[PATCH 1/2] drm/amdkfd: Use bitmap_zalloc() when applicable

[PATCH 1/2] drm/amdkfd: Use bitmap_zalloc() when applicable

[Christophe JAILLET]

'doorbell_bitmap' is a bitmap. So use 'bitmap_zalloc()' to simplify code,
improve the semantic and avoid some open-coded arithmetic in allocator
arguments.

Also change the corresponding 'kfree()' into 'bitmap_free()' to keep
consistency.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index f29b3932e3dc..172ee8763523 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1011,7 +1011,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 			free_pages((unsigned long)pdd->qpd.cwsr_kaddr,
 				get_order(KFD_CWSR_TBA_TMA_SIZE));
 
-		kfree(pdd->qpd.doorbell_bitmap);
+		bitmap_free(pdd->qpd.doorbell_bitmap);
 		idr_destroy(&pdd->alloc_idr);
 
 		kfd_free_process_doorbells(pdd->dev, pdd->doorbell_index);
@@ -1434,9 +1434,8 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
 	if (!KFD_IS_SOC15(dev->device_info->asic_family))
 		return 0;
 
-	qpd->doorbell_bitmap =
-		kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
-				     BITS_PER_BYTE), GFP_KERNEL);
+	qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS
+					     GFP_KERNEL);
 	if (!qpd->doorbell_bitmap)
 		return -ENOMEM;
 
-- 
2.30.2

[PATCH 2/2] drm/amdkfd: Slighly optimize 'init_doorbell_bitmap()'

[Christophe JAILLET]

The 'doorbell_bitmap' bitmap has just been allocated. So we can use the
non-atomic '__set_bit()' function to save a few cycles as no concurrent
access can happen.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
bitmap_set() could certainly also be use, but range checking would be
tricky.
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 172ee8763523..2e9d341062c4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1447,9 +1447,9 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
 
 	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
 		if (i >= range_start && i <= range_end) {
-			set_bit(i, qpd->doorbell_bitmap);
-			set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
-				qpd->doorbell_bitmap);
+			__set_bit(i, qpd->doorbell_bitmap);
+			__set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
+				  qpd->doorbell_bitmap);
 		}
 	}
 
-- 
2.30.2

Re: [PATCH 2/2] drm/amdkfd: Slighly optimize 'init_doorbell_bitmap()'

[Felix Kuehling]

Am 2021-11-21 um 12:41 p.m. schrieb Christophe JAILLET:
> The 'doorbell_bitmap' bitmap has just been allocated. So we can use the
> non-atomic '__set_bit()' function to save a few cycles as no concurrent
> access can happen.
>
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>

Thank you for the patches. I think the same sort of change (at least the
allocation/freeing part) could be applied to the queue_slot_bitmap in
kfd_process_queue_manager.c. Would you like to submit another revision
of this patch series that handles that as well?

Either way, this series is

Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>


> ---
> bitmap_set() could certainly also be use, but range checking would be
> tricky.
> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> index 172ee8763523..2e9d341062c4 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
> @@ -1447,9 +1447,9 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
>  
>  	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
>  		if (i >= range_start && i <= range_end) {
> -			set_bit(i, qpd->doorbell_bitmap);
> -			set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
> -				qpd->doorbell_bitmap);
> +			__set_bit(i, qpd->doorbell_bitmap);
> +			__set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
> +				  qpd->doorbell_bitmap);
>  		}
>  	}
>  

Re: [PATCH 1/2] drm/amdkfd: Use bitmap_zalloc() when applicable

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 3269 bytes --]

Hi Christophe,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on drm/drm-next]
[also build test ERROR on v5.16-rc2 next-20211118]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Christophe-JAILLET/drm-amdkfd-Use-bitmap_zalloc-when-applicable/20211122-014305
base:   git://anongit.freedesktop.org/drm/drm drm-next
config: x86_64-randconfig-a015-20211121 (attached as .config)
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # https://github.com/0day-ci/linux/commit/0f65744094ce515a9dfe6b5e655ad59861eca35b
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Christophe-JAILLET/drm-amdkfd-Use-bitmap_zalloc-when-applicable/20211122-014305
        git checkout 0f65744094ce515a9dfe6b5e655ad59861eca35b
        # save the attached .config to linux build tree
        mkdir build_dir
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

>> drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:1438:11: error: called object type 'int' is not a function or function pointer
                                                GFP_KERNEL);
                                                ^
   include/linux/gfp.h:324:20: note: expanded from macro 'GFP_KERNEL'
   #define GFP_KERNEL      (__GFP_RECLAIM | __GFP_IO | __GFP_FS)
                           ^
   1 error generated.


vim +/int +1438 drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c

  1426	
  1427	static int init_doorbell_bitmap(struct qcm_process_device *qpd,
  1428				struct kfd_dev *dev)
  1429	{
  1430		unsigned int i;
  1431		int range_start = dev->shared_resources.non_cp_doorbells_start;
  1432		int range_end = dev->shared_resources.non_cp_doorbells_end;
  1433	
  1434		if (!KFD_IS_SOC15(dev->device_info->asic_family))
  1435			return 0;
  1436	
  1437		qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS
> 1438						     GFP_KERNEL);
  1439		if (!qpd->doorbell_bitmap)
  1440			return -ENOMEM;
  1441	
  1442		/* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */
  1443		pr_debug("reserved doorbell 0x%03x - 0x%03x\n", range_start, range_end);
  1444		pr_debug("reserved doorbell 0x%03x - 0x%03x\n",
  1445				range_start + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
  1446				range_end + KFD_QUEUE_DOORBELL_MIRROR_OFFSET);
  1447	
  1448		for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
  1449			if (i >= range_start && i <= range_end) {
  1450				set_bit(i, qpd->doorbell_bitmap);
  1451				set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
  1452					qpd->doorbell_bitmap);
  1453			}
  1454		}
  1455	
  1456		return 0;
  1457	}
  1458	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 36177 bytes --]

Re: [PATCH 2/2] drm/amdkfd: Slighly optimize 'init_doorbell_bitmap()'

[Christophe JAILLET]

Le 22/11/2021 à 22:44, Felix Kuehling a écrit :
> Am 2021-11-21 um 12:41 p.m. schrieb Christophe JAILLET:
>> The 'doorbell_bitmap' bitmap has just been allocated. So we can use the
>> non-atomic '__set_bit()' function to save a few cycles as no concurrent
>> access can happen.
>>
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> 
> Thank you for the patches. I think the same sort of change (at least the
> allocation/freeing part) could be applied to the queue_slot_bitmap in
> kfd_process_queue_manager.c. Would you like to submit another revision
> of this patch series that handles that as well?

I'll send a v2 which will fix the missing ',' spotted by the kernel test 
robot and include kfd_process_queue_manager.c.

All my patches are compile tested (otherwise it is said bellow the ---). 
Looks like I missed this one :(.

CJ

> 
> Either way, this series is
> 
> Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
> 
> 
>> ---
>> bitmap_set() could certainly also be use, but range checking would be
>> tricky.
>> ---
>>   drivers/gpu/drm/amd/amdkfd/kfd_process.c | 6 +++---
>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> index 172ee8763523..2e9d341062c4 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> @@ -1447,9 +1447,9 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
>>   
>>   	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
>>   		if (i >= range_start && i <= range_end) {
>> -			set_bit(i, qpd->doorbell_bitmap);
>> -			set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
>> -				qpd->doorbell_bitmap);
>> +			__set_bit(i, qpd->doorbell_bitmap);
>> +			__set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
>> +				  qpd->doorbell_bitmap);
>>   		}
>>   	}
>>   
>