lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <f0b85a37-ead0-4954-b7e0-dd09cbd9ab09@intel.com>
Date: Thu, 18 Apr 2024 13:53:50 +0800
From: "Li, Ming" <ming4.li@...el.com>
To: Terry Bowman <Terry.Bowman@....com>, <dan.j.williams@...el.com>,
	<rrichter@....com>
CC: <linux-cxl@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [RFC PATCH 3/6] PCI/AER: Enable RCEC to report internal error for
 CXL root port

On 4/16/2024 10:46 PM, Terry Bowman wrote:
> Hi Ming,
> 
> On 4/16/24 02:27, Li, Ming wrote:
>> On 3/26/2024 3:42 AM, Terry Bowman wrote:
>>> Hi Li, 
>>>
>>> I added comments below.
>>>
>>> On 3/13/24 03:35, Li Ming wrote:
>>>> Per CXl r3.1 section 12.2.2, CXL.cachemem protocol erros detected by CXL
>>>> root port could be logged in RCEC AER Extended Capability as
>>>> PCI_ERR_UNC_INTN or PCI_ERR_COR_INTERNAL. Unmask these errors for that
>>>> case.
>>>>
>>>> Signed-off-by: Li Ming <ming4.li@...el.com>
>>>> ---
>>>>  drivers/pci/pcie/aer.c | 24 +++++++++++++++++-------
>>>>  1 file changed, 17 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/drivers/pci/pcie/aer.c b/drivers/pci/pcie/aer.c
>>>> index 42a3bd35a3e1..364c74e47273 100644
>>>> --- a/drivers/pci/pcie/aer.c
>>>> +++ b/drivers/pci/pcie/aer.c
>>>> @@ -985,7 +985,7 @@ static bool cxl_error_is_native(struct pci_dev *dev)
>>>>  {
>>>>  	struct pci_host_bridge *host = pci_find_host_bridge(dev->bus);
>>>>  
>>>> -	return (pcie_ports_native || host->native_aer);
>>>> +	return (pcie_ports_native || host->native_aer) && host->is_cxl;
>>>>  }
>>>>  
>>>>  static bool is_internal_error(struct aer_err_info *info)
>>>> @@ -1041,8 +1041,13 @@ static int handles_cxl_error_iter(struct pci_dev *dev, void *data)
>>>>  {
>>>>  	bool *handles_cxl = data;
>>>>  
>>>> -	if (!*handles_cxl)
>>>> -		*handles_cxl = is_cxl_mem_dev(dev) && cxl_error_is_native(dev);
>>>> +	if (!*handles_cxl && cxl_error_is_native(dev)) {
>>>> +		if (pci_pcie_type(dev) == PCI_EXP_TYPE_RC_END &&
>>>> +		    dev->rcec && is_cxl_mem_dev(dev))
>>>> +			*handles_cxl = true;
>>>> +		if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT)
>>>> +			*handles_cxl = true;
>>>> +	}
>>> I understand a root port can be found under an RCEC. It's possible. But, does the downstream 
>>> root port forward AER to the upstream RCEC? My understanding is AER is handled and processed
>>> at the first root port/RCEC upstream from the device/RCH/USP/DSP.
>>>  
>>> Regards,
>>> Terry
>>>
>>
>> CXL r3.1 section 12.2.2 mentions this:
>>
>> "If the CXL.cachemem protocol errors detected by a CXL root port are logged as
>> CIEs or UIEs in an RCEC’s AER Extended Capability, it is recommended that the System
>> Firmware populate an RDPAS record (see Section 9.18.1.5) to establish the association
>> between the RCEC and the root port."
>>
>> I think it means that CXL root port is possible to forward its AER to RCEC.
>>
>> Thanks
>> Ming
>>
> 
> Thanks for pointing to spec details. 
> 
> In testing here, we used root port as agent to consume root port CXL protocol errors.
> The logic to handle the root port errors requires little to no AER driver changes.
> This results in a root port consuming VH protocol errors and RCEC consuming RCD 
> protocol errors. The RCEC and root port both use the PCIe port bus driver's AER service
> driver in separate instances for RCEC-RCD and root-port-VH.
> 
> The driver support is much simpler if RCEC does not handle VH protocol errors. Is there 
> a reason to forward root port VH mode protocol errors to an RCEC rather than consume 
> in the root port's AER driver and forward to CXL error handler? 
> 
> Regards,
> Terry

I agree that is simpler if only root port handle VH protocol errors, but I think that software has no chance to choose if VH protocol errors reported to RCEC or root port, it depends on platform implementation. So I think we should support both cases.


Thanks
Ming

> 
>>>>  
>>>>  	/* Non-zero terminates iteration */
>>>>  	return *handles_cxl;
>>>> @@ -1054,13 +1059,18 @@ static bool handles_cxl_errors(struct pci_dev *rcec)
>>>>  
>>>>  	if (pci_pcie_type(rcec) == PCI_EXP_TYPE_RC_EC &&
>>>>  	    pcie_aer_is_native(rcec))
>>>> -		pcie_walk_rcec(rcec, handles_cxl_error_iter, &handles_cxl);
>>>> +		pcie_walk_rcec_all(rcec, handles_cxl_error_iter, &handles_cxl);
>>>>  
>>>>  	return handles_cxl;
>>>>  }
>>>>  
>>>> -static void cxl_rch_enable_rcec(struct pci_dev *rcec)
>>>> +static void cxl_enable_rcec(struct pci_dev *rcec)
>>>>  {
>>>> +	/*
>>>> +	 * Enable RCEC's internal error report for two cases:
>>>> +	 * 1. RCiEP detected CXL.cachemem protocol errors
>>>> +	 * 2. CXL root port detected CXL.cachemem protocol errors.
>>>> +	 */
>>>>  	if (!handles_cxl_errors(rcec))
>>>>  		return;
>>>>  
>>>> @@ -1069,7 +1079,7 @@ static void cxl_rch_enable_rcec(struct pci_dev *rcec)
>>>>  }
>>>>  
>>>>  #else
>>>> -static inline void cxl_rch_enable_rcec(struct pci_dev *dev) { }
>>>> +static inline void cxl_enable_rcec(struct pci_dev *dev) { }
>>>>  static inline void cxl_rch_handle_error(struct pci_dev *dev,
>>>>  					struct aer_err_info *info) { }
>>>>  #endif
>>>> @@ -1494,7 +1504,7 @@ static int aer_probe(struct pcie_device *dev)
>>>>  		return status;
>>>>  	}
>>>>  
>>>> -	cxl_rch_enable_rcec(port);
>>>> +	cxl_enable_rcec(port);
>>>>  	aer_enable_rootport(rpc);
>>>>  	pci_info(port, "enabled with IRQ %d\n", dev->irq);
>>>>  	return 0;
>>


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ