add the "minimum viable functionality", e.g., enable TPH
> just for Processing Hints, with no Steering Tag support at all.  Would
> also include "pci=notph".

Will do

> 
>   PCI/TPH: Add pcie_tph_set_st_entry() to set ST tag
>   PCI/TPH: Add pcie_tph_get_cpu_st() to get ST tag
> 
> And squash these also to add Steering Tag support in a single commit,
> including enhancing the save/restore.

Can you elaborate on save/restore enhancement? Assuming that the first
combined patch will have save/restore support as suggested. Are you
talking about the ST entries save/restore as the enhancements (see
below), because the second combined patch deals with ST?

        st_entry = (u16 *)cap;
        offset = PCI_TPH_BASE_SIZEOF;
	num_entries = get_st_table_size(pdev);
        for (i = 0; i < num_entries; i++) {
                pci_write_config_word(pdev, pdev->tph_cap + offset,
 	                              *st_entry++);
                offset += sizeof(u16);
	}

> 
>   PCI/TPH: Add pcie_tph_modes() to query TPH modes
>   PCI/TPH: Add pcie_tph_enabled() to check TPH state
> 
> And maybe we can get away without these altogether.  I mentioned
> pcie_tph_modes() elsewhere; seems possibly unnecessary since drivers
> can just request the mode they want and we'll fail if it's not
> supported.
> 
> Drivers should also be able to remember whether they enabled TPH
> successfully without us having to remind them.