On 8/27/2025 3:04 AM, Lukas Wunner wrote:
> On Tue, Aug 26, 2025 at 08:35:35PM -0500, Terry Bowman wrote:
>> +++ b/include/linux/pci.h
>> @@ -2760,6 +2760,17 @@ static inline bool pci_is_thunderbolt_attached(struct pci_dev *pdev)
>> void pci_uevent_ers(struct pci_dev *pdev, enum pci_ers_result err_type);
>> #endif
>>
>> +#if defined(CONFIG_PCIEAER)
>> +pci_ers_result_t pci_ers_merge_result(enum pci_ers_result orig,
>> + enum pci_ers_result new);
>> +#else
>> +static inline pci_ers_result_t pci_ers_merge_result(enum pci_ers_result orig,
>> + enum pci_ers_result new)
>> +{
>> + return PCI_ERS_RESULT_NONE;
>> +}
>> +#endif
>> +
>> #include <linux/dma-mapping.h>
>>
>> #define pci_emerg(pdev, fmt, arg...) dev_emerg(&(pdev)->dev, fmt, ##arg)
> Would it be possible for you to just declare a local version of
> pci_ers_merge_result() within drivers/cxl/ which is encapsulated by
> "#ifndef CONFIG_PCIEAER"?
>
> That would avoid the need to make this public in include/linux/pci.h.
>
> Thanks,
>
> Lukas
Hi Lukas,
The move of local merge_result() to exported pci_merge_result() was requested
by Jonathan Cameron:
https://lore.kernel.org/linux-cxl/20250627120541.00003a14@xxxxxxxxxx/
I believe the intent was to make reuse of the PCI merge function to keep the PCI
and CXL UCE flows somewhat similar.
Jonathan may have a better explanation. I'm not opposed to either solution but
adding details here for discussion with Jonathan.
Regards,
Terry
