[PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Dave Jiang]

When a PCI reset happens, the lock and enable flags of the CXL device
should be cleared to avoid stale state flags after reset. Add flag
clearing during cxl_reset_done() to clear the relevant endpoint
decoder flags for all decoders of the endpoint device.

Reported-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
---
v2:
- Rename CLEAR_MASK to RESET_MASK (Alison)
- Make the helper function take a mask argument to specify which flags
  to clear (Alison)
---
 drivers/cxl/cxl.h |  1 +
 drivers/cxl/pci.c | 19 +++++++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 9b947286eb9b..d09c84bcc015 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -333,6 +333,7 @@ int cxl_dport_map_rcd_linkcap(struct pci_dev *pdev, struct cxl_dport *dport);
 #define CXL_DECODER_F_LOCK  BIT(4)
 #define CXL_DECODER_F_ENABLE    BIT(5)
 #define CXL_DECODER_F_NORMALIZED_ADDRESSING BIT(6)
+#define CXL_DECODER_F_RESET_MASK (CXL_DECODER_F_ENABLE | CXL_DECODER_F_LOCK)
 
 enum cxl_decoder_type {
 	CXL_DECODER_DEVMEM = 2,
diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
index fbb300a01830..a0bba4b4db92 100644
--- a/drivers/cxl/pci.c
+++ b/drivers/cxl/pci.c
@@ -1030,6 +1030,20 @@ static void cxl_error_resume(struct pci_dev *pdev)
 		 dev->driver ? "successful" : "failed");
 }
 
+static int cxl_decoder_clear_flags(struct device *dev, void *data)
+{
+	struct cxl_endpoint_decoder *cxled;
+	unsigned long *flags = data;
+
+	if (!is_endpoint_decoder(dev))
+		return 0;
+
+	cxled = to_cxl_endpoint_decoder(dev);
+	cxled->cxld.flags &= ~(*flags);
+
+	return 0;
+}
+
 static void cxl_reset_done(struct pci_dev *pdev)
 {
 	struct cxl_dev_state *cxlds = pci_get_drvdata(pdev);
@@ -1045,6 +1059,11 @@ static void cxl_reset_done(struct pci_dev *pdev)
 	guard(device)(&cxlmd->dev);
 	if (cxlmd->endpoint &&
 	    cxl_endpoint_decoder_reset_detected(cxlmd->endpoint)) {
+		unsigned long flags = CXL_DECODER_F_RESET_MASK;
+
+		device_for_each_child(&cxlmd->endpoint->dev, &flags,
+				      cxl_decoder_clear_flags);
+
 		dev_crit(dev, "SBR happened without memory regions removal.\n");
 		dev_crit(dev, "System may be unstable if regions hosted system memory.\n");
 		add_taint(TAINT_USER, LOCKDEP_STILL_OK);

base-commit: f338e77383789c0cae23ca3d48adcc5e9e137e3c
-- 
2.53.0

Re: [PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Cheatham, Benjamin]

On 3/17/2026 6:49 PM, Dave Jiang wrote:
> When a PCI reset happens, the lock and enable flags of the CXL device
> should be cleared to avoid stale state flags after reset. Add flag
> clearing during cxl_reset_done() to clear the relevant endpoint
> decoder flags for all decoders of the endpoint device.
> 
> Reported-by: Dan Williams <dan.j.williams@intel.com>
> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
> ---
> v2:
> - Rename CLEAR_MASK to RESET_MASK (Alison)
> - Make the helper function take a mask argument to specify which flags
>   to clear (Alison)
> ---
>  drivers/cxl/cxl.h |  1 +
>  drivers/cxl/pci.c | 19 +++++++++++++++++++
>  2 files changed, 20 insertions(+)
> 
> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
> index 9b947286eb9b..d09c84bcc015 100644
> --- a/drivers/cxl/cxl.h
> +++ b/drivers/cxl/cxl.h
> @@ -333,6 +333,7 @@ int cxl_dport_map_rcd_linkcap(struct pci_dev *pdev, struct cxl_dport *dport);
>  #define CXL_DECODER_F_LOCK  BIT(4)
>  #define CXL_DECODER_F_ENABLE    BIT(5)
>  #define CXL_DECODER_F_NORMALIZED_ADDRESSING BIT(6)
> +#define CXL_DECODER_F_RESET_MASK (CXL_DECODER_F_ENABLE | CXL_DECODER_F_LOCK)
>  
>  enum cxl_decoder_type {
>  	CXL_DECODER_DEVMEM = 2,
> diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
> index fbb300a01830..a0bba4b4db92 100644
> --- a/drivers/cxl/pci.c
> +++ b/drivers/cxl/pci.c
> @@ -1030,6 +1030,20 @@ static void cxl_error_resume(struct pci_dev *pdev)
>  		 dev->driver ? "successful" : "failed");
>  }
>  
> +static int cxl_decoder_clear_flags(struct device *dev, void *data)
> +{
> +	struct cxl_endpoint_decoder *cxled;
> +	unsigned long *flags = data;
> +
> +	if (!is_endpoint_decoder(dev))
> +		return 0;

I would either a) rename this function to cxl_endpoint_decoder_clear_flags() or b) change
it to use struct cxl_decoder devices directly. The generality of option b is nicer imo,
but you'd have to throw in an extra check for switch decoders (or add a is_cxl_decoder() helper).

With either of those:
Reviewed-by: Ben Cheatham <benjamin.cheatham@amd.com>

Thanks,
Ben
> +
> +	cxled = to_cxl_endpoint_decoder(dev);
> +	cxled->cxld.flags &= ~(*flags);
> +
> +	return 0;
> +}
> +
>  static void cxl_reset_done(struct pci_dev *pdev)
>  {
>  	struct cxl_dev_state *cxlds = pci_get_drvdata(pdev);
> @@ -1045,6 +1059,11 @@ static void cxl_reset_done(struct pci_dev *pdev)
>  	guard(device)(&cxlmd->dev);
>  	if (cxlmd->endpoint &&
>  	    cxl_endpoint_decoder_reset_detected(cxlmd->endpoint)) {
> +		unsigned long flags = CXL_DECODER_F_RESET_MASK;
> +
> +		device_for_each_child(&cxlmd->endpoint->dev, &flags,
> +				      cxl_decoder_clear_flags);
> +
>  		dev_crit(dev, "SBR happened without memory regions removal.\n");
>  		dev_crit(dev, "System may be unstable if regions hosted system memory.\n");
>  		add_taint(TAINT_USER, LOCKDEP_STILL_OK);
> 
> base-commit: f338e77383789c0cae23ca3d48adcc5e9e137e3c

Re: [PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Dave Jiang]

On 3/18/26 7:12 AM, Cheatham, Benjamin wrote:
> On 3/17/2026 6:49 PM, Dave Jiang wrote:
>> When a PCI reset happens, the lock and enable flags of the CXL device
>> should be cleared to avoid stale state flags after reset. Add flag
>> clearing during cxl_reset_done() to clear the relevant endpoint
>> decoder flags for all decoders of the endpoint device.
>>
>> Reported-by: Dan Williams <dan.j.williams@intel.com>
>> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
>> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
>> ---
>> v2:
>> - Rename CLEAR_MASK to RESET_MASK (Alison)
>> - Make the helper function take a mask argument to specify which flags
>>   to clear (Alison)
>> ---
>>  drivers/cxl/cxl.h |  1 +
>>  drivers/cxl/pci.c | 19 +++++++++++++++++++
>>  2 files changed, 20 insertions(+)
>>
>> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
>> index 9b947286eb9b..d09c84bcc015 100644
>> --- a/drivers/cxl/cxl.h
>> +++ b/drivers/cxl/cxl.h
>> @@ -333,6 +333,7 @@ int cxl_dport_map_rcd_linkcap(struct pci_dev *pdev, struct cxl_dport *dport);
>>  #define CXL_DECODER_F_LOCK  BIT(4)
>>  #define CXL_DECODER_F_ENABLE    BIT(5)
>>  #define CXL_DECODER_F_NORMALIZED_ADDRESSING BIT(6)
>> +#define CXL_DECODER_F_RESET_MASK (CXL_DECODER_F_ENABLE | CXL_DECODER_F_LOCK)
>>  
>>  enum cxl_decoder_type {
>>  	CXL_DECODER_DEVMEM = 2,
>> diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
>> index fbb300a01830..a0bba4b4db92 100644
>> --- a/drivers/cxl/pci.c
>> +++ b/drivers/cxl/pci.c
>> @@ -1030,6 +1030,20 @@ static void cxl_error_resume(struct pci_dev *pdev)
>>  		 dev->driver ? "successful" : "failed");
>>  }
>>  
>> +static int cxl_decoder_clear_flags(struct device *dev, void *data)
>> +{
>> +	struct cxl_endpoint_decoder *cxled;
>> +	unsigned long *flags = data;
>> +
>> +	if (!is_endpoint_decoder(dev))
>> +		return 0;
> 
> I would either a) rename this function to cxl_endpoint_decoder_clear_flags() or b) change
> it to use struct cxl_decoder devices directly. The generality of option b is nicer imo,
> but you'd have to throw in an extra check for switch decoders (or add a is_cxl_decoder() helper).

Thanks for the review Ben. We can probably do both to make it extra clear.

DJ

> 
> With either of those:
> Reviewed-by: Ben Cheatham <benjamin.cheatham@amd.com>
> 
> Thanks,
> Ben
>> +
>> +	cxled = to_cxl_endpoint_decoder(dev);
>> +	cxled->cxld.flags &= ~(*flags);
>> +
>> +	return 0;
>> +}
>> +
>>  static void cxl_reset_done(struct pci_dev *pdev)
>>  {
>>  	struct cxl_dev_state *cxlds = pci_get_drvdata(pdev);
>> @@ -1045,6 +1059,11 @@ static void cxl_reset_done(struct pci_dev *pdev)
>>  	guard(device)(&cxlmd->dev);
>>  	if (cxlmd->endpoint &&
>>  	    cxl_endpoint_decoder_reset_detected(cxlmd->endpoint)) {
>> +		unsigned long flags = CXL_DECODER_F_RESET_MASK;
>> +
>> +		device_for_each_child(&cxlmd->endpoint->dev, &flags,
>> +				      cxl_decoder_clear_flags);
>> +
>>  		dev_crit(dev, "SBR happened without memory regions removal.\n");
>>  		dev_crit(dev, "System may be unstable if regions hosted system memory.\n");
>>  		add_taint(TAINT_USER, LOCKDEP_STILL_OK);
>>
>> base-commit: f338e77383789c0cae23ca3d48adcc5e9e137e3c
> 

Re: [PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Dave Jiang]

On 3/18/26 8:22 AM, Dave Jiang wrote:
> 
> 
> On 3/18/26 7:12 AM, Cheatham, Benjamin wrote:
>> On 3/17/2026 6:49 PM, Dave Jiang wrote:
>>> When a PCI reset happens, the lock and enable flags of the CXL device
>>> should be cleared to avoid stale state flags after reset. Add flag
>>> clearing during cxl_reset_done() to clear the relevant endpoint
>>> decoder flags for all decoders of the endpoint device.
>>>
>>> Reported-by: Dan Williams <dan.j.williams@intel.com>
>>> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
>>> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
>>> ---
>>> v2:
>>> - Rename CLEAR_MASK to RESET_MASK (Alison)
>>> - Make the helper function take a mask argument to specify which flags
>>>   to clear (Alison)
>>> ---
>>>  drivers/cxl/cxl.h |  1 +
>>>  drivers/cxl/pci.c | 19 +++++++++++++++++++
>>>  2 files changed, 20 insertions(+)
>>>
>>> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
>>> index 9b947286eb9b..d09c84bcc015 100644
>>> --- a/drivers/cxl/cxl.h
>>> +++ b/drivers/cxl/cxl.h
>>> @@ -333,6 +333,7 @@ int cxl_dport_map_rcd_linkcap(struct pci_dev *pdev, struct cxl_dport *dport);
>>>  #define CXL_DECODER_F_LOCK  BIT(4)
>>>  #define CXL_DECODER_F_ENABLE    BIT(5)
>>>  #define CXL_DECODER_F_NORMALIZED_ADDRESSING BIT(6)
>>> +#define CXL_DECODER_F_RESET_MASK (CXL_DECODER_F_ENABLE | CXL_DECODER_F_LOCK)
>>>  
>>>  enum cxl_decoder_type {
>>>  	CXL_DECODER_DEVMEM = 2,
>>> diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
>>> index fbb300a01830..a0bba4b4db92 100644
>>> --- a/drivers/cxl/pci.c
>>> +++ b/drivers/cxl/pci.c
>>> @@ -1030,6 +1030,20 @@ static void cxl_error_resume(struct pci_dev *pdev)
>>>  		 dev->driver ? "successful" : "failed");
>>>  }
>>>  
>>> +static int cxl_decoder_clear_flags(struct device *dev, void *data)
>>> +{
>>> +	struct cxl_endpoint_decoder *cxled;
>>> +	unsigned long *flags = data;
>>> +
>>> +	if (!is_endpoint_decoder(dev))
>>> +		return 0;
>>
>> I would either a) rename this function to cxl_endpoint_decoder_clear_flags() or b) change
>> it to use struct cxl_decoder devices directly. The generality of option b is nicer imo,
>> but you'd have to throw in an extra check for switch decoders (or add a is_cxl_decoder() helper).
> 
> Thanks for the review Ben. We can probably do both to make it extra clear.

Actually we can't go with option B because the decoders come in as the child device and not passed in as we iterate through device_for_each_child(). So I'll just go with option a. 

> 
> DJ
> 
>>
>> With either of those:
>> Reviewed-by: Ben Cheatham <benjamin.cheatham@amd.com>
>>
>> Thanks,
>> Ben
>>> +
>>> +	cxled = to_cxl_endpoint_decoder(dev);
>>> +	cxled->cxld.flags &= ~(*flags);
>>> +
>>> +	return 0;
>>> +}
>>> +
>>>  static void cxl_reset_done(struct pci_dev *pdev)
>>>  {
>>>  	struct cxl_dev_state *cxlds = pci_get_drvdata(pdev);
>>> @@ -1045,6 +1059,11 @@ static void cxl_reset_done(struct pci_dev *pdev)
>>>  	guard(device)(&cxlmd->dev);
>>>  	if (cxlmd->endpoint &&
>>>  	    cxl_endpoint_decoder_reset_detected(cxlmd->endpoint)) {
>>> +		unsigned long flags = CXL_DECODER_F_RESET_MASK;
>>> +
>>> +		device_for_each_child(&cxlmd->endpoint->dev, &flags,
>>> +				      cxl_decoder_clear_flags);
>>> +
>>>  		dev_crit(dev, "SBR happened without memory regions removal.\n");
>>>  		dev_crit(dev, "System may be unstable if regions hosted system memory.\n");
>>>  		add_taint(TAINT_USER, LOCKDEP_STILL_OK);
>>>
>>> base-commit: f338e77383789c0cae23ca3d48adcc5e9e137e3c
>>
> 

Re: [PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Jonathan Cameron]

On Tue, 17 Mar 2026 16:49:26 -0700
Dave Jiang <dave.jiang@intel.com> wrote:

> When a PCI reset happens, the lock and enable flags of the CXL device
> should be cleared to avoid stale state flags after reset. Add flag
> clearing during cxl_reset_done() to clear the relevant endpoint
> decoder flags for all decoders of the endpoint device.
> 
> Reported-by: Dan Williams <dan.j.williams@intel.com>
> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
> ---
> v2:
> - Rename CLEAR_MASK to RESET_MASK (Alison)
> - Make the helper function take a mask argument to specify which flags
>   to clear (Alison)
Unless we know of another use for this helper, I'd have gone with
Alison's other alternative of just giving it a more specific name
and not passing in a flags parameter.  Feels like a bit of premature
generalization!

Re: [PATCH v2] cxl: Add endpoint decoder flags clear when PCI reset happens

[Dave Jiang]

On 3/19/26 7:46 AM, Jonathan Cameron wrote:
> On Tue, 17 Mar 2026 16:49:26 -0700
> Dave Jiang <dave.jiang@intel.com> wrote:
> 
>> When a PCI reset happens, the lock and enable flags of the CXL device
>> should be cleared to avoid stale state flags after reset. Add flag
>> clearing during cxl_reset_done() to clear the relevant endpoint
>> decoder flags for all decoders of the endpoint device.
>>
>> Reported-by: Dan Williams <dan.j.williams@intel.com>
>> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
>> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
>> ---
>> v2:
>> - Rename CLEAR_MASK to RESET_MASK (Alison)
>> - Make the helper function take a mask argument to specify which flags
>>   to clear (Alison)
> Unless we know of another use for this helper, I'd have gone with
> Alison's other alternative of just giving it a more specific name
> and not passing in a flags parameter.  Feels like a bit of premature
> generalization!
> 
Ok will do.