[PATCH v3 2/3] drm/xe: Allow to disable engines

[Lucas De Marchi]

On Fri, May 23, 2025 at 02:02:11PM -0700, Matt Roper wrote:
>On Fri, May 23, 2025 at 10:42:32AM -0700, Lucas De Marchi wrote:
>> Sometimes it's useful to load the driver with a smaller set of engines
>> to allow more targeted debugging, particularly on early enabling.
>>
>> Besides checking what is fused off in hardware, add similar logic to
>> disable engines in software. This will use configfs to allow users
>> to set what engine to disable, so already add prepare for that. The
>> exact configfs interface will be added later.
>>
>> Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
>> ---
>>  drivers/gpu/drm/xe/xe_configfs.c  | 15 +++++++++++++++
>>  drivers/gpu/drm/xe/xe_configfs.h  |  2 ++
>>  drivers/gpu/drm/xe/xe_hw_engine.c | 20 ++++++++++++++++++++
>>  3 files changed, 37 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_configfs.c b/drivers/gpu/drm/xe/xe_configfs.c
>> index cb9f175c89a1c..11ca36f194bfc 100644
>> --- a/drivers/gpu/drm/xe/xe_configfs.c
>> +++ b/drivers/gpu/drm/xe/xe_configfs.c
>> @@ -226,6 +226,21 @@ void xe_configfs_clear_survivability_mode(struct pci_dev *pdev)
>>  	config_item_put(&dev->group.cg_item);
>>  }
>>
>> +/**
>> + * xe_configfs_get_engine_allowed - get engine allowed mask from configfs
>> + * @pdev: pci device
>> + *
>> + * Find the configfs group that belongs to the pci device and return
>> + * the mask of engines allowed to be used.
>> + *
>> + * Return: engine mask with allowed engines
>> + */
>> +u64 xe_configfs_get_engine_allowed(struct pci_dev *pdev)
>
>Nitpick:  should this be plural engine*s*_allowed?  Alternatively,
>engine_mask_allowed, although the function name is starting to get a bit
>long...
>
>> +{
>> +	/* dummy implementation */
>> +	return U64_MAX;
>> +}
>> +
>>  int __init xe_configfs_init(void)
>>  {
>>  	struct config_group *root = &xe_configfs.su_group;
>> diff --git a/drivers/gpu/drm/xe/xe_configfs.h b/drivers/gpu/drm/xe/xe_configfs.h
>> index ef6d231b3024b..050da4689d653 100644
>> --- a/drivers/gpu/drm/xe/xe_configfs.h
>> +++ b/drivers/gpu/drm/xe/xe_configfs.h
>> @@ -14,11 +14,13 @@ int xe_configfs_init(void);
>>  void xe_configfs_exit(void);
>>  bool xe_configfs_get_survivability_mode(struct pci_dev *pdev);
>>  void xe_configfs_clear_survivability_mode(struct pci_dev *pdev);
>> +u64 xe_configfs_get_engine_allowed(struct pci_dev *pdev);
>>  #else
>>  static inline int xe_configfs_init(void) { return 0; }
>>  static inline void xe_configfs_exit(void) { }
>>  static inline bool xe_configfs_get_survivability_mode(struct pci_dev *pdev) { return false; }
>>  static inline void xe_configfs_clear_survivability_mode(struct pci_dev *pdev) { }
>> +static inline u64 xe_configfs_get_engine_allowed(struct pci_dev *pdev) { return U64_MAX; }
>>  #endif
>>
>>  #endif
>> diff --git a/drivers/gpu/drm/xe/xe_hw_engine.c b/drivers/gpu/drm/xe/xe_hw_engine.c
>> index 93241fd0a4ba3..8e7f580db86d8 100644
>> --- a/drivers/gpu/drm/xe/xe_hw_engine.c
>> +++ b/drivers/gpu/drm/xe/xe_hw_engine.c
>> @@ -17,6 +17,7 @@
>>  #include "regs/xe_irq_regs.h"
>>  #include "xe_assert.h"
>>  #include "xe_bo.h"
>> +#include "xe_configfs.h"
>>  #include "xe_device.h"
>>  #include "xe_execlist.h"
>>  #include "xe_force_wake.h"
>> @@ -810,6 +811,24 @@ static void check_gsc_availability(struct xe_gt *gt)
>>  	}
>>  }
>>
>> +static void check_sw_disable(struct xe_gt *gt)
>> +{
>> +	struct xe_device *xe = gt_to_xe(gt);
>> +	u64 sw_allowed = xe_configfs_get_engine_allowed(to_pci_dev(xe->drm.dev));
>> +	enum xe_hw_engine_id id;
>> +
>> +	for (id = 0; id < XE_NUM_HW_ENGINES; ++id) {
>> +		if (!(gt->info.engine_mask & BIT(id)))
>> +			continue;
>> +
>> +		if (!(sw_allowed & BIT(id))) {
>> +			gt->info.engine_mask &= ~BIT(id);
>> +			drm_info(&xe->drm, "%s disabled via configfs\n",
>
>We should probably use xe_gt_info() so that it's clear what these
>messages apply to, especially on a multi-tile platform where the
>different tiles may be fused differently and leave a different set of
>remaining engines that could be potentially disabled by software.

I had it as gt_info here, but that wasn't matching the log message on
the other functions in this file. I will add a patch to convert logging
messages

Lucas De Marchi