drm/amdkfd: Hardware DWORD size is 4 bytes

Don't use sizeof(uint32_t) or similar types for hardware or firmware
DWORD size. The hardware and firmware don't care about Linux types.
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c

@@ -95,7 +95,7 @@ static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev,
 	ib_packet->bitfields3.ib_base_hi = largep->u.high_part;
 
 	ib_packet->control = (1 << 23) | (1 << 31) |
-			((size_in_bytes / sizeof(uint32_t)) & 0xfffff);
+			((size_in_bytes / 4) & 0xfffff);
 
 	ib_packet->bitfields5.pasid = pasid;
 
@@ -126,8 +126,7 @@ static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev,
 
 	rm_packet->header.opcode = IT_RELEASE_MEM;
 	rm_packet->header.type = PM4_TYPE_3;
-	rm_packet->header.count = sizeof(struct pm4__release_mem) /
-					sizeof(unsigned int) - 2;
+	rm_packet->header.count = sizeof(struct pm4__release_mem) / 4 - 2;
 
 	rm_packet->bitfields2.event_type = CACHE_FLUSH_AND_INV_TS_EVENT;
 	rm_packet->bitfields2.event_index =
@@ -652,8 +651,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev,
 	packets_vec[0].header.opcode = IT_SET_UCONFIG_REG;
 	packets_vec[0].header.type = PM4_TYPE_3;
 	packets_vec[0].bitfields2.reg_offset =
-			GRBM_GFX_INDEX / (sizeof(uint32_t)) -
-				USERCONFIG_REG_BASE;
+			GRBM_GFX_INDEX / 4 - USERCONFIG_REG_BASE;
 
 	packets_vec[0].bitfields2.insert_vmid = 0;
 	packets_vec[0].reg_data[0] = reg_gfx_index.u32All;
@@ -661,8 +659,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev,
 	packets_vec[1].header.count = 1;
 	packets_vec[1].header.opcode = IT_SET_CONFIG_REG;
 	packets_vec[1].header.type = PM4_TYPE_3;
-	packets_vec[1].bitfields2.reg_offset = SQ_CMD / (sizeof(uint32_t)) -
-						AMD_CONFIG_REG_BASE;
+	packets_vec[1].bitfields2.reg_offset = SQ_CMD / 4 - AMD_CONFIG_REG_BASE;
 
 	packets_vec[1].bitfields2.vmid_shift = SQ_CMD_VMID_OFFSET;
 	packets_vec[1].bitfields2.insert_vmid = 1;
@@ -678,8 +675,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev,
 
 	packets_vec[2].ordinal1 = packets_vec[0].ordinal1;
 	packets_vec[2].bitfields2.reg_offset =
-				GRBM_GFX_INDEX / (sizeof(uint32_t)) -
-					USERCONFIG_REG_BASE;
+				GRBM_GFX_INDEX / 4 - USERCONFIG_REG_BASE;
 
 	packets_vec[2].bitfields2.insert_vmid = 0;
 	packets_vec[2].reg_data[0] = reg_gfx_index.u32All;

drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c

@@ -218,7 +218,7 @@ static int acquire_packet_buffer(struct kernel_queue *kq,
 	rptr = *kq->rptr_kernel;
 	wptr = *kq->wptr_kernel;
 	queue_address = (unsigned int *)kq->pq_kernel_addr;
-	queue_size_dwords = kq->queue->properties.queue_size / sizeof(uint32_t);
+	queue_size_dwords = kq->queue->properties.queue_size / 4;
 
 	pr_debug("rptr: %d\n", rptr);
 	pr_debug("wptr: %d\n", wptr);

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c

@@ -154,7 +154,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id,
 {
 	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
-	uint32_t wptr_mask = (uint32_t)((p->queue_size / sizeof(uint32_t)) - 1);
+	uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1);
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
@@ -183,8 +183,7 @@ static int update_mqd(struct mqd_manager *mm, void *mqd,
 	 * Calculating queue size which is log base 2 of actual queue size -1
 	 * dwords and another -1 for ffs
 	 */
-	m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int))
-								- 1 - 1;
+	m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1;
 	m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8);
 	m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8);
 	m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr);
@@ -209,7 +208,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd,
 	struct cik_sdma_rlc_registers *m;
 
 	m = get_sdma_mqd(mqd);
-	m->sdma_rlc_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1)
+	m->sdma_rlc_rb_cntl = (ffs(q->queue_size / 4) - 1)
 			<< SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT |
 			q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT |
 			1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT |
@@ -350,8 +349,7 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd,
 	 * Calculating queue size which is log base 2 of actual queue
 	 * size -1 dwords
 	 */
-	m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int))
-								- 1 - 1;
+	m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1;
 	m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8);
 	m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8);
 	m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr);

drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c

@@ -103,7 +103,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 {
 	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
-	uint32_t wptr_mask = (uint32_t)((p->queue_size / sizeof(uint32_t)) - 1);
+	uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1);
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
@@ -121,8 +121,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd,
 	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT |
 			atc_bit << CP_HQD_PQ_CONTROL__PQ_ATC__SHIFT |
 			mtype << CP_HQD_PQ_CONTROL__MTYPE__SHIFT;
-	m->cp_hqd_pq_control |=
-			ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1;
+	m->cp_hqd_pq_control |=	ffs(q->queue_size / 4) - 1 - 1;
 	pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control);
 
 	m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8);
@@ -152,7 +151,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd,
 	 * is safe, giving a maximum field value of 0xA.
 	 */
 	m->cp_hqd_eop_control |= min(0xA,
-		ffs(q->eop_ring_buffer_size / sizeof(unsigned int)) - 1 - 1);
+		ffs(q->eop_ring_buffer_size / 4) - 1 - 1);
 	m->cp_hqd_eop_base_addr_lo =
 			lower_32_bits(q->eop_ring_buffer_address >> 8);
 	m->cp_hqd_eop_base_addr_hi =
@@ -288,7 +287,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd,
 	struct vi_sdma_mqd *m;
 
 	m = get_sdma_mqd(mqd);
-	m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1)
+	m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / 4) - 1)
 		<< SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT |
 		q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT |
 		1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT |

drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c

@@ -45,7 +45,7 @@ static unsigned int build_pm4_header(unsigned int opcode, size_t packet_size)
 
 	header.u32All = 0;
 	header.opcode = opcode;
-	header.count = packet_size/sizeof(uint32_t) - 2;
+	header.count = packet_size / 4 - 2;
 	header.type = PM4_TYPE_3;
 
 	return header.u32All;