Re: [PATCH v4 4/7] cxl/acpi: Add background worker to wait for cxl_pci and cxl_mem probe

["Koralahalli Channabasappa, Smita" <Smita.KoralahalliChannabasappa@amd.com>]

On 6/3/2025 4:45 PM, Dave Jiang wrote:
> 
> 
> On 6/3/25 3:19 PM, Smita Koralahalli wrote:
>> Introduce a waitqueue mechanism to coordinate initialization between the
>> cxl_pci and cxl_mem drivers.
>>
>> Launch a background worker from cxl_acpi_probe() that waits for both
>> drivers to complete initialization before invoking wait_for_device_probe().
>> Without this, the probe completion wait could begin prematurely, before
>> the drivers are present, leading to missed updates.
>>
>> Co-developed-by: Nathan Fontenot <Nathan.Fontenot@amd.com>
>> Signed-off-by: Nathan Fontenot <Nathan.Fontenot@amd.com>
>> Co-developed-by: Terry Bowman <terry.bowman@amd.com>
>> Signed-off-by: Terry Bowman <terry.bowman@amd.com>
>> Signed-off-by: Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
>> ---
>>   drivers/cxl/acpi.c         | 23 +++++++++++++++++++++++
>>   drivers/cxl/core/suspend.c | 21 +++++++++++++++++++++
>>   drivers/cxl/cxl.h          |  2 ++
>>   3 files changed, 46 insertions(+)
>>
>> diff --git a/drivers/cxl/acpi.c b/drivers/cxl/acpi.c
>> index cb14829bb9be..978f63b32b41 100644
>> --- a/drivers/cxl/acpi.c
>> +++ b/drivers/cxl/acpi.c
>> @@ -813,6 +813,24 @@ static int pair_cxl_resource(struct device *dev, void *data)
>>   	return 0;
>>   }
>>   
>> +static void cxl_softreserv_mem_work_fn(struct work_struct *work)
>> +{
>> +	/* Wait for cxl_pci and cxl_mem drivers to load */
>> +	cxl_wait_for_pci_mem();
>> +
>> +	/*
>> +	 * Wait for the driver probe routines to complete after cxl_pci
>> +	 * and cxl_mem drivers are loaded.
>> +	 */
>> +	wait_for_device_probe();
>> +}
>> +static DECLARE_WORK(cxl_sr_work, cxl_softreserv_mem_work_fn);
>> +
>> +static void cxl_softreserv_mem_update(void)
>> +{
>> +	schedule_work(&cxl_sr_work);
>> +}
>> +
>>   static int cxl_acpi_probe(struct platform_device *pdev)
>>   {
>>   	int rc;
>> @@ -887,6 +905,10 @@ static int cxl_acpi_probe(struct platform_device *pdev)
>>   
>>   	/* In case PCI is scanned before ACPI re-trigger memdev attach */
>>   	cxl_bus_rescan();
>> +
>> +	/* Update SOFT RESERVE resources that intersect with CXL regions */
>> +	cxl_softreserv_mem_update();
>> +
>>   	return 0;
>>   }
>>   
>> @@ -918,6 +940,7 @@ static int __init cxl_acpi_init(void)
>>   
>>   static void __exit cxl_acpi_exit(void)
>>   {
>> +	cancel_work_sync(&cxl_sr_work);
>>   	platform_driver_unregister(&cxl_acpi_driver);
>>   	cxl_bus_drain();
>>   }
>> diff --git a/drivers/cxl/core/suspend.c b/drivers/cxl/core/suspend.c
>> index 72818a2c8ec8..c0d8f70aed56 100644
>> --- a/drivers/cxl/core/suspend.c
>> +++ b/drivers/cxl/core/suspend.c
>> @@ -2,12 +2,15 @@
>>   /* Copyright(c) 2022 Intel Corporation. All rights reserved. */
>>   #include <linux/atomic.h>
>>   #include <linux/export.h>
>> +#include <linux/wait.h>
>>   #include "cxlmem.h"
>>   #include "cxlpci.h"
>>   
>>   static atomic_t mem_active;
>>   static atomic_t pci_loaded;
>>   
>> +static DECLARE_WAIT_QUEUE_HEAD(cxl_wait_queue);
>> +
>>   bool cxl_mem_active(void)
>>   {
>>   	if (IS_ENABLED(CONFIG_CXL_MEM))
>> @@ -19,6 +22,7 @@ bool cxl_mem_active(void)
>>   void cxl_mem_active_inc(void)
>>   {
>>   	atomic_inc(&mem_active);
>> +	wake_up(&cxl_wait_queue);
>>   }
>>   EXPORT_SYMBOL_NS_GPL(cxl_mem_active_inc, "CXL");
>>   
>> @@ -28,8 +32,25 @@ void cxl_mem_active_dec(void)
>>   }
>>   EXPORT_SYMBOL_NS_GPL(cxl_mem_active_dec, "CXL");
>>   
>> +static bool cxl_pci_loaded(void)
>> +{
>> +	if (IS_ENABLED(CONFIG_CXL_PCI))
>> +		return atomic_read(&pci_loaded) != 0;
>> +
>> +	return false;
>> +}
>> +
>>   void mark_cxl_pci_loaded(void)
>>   {
>>   	atomic_inc(&pci_loaded);
>> +	wake_up(&cxl_wait_queue);
>>   }
>>   EXPORT_SYMBOL_NS_GPL(mark_cxl_pci_loaded, "CXL");
>> +
>> +void cxl_wait_for_pci_mem(void)
>> +{
>> +	if (!wait_event_timeout(cxl_wait_queue, cxl_pci_loaded() &&
>> +				cxl_mem_active(), 30 * HZ))
> 
> I'm trying to understand why cxl_pci_loaded() is needed. cxl_mem_active() goes above 0 when a cxl_mem_probe() instance succeeds. cxl_mem_probe() being triggered implies that an instance of cxl_pci_probe() has been called since cxl_mem_probe() is triggered from devm_cxl_add_memdev() with memdev being added and cxl_mem driver also have been loaded. So does cxl_mem_active() not imply cxl_pci_loaded() and makes it unnecessary?

Yeah you are right. I will remove this check.

Thanks
Smita
> 
> DJ
> 
> 
>> +		pr_debug("Timeout waiting for cxl_pci or cxl_mem probing\n");
>> +}
>> +EXPORT_SYMBOL_NS_GPL(cxl_wait_for_pci_mem, "CXL");
>> diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
>> index a9ab46eb0610..1ba7d39c2991 100644
>> --- a/drivers/cxl/cxl.h
>> +++ b/drivers/cxl/cxl.h
>> @@ -902,6 +902,8 @@ void cxl_coordinates_combine(struct access_coordinate *out,
>>   
>>   bool cxl_endpoint_decoder_reset_detected(struct cxl_port *port);
>>   
>> +void cxl_wait_for_pci_mem(void);
>> +
>>   /*
>>    * Unit test builds overrides this to __weak, find the 'strong' version
>>    * of these symbols in tools/testing/cxl/.
>