Re: [PATCH v20 1/9] PCI/AER: Introduce AER-CXL protocol error kfifo

[Bowman, Terry]

On 9/9/2026 5:02 PM, Jonathan Cameron wrote:
> Hi Terry,
> 
>>>>
>>>> A subsequent patch wires cxl_forward_error() into handle_error_source().
>>>>
>>>> Add MAINTAINERS entries for aer_cxl_vh.c and aer_cxl_rch.c under the CXL
>>>> entry.  
>>>
>>> I couldn't immediately find any discussion about switching away from panic
>>> on a kfifo overflow.  Was there a reply to an earlier version with a
>>> discussion of that?  I'm not against the change but a 'why' 
>>> here would be good to have.
>>>   
>>
>> The driver panics on kfifo full error during UCE enqueue. This was recommended by you and
>> Richard. Changes are at the link below in cxl_forward_error():
>> https://lore.kernel.org/linux-cxl/20260902133933.2992457-2-terry.bowman@xxxxxxx/
>>
> Panic on UCE is a given and correct. I ran the not panic on correctable via some
> of our folk and they were fine with it (at least so far!).  I just saw that
> you'd stopped panic on this as part of the change log and wondered why.
> 
> Not a problem, just following the breadcrumbs!
> 
> Jonathan

Hi Jonathan,

This area was raised during your and Richard's v18 reviews. The comments:
https://lkml.org/lkml/2026/7/23/446

v19 already implemented the fix from that review: panic on UCE, drop (log) on CE. The 
v18 code had dropped every severity on overflow without panicking, and you and 
Richard pointed out a dropped UCE must not be silently lost.

v20 only tidied the CE branch (early return + clearer log message). The panic policy 
itself is unchanged since v19.

-Terry