drm/amdkfd: drop IOMMUv2 support

Now that we use the dGPU path for all APUs, drop the
IOMMUv2 support.

v2: drop the now unused queue manager functions for gfx7/8 APUs

Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: alexdeucher

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c

@@ -226,16 +226,6 @@ void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm)
 		kgd2kfd_suspend(adev->kfd.dev, run_pm);
 }
 
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev)
-{
-	int r = 0;
-
-	if (adev->kfd.dev)
-		r = kgd2kfd_resume_iommu(adev->kfd.dev);
-
-	return r;
-}
-
 int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm)
 {
 	int r = 0;

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

@@ -148,7 +148,6 @@ int amdgpu_amdkfd_init(void);
 void amdgpu_amdkfd_fini(void);
 
 void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev);
 int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
 void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 			const void *ih_ring_entry);
@@ -399,7 +398,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 const struct kgd2kfd_shared_resources *gpu_resources);
 void kgd2kfd_device_exit(struct kfd_dev *kfd);
 void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm);
-int kgd2kfd_resume_iommu(struct kfd_dev *kfd);
 int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
 int kgd2kfd_pre_reset(struct kfd_dev *kfd);
 int kgd2kfd_post_reset(struct kfd_dev *kfd);
@@ -439,11 +437,6 @@ static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
 {
 }
 
-static int __maybe_unused kgd2kfd_resume_iommu(struct kfd_dev *kfd)
-{
-	return 0;
-}
-
 static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
 {
 	return 0;

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -3236,12 +3236,6 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (!adev->in_s0ix) {
-		r = amdgpu_amdkfd_resume_iommu(adev);
-		if (r)
-			return r;
-	}
-
 	r = amdgpu_device_ip_resume_phase1(adev);
 	if (r)
 		return r;
@@ -4946,9 +4940,6 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				dev_warn(tmp_adev->dev, "asic atom init failed!");
 			} else {
 				dev_info(tmp_adev->dev, "GPU reset succeeded, trying to resume\n");
-				r = amdgpu_amdkfd_resume_iommu(tmp_adev);
-				if (r)
-					goto out;
 
 				r = amdgpu_device_ip_resume_phase1(tmp_adev);
 				if (r)

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -755,20 +755,6 @@ module_param(debug_largebar, int, 0444);
 MODULE_PARM_DESC(debug_largebar,
 	"Debug large-bar flag used to simulate large-bar capability on non-large bar machine (0 = disable, 1 = enable)");
 
-/**
- * DOC: ignore_crat (int)
- * Ignore CRAT table during KFD initialization. By default, KFD uses the ACPI CRAT
- * table to get information about AMD APUs. This option can serve as a workaround on
- * systems with a broken CRAT table.
- *
- * Default is auto (according to asic type, iommu_v2, and crat table, to decide
- * whether use CRAT)
- */
-int ignore_crat;
-module_param(ignore_crat, int, 0444);
-MODULE_PARM_DESC(ignore_crat,
-	"Ignore CRAT table during KFD initialization (0 = auto (default), 1 = ignore CRAT)");
-
 /**
  * DOC: halt_if_hws_hang (int)
  * Halt if HWS hang is detected. Default value, 0, disables the halt on hang.

drivers/gpu/drm/amd/amdkfd/Kconfig

@@ -6,7 +6,6 @@
 config HSA_AMD
 	bool "HSA kernel driver for AMD GPU devices"
 	depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64)
-	imply AMD_IOMMU_V2 if X86_64
 	select HMM_MIRROR
 	select MMU_NOTIFIER
 	select DRM_AMDGPU_USERPTR

drivers/gpu/drm/amd/amdkfd/Makefile

@@ -59,10 +59,6 @@ AMDKFD_FILES	:= $(AMDKFD_PATH)/kfd_module.o \
 		$(AMDKFD_PATH)/kfd_crat.o \
 		$(AMDKFD_PATH)/kfd_debug.o
 
-ifneq ($(CONFIG_AMD_IOMMU_V2),)
-AMDKFD_FILES += $(AMDKFD_PATH)/kfd_iommu.o
-endif
-
 ifneq ($(CONFIG_DEBUG_FS),)
 AMDKFD_FILES += $(AMDKFD_PATH)/kfd_debugfs.o
 endif

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

@@ -1026,9 +1026,6 @@ bool kfd_dev_is_large_bar(struct kfd_node *dev)
 		return true;
 	}
 
-	if (dev->kfd->use_iommu_v2)
-		return false;
-
 	if (dev->local_mem_info.local_mem_size_private == 0 &&
 	    dev->local_mem_info.local_mem_size_public > 0)
 		return true;

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

@@ -26,7 +26,6 @@
 #include "kfd_crat.h"
 #include "kfd_priv.h"
 #include "kfd_topology.h"
-#include "kfd_iommu.h"
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
 
@@ -1536,72 +1535,6 @@ int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pc
 	return num_of_cache_types;
 }
 
-static bool kfd_ignore_crat(void)
-{
-	bool ret;
-
-	if (ignore_crat)
-		return true;
-
-	ret = true;
-
-	return ret;
-}
-
-/*
- * kfd_create_crat_image_acpi - Allocates memory for CRAT image and
- * copies CRAT from ACPI (if available).
- * NOTE: Call kfd_destroy_crat_image to free CRAT image memory
- *
- *	@crat_image: CRAT read from ACPI. If no CRAT in ACPI then
- *		     crat_image will be NULL
- *	@size: [OUT] size of crat_image
- *
- *	Return 0 if successful else return error code
- */
-int kfd_create_crat_image_acpi(void **crat_image, size_t *size)
-{
-	struct acpi_table_header *crat_table;
-	acpi_status status;
-	void *pcrat_image;
-	int rc = 0;
-
-	if (!crat_image)
-		return -EINVAL;
-
-	*crat_image = NULL;
-
-	if (kfd_ignore_crat()) {
-		pr_info("CRAT table disabled by module option\n");
-		return -ENODATA;
-	}
-
-	/* Fetch the CRAT table from ACPI */
-	status = acpi_get_table(CRAT_SIGNATURE, 0, &crat_table);
-	if (status == AE_NOT_FOUND) {
-		pr_info("CRAT table not found\n");
-		return -ENODATA;
-	} else if (ACPI_FAILURE(status)) {
-		const char *err = acpi_format_exception(status);
-
-		pr_err("CRAT table error: %s\n", err);
-		return -EINVAL;
-	}
-
-	pcrat_image = kvmalloc(crat_table->length, GFP_KERNEL);
-	if (!pcrat_image) {
-		rc = -ENOMEM;
-		goto out;
-	}
-
-	memcpy(pcrat_image, crat_table, crat_table->length);
-	*crat_image = pcrat_image;
-	*size = crat_table->length;
-out:
-	acpi_put_table(crat_table);
-	return rc;
-}
-
 /* Memory required to create Virtual CRAT.
  * Since there is no easy way to predict the amount of memory required, the
  * following amount is allocated for GPU Virtual CRAT. This is
@@ -2169,12 +2102,6 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 
 	cu->hsa_capability = 0;
 
-	/* Check if this node supports IOMMU. During parsing this flag will
-	 * translate to HSA_CAP_ATS_PRESENT
-	 */
-	if (!kfd_iommu_check_device(kdev->kfd))
-		cu->hsa_capability |= CRAT_CU_FLAGS_IOMMU_PRESENT;
-
 	crat_table->length += sub_type_hdr->length;
 	crat_table->total_entries++;
 

drivers/gpu/drm/amd/amdkfd/kfd_crat.h

@@ -307,7 +307,6 @@ struct kfd_gpu_cache_info {
 };
 int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pcache_info);
 
-int kfd_create_crat_image_acpi(void **crat_image, size_t *size);
 void kfd_destroy_crat_image(void *crat_image);
 int kfd_parse_crat_table(void *crat_image, struct list_head *device_list,
 			 uint32_t proximity_domain);

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

@@ -29,7 +29,6 @@
 #include "kfd_pm4_headers_vi.h"
 #include "kfd_pm4_headers_aldebaran.h"
 #include "cwsr_trap_handler.h"
-#include "kfd_iommu.h"
 #include "amdgpu_amdkfd.h"
 #include "kfd_smi_events.h"
 #include "kfd_svm.h"
@@ -62,7 +61,6 @@ static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size,
 				unsigned int chunk_size);
 static void kfd_gtt_sa_fini(struct kfd_dev *kfd);
 
-static int kfd_resume_iommu(struct kfd_dev *kfd);
 static int kfd_resume(struct kfd_node *kfd);
 
 static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
@@ -755,15 +753,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 
 	kfd->noretry = kfd->adev->gmc.noretry;
 
-	/* If CRAT is broken, won't set iommu enabled */
-	kfd_double_confirm_iommu_support(kfd);
-
-	if (kfd_iommu_device_init(kfd)) {
-		kfd->use_iommu_v2 = false;
-		dev_err(kfd_device, "Error initializing iommuv2\n");
-		goto device_iommu_error;
-	}
-
 	kfd_cwsr_init(kfd);
 
 	dev_info(kfd_device, "Total number of KFD nodes to be created: %d\n",
@@ -838,9 +827,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 
 	svm_range_set_max_pages(kfd->adev);
 
-	if (kfd_resume_iommu(kfd))
-		goto kfd_resume_iommu_error;
-
 	spin_lock_init(&kfd->watch_points_lock);
 
 	kfd->init_complete = true;
@@ -852,11 +838,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 
 	goto out;
 
-kfd_resume_iommu_error:
 node_init_error:
 node_alloc_error:
 	kfd_cleanup_nodes(kfd, i);
-device_iommu_error:
 	kfd_doorbell_fini(kfd);
 kfd_doorbell_error:
 	kfd_gtt_sa_fini(kfd);
@@ -971,7 +955,6 @@ void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
 		node = kfd->nodes[i];
 		node->dqm->ops.stop(node->dqm);
 	}
-	kfd_iommu_suspend(kfd);
 }
 
 int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
@@ -1001,26 +984,6 @@ int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
 	return ret;
 }
 
-int kgd2kfd_resume_iommu(struct kfd_dev *kfd)
-{
-	if (!kfd->init_complete)
-		return 0;
-
-	return kfd_resume_iommu(kfd);
-}
-
-static int kfd_resume_iommu(struct kfd_dev *kfd)
-{
-	int err = 0;
-
-	err = kfd_iommu_resume(kfd);
-	if (err)
-		dev_err(kfd_device,
-			"Failed to resume IOMMU for device %x:%x\n",
-			kfd->adev->pdev->vendor, kfd->adev->pdev->device);
-	return err;
-}
-
 static int kfd_resume(struct kfd_node *node)
 {
 	int err = 0;