Re: [RFC PATCH 3/5] cxl/core: introduce cxl_mem_report_poison()

[Shiyang Ruan <ruansy.fnst@fujitsu.com>]

在 2024/2/10 14:46, Dan Williams 写道:
> Shiyang Ruan wrote:
>> If poison is detected(reported from cxl memdev), OS should be notified to
>> handle it.  Introduce this function:
>>    1. translate DPA to HPA;
>>    2. construct a MCE instance; (TODO: more details need to be filled)
>>    3. log it into MCE event queue;
>>
>> After that, MCE mechanism can walk over its notifier chain to execute
>> specific handlers.
>>
>> Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
>> ---
>>   arch/x86/kernel/cpu/mce/core.c |  1 +
>>   drivers/cxl/core/mbox.c        | 33 +++++++++++++++++++++++++++++++++
>>   2 files changed, 34 insertions(+)
>>
>> diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
>> index bc39252bc54f..a64c0aceb7e0 100644
>> --- a/arch/x86/kernel/cpu/mce/core.c
>> +++ b/arch/x86/kernel/cpu/mce/core.c
>> @@ -131,6 +131,7 @@ void mce_setup(struct mce *m)
>>   	m->ppin = cpu_data(m->extcpu).ppin;
>>   	m->microcode = boot_cpu_data.microcode;
>>   }
>> +EXPORT_SYMBOL_GPL(mce_setup);
> 
> No, mce_setup() is x86 specific and the CXL subsystem is CPU
> architecture independent. My expectation is that CXL should translate
> errors for edac similar to how the ACPI GHES code does it. See usage of
> edac_raw_mc_handle_error() and memory_failure_queue().
> 
> Otherwise an MCE is a CPU consumption of poison event, and CXL is
> reporting device-side discovery of poison.

Yes, I misunderstood here.  I was mean to use MCE to finally call 
memory_failure(). I think memory_failure_queue() is what I need.

   void memory_failure_queue(unsigned long pfn, int flags)

But it can only queue one PFN at a time, we may need to make it support 
queuing a range of PFN.


--
Thanks,
Ruan.