Re: [PATCH v11 21/23] CXL/PCI: Introduce CXL uncorrectable protocol error recovery

[Dave Jiang <dave.jiang@intel.com>]

On 9/11/25 12:19 PM, Bowman, Terry wrote:
> 
> 
> On 9/3/2025 5:30 PM, Dave Jiang wrote:
>>
>> On 8/26/25 6:35 PM, Terry Bowman wrote:
>>> Populate the cxl_do_recovery() function with uncorrectable protocol error (UCE)
>>> handling. Follow similar design as found in PCIe error driver,
>>> pcie_do_recovery(). One difference is cxl_do_recovery() will treat all UCEs
>>> as fatal with a kernel panic. This is to prevent corruption on CXL memory.
>>>
>>> Introduce cxl_walk_port(). Make this analogous to pci_walk_bridge() but walking
>>> CXL ports instead. This will iterate through the CXL topology from the
>>> erroring device through the downstream CXL Ports and Endpoints.
>>>
>>> Export pci_aer_clear_fatal_status() for CXL to use if a UCE is not found.
>>>
>>> Signed-off-by: Terry Bowman <terry.bowman@amd.com>
>>>
>>> ---
>>> Changes in v10->v11:
>>> - pci_ers_merge_results() - Move to earlier patch
>>> ---
>>>  drivers/cxl/core/port.c |  1 +
>>>  drivers/cxl/core/ras.c  | 94 +++++++++++++++++++++++++++++++++++++++++
>>>  drivers/pci/pci.h       |  2 -
>>>  include/linux/aer.h     |  2 +
>>>  4 files changed, 97 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
>>> index 758fb73374c1..085c8620a797 100644
>>> --- a/drivers/cxl/core/port.c
>>> +++ b/drivers/cxl/core/port.c
>>> @@ -1347,6 +1347,7 @@ struct cxl_port *find_cxl_port(struct device *dport_dev,
>>>  	port = __find_cxl_port(&ctx);
>>>  	return port;
>>>  }
>>> +EXPORT_SYMBOL_NS_GPL(find_cxl_port, "CXL");
>>>  
>>>  static struct cxl_port *find_cxl_port_at(struct cxl_port *parent_port,
>>>  					 struct device *dport_dev,
>>> diff --git a/drivers/cxl/core/ras.c b/drivers/cxl/core/ras.c
>>> index 536ca9c815ce..3da675f72616 100644
>>> --- a/drivers/cxl/core/ras.c
>>> +++ b/drivers/cxl/core/ras.c
>>> @@ -6,6 +6,7 @@
>>>  #include <cxl/event.h>
>>>  #include <cxlmem.h>
>>>  #include <cxlpci.h>
>>> +#include <cxl.h>
>>>  #include "trace.h"
>>>  
>>>  static void cxl_cper_trace_corr_port_prot_err(struct pci_dev *pdev,
>>> @@ -468,8 +469,101 @@ void cxl_endpoint_port_init_ras(struct cxl_port *ep)
>>>  }
>>>  EXPORT_SYMBOL_NS_GPL(cxl_endpoint_port_init_ras, "CXL");
>>>  
>>> +static int cxl_report_error_detected(struct device *dev, void *data)
>>> +{
>>> +	struct pci_dev *pdev = to_pci_dev(dev);
>>> +	pci_ers_result_t vote, *result = data;
>>> +
>>> +	guard(device)(dev);
>>> +
>>> +	if (pci_pcie_type(pdev) == PCI_EXP_TYPE_ENDPOINT)
>>> +		vote = cxl_error_detected(dev);
>>> +	else
>>> +		vote = cxl_port_error_detected(dev);
>>> +
>>> +	vote = cxl_error_detected(dev);
>>> +	*result = pci_ers_merge_result(*result, vote);
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static int match_port_by_parent_dport(struct device *dev, const void *dport_dev)
>>> +{
>>> +	struct cxl_port *port;
>>> +
>>> +	if (!is_cxl_port(dev))
>>> +		return 0;
>>> +
>>> +	port = to_cxl_port(dev);
>>> +
>>> +	return port->parent_dport->dport_dev == dport_dev;
>>> +}
>>> +
>>> +static void cxl_walk_port(struct device *port_dev,
>>> +			  int (*cb)(struct device *, void *),
>>> +			  void *userdata)
>>> +{
>>> +	struct cxl_dport *dport = NULL;
>>> +	struct cxl_port *port;
>>> +	unsigned long index;
>>> +
>>> +	if (!port_dev)
>>> +		return;
>>> +
>>> +	port = to_cxl_port(port_dev);
>>> +	if (port->uport_dev && dev_is_pci(port->uport_dev))
>>> +		cb(port->uport_dev, userdata);
>>> +
>>> +	xa_for_each(&port->dports, index, dport)
>>> +	{
>>> +		struct device *child_port_dev __free(put_device) =
>>> +			bus_find_device(&cxl_bus_type, &port->dev, dport,
>>> +					match_port_by_parent_dport);
>>> +
>>> +		cb(dport->dport_dev, userdata);
>>> +
>>> +		cxl_walk_port(child_port_dev, cxl_report_error_detected, userdata);
>>> +	}
>>> +
>>> +	if (is_cxl_endpoint(port))
>>> +		cb(port->uport_dev->parent, userdata);
>>> +}
>>> +
>>>  static void cxl_do_recovery(struct device *dev)
>>>  {
>>> +	pci_ers_result_t status = PCI_ERS_RESULT_CAN_RECOVER;
>>> +	struct pci_dev *pdev = to_pci_dev(dev);
>>> +	struct cxl_dport *dport;
>>> +	struct cxl_port *port;
>>> +
>>> +	if ((pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT) ||
>>> +	    (pci_pcie_type(pdev) == PCI_EXP_TYPE_DOWNSTREAM)) {
>>> +		port = find_cxl_port(&pdev->dev, &dport);
>>> +	} else	if (pci_pcie_type(pdev) == PCI_EXP_TYPE_UPSTREAM) {
>>> +		struct device *port_dev = bus_find_device(&cxl_bus_type, NULL,
>>> +							  &pdev->dev, match_uport);
>>> +		port = to_cxl_port(port_dev);
>>> +	}
>> Do we not attempt recovery if the device is an endpoint? Is it because it is handled directly by AER callback of the cxl_pci driver? Should endpoint error just not be forwarded from the AER kfifo producer instead of being checked on the consumer end after going through the kfifo mechanism?
>>
>> DJ
> 
> The UCE fatal case is handled in the PCIe AER handling callback which is 
> what I used for testing. I need to add EP support here for use in the EP 
> UCE non-fatal case. 

I think that's what's missing. My point is why forward the information to just be dropped. Maybe a comment to make note of that if the support is not there yet.

DJ

> 
> I don't think bypassing the kfifo for EPs is ideal. The only headache with 
> the EPs is EP fatal UCE is handled in PCIe AER callbacks. We may be able to 
> improve with future series by adding logic to detect the link health and then 
> access if ok in the UCE fatal case. 
> 
> Terry
> 
>>> +
>>> +	if (!port)
>>> +		return;
>>> +
>>> +	cxl_walk_port(&port->dev, cxl_report_error_detected, &status);
>>> +	if (status == PCI_ERS_RESULT_PANIC)
>>> +		panic("CXL cachemem error.");
>>> +
>>> +	/*
>>> +	 * If we have native control of AER, clear error status in the device
>>> +	 * that detected the error.  If the platform retained control of AER,
>>> +	 * it is responsible for clearing this status.  In that case, the
>>> +	 * signaling device may not even be visible to the OS.
>>> +	 */
>>> +	if (cxl_error_is_native(pdev)) {
>>> +		pcie_clear_device_status(pdev);
>>> +		pci_aer_clear_nonfatal_status(pdev);
>>> +		pci_aer_clear_fatal_status(pdev);
>>> +	}
>>> +	put_device(&port->dev);
>>>  }
>>>  
>>>  static void cxl_handle_cor_ras(struct device *dev, u64 serial, void __iomem *ras_base)
>>> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
>>> index 69ff7c2d214f..0c4f73dd645f 100644
>>> --- a/drivers/pci/pci.h
>>> +++ b/drivers/pci/pci.h
>>> @@ -1170,13 +1170,11 @@ static inline void cxl_rch_enable_rcec(struct pci_dev *rcec) { }
>>>  
>>>  #ifdef CONFIG_CXL_RAS
>>>  void pci_aer_unmask_internal_errors(struct pci_dev *dev);
>>> -bool cxl_error_is_native(struct pci_dev *dev);
>>>  bool is_internal_error(struct aer_err_info *info);
>>>  bool is_cxl_error(struct pci_dev *pdev, struct aer_err_info *info);
>>>  void cxl_forward_error(struct pci_dev *pdev, struct aer_err_info *info);
>>>  #else
>>>  static inline void pci_aer_unmask_internal_errors(struct pci_dev *dev) { }
>>> -static inline bool cxl_error_is_native(struct pci_dev *dev) { return false; }
>>>  static inline bool is_internal_error(struct aer_err_info *info) { return false; }
>>>  static inline bool is_cxl_error(struct pci_dev *pdev, struct aer_err_info *info) { return false; }
>>>  static inline void cxl_forward_error(struct pci_dev *pdev, struct aer_err_info *info) { }
>>> diff --git a/include/linux/aer.h b/include/linux/aer.h
>>> index 1f79f0be4bf7..751a026fea73 100644
>>> --- a/include/linux/aer.h
>>> +++ b/include/linux/aer.h
>>> @@ -81,10 +81,12 @@ static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; }
>>>  int cxl_proto_err_kfifo_get(struct cxl_proto_err_work_data *wd);
>>>  void cxl_register_proto_err_work(struct work_struct *work);
>>>  void cxl_unregister_proto_err_work(void);
>>> +bool cxl_error_is_native(struct pci_dev *dev);
>>>  #else
>>>  static inline int cxl_proto_err_kfifo_get(struct cxl_proto_err_work_data *wd) { return 0; }
>>>  static inline void cxl_register_proto_err_work(struct work_struct *work) { }
>>>  static inline void cxl_unregister_proto_err_work(void) { }
>>> +static inline bool cxl_error_is_native(struct pci_dev *dev) { return false; }
>>>  #endif
>>>  
>>>  void pci_print_aer(struct pci_dev *dev, int aer_severity,
>