drm/radeon: deprecate and remove KFD interface

Submitted by Christian König on Oct. 30, 2017, 1:16 p.m.

Details

Message ID 1509369381-16062-1-git-send-email-deathsimple@vodafone.de
State New
Headers show
Series "drm/radeon: deprecate and remove KFD interface" ( rev: 1 ) in AMD X.Org drivers

Not browsing as part of any series.

Commit Message

Christian König Oct. 30, 2017, 1:16 p.m.
From: Christian König <christian.koenig@amd.com>

To quote Felix: "For testing KV with current user mode stack, please use
amdgpu. I don't expect this to work with radeon and I'm not planning to spend
any effort on making radeon work with a current user mode stack."

Only compile tested, but should be straight forward.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/Kconfig  |   2 +-
 drivers/gpu/drm/radeon/Makefile     |   3 +-
 drivers/gpu/drm/radeon/cik.c        |  14 +-
 drivers/gpu/drm/radeon/cikd.h       |   2 -
 drivers/gpu/drm/radeon/radeon.h     |   3 -
 drivers/gpu/drm/radeon/radeon_drv.c |  10 -
 drivers/gpu/drm/radeon/radeon_kfd.c | 901 ------------------------------------
 drivers/gpu/drm/radeon/radeon_kfd.h |  47 --
 drivers/gpu/drm/radeon/radeon_kms.c |   7 -
 9 files changed, 4 insertions(+), 985 deletions(-)
 delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.c
 delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.h

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig
index e13c67c..bc5a294 100644
--- a/drivers/gpu/drm/amd/amdkfd/Kconfig
+++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
@@ -4,6 +4,6 @@ 
 
 config HSA_AMD
 	tristate "HSA kernel driver for AMD GPU devices"
-	depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64
+	depends on DRM_AMDGPU && AMD_IOMMU_V2 && X86_64
 	help
 	  Enable this if you want to use HSA features on AMD GPU devices.
diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile
index be16c63..cf3e598 100644
--- a/drivers/gpu/drm/radeon/Makefile
+++ b/drivers/gpu/drm/radeon/Makefile
@@ -102,8 +102,7 @@  radeon-y += \
 radeon-y += \
 	radeon_vce.o \
 	vce_v1_0.o \
-	vce_v2_0.o \
-	radeon_kfd.o
+	vce_v2_0.o
 
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
 radeon-$(CONFIG_ACPI) += radeon_acpi.o
diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c
index 3cb6c55..898f9a0 100644
--- a/drivers/gpu/drm/radeon/cik.c
+++ b/drivers/gpu/drm/radeon/cik.c
@@ -33,7 +33,6 @@ 
 #include "cik_blit_shaders.h"
 #include "radeon_ucode.h"
 #include "clearstate_ci.h"
-#include "radeon_kfd.h"
 
 #define SH_MEM_CONFIG_GFX_DEFAULT \
 	ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
@@ -5684,10 +5683,9 @@  int cik_vm_init(struct radeon_device *rdev)
 	/*
 	 * number of VMs
 	 * VMID 0 is reserved for System
-	 * radeon graphics/compute will use VMIDs 1-7
-	 * amdkfd will use VMIDs 8-15
+	 * radeon graphics/compute will use VMIDs 1-15
 	 */
-	rdev->vm_manager.nvm = RADEON_NUM_OF_VMIDS;
+	rdev->vm_manager.nvm = 16;
 	/* base offset of vram pages */
 	if (rdev->flags & RADEON_IS_IGP) {
 		u64 tmp = RREG32(MC_VM_FB_OFFSET);
@@ -7589,9 +7587,6 @@  int cik_irq_process(struct radeon_device *rdev)
 		/* wptr/rptr are in bytes! */
 		ring_index = rptr / 4;
 
-		radeon_kfd_interrupt(rdev,
-				(const void *) &rdev->ih.ring[ring_index]);
-
 		src_id =  le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
 		src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
 		ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff;
@@ -8486,10 +8481,6 @@  static int cik_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
-	r = radeon_kfd_resume(rdev);
-	if (r)
-		return r;
-
 	return 0;
 }
 
@@ -8538,7 +8529,6 @@  int cik_resume(struct radeon_device *rdev)
  */
 int cik_suspend(struct radeon_device *rdev)
 {
-	radeon_kfd_suspend(rdev);
 	radeon_pm_suspend(rdev);
 	radeon_audio_fini(rdev);
 	radeon_vm_manager_fini(rdev);
diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h
index e210154..cda16fc 100644
--- a/drivers/gpu/drm/radeon/cikd.h
+++ b/drivers/gpu/drm/radeon/cikd.h
@@ -30,8 +30,6 @@ 
 #define CIK_RB_BITMAP_WIDTH_PER_SH     2
 #define HAWAII_RB_BITMAP_WIDTH_PER_SH  4
 
-#define RADEON_NUM_OF_VMIDS	8
-
 /* DIDT IND registers */
 #define DIDT_SQ_CTRL0                                     0x0
 #       define DIDT_CTRL_EN                               (1 << 0)
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index ec63bc5..d94741b 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -2456,9 +2456,6 @@  struct radeon_device {
 	u64 vram_pin_size;
 	u64 gart_pin_size;
 
-	/* amdkfd interface */
-	struct kfd_dev		*kfd;
-
 	struct mutex	mn_lock;
 	DECLARE_HASHTABLE(mn_hash, 7);
 };
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index f4becad..31dd04f 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -43,7 +43,6 @@ 
 #include <drm/drm_fb_helper.h>
 
 #include <drm/drm_crtc_helper.h>
-#include "radeon_kfd.h"
 
 /*
  * KMS wrapper.
@@ -338,14 +337,6 @@  static int radeon_pci_probe(struct pci_dev *pdev,
 {
 	int ret;
 
-	/*
-	 * Initialize amdkfd before starting radeon. If it was not loaded yet,
-	 * defer radeon probing
-	 */
-	ret = radeon_kfd_init();
-	if (ret == -EPROBE_DEFER)
-		return ret;
-
 	if (vga_switcheroo_client_probe_defer(pdev))
 		return -EPROBE_DEFER;
 
@@ -645,7 +636,6 @@  static int __init radeon_init(void)
 
 static void __exit radeon_exit(void)
 {
-	radeon_kfd_fini();
 	pci_unregister_driver(pdriver);
 	radeon_unregister_atpx_handler();
 }
diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c b/drivers/gpu/drm/radeon/radeon_kfd.c
deleted file mode 100644
index 385b4d7..0000000
--- a/drivers/gpu/drm/radeon/radeon_kfd.c
+++ /dev/null
@@ -1,901 +0,0 @@ 
-/*
- * Copyright 2014 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#include <linux/module.h>
-#include <linux/fdtable.h>
-#include <linux/uaccess.h>
-#include <drm/drmP.h>
-#include "radeon.h"
-#include "cikd.h"
-#include "cik_reg.h"
-#include "radeon_kfd.h"
-#include "radeon_ucode.h"
-#include <linux/firmware.h>
-#include "cik_structs.h"
-
-#define CIK_PIPE_PER_MEC	(4)
-
-static const uint32_t watchRegs[MAX_WATCH_ADDRESSES * ADDRESS_WATCH_REG_MAX] = {
-	TCP_WATCH0_ADDR_H, TCP_WATCH0_ADDR_L, TCP_WATCH0_CNTL,
-	TCP_WATCH1_ADDR_H, TCP_WATCH1_ADDR_L, TCP_WATCH1_CNTL,
-	TCP_WATCH2_ADDR_H, TCP_WATCH2_ADDR_L, TCP_WATCH2_CNTL,
-	TCP_WATCH3_ADDR_H, TCP_WATCH3_ADDR_L, TCP_WATCH3_CNTL
-};
-
-struct kgd_mem {
-	struct radeon_bo *bo;
-	uint64_t gpu_addr;
-	void *cpu_ptr;
-};
-
-
-static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
-			void **mem_obj, uint64_t *gpu_addr,
-			void **cpu_ptr);
-
-static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
-
-static uint64_t get_vmem_size(struct kgd_dev *kgd);
-static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd);
-
-static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
-
-static int alloc_pasid(unsigned int bits);
-static void free_pasid(unsigned int pasid);
-
-static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
-
-/*
- * Register access functions
- */
-
-static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t sh_mem_config,	uint32_t sh_mem_ape1_base,
-		uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases);
-
-static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
-					unsigned int vmid);
-
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t hpd_size, uint64_t hpd_gpu_addr);
-static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
-static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm);
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
-static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id);
-
-static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
-				uint32_t queue_id);
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout);
-static int kgd_address_watch_disable(struct kgd_dev *kgd);
-static int kgd_address_watch_execute(struct kgd_dev *kgd,
-					unsigned int watch_point_id,
-					uint32_t cntl_val,
-					uint32_t addr_hi,
-					uint32_t addr_lo);
-static int kgd_wave_control_execute(struct kgd_dev *kgd,
-					uint32_t gfx_index_val,
-					uint32_t sq_cmd);
-static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
-					unsigned int watch_point_id,
-					unsigned int reg_offset);
-
-static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid);
-static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
-							uint8_t vmid);
-static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid);
-
-static const struct kfd2kgd_calls kfd2kgd = {
-	.init_gtt_mem_allocation = alloc_gtt_mem,
-	.free_gtt_mem = free_gtt_mem,
-	.get_vmem_size = get_vmem_size,
-	.get_gpu_clock_counter = get_gpu_clock_counter,
-	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
-	.alloc_pasid = alloc_pasid,
-	.free_pasid = free_pasid,
-	.program_sh_mem_settings = kgd_program_sh_mem_settings,
-	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
-	.init_pipeline = kgd_init_pipeline,
-	.init_interrupts = kgd_init_interrupts,
-	.hqd_load = kgd_hqd_load,
-	.hqd_sdma_load = kgd_hqd_sdma_load,
-	.hqd_is_occupied = kgd_hqd_is_occupied,
-	.hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
-	.hqd_destroy = kgd_hqd_destroy,
-	.hqd_sdma_destroy = kgd_hqd_sdma_destroy,
-	.address_watch_disable = kgd_address_watch_disable,
-	.address_watch_execute = kgd_address_watch_execute,
-	.wave_control_execute = kgd_wave_control_execute,
-	.address_watch_get_offset = kgd_address_watch_get_offset,
-	.get_atc_vmid_pasid_mapping_pasid = get_atc_vmid_pasid_mapping_pasid,
-	.get_atc_vmid_pasid_mapping_valid = get_atc_vmid_pasid_mapping_valid,
-	.write_vmid_invalidate_request = write_vmid_invalidate_request,
-	.get_fw_version = get_fw_version
-};
-
-static const struct kgd2kfd_calls *kgd2kfd;
-
-int radeon_kfd_init(void)
-{
-	int ret;
-
-#if defined(CONFIG_HSA_AMD_MODULE)
-	int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
-
-	kgd2kfd_init_p = symbol_request(kgd2kfd_init);
-
-	if (kgd2kfd_init_p == NULL)
-		return -ENOENT;
-
-	ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd);
-	if (ret) {
-		symbol_put(kgd2kfd_init);
-		kgd2kfd = NULL;
-	}
-
-#elif defined(CONFIG_HSA_AMD)
-	ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
-	if (ret)
-		kgd2kfd = NULL;
-
-#else
-	ret = -ENOENT;
-#endif
-
-	return ret;
-}
-
-void radeon_kfd_fini(void)
-{
-	if (kgd2kfd) {
-		kgd2kfd->exit();
-		symbol_put(kgd2kfd_init);
-	}
-}
-
-void radeon_kfd_device_probe(struct radeon_device *rdev)
-{
-	if (kgd2kfd)
-		rdev->kfd = kgd2kfd->probe((struct kgd_dev *)rdev,
-			rdev->pdev, &kfd2kgd);
-}
-
-void radeon_kfd_device_init(struct radeon_device *rdev)
-{
-	int i, queue, pipe, mec;
-
-	if (rdev->kfd) {
-		struct kgd2kfd_shared_resources gpu_resources = {
-			.compute_vmid_bitmap = 0xFF00,
-			.num_pipe_per_mec = 4,
-			.num_queue_per_pipe = 8
-		};
-
-		bitmap_zero(gpu_resources.queue_bitmap, KGD_MAX_QUEUES);
-
-		for (i = 0; i < KGD_MAX_QUEUES; ++i) {
-			queue = i % gpu_resources.num_queue_per_pipe;
-			pipe = (i / gpu_resources.num_queue_per_pipe)
-				% gpu_resources.num_pipe_per_mec;
-			mec = (i / gpu_resources.num_queue_per_pipe)
-				/ gpu_resources.num_pipe_per_mec;
-
-			if (mec == 0 && pipe > 0)
-				set_bit(i, gpu_resources.queue_bitmap);
-		}
-
-		radeon_doorbell_get_kfd_info(rdev,
-				&gpu_resources.doorbell_physical_address,
-				&gpu_resources.doorbell_aperture_size,
-				&gpu_resources.doorbell_start_offset);
-
-		kgd2kfd->device_init(rdev->kfd, &gpu_resources);
-	}
-}
-
-void radeon_kfd_device_fini(struct radeon_device *rdev)
-{
-	if (rdev->kfd) {
-		kgd2kfd->device_exit(rdev->kfd);
-		rdev->kfd = NULL;
-	}
-}
-
-void radeon_kfd_interrupt(struct radeon_device *rdev, const void *ih_ring_entry)
-{
-	if (rdev->kfd)
-		kgd2kfd->interrupt(rdev->kfd, ih_ring_entry);
-}
-
-void radeon_kfd_suspend(struct radeon_device *rdev)
-{
-	if (rdev->kfd)
-		kgd2kfd->suspend(rdev->kfd);
-}
-
-int radeon_kfd_resume(struct radeon_device *rdev)
-{
-	int r = 0;
-
-	if (rdev->kfd)
-		r = kgd2kfd->resume(rdev->kfd);
-
-	return r;
-}
-
-static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
-			void **mem_obj, uint64_t *gpu_addr,
-			void **cpu_ptr)
-{
-	struct radeon_device *rdev = (struct radeon_device *)kgd;
-	struct kgd_mem **mem = (struct kgd_mem **) mem_obj;
-	int r;
-
-	BUG_ON(kgd == NULL);
-	BUG_ON(gpu_addr == NULL);
-	BUG_ON(cpu_ptr == NULL);
-
-	*mem = kmalloc(sizeof(struct kgd_mem), GFP_KERNEL);
-	if ((*mem) == NULL)
-		return -ENOMEM;
-
-	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_GTT,
-				RADEON_GEM_GTT_WC, NULL, NULL, &(*mem)->bo);
-	if (r) {
-		dev_err(rdev->dev,
-			"failed to allocate BO for amdkfd (%d)\n", r);
-		return r;
-	}
-
-	/* map the buffer */
-	r = radeon_bo_reserve((*mem)->bo, true);
-	if (r) {
-		dev_err(rdev->dev, "(%d) failed to reserve bo for amdkfd\n", r);
-		goto allocate_mem_reserve_bo_failed;
-	}
-
-	r = radeon_bo_pin((*mem)->bo, RADEON_GEM_DOMAIN_GTT,
-				&(*mem)->gpu_addr);
-	if (r) {
-		dev_err(rdev->dev, "(%d) failed to pin bo for amdkfd\n", r);
-		goto allocate_mem_pin_bo_failed;
-	}
-	*gpu_addr = (*mem)->gpu_addr;
-
-	r = radeon_bo_kmap((*mem)->bo, &(*mem)->cpu_ptr);
-	if (r) {
-		dev_err(rdev->dev,
-			"(%d) failed to map bo to kernel for amdkfd\n", r);
-		goto allocate_mem_kmap_bo_failed;
-	}
-	*cpu_ptr = (*mem)->cpu_ptr;
-
-	radeon_bo_unreserve((*mem)->bo);
-
-	return 0;
-
-allocate_mem_kmap_bo_failed:
-	radeon_bo_unpin((*mem)->bo);
-allocate_mem_pin_bo_failed:
-	radeon_bo_unreserve((*mem)->bo);
-allocate_mem_reserve_bo_failed:
-	radeon_bo_unref(&(*mem)->bo);
-
-	return r;
-}
-
-static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
-{
-	struct kgd_mem *mem = (struct kgd_mem *) mem_obj;
-
-	BUG_ON(mem == NULL);
-
-	radeon_bo_reserve(mem->bo, true);
-	radeon_bo_kunmap(mem->bo);
-	radeon_bo_unpin(mem->bo);
-	radeon_bo_unreserve(mem->bo);
-	radeon_bo_unref(&(mem->bo));
-	kfree(mem);
-}
-
-static uint64_t get_vmem_size(struct kgd_dev *kgd)
-{
-	struct radeon_device *rdev = (struct radeon_device *)kgd;
-
-	BUG_ON(kgd == NULL);
-
-	return rdev->mc.real_vram_size;
-}
-
-static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
-{
-	struct radeon_device *rdev = (struct radeon_device *)kgd;
-
-	return rdev->asic->get_gpu_clock_counter(rdev);
-}
-
-static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
-{
-	struct radeon_device *rdev = (struct radeon_device *)kgd;
-
-	/* The sclk is in quantas of 10kHz */
-	return rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk / 100;
-}
-
-/*
- * PASID manager
- */
-static DEFINE_IDA(pasid_ida);
-
-static int alloc_pasid(unsigned int bits)
-{
-	int pasid = -EINVAL;
-
-	for (bits = min(bits, 31U); bits > 0; bits--) {
-		pasid = ida_simple_get(&pasid_ida,
-				       1U << (bits - 1), 1U << bits,
-				       GFP_KERNEL);
-		if (pasid != -ENOSPC)
-			break;
-	}
-
-	return pasid;
-}
-
-static void free_pasid(unsigned int pasid)
-{
-	ida_simple_remove(&pasid_ida, pasid);
-}
-
-static inline struct radeon_device *get_radeon_device(struct kgd_dev *kgd)
-{
-	return (struct radeon_device *)kgd;
-}
-
-static void write_register(struct kgd_dev *kgd, uint32_t offset, uint32_t value)
-{
-	struct radeon_device *rdev = get_radeon_device(kgd);
-
-	writel(value, (void __iomem *)(rdev->rmmio + offset));
-}
-
-static uint32_t read_register(struct kgd_dev *kgd, uint32_t offset)
-{
-	struct radeon_device *rdev = get_radeon_device(kgd);
-
-	return readl((void __iomem *)(rdev->rmmio + offset));
-}
-
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
-			uint32_t queue, uint32_t vmid)
-{
-	struct radeon_device *rdev = get_radeon_device(kgd);
-	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
-
-	mutex_lock(&rdev->srbm_mutex);
-	write_register(kgd, SRBM_GFX_CNTL, value);
-}
-
-static void unlock_srbm(struct kgd_dev *kgd)
-{
-	struct radeon_device *rdev = get_radeon_device(kgd);
-
-	write_register(kgd, SRBM_GFX_CNTL, 0);
-	mutex_unlock(&rdev->srbm_mutex);
-}
-
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t queue_id)
-{
-	uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1;
-	uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC);
-
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
-}
-
-static void release_queue(struct kgd_dev *kgd)
-{
-	unlock_srbm(kgd);
-}
-
-static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
-					uint32_t sh_mem_config,
-					uint32_t sh_mem_ape1_base,
-					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
-{
-	lock_srbm(kgd, 0, 0, 0, vmid);
-
-	write_register(kgd, SH_MEM_CONFIG, sh_mem_config);
-	write_register(kgd, SH_MEM_APE1_BASE, sh_mem_ape1_base);
-	write_register(kgd, SH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
-	write_register(kgd, SH_MEM_BASES, sh_mem_bases);
-
-	unlock_srbm(kgd);
-}
-
-static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
-					unsigned int vmid)
-{
-	/*
-	 * We have to assume that there is no outstanding mapping.
-	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0
-	 * because a mapping is in progress or because a mapping finished and
-	 * the SW cleared it.
-	 * So the protocol is to always wait & clear.
-	 */
-	uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
-					ATC_VMID_PASID_MAPPING_VALID_MASK;
-
-	write_register(kgd, ATC_VMID0_PASID_MAPPING + vmid*sizeof(uint32_t),
-			pasid_mapping);
-
-	while (!(read_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS) &
-								(1U << vmid)))
-		cpu_relax();
-	write_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U << vmid);
-
-	/* Mapping vmid to pasid also for IH block */
-	write_register(kgd, IH_VMID_0_LUT + vmid * sizeof(uint32_t),
-			pasid_mapping);
-
-	return 0;
-}
-
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t hpd_size, uint64_t hpd_gpu_addr)
-{
-	/* nothing to do here */
-	return 0;
-}
-
-static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
-{
-	uint32_t mec;
-	uint32_t pipe;
-
-	mec = (pipe_id / CIK_PIPE_PER_MEC) + 1;
-	pipe = (pipe_id % CIK_PIPE_PER_MEC);
-
-	lock_srbm(kgd, mec, pipe, 0, 0);
-
-	write_register(kgd, CPC_INT_CNTL,
-			TIME_STAMP_INT_ENABLE | OPCODE_ERROR_INT_ENABLE);
-
-	unlock_srbm(kgd);
-
-	return 0;
-}
-
-static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers *m)
-{
-	uint32_t retval;
-
-	retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET +
-			m->sdma_queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
-
-	pr_debug("kfd: sdma base address: 0x%x\n", retval);
-
-	return retval;
-}
-
-static inline struct cik_mqd *get_mqd(void *mqd)
-{
-	return (struct cik_mqd *)mqd;
-}
-
-static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
-{
-	return (struct cik_sdma_rlc_registers *)mqd;
-}
-
-static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
-{
-	uint32_t wptr_shadow, is_wptr_shadow_valid;
-	struct cik_mqd *m;
-
-	m = get_mqd(mqd);
-
-	is_wptr_shadow_valid = !get_user(wptr_shadow, wptr);
-
-	acquire_queue(kgd, pipe_id, queue_id);
-	write_register(kgd, CP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo);
-	write_register(kgd, CP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi);
-	write_register(kgd, CP_MQD_CONTROL, m->cp_mqd_control);
-
-	write_register(kgd, CP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo);
-	write_register(kgd, CP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi);
-	write_register(kgd, CP_HQD_PQ_CONTROL, m->cp_hqd_pq_control);
-
-	write_register(kgd, CP_HQD_IB_CONTROL, m->cp_hqd_ib_control);
-	write_register(kgd, CP_HQD_IB_BASE_ADDR, m->cp_hqd_ib_base_addr_lo);
-	write_register(kgd, CP_HQD_IB_BASE_ADDR_HI, m->cp_hqd_ib_base_addr_hi);
-
-	write_register(kgd, CP_HQD_IB_RPTR, m->cp_hqd_ib_rptr);
-
-	write_register(kgd, CP_HQD_PERSISTENT_STATE,
-			m->cp_hqd_persistent_state);
-	write_register(kgd, CP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd);
-	write_register(kgd, CP_HQD_MSG_TYPE, m->cp_hqd_msg_type);
-
-	write_register(kgd, CP_HQD_ATOMIC0_PREOP_LO,
-			m->cp_hqd_atomic0_preop_lo);
-
-	write_register(kgd, CP_HQD_ATOMIC0_PREOP_HI,
-			m->cp_hqd_atomic0_preop_hi);
-
-	write_register(kgd, CP_HQD_ATOMIC1_PREOP_LO,
-			m->cp_hqd_atomic1_preop_lo);
-
-	write_register(kgd, CP_HQD_ATOMIC1_PREOP_HI,
-			m->cp_hqd_atomic1_preop_hi);
-
-	write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR,
-			m->cp_hqd_pq_rptr_report_addr_lo);
-
-	write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR_HI,
-			m->cp_hqd_pq_rptr_report_addr_hi);
-
-	write_register(kgd, CP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr);
-
-	write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR,
-			m->cp_hqd_pq_wptr_poll_addr_lo);
-
-	write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR_HI,
-			m->cp_hqd_pq_wptr_poll_addr_hi);
-
-	write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL,
-			m->cp_hqd_pq_doorbell_control);
-
-	write_register(kgd, CP_HQD_VMID, m->cp_hqd_vmid);
-
-	write_register(kgd, CP_HQD_QUANTUM, m->cp_hqd_quantum);
-
-	write_register(kgd, CP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority);
-	write_register(kgd, CP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority);
-
-	write_register(kgd, CP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr);
-
-	if (is_wptr_shadow_valid)
-		write_register(kgd, CP_HQD_PQ_WPTR, wptr_shadow);
-
-	write_register(kgd, CP_HQD_ACTIVE, m->cp_hqd_active);
-	release_queue(kgd);
-
-	return 0;
-}
-
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
-{
-	struct cik_sdma_rlc_registers *m;
-	uint32_t sdma_base_addr;
-
-	m = get_sdma_mqd(mqd);
-	sdma_base_addr = get_sdma_base_addr(m);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_VIRTUAL_ADDR,
-			m->sdma_rlc_virtual_addr);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_RB_BASE,
-			m->sdma_rlc_rb_base);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_RB_BASE_HI,
-			m->sdma_rlc_rb_base_hi);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_LO,
-			m->sdma_rlc_rb_rptr_addr_lo);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_HI,
-			m->sdma_rlc_rb_rptr_addr_hi);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_DOORBELL,
-			m->sdma_rlc_doorbell);
-
-	write_register(kgd,
-			sdma_base_addr + SDMA0_RLC0_RB_CNTL,
-			m->sdma_rlc_rb_cntl);
-
-	return 0;
-}
-
-static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
-{
-	uint32_t act;
-	bool retval = false;
-	uint32_t low, high;
-
-	acquire_queue(kgd, pipe_id, queue_id);
-	act = read_register(kgd, CP_HQD_ACTIVE);
-	if (act) {
-		low = lower_32_bits(queue_address >> 8);
-		high = upper_32_bits(queue_address >> 8);
-
-		if (low == read_register(kgd, CP_HQD_PQ_BASE) &&
-				high == read_register(kgd, CP_HQD_PQ_BASE_HI))
-			retval = true;
-	}
-	release_queue(kgd);
-	return retval;
-}
-
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
-{
-	struct cik_sdma_rlc_registers *m;
-	uint32_t sdma_base_addr;
-	uint32_t sdma_rlc_rb_cntl;
-
-	m = get_sdma_mqd(mqd);
-	sdma_base_addr = get_sdma_base_addr(m);
-
-	sdma_rlc_rb_cntl = read_register(kgd,
-					sdma_base_addr + SDMA0_RLC0_RB_CNTL);
-
-	if (sdma_rlc_rb_cntl & SDMA_RB_ENABLE)
-		return true;
-
-	return false;
-}
-
-static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
-				uint32_t queue_id)
-{
-	uint32_t temp;
-
-	acquire_queue(kgd, pipe_id, queue_id);
-	write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, 0);
-
-	write_register(kgd, CP_HQD_DEQUEUE_REQUEST, reset_type);
-
-	while (true) {
-		temp = read_register(kgd, CP_HQD_ACTIVE);
-		if (temp & 0x1)
-			break;
-		if (timeout == 0) {
-			pr_err("kfd: cp queue preemption time out (%dms)\n",
-				temp);
-			release_queue(kgd);
-			return -ETIME;
-		}
-		msleep(20);
-		timeout -= 20;
-	}
-
-	release_queue(kgd);
-	return 0;
-}
-
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
-				unsigned int timeout)
-{
-	struct cik_sdma_rlc_registers *m;
-	uint32_t sdma_base_addr;
-	uint32_t temp;
-
-	m = get_sdma_mqd(mqd);
-	sdma_base_addr = get_sdma_base_addr(m);
-
-	temp = read_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL);
-	temp = temp & ~SDMA_RB_ENABLE;
-	write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL, temp);
-
-	while (true) {
-		temp = read_register(kgd, sdma_base_addr +
-						SDMA0_RLC0_CONTEXT_STATUS);
-		if (temp & SDMA_RLC_IDLE)
-			break;
-		if (timeout == 0)
-			return -ETIME;
-		msleep(20);
-		timeout -= 20;
-	}
-
-	write_register(kgd, sdma_base_addr + SDMA0_RLC0_DOORBELL, 0);
-	write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_RPTR, 0);
-	write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_WPTR, 0);
-	write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_BASE, 0);
-
-	return 0;
-}
-
-static int kgd_address_watch_disable(struct kgd_dev *kgd)
-{
-	union TCP_WATCH_CNTL_BITS cntl;
-	unsigned int i;
-
-	cntl.u32All = 0;
-
-	cntl.bitfields.valid = 0;
-	cntl.bitfields.mask = ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK;
-	cntl.bitfields.atc = 1;
-
-	/* Turning off this address until we set all the registers */
-	for (i = 0; i < MAX_WATCH_ADDRESSES; i++)
-		write_register(kgd,
-				watchRegs[i * ADDRESS_WATCH_REG_MAX +
-					ADDRESS_WATCH_REG_CNTL],
-				cntl.u32All);
-
-	return 0;
-}
-
-static int kgd_address_watch_execute(struct kgd_dev *kgd,
-					unsigned int watch_point_id,
-					uint32_t cntl_val,
-					uint32_t addr_hi,
-					uint32_t addr_lo)
-{
-	union TCP_WATCH_CNTL_BITS cntl;
-
-	cntl.u32All = cntl_val;
-
-	/* Turning off this watch point until we set all the registers */
-	cntl.bitfields.valid = 0;
-	write_register(kgd,
-			watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
-				ADDRESS_WATCH_REG_CNTL],
-			cntl.u32All);
-
-	write_register(kgd,
-			watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
-				ADDRESS_WATCH_REG_ADDR_HI],
-			addr_hi);
-
-	write_register(kgd,
-			watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
-				ADDRESS_WATCH_REG_ADDR_LO],
-			addr_lo);
-
-	/* Enable the watch point */
-	cntl.bitfields.valid = 1;
-
-	write_register(kgd,
-			watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
-				ADDRESS_WATCH_REG_CNTL],
-			cntl.u32All);
-
-	return 0;
-}
-
-static int kgd_wave_control_execute(struct kgd_dev *kgd,
-					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
-{
-	struct radeon_device *rdev = get_radeon_device(kgd);
-	uint32_t data;
-
-	mutex_lock(&rdev->grbm_idx_mutex);
-
-	write_register(kgd, GRBM_GFX_INDEX, gfx_index_val);
-	write_register(kgd, SQ_CMD, sq_cmd);
-
-	/*  Restore the GRBM_GFX_INDEX register  */
-
-	data = INSTANCE_BROADCAST_WRITES | SH_BROADCAST_WRITES |
-		SE_BROADCAST_WRITES;
-
-	write_register(kgd, GRBM_GFX_INDEX, data);
-
-	mutex_unlock(&rdev->grbm_idx_mutex);
-
-	return 0;
-}
-
-static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
-					unsigned int watch_point_id,
-					unsigned int reg_offset)
-{
-	return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset]
-		/ 4;
-}
-
-static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid)
-{
-	uint32_t reg;
-	struct radeon_device *rdev = (struct radeon_device *) kgd;
-
-	reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
-	return reg & ATC_VMID_PASID_MAPPING_VALID_MASK;
-}
-
-static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
-							uint8_t vmid)
-{
-	uint32_t reg;
-	struct radeon_device *rdev = (struct radeon_device *) kgd;
-
-	reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
-	return reg & ATC_VMID_PASID_MAPPING_PASID_MASK;
-}
-
-static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
-{
-	struct radeon_device *rdev = (struct radeon_device *) kgd;
-
-	return WREG32(VM_INVALIDATE_REQUEST, 1 << vmid);
-}
-
-static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
-{
-	struct radeon_device *rdev = (struct radeon_device *) kgd;
-	const union radeon_firmware_header *hdr;
-
-	BUG_ON(kgd == NULL || rdev->mec_fw == NULL);
-
-	switch (type) {
-	case KGD_ENGINE_PFP:
-		hdr = (const union radeon_firmware_header *) rdev->pfp_fw->data;
-		break;
-
-	case KGD_ENGINE_ME:
-		hdr = (const union radeon_firmware_header *) rdev->me_fw->data;
-		break;
-
-	case KGD_ENGINE_CE:
-		hdr = (const union radeon_firmware_header *) rdev->ce_fw->data;
-		break;
-
-	case KGD_ENGINE_MEC1:
-		hdr = (const union radeon_firmware_header *) rdev->mec_fw->data;
-		break;
-
-	case KGD_ENGINE_MEC2:
-		hdr = (const union radeon_firmware_header *)
-							rdev->mec2_fw->data;
-		break;
-
-	case KGD_ENGINE_RLC:
-		hdr = (const union radeon_firmware_header *) rdev->rlc_fw->data;
-		break;
-
-	case KGD_ENGINE_SDMA1:
-	case KGD_ENGINE_SDMA2:
-		hdr = (const union radeon_firmware_header *)
-							rdev->sdma_fw->data;
-		break;
-
-	default:
-		return 0;
-	}
-
-	if (hdr == NULL)
-		return 0;
-
-	/* Only 12 bit in use*/
-	return hdr->common.ucode_version;
-}
diff --git a/drivers/gpu/drm/radeon/radeon_kfd.h b/drivers/gpu/drm/radeon/radeon_kfd.h
deleted file mode 100644
index 9df1fea..0000000
--- a/drivers/gpu/drm/radeon/radeon_kfd.h
+++ /dev/null
@@ -1,47 +0,0 @@ 
-/*
- * Copyright 2014 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/*
- * radeon_kfd.h defines the private interface between the
- * AMD kernel graphics drivers and the AMD KFD.
- */
-
-#ifndef RADEON_KFD_H_INCLUDED
-#define RADEON_KFD_H_INCLUDED
-
-#include <linux/types.h>
-#include "kgd_kfd_interface.h"
-
-struct radeon_device;
-
-int radeon_kfd_init(void);
-void radeon_kfd_fini(void);
-
-void radeon_kfd_suspend(struct radeon_device *rdev);
-int radeon_kfd_resume(struct radeon_device *rdev);
-void radeon_kfd_interrupt(struct radeon_device *rdev,
-			const void *ih_ring_entry);
-void radeon_kfd_device_probe(struct radeon_device *rdev);
-void radeon_kfd_device_init(struct radeon_device *rdev);
-void radeon_kfd_device_fini(struct radeon_device *rdev);
-
-#endif /* RADEON_KFD_H_INCLUDED */
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index dfee8f7..cde037f 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -34,8 +34,6 @@ 
 #include <linux/slab.h>
 #include <linux/pm_runtime.h>
 
-#include "radeon_kfd.h"
-
 #if defined(CONFIG_VGA_SWITCHEROO)
 bool radeon_has_atpx(void);
 #else
@@ -68,8 +66,6 @@  void radeon_driver_unload_kms(struct drm_device *dev)
 		pm_runtime_forbid(dev->dev);
 	}
 
-	radeon_kfd_device_fini(rdev);
-
 	radeon_acpi_fini(rdev);
 	
 	radeon_modeset_fini(rdev);
@@ -174,9 +170,6 @@  int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
 				"Error during ACPI methods call\n");
 	}
 
-	radeon_kfd_device_probe(rdev);
-	radeon_kfd_device_init(rdev);
-
 	if (radeon_is_px(dev)) {
 		pm_runtime_use_autosuspend(dev->dev);
 		pm_runtime_set_autosuspend_delay(dev->dev, 5000);

Comments

Don't have any strong objection, but I just want to ask if current
users can just move to using amdgpu on KV and their current usermode
stack will work as usual.
btw, are there any "current users" that you are aware of ?


On Mon, Oct 30, 2017 at 3:16 PM, Christian König
<deathsimple@vodafone.de> wrote:
> From: Christian König <christian.koenig@amd.com>
>
> To quote Felix: "For testing KV with current user mode stack, please use
> amdgpu. I don't expect this to work with radeon and I'm not planning to spend
> any effort on making radeon work with a current user mode stack."
>
> Only compile tested, but should be straight forward.
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
>  drivers/gpu/drm/amd/amdkfd/Kconfig  |   2 +-
>  drivers/gpu/drm/radeon/Makefile     |   3 +-
>  drivers/gpu/drm/radeon/cik.c        |  14 +-
>  drivers/gpu/drm/radeon/cikd.h       |   2 -
>  drivers/gpu/drm/radeon/radeon.h     |   3 -
>  drivers/gpu/drm/radeon/radeon_drv.c |  10 -
>  drivers/gpu/drm/radeon/radeon_kfd.c | 901 ------------------------------------
>  drivers/gpu/drm/radeon/radeon_kfd.h |  47 --
>  drivers/gpu/drm/radeon/radeon_kms.c |   7 -
>  9 files changed, 4 insertions(+), 985 deletions(-)
>  delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.c
>  delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.h
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig
> index e13c67c..bc5a294 100644
> --- a/drivers/gpu/drm/amd/amdkfd/Kconfig
> +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
> @@ -4,6 +4,6 @@
>
>  config HSA_AMD
>         tristate "HSA kernel driver for AMD GPU devices"
> -       depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64
> +       depends on DRM_AMDGPU && AMD_IOMMU_V2 && X86_64
>         help
>           Enable this if you want to use HSA features on AMD GPU devices.
> diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile
> index be16c63..cf3e598 100644
> --- a/drivers/gpu/drm/radeon/Makefile
> +++ b/drivers/gpu/drm/radeon/Makefile
> @@ -102,8 +102,7 @@ radeon-y += \
>  radeon-y += \
>         radeon_vce.o \
>         vce_v1_0.o \
> -       vce_v2_0.o \
> -       radeon_kfd.o
> +       vce_v2_0.o
>
>  radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
>  radeon-$(CONFIG_ACPI) += radeon_acpi.o
> diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c
> index 3cb6c55..898f9a0 100644
> --- a/drivers/gpu/drm/radeon/cik.c
> +++ b/drivers/gpu/drm/radeon/cik.c
> @@ -33,7 +33,6 @@
>  #include "cik_blit_shaders.h"
>  #include "radeon_ucode.h"
>  #include "clearstate_ci.h"
> -#include "radeon_kfd.h"
>
>  #define SH_MEM_CONFIG_GFX_DEFAULT \
>         ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
> @@ -5684,10 +5683,9 @@ int cik_vm_init(struct radeon_device *rdev)
>         /*
>          * number of VMs
>          * VMID 0 is reserved for System
> -        * radeon graphics/compute will use VMIDs 1-7
> -        * amdkfd will use VMIDs 8-15
> +        * radeon graphics/compute will use VMIDs 1-15
>          */
> -       rdev->vm_manager.nvm = RADEON_NUM_OF_VMIDS;
> +       rdev->vm_manager.nvm = 16;
>         /* base offset of vram pages */
>         if (rdev->flags & RADEON_IS_IGP) {
>                 u64 tmp = RREG32(MC_VM_FB_OFFSET);
> @@ -7589,9 +7587,6 @@ int cik_irq_process(struct radeon_device *rdev)
>                 /* wptr/rptr are in bytes! */
>                 ring_index = rptr / 4;
>
> -               radeon_kfd_interrupt(rdev,
> -                               (const void *) &rdev->ih.ring[ring_index]);
> -
>                 src_id =  le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
>                 src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
>                 ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff;
> @@ -8486,10 +8481,6 @@ static int cik_startup(struct radeon_device *rdev)
>         if (r)
>                 return r;
>
> -       r = radeon_kfd_resume(rdev);
> -       if (r)
> -               return r;
> -
>         return 0;
>  }
>
> @@ -8538,7 +8529,6 @@ int cik_resume(struct radeon_device *rdev)
>   */
>  int cik_suspend(struct radeon_device *rdev)
>  {
> -       radeon_kfd_suspend(rdev);
>         radeon_pm_suspend(rdev);
>         radeon_audio_fini(rdev);
>         radeon_vm_manager_fini(rdev);
> diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h
> index e210154..cda16fc 100644
> --- a/drivers/gpu/drm/radeon/cikd.h
> +++ b/drivers/gpu/drm/radeon/cikd.h
> @@ -30,8 +30,6 @@
>  #define CIK_RB_BITMAP_WIDTH_PER_SH     2
>  #define HAWAII_RB_BITMAP_WIDTH_PER_SH  4
>
> -#define RADEON_NUM_OF_VMIDS    8
> -
>  /* DIDT IND registers */
>  #define DIDT_SQ_CTRL0                                     0x0
>  #       define DIDT_CTRL_EN                               (1 << 0)
> diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
> index ec63bc5..d94741b 100644
> --- a/drivers/gpu/drm/radeon/radeon.h
> +++ b/drivers/gpu/drm/radeon/radeon.h
> @@ -2456,9 +2456,6 @@ struct radeon_device {
>         u64 vram_pin_size;
>         u64 gart_pin_size;
>
> -       /* amdkfd interface */
> -       struct kfd_dev          *kfd;
> -
>         struct mutex    mn_lock;
>         DECLARE_HASHTABLE(mn_hash, 7);
>  };
> diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
> index f4becad..31dd04f 100644
> --- a/drivers/gpu/drm/radeon/radeon_drv.c
> +++ b/drivers/gpu/drm/radeon/radeon_drv.c
> @@ -43,7 +43,6 @@
>  #include <drm/drm_fb_helper.h>
>
>  #include <drm/drm_crtc_helper.h>
> -#include "radeon_kfd.h"
>
>  /*
>   * KMS wrapper.
> @@ -338,14 +337,6 @@ static int radeon_pci_probe(struct pci_dev *pdev,
>  {
>         int ret;
>
> -       /*
> -        * Initialize amdkfd before starting radeon. If it was not loaded yet,
> -        * defer radeon probing
> -        */
> -       ret = radeon_kfd_init();
> -       if (ret == -EPROBE_DEFER)
> -               return ret;
> -
>         if (vga_switcheroo_client_probe_defer(pdev))
>                 return -EPROBE_DEFER;
>
> @@ -645,7 +636,6 @@ static int __init radeon_init(void)
>
>  static void __exit radeon_exit(void)
>  {
> -       radeon_kfd_fini();
>         pci_unregister_driver(pdriver);
>         radeon_unregister_atpx_handler();
>  }
> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c b/drivers/gpu/drm/radeon/radeon_kfd.c
> deleted file mode 100644
> index 385b4d7..0000000
> --- a/drivers/gpu/drm/radeon/radeon_kfd.c
> +++ /dev/null
> @@ -1,901 +0,0 @@
> -/*
> - * Copyright 2014 Advanced Micro Devices, Inc.
> - *
> - * Permission is hereby granted, free of charge, to any person obtaining a
> - * copy of this software and associated documentation files (the "Software"),
> - * to deal in the Software without restriction, including without limitation
> - * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> - * and/or sell copies of the Software, and to permit persons to whom the
> - * Software is furnished to do so, subject to the following conditions:
> - *
> - * The above copyright notice and this permission notice shall be included in
> - * all copies or substantial portions of the Software.
> - *
> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> - * OTHER DEALINGS IN THE SOFTWARE.
> - */
> -
> -#include <linux/module.h>
> -#include <linux/fdtable.h>
> -#include <linux/uaccess.h>
> -#include <drm/drmP.h>
> -#include "radeon.h"
> -#include "cikd.h"
> -#include "cik_reg.h"
> -#include "radeon_kfd.h"
> -#include "radeon_ucode.h"
> -#include <linux/firmware.h>
> -#include "cik_structs.h"
> -
> -#define CIK_PIPE_PER_MEC       (4)
> -
> -static const uint32_t watchRegs[MAX_WATCH_ADDRESSES * ADDRESS_WATCH_REG_MAX] = {
> -       TCP_WATCH0_ADDR_H, TCP_WATCH0_ADDR_L, TCP_WATCH0_CNTL,
> -       TCP_WATCH1_ADDR_H, TCP_WATCH1_ADDR_L, TCP_WATCH1_CNTL,
> -       TCP_WATCH2_ADDR_H, TCP_WATCH2_ADDR_L, TCP_WATCH2_CNTL,
> -       TCP_WATCH3_ADDR_H, TCP_WATCH3_ADDR_L, TCP_WATCH3_CNTL
> -};
> -
> -struct kgd_mem {
> -       struct radeon_bo *bo;
> -       uint64_t gpu_addr;
> -       void *cpu_ptr;
> -};
> -
> -
> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
> -                       void **mem_obj, uint64_t *gpu_addr,
> -                       void **cpu_ptr);
> -
> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
> -
> -static uint64_t get_vmem_size(struct kgd_dev *kgd);
> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd);
> -
> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
> -
> -static int alloc_pasid(unsigned int bits);
> -static void free_pasid(unsigned int pasid);
> -
> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
> -
> -/*
> - * Register access functions
> - */
> -
> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
> -               uint32_t sh_mem_config, uint32_t sh_mem_ape1_base,
> -               uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases);
> -
> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
> -                                       unsigned int vmid);
> -
> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
> -                               uint32_t hpd_size, uint64_t hpd_gpu_addr);
> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
> -                       uint32_t queue_id, uint32_t __user *wptr,
> -                       uint32_t wptr_shift, uint32_t wptr_mask,
> -                       struct mm_struct *mm);
> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
> -                               uint32_t pipe_id, uint32_t queue_id);
> -
> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
> -                               unsigned int timeout, uint32_t pipe_id,
> -                               uint32_t queue_id);
> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
> -                               unsigned int timeout);
> -static int kgd_address_watch_disable(struct kgd_dev *kgd);
> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
> -                                       unsigned int watch_point_id,
> -                                       uint32_t cntl_val,
> -                                       uint32_t addr_hi,
> -                                       uint32_t addr_lo);
> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
> -                                       uint32_t gfx_index_val,
> -                                       uint32_t sq_cmd);
> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
> -                                       unsigned int watch_point_id,
> -                                       unsigned int reg_offset);
> -
> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid);
> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
> -                                                       uint8_t vmid);
> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid);
> -
> -static const struct kfd2kgd_calls kfd2kgd = {
> -       .init_gtt_mem_allocation = alloc_gtt_mem,
> -       .free_gtt_mem = free_gtt_mem,
> -       .get_vmem_size = get_vmem_size,
> -       .get_gpu_clock_counter = get_gpu_clock_counter,
> -       .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
> -       .alloc_pasid = alloc_pasid,
> -       .free_pasid = free_pasid,
> -       .program_sh_mem_settings = kgd_program_sh_mem_settings,
> -       .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
> -       .init_pipeline = kgd_init_pipeline,
> -       .init_interrupts = kgd_init_interrupts,
> -       .hqd_load = kgd_hqd_load,
> -       .hqd_sdma_load = kgd_hqd_sdma_load,
> -       .hqd_is_occupied = kgd_hqd_is_occupied,
> -       .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
> -       .hqd_destroy = kgd_hqd_destroy,
> -       .hqd_sdma_destroy = kgd_hqd_sdma_destroy,
> -       .address_watch_disable = kgd_address_watch_disable,
> -       .address_watch_execute = kgd_address_watch_execute,
> -       .wave_control_execute = kgd_wave_control_execute,
> -       .address_watch_get_offset = kgd_address_watch_get_offset,
> -       .get_atc_vmid_pasid_mapping_pasid = get_atc_vmid_pasid_mapping_pasid,
> -       .get_atc_vmid_pasid_mapping_valid = get_atc_vmid_pasid_mapping_valid,
> -       .write_vmid_invalidate_request = write_vmid_invalidate_request,
> -       .get_fw_version = get_fw_version
> -};
> -
> -static const struct kgd2kfd_calls *kgd2kfd;
> -
> -int radeon_kfd_init(void)
> -{
> -       int ret;
> -
> -#if defined(CONFIG_HSA_AMD_MODULE)
> -       int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
> -
> -       kgd2kfd_init_p = symbol_request(kgd2kfd_init);
> -
> -       if (kgd2kfd_init_p == NULL)
> -               return -ENOENT;
> -
> -       ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd);
> -       if (ret) {
> -               symbol_put(kgd2kfd_init);
> -               kgd2kfd = NULL;
> -       }
> -
> -#elif defined(CONFIG_HSA_AMD)
> -       ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
> -       if (ret)
> -               kgd2kfd = NULL;
> -
> -#else
> -       ret = -ENOENT;
> -#endif
> -
> -       return ret;
> -}
> -
> -void radeon_kfd_fini(void)
> -{
> -       if (kgd2kfd) {
> -               kgd2kfd->exit();
> -               symbol_put(kgd2kfd_init);
> -       }
> -}
> -
> -void radeon_kfd_device_probe(struct radeon_device *rdev)
> -{
> -       if (kgd2kfd)
> -               rdev->kfd = kgd2kfd->probe((struct kgd_dev *)rdev,
> -                       rdev->pdev, &kfd2kgd);
> -}
> -
> -void radeon_kfd_device_init(struct radeon_device *rdev)
> -{
> -       int i, queue, pipe, mec;
> -
> -       if (rdev->kfd) {
> -               struct kgd2kfd_shared_resources gpu_resources = {
> -                       .compute_vmid_bitmap = 0xFF00,
> -                       .num_pipe_per_mec = 4,
> -                       .num_queue_per_pipe = 8
> -               };
> -
> -               bitmap_zero(gpu_resources.queue_bitmap, KGD_MAX_QUEUES);
> -
> -               for (i = 0; i < KGD_MAX_QUEUES; ++i) {
> -                       queue = i % gpu_resources.num_queue_per_pipe;
> -                       pipe = (i / gpu_resources.num_queue_per_pipe)
> -                               % gpu_resources.num_pipe_per_mec;
> -                       mec = (i / gpu_resources.num_queue_per_pipe)
> -                               / gpu_resources.num_pipe_per_mec;
> -
> -                       if (mec == 0 && pipe > 0)
> -                               set_bit(i, gpu_resources.queue_bitmap);
> -               }
> -
> -               radeon_doorbell_get_kfd_info(rdev,
> -                               &gpu_resources.doorbell_physical_address,
> -                               &gpu_resources.doorbell_aperture_size,
> -                               &gpu_resources.doorbell_start_offset);
> -
> -               kgd2kfd->device_init(rdev->kfd, &gpu_resources);
> -       }
> -}
> -
> -void radeon_kfd_device_fini(struct radeon_device *rdev)
> -{
> -       if (rdev->kfd) {
> -               kgd2kfd->device_exit(rdev->kfd);
> -               rdev->kfd = NULL;
> -       }
> -}
> -
> -void radeon_kfd_interrupt(struct radeon_device *rdev, const void *ih_ring_entry)
> -{
> -       if (rdev->kfd)
> -               kgd2kfd->interrupt(rdev->kfd, ih_ring_entry);
> -}
> -
> -void radeon_kfd_suspend(struct radeon_device *rdev)
> -{
> -       if (rdev->kfd)
> -               kgd2kfd->suspend(rdev->kfd);
> -}
> -
> -int radeon_kfd_resume(struct radeon_device *rdev)
> -{
> -       int r = 0;
> -
> -       if (rdev->kfd)
> -               r = kgd2kfd->resume(rdev->kfd);
> -
> -       return r;
> -}
> -
> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
> -                       void **mem_obj, uint64_t *gpu_addr,
> -                       void **cpu_ptr)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
> -       struct kgd_mem **mem = (struct kgd_mem **) mem_obj;
> -       int r;
> -
> -       BUG_ON(kgd == NULL);
> -       BUG_ON(gpu_addr == NULL);
> -       BUG_ON(cpu_ptr == NULL);
> -
> -       *mem = kmalloc(sizeof(struct kgd_mem), GFP_KERNEL);
> -       if ((*mem) == NULL)
> -               return -ENOMEM;
> -
> -       r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_GTT,
> -                               RADEON_GEM_GTT_WC, NULL, NULL, &(*mem)->bo);
> -       if (r) {
> -               dev_err(rdev->dev,
> -                       "failed to allocate BO for amdkfd (%d)\n", r);
> -               return r;
> -       }
> -
> -       /* map the buffer */
> -       r = radeon_bo_reserve((*mem)->bo, true);
> -       if (r) {
> -               dev_err(rdev->dev, "(%d) failed to reserve bo for amdkfd\n", r);
> -               goto allocate_mem_reserve_bo_failed;
> -       }
> -
> -       r = radeon_bo_pin((*mem)->bo, RADEON_GEM_DOMAIN_GTT,
> -                               &(*mem)->gpu_addr);
> -       if (r) {
> -               dev_err(rdev->dev, "(%d) failed to pin bo for amdkfd\n", r);
> -               goto allocate_mem_pin_bo_failed;
> -       }
> -       *gpu_addr = (*mem)->gpu_addr;
> -
> -       r = radeon_bo_kmap((*mem)->bo, &(*mem)->cpu_ptr);
> -       if (r) {
> -               dev_err(rdev->dev,
> -                       "(%d) failed to map bo to kernel for amdkfd\n", r);
> -               goto allocate_mem_kmap_bo_failed;
> -       }
> -       *cpu_ptr = (*mem)->cpu_ptr;
> -
> -       radeon_bo_unreserve((*mem)->bo);
> -
> -       return 0;
> -
> -allocate_mem_kmap_bo_failed:
> -       radeon_bo_unpin((*mem)->bo);
> -allocate_mem_pin_bo_failed:
> -       radeon_bo_unreserve((*mem)->bo);
> -allocate_mem_reserve_bo_failed:
> -       radeon_bo_unref(&(*mem)->bo);
> -
> -       return r;
> -}
> -
> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
> -{
> -       struct kgd_mem *mem = (struct kgd_mem *) mem_obj;
> -
> -       BUG_ON(mem == NULL);
> -
> -       radeon_bo_reserve(mem->bo, true);
> -       radeon_bo_kunmap(mem->bo);
> -       radeon_bo_unpin(mem->bo);
> -       radeon_bo_unreserve(mem->bo);
> -       radeon_bo_unref(&(mem->bo));
> -       kfree(mem);
> -}
> -
> -static uint64_t get_vmem_size(struct kgd_dev *kgd)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
> -
> -       BUG_ON(kgd == NULL);
> -
> -       return rdev->mc.real_vram_size;
> -}
> -
> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
> -
> -       return rdev->asic->get_gpu_clock_counter(rdev);
> -}
> -
> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
> -
> -       /* The sclk is in quantas of 10kHz */
> -       return rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk / 100;
> -}
> -
> -/*
> - * PASID manager
> - */
> -static DEFINE_IDA(pasid_ida);
> -
> -static int alloc_pasid(unsigned int bits)
> -{
> -       int pasid = -EINVAL;
> -
> -       for (bits = min(bits, 31U); bits > 0; bits--) {
> -               pasid = ida_simple_get(&pasid_ida,
> -                                      1U << (bits - 1), 1U << bits,
> -                                      GFP_KERNEL);
> -               if (pasid != -ENOSPC)
> -                       break;
> -       }
> -
> -       return pasid;
> -}
> -
> -static void free_pasid(unsigned int pasid)
> -{
> -       ida_simple_remove(&pasid_ida, pasid);
> -}
> -
> -static inline struct radeon_device *get_radeon_device(struct kgd_dev *kgd)
> -{
> -       return (struct radeon_device *)kgd;
> -}
> -
> -static void write_register(struct kgd_dev *kgd, uint32_t offset, uint32_t value)
> -{
> -       struct radeon_device *rdev = get_radeon_device(kgd);
> -
> -       writel(value, (void __iomem *)(rdev->rmmio + offset));
> -}
> -
> -static uint32_t read_register(struct kgd_dev *kgd, uint32_t offset)
> -{
> -       struct radeon_device *rdev = get_radeon_device(kgd);
> -
> -       return readl((void __iomem *)(rdev->rmmio + offset));
> -}
> -
> -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
> -                       uint32_t queue, uint32_t vmid)
> -{
> -       struct radeon_device *rdev = get_radeon_device(kgd);
> -       uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
> -
> -       mutex_lock(&rdev->srbm_mutex);
> -       write_register(kgd, SRBM_GFX_CNTL, value);
> -}
> -
> -static void unlock_srbm(struct kgd_dev *kgd)
> -{
> -       struct radeon_device *rdev = get_radeon_device(kgd);
> -
> -       write_register(kgd, SRBM_GFX_CNTL, 0);
> -       mutex_unlock(&rdev->srbm_mutex);
> -}
> -
> -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
> -                               uint32_t queue_id)
> -{
> -       uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1;
> -       uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC);
> -
> -       lock_srbm(kgd, mec, pipe, queue_id, 0);
> -}
> -
> -static void release_queue(struct kgd_dev *kgd)
> -{
> -       unlock_srbm(kgd);
> -}
> -
> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
> -                                       uint32_t sh_mem_config,
> -                                       uint32_t sh_mem_ape1_base,
> -                                       uint32_t sh_mem_ape1_limit,
> -                                       uint32_t sh_mem_bases)
> -{
> -       lock_srbm(kgd, 0, 0, 0, vmid);
> -
> -       write_register(kgd, SH_MEM_CONFIG, sh_mem_config);
> -       write_register(kgd, SH_MEM_APE1_BASE, sh_mem_ape1_base);
> -       write_register(kgd, SH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
> -       write_register(kgd, SH_MEM_BASES, sh_mem_bases);
> -
> -       unlock_srbm(kgd);
> -}
> -
> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
> -                                       unsigned int vmid)
> -{
> -       /*
> -        * We have to assume that there is no outstanding mapping.
> -        * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0
> -        * because a mapping is in progress or because a mapping finished and
> -        * the SW cleared it.
> -        * So the protocol is to always wait & clear.
> -        */
> -       uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
> -                                       ATC_VMID_PASID_MAPPING_VALID_MASK;
> -
> -       write_register(kgd, ATC_VMID0_PASID_MAPPING + vmid*sizeof(uint32_t),
> -                       pasid_mapping);
> -
> -       while (!(read_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS) &
> -                                                               (1U << vmid)))
> -               cpu_relax();
> -       write_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U << vmid);
> -
> -       /* Mapping vmid to pasid also for IH block */
> -       write_register(kgd, IH_VMID_0_LUT + vmid * sizeof(uint32_t),
> -                       pasid_mapping);
> -
> -       return 0;
> -}
> -
> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
> -                               uint32_t hpd_size, uint64_t hpd_gpu_addr)
> -{
> -       /* nothing to do here */
> -       return 0;
> -}
> -
> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
> -{
> -       uint32_t mec;
> -       uint32_t pipe;
> -
> -       mec = (pipe_id / CIK_PIPE_PER_MEC) + 1;
> -       pipe = (pipe_id % CIK_PIPE_PER_MEC);
> -
> -       lock_srbm(kgd, mec, pipe, 0, 0);
> -
> -       write_register(kgd, CPC_INT_CNTL,
> -                       TIME_STAMP_INT_ENABLE | OPCODE_ERROR_INT_ENABLE);
> -
> -       unlock_srbm(kgd);
> -
> -       return 0;
> -}
> -
> -static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers *m)
> -{
> -       uint32_t retval;
> -
> -       retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET +
> -                       m->sdma_queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
> -
> -       pr_debug("kfd: sdma base address: 0x%x\n", retval);
> -
> -       return retval;
> -}
> -
> -static inline struct cik_mqd *get_mqd(void *mqd)
> -{
> -       return (struct cik_mqd *)mqd;
> -}
> -
> -static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
> -{
> -       return (struct cik_sdma_rlc_registers *)mqd;
> -}
> -
> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
> -                       uint32_t queue_id, uint32_t __user *wptr,
> -                       uint32_t wptr_shift, uint32_t wptr_mask,
> -                       struct mm_struct *mm)
> -{
> -       uint32_t wptr_shadow, is_wptr_shadow_valid;
> -       struct cik_mqd *m;
> -
> -       m = get_mqd(mqd);
> -
> -       is_wptr_shadow_valid = !get_user(wptr_shadow, wptr);
> -
> -       acquire_queue(kgd, pipe_id, queue_id);
> -       write_register(kgd, CP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo);
> -       write_register(kgd, CP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi);
> -       write_register(kgd, CP_MQD_CONTROL, m->cp_mqd_control);
> -
> -       write_register(kgd, CP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo);
> -       write_register(kgd, CP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi);
> -       write_register(kgd, CP_HQD_PQ_CONTROL, m->cp_hqd_pq_control);
> -
> -       write_register(kgd, CP_HQD_IB_CONTROL, m->cp_hqd_ib_control);
> -       write_register(kgd, CP_HQD_IB_BASE_ADDR, m->cp_hqd_ib_base_addr_lo);
> -       write_register(kgd, CP_HQD_IB_BASE_ADDR_HI, m->cp_hqd_ib_base_addr_hi);
> -
> -       write_register(kgd, CP_HQD_IB_RPTR, m->cp_hqd_ib_rptr);
> -
> -       write_register(kgd, CP_HQD_PERSISTENT_STATE,
> -                       m->cp_hqd_persistent_state);
> -       write_register(kgd, CP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd);
> -       write_register(kgd, CP_HQD_MSG_TYPE, m->cp_hqd_msg_type);
> -
> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_LO,
> -                       m->cp_hqd_atomic0_preop_lo);
> -
> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_HI,
> -                       m->cp_hqd_atomic0_preop_hi);
> -
> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_LO,
> -                       m->cp_hqd_atomic1_preop_lo);
> -
> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_HI,
> -                       m->cp_hqd_atomic1_preop_hi);
> -
> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR,
> -                       m->cp_hqd_pq_rptr_report_addr_lo);
> -
> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR_HI,
> -                       m->cp_hqd_pq_rptr_report_addr_hi);
> -
> -       write_register(kgd, CP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr);
> -
> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR,
> -                       m->cp_hqd_pq_wptr_poll_addr_lo);
> -
> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR_HI,
> -                       m->cp_hqd_pq_wptr_poll_addr_hi);
> -
> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL,
> -                       m->cp_hqd_pq_doorbell_control);
> -
> -       write_register(kgd, CP_HQD_VMID, m->cp_hqd_vmid);
> -
> -       write_register(kgd, CP_HQD_QUANTUM, m->cp_hqd_quantum);
> -
> -       write_register(kgd, CP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority);
> -       write_register(kgd, CP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority);
> -
> -       write_register(kgd, CP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr);
> -
> -       if (is_wptr_shadow_valid)
> -               write_register(kgd, CP_HQD_PQ_WPTR, wptr_shadow);
> -
> -       write_register(kgd, CP_HQD_ACTIVE, m->cp_hqd_active);
> -       release_queue(kgd);
> -
> -       return 0;
> -}
> -
> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
> -{
> -       struct cik_sdma_rlc_registers *m;
> -       uint32_t sdma_base_addr;
> -
> -       m = get_sdma_mqd(mqd);
> -       sdma_base_addr = get_sdma_base_addr(m);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_VIRTUAL_ADDR,
> -                       m->sdma_rlc_virtual_addr);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE,
> -                       m->sdma_rlc_rb_base);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE_HI,
> -                       m->sdma_rlc_rb_base_hi);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_LO,
> -                       m->sdma_rlc_rb_rptr_addr_lo);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_HI,
> -                       m->sdma_rlc_rb_rptr_addr_hi);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_DOORBELL,
> -                       m->sdma_rlc_doorbell);
> -
> -       write_register(kgd,
> -                       sdma_base_addr + SDMA0_RLC0_RB_CNTL,
> -                       m->sdma_rlc_rb_cntl);
> -
> -       return 0;
> -}
> -
> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
> -                               uint32_t pipe_id, uint32_t queue_id)
> -{
> -       uint32_t act;
> -       bool retval = false;
> -       uint32_t low, high;
> -
> -       acquire_queue(kgd, pipe_id, queue_id);
> -       act = read_register(kgd, CP_HQD_ACTIVE);
> -       if (act) {
> -               low = lower_32_bits(queue_address >> 8);
> -               high = upper_32_bits(queue_address >> 8);
> -
> -               if (low == read_register(kgd, CP_HQD_PQ_BASE) &&
> -                               high == read_register(kgd, CP_HQD_PQ_BASE_HI))
> -                       retval = true;
> -       }
> -       release_queue(kgd);
> -       return retval;
> -}
> -
> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
> -{
> -       struct cik_sdma_rlc_registers *m;
> -       uint32_t sdma_base_addr;
> -       uint32_t sdma_rlc_rb_cntl;
> -
> -       m = get_sdma_mqd(mqd);
> -       sdma_base_addr = get_sdma_base_addr(m);
> -
> -       sdma_rlc_rb_cntl = read_register(kgd,
> -                                       sdma_base_addr + SDMA0_RLC0_RB_CNTL);
> -
> -       if (sdma_rlc_rb_cntl & SDMA_RB_ENABLE)
> -               return true;
> -
> -       return false;
> -}
> -
> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
> -                               unsigned int timeout, uint32_t pipe_id,
> -                               uint32_t queue_id)
> -{
> -       uint32_t temp;
> -
> -       acquire_queue(kgd, pipe_id, queue_id);
> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, 0);
> -
> -       write_register(kgd, CP_HQD_DEQUEUE_REQUEST, reset_type);
> -
> -       while (true) {
> -               temp = read_register(kgd, CP_HQD_ACTIVE);
> -               if (temp & 0x1)
> -                       break;
> -               if (timeout == 0) {
> -                       pr_err("kfd: cp queue preemption time out (%dms)\n",
> -                               temp);
> -                       release_queue(kgd);
> -                       return -ETIME;
> -               }
> -               msleep(20);
> -               timeout -= 20;
> -       }
> -
> -       release_queue(kgd);
> -       return 0;
> -}
> -
> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
> -                               unsigned int timeout)
> -{
> -       struct cik_sdma_rlc_registers *m;
> -       uint32_t sdma_base_addr;
> -       uint32_t temp;
> -
> -       m = get_sdma_mqd(mqd);
> -       sdma_base_addr = get_sdma_base_addr(m);
> -
> -       temp = read_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL);
> -       temp = temp & ~SDMA_RB_ENABLE;
> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL, temp);
> -
> -       while (true) {
> -               temp = read_register(kgd, sdma_base_addr +
> -                                               SDMA0_RLC0_CONTEXT_STATUS);
> -               if (temp & SDMA_RLC_IDLE)
> -                       break;
> -               if (timeout == 0)
> -                       return -ETIME;
> -               msleep(20);
> -               timeout -= 20;
> -       }
> -
> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_DOORBELL, 0);
> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_RPTR, 0);
> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_WPTR, 0);
> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_BASE, 0);
> -
> -       return 0;
> -}
> -
> -static int kgd_address_watch_disable(struct kgd_dev *kgd)
> -{
> -       union TCP_WATCH_CNTL_BITS cntl;
> -       unsigned int i;
> -
> -       cntl.u32All = 0;
> -
> -       cntl.bitfields.valid = 0;
> -       cntl.bitfields.mask = ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK;
> -       cntl.bitfields.atc = 1;
> -
> -       /* Turning off this address until we set all the registers */
> -       for (i = 0; i < MAX_WATCH_ADDRESSES; i++)
> -               write_register(kgd,
> -                               watchRegs[i * ADDRESS_WATCH_REG_MAX +
> -                                       ADDRESS_WATCH_REG_CNTL],
> -                               cntl.u32All);
> -
> -       return 0;
> -}
> -
> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
> -                                       unsigned int watch_point_id,
> -                                       uint32_t cntl_val,
> -                                       uint32_t addr_hi,
> -                                       uint32_t addr_lo)
> -{
> -       union TCP_WATCH_CNTL_BITS cntl;
> -
> -       cntl.u32All = cntl_val;
> -
> -       /* Turning off this watch point until we set all the registers */
> -       cntl.bitfields.valid = 0;
> -       write_register(kgd,
> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
> -                               ADDRESS_WATCH_REG_CNTL],
> -                       cntl.u32All);
> -
> -       write_register(kgd,
> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
> -                               ADDRESS_WATCH_REG_ADDR_HI],
> -                       addr_hi);
> -
> -       write_register(kgd,
> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
> -                               ADDRESS_WATCH_REG_ADDR_LO],
> -                       addr_lo);
> -
> -       /* Enable the watch point */
> -       cntl.bitfields.valid = 1;
> -
> -       write_register(kgd,
> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
> -                               ADDRESS_WATCH_REG_CNTL],
> -                       cntl.u32All);
> -
> -       return 0;
> -}
> -
> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
> -                                       uint32_t gfx_index_val,
> -                                       uint32_t sq_cmd)
> -{
> -       struct radeon_device *rdev = get_radeon_device(kgd);
> -       uint32_t data;
> -
> -       mutex_lock(&rdev->grbm_idx_mutex);
> -
> -       write_register(kgd, GRBM_GFX_INDEX, gfx_index_val);
> -       write_register(kgd, SQ_CMD, sq_cmd);
> -
> -       /*  Restore the GRBM_GFX_INDEX register  */
> -
> -       data = INSTANCE_BROADCAST_WRITES | SH_BROADCAST_WRITES |
> -               SE_BROADCAST_WRITES;
> -
> -       write_register(kgd, GRBM_GFX_INDEX, data);
> -
> -       mutex_unlock(&rdev->grbm_idx_mutex);
> -
> -       return 0;
> -}
> -
> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
> -                                       unsigned int watch_point_id,
> -                                       unsigned int reg_offset)
> -{
> -       return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset]
> -               / 4;
> -}
> -
> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid)
> -{
> -       uint32_t reg;
> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
> -
> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
> -       return reg & ATC_VMID_PASID_MAPPING_VALID_MASK;
> -}
> -
> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
> -                                                       uint8_t vmid)
> -{
> -       uint32_t reg;
> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
> -
> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
> -       return reg & ATC_VMID_PASID_MAPPING_PASID_MASK;
> -}
> -
> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
> -
> -       return WREG32(VM_INVALIDATE_REQUEST, 1 << vmid);
> -}
> -
> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
> -{
> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
> -       const union radeon_firmware_header *hdr;
> -
> -       BUG_ON(kgd == NULL || rdev->mec_fw == NULL);
> -
> -       switch (type) {
> -       case KGD_ENGINE_PFP:
> -               hdr = (const union radeon_firmware_header *) rdev->pfp_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_ME:
> -               hdr = (const union radeon_firmware_header *) rdev->me_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_CE:
> -               hdr = (const union radeon_firmware_header *) rdev->ce_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_MEC1:
> -               hdr = (const union radeon_firmware_header *) rdev->mec_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_MEC2:
> -               hdr = (const union radeon_firmware_header *)
> -                                                       rdev->mec2_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_RLC:
> -               hdr = (const union radeon_firmware_header *) rdev->rlc_fw->data;
> -               break;
> -
> -       case KGD_ENGINE_SDMA1:
> -       case KGD_ENGINE_SDMA2:
> -               hdr = (const union radeon_firmware_header *)
> -                                                       rdev->sdma_fw->data;
> -               break;
> -
> -       default:
> -               return 0;
> -       }
> -
> -       if (hdr == NULL)
> -               return 0;
> -
> -       /* Only 12 bit in use*/
> -       return hdr->common.ucode_version;
> -}
> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.h b/drivers/gpu/drm/radeon/radeon_kfd.h
> deleted file mode 100644
> index 9df1fea..0000000
> --- a/drivers/gpu/drm/radeon/radeon_kfd.h
> +++ /dev/null
> @@ -1,47 +0,0 @@
> -/*
> - * Copyright 2014 Advanced Micro Devices, Inc.
> - *
> - * Permission is hereby granted, free of charge, to any person obtaining a
> - * copy of this software and associated documentation files (the "Software"),
> - * to deal in the Software without restriction, including without limitation
> - * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> - * and/or sell copies of the Software, and to permit persons to whom the
> - * Software is furnished to do so, subject to the following conditions:
> - *
> - * The above copyright notice and this permission notice shall be included in
> - * all copies or substantial portions of the Software.
> - *
> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> - * OTHER DEALINGS IN THE SOFTWARE.
> - */
> -
> -/*
> - * radeon_kfd.h defines the private interface between the
> - * AMD kernel graphics drivers and the AMD KFD.
> - */
> -
> -#ifndef RADEON_KFD_H_INCLUDED
> -#define RADEON_KFD_H_INCLUDED
> -
> -#include <linux/types.h>
> -#include "kgd_kfd_interface.h"
> -
> -struct radeon_device;
> -
> -int radeon_kfd_init(void);
> -void radeon_kfd_fini(void);
> -
> -void radeon_kfd_suspend(struct radeon_device *rdev);
> -int radeon_kfd_resume(struct radeon_device *rdev);
> -void radeon_kfd_interrupt(struct radeon_device *rdev,
> -                       const void *ih_ring_entry);
> -void radeon_kfd_device_probe(struct radeon_device *rdev);
> -void radeon_kfd_device_init(struct radeon_device *rdev);
> -void radeon_kfd_device_fini(struct radeon_device *rdev);
> -
> -#endif /* RADEON_KFD_H_INCLUDED */
> diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
> index dfee8f7..cde037f 100644
> --- a/drivers/gpu/drm/radeon/radeon_kms.c
> +++ b/drivers/gpu/drm/radeon/radeon_kms.c
> @@ -34,8 +34,6 @@
>  #include <linux/slab.h>
>  #include <linux/pm_runtime.h>
>
> -#include "radeon_kfd.h"
> -
>  #if defined(CONFIG_VGA_SWITCHEROO)
>  bool radeon_has_atpx(void);
>  #else
> @@ -68,8 +66,6 @@ void radeon_driver_unload_kms(struct drm_device *dev)
>                 pm_runtime_forbid(dev->dev);
>         }
>
> -       radeon_kfd_device_fini(rdev);
> -
>         radeon_acpi_fini(rdev);
>
>         radeon_modeset_fini(rdev);
> @@ -174,9 +170,6 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
>                                 "Error during ACPI methods call\n");
>         }
>
> -       radeon_kfd_device_probe(rdev);
> -       radeon_kfd_device_init(rdev);
> -
>         if (radeon_is_px(dev)) {
>                 pm_runtime_use_autosuspend(dev->dev);
>                 pm_runtime_set_autosuspend_delay(dev->dev, 5000);
> --
> 2.7.4
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Am 31.10.2017 um 11:44 schrieb Oded Gabbay:
> Don't have any strong objection, but I just want to ask if current
> users can just move to using amdgpu on KV and their current usermode
> stack will work as usual.

Yes, I think so.

> btw, are there any "current users" that you are aware of ?

Not the slightest idea, but from Felix comments at least the current 
user mode stack + old kernel won't work.

Regards,
Christian.

>
>
> On Mon, Oct 30, 2017 at 3:16 PM, Christian König
> <deathsimple@vodafone.de> wrote:
>> From: Christian König <christian.koenig@amd.com>
>>
>> To quote Felix: "For testing KV with current user mode stack, please use
>> amdgpu. I don't expect this to work with radeon and I'm not planning to spend
>> any effort on making radeon work with a current user mode stack."
>>
>> Only compile tested, but should be straight forward.
>>
>> Signed-off-by: Christian König <christian.koenig@amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdkfd/Kconfig  |   2 +-
>>   drivers/gpu/drm/radeon/Makefile     |   3 +-
>>   drivers/gpu/drm/radeon/cik.c        |  14 +-
>>   drivers/gpu/drm/radeon/cikd.h       |   2 -
>>   drivers/gpu/drm/radeon/radeon.h     |   3 -
>>   drivers/gpu/drm/radeon/radeon_drv.c |  10 -
>>   drivers/gpu/drm/radeon/radeon_kfd.c | 901 ------------------------------------
>>   drivers/gpu/drm/radeon/radeon_kfd.h |  47 --
>>   drivers/gpu/drm/radeon/radeon_kms.c |   7 -
>>   9 files changed, 4 insertions(+), 985 deletions(-)
>>   delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.c
>>   delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.h
>>
>> diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig
>> index e13c67c..bc5a294 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/Kconfig
>> +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
>> @@ -4,6 +4,6 @@
>>
>>   config HSA_AMD
>>          tristate "HSA kernel driver for AMD GPU devices"
>> -       depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64
>> +       depends on DRM_AMDGPU && AMD_IOMMU_V2 && X86_64
>>          help
>>            Enable this if you want to use HSA features on AMD GPU devices.
>> diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile
>> index be16c63..cf3e598 100644
>> --- a/drivers/gpu/drm/radeon/Makefile
>> +++ b/drivers/gpu/drm/radeon/Makefile
>> @@ -102,8 +102,7 @@ radeon-y += \
>>   radeon-y += \
>>          radeon_vce.o \
>>          vce_v1_0.o \
>> -       vce_v2_0.o \
>> -       radeon_kfd.o
>> +       vce_v2_0.o
>>
>>   radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
>>   radeon-$(CONFIG_ACPI) += radeon_acpi.o
>> diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c
>> index 3cb6c55..898f9a0 100644
>> --- a/drivers/gpu/drm/radeon/cik.c
>> +++ b/drivers/gpu/drm/radeon/cik.c
>> @@ -33,7 +33,6 @@
>>   #include "cik_blit_shaders.h"
>>   #include "radeon_ucode.h"
>>   #include "clearstate_ci.h"
>> -#include "radeon_kfd.h"
>>
>>   #define SH_MEM_CONFIG_GFX_DEFAULT \
>>          ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
>> @@ -5684,10 +5683,9 @@ int cik_vm_init(struct radeon_device *rdev)
>>          /*
>>           * number of VMs
>>           * VMID 0 is reserved for System
>> -        * radeon graphics/compute will use VMIDs 1-7
>> -        * amdkfd will use VMIDs 8-15
>> +        * radeon graphics/compute will use VMIDs 1-15
>>           */
>> -       rdev->vm_manager.nvm = RADEON_NUM_OF_VMIDS;
>> +       rdev->vm_manager.nvm = 16;
>>          /* base offset of vram pages */
>>          if (rdev->flags & RADEON_IS_IGP) {
>>                  u64 tmp = RREG32(MC_VM_FB_OFFSET);
>> @@ -7589,9 +7587,6 @@ int cik_irq_process(struct radeon_device *rdev)
>>                  /* wptr/rptr are in bytes! */
>>                  ring_index = rptr / 4;
>>
>> -               radeon_kfd_interrupt(rdev,
>> -                               (const void *) &rdev->ih.ring[ring_index]);
>> -
>>                  src_id =  le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
>>                  src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
>>                  ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff;
>> @@ -8486,10 +8481,6 @@ static int cik_startup(struct radeon_device *rdev)
>>          if (r)
>>                  return r;
>>
>> -       r = radeon_kfd_resume(rdev);
>> -       if (r)
>> -               return r;
>> -
>>          return 0;
>>   }
>>
>> @@ -8538,7 +8529,6 @@ int cik_resume(struct radeon_device *rdev)
>>    */
>>   int cik_suspend(struct radeon_device *rdev)
>>   {
>> -       radeon_kfd_suspend(rdev);
>>          radeon_pm_suspend(rdev);
>>          radeon_audio_fini(rdev);
>>          radeon_vm_manager_fini(rdev);
>> diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h
>> index e210154..cda16fc 100644
>> --- a/drivers/gpu/drm/radeon/cikd.h
>> +++ b/drivers/gpu/drm/radeon/cikd.h
>> @@ -30,8 +30,6 @@
>>   #define CIK_RB_BITMAP_WIDTH_PER_SH     2
>>   #define HAWAII_RB_BITMAP_WIDTH_PER_SH  4
>>
>> -#define RADEON_NUM_OF_VMIDS    8
>> -
>>   /* DIDT IND registers */
>>   #define DIDT_SQ_CTRL0                                     0x0
>>   #       define DIDT_CTRL_EN                               (1 << 0)
>> diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
>> index ec63bc5..d94741b 100644
>> --- a/drivers/gpu/drm/radeon/radeon.h
>> +++ b/drivers/gpu/drm/radeon/radeon.h
>> @@ -2456,9 +2456,6 @@ struct radeon_device {
>>          u64 vram_pin_size;
>>          u64 gart_pin_size;
>>
>> -       /* amdkfd interface */
>> -       struct kfd_dev          *kfd;
>> -
>>          struct mutex    mn_lock;
>>          DECLARE_HASHTABLE(mn_hash, 7);
>>   };
>> diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
>> index f4becad..31dd04f 100644
>> --- a/drivers/gpu/drm/radeon/radeon_drv.c
>> +++ b/drivers/gpu/drm/radeon/radeon_drv.c
>> @@ -43,7 +43,6 @@
>>   #include <drm/drm_fb_helper.h>
>>
>>   #include <drm/drm_crtc_helper.h>
>> -#include "radeon_kfd.h"
>>
>>   /*
>>    * KMS wrapper.
>> @@ -338,14 +337,6 @@ static int radeon_pci_probe(struct pci_dev *pdev,
>>   {
>>          int ret;
>>
>> -       /*
>> -        * Initialize amdkfd before starting radeon. If it was not loaded yet,
>> -        * defer radeon probing
>> -        */
>> -       ret = radeon_kfd_init();
>> -       if (ret == -EPROBE_DEFER)
>> -               return ret;
>> -
>>          if (vga_switcheroo_client_probe_defer(pdev))
>>                  return -EPROBE_DEFER;
>>
>> @@ -645,7 +636,6 @@ static int __init radeon_init(void)
>>
>>   static void __exit radeon_exit(void)
>>   {
>> -       radeon_kfd_fini();
>>          pci_unregister_driver(pdriver);
>>          radeon_unregister_atpx_handler();
>>   }
>> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c b/drivers/gpu/drm/radeon/radeon_kfd.c
>> deleted file mode 100644
>> index 385b4d7..0000000
>> --- a/drivers/gpu/drm/radeon/radeon_kfd.c
>> +++ /dev/null
>> @@ -1,901 +0,0 @@
>> -/*
>> - * Copyright 2014 Advanced Micro Devices, Inc.
>> - *
>> - * Permission is hereby granted, free of charge, to any person obtaining a
>> - * copy of this software and associated documentation files (the "Software"),
>> - * to deal in the Software without restriction, including without limitation
>> - * the rights to use, copy, modify, merge, publish, distribute, sublicense,
>> - * and/or sell copies of the Software, and to permit persons to whom the
>> - * Software is furnished to do so, subject to the following conditions:
>> - *
>> - * The above copyright notice and this permission notice shall be included in
>> - * all copies or substantial portions of the Software.
>> - *
>> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
>> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
>> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
>> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>> - * OTHER DEALINGS IN THE SOFTWARE.
>> - */
>> -
>> -#include <linux/module.h>
>> -#include <linux/fdtable.h>
>> -#include <linux/uaccess.h>
>> -#include <drm/drmP.h>
>> -#include "radeon.h"
>> -#include "cikd.h"
>> -#include "cik_reg.h"
>> -#include "radeon_kfd.h"
>> -#include "radeon_ucode.h"
>> -#include <linux/firmware.h>
>> -#include "cik_structs.h"
>> -
>> -#define CIK_PIPE_PER_MEC       (4)
>> -
>> -static const uint32_t watchRegs[MAX_WATCH_ADDRESSES * ADDRESS_WATCH_REG_MAX] = {
>> -       TCP_WATCH0_ADDR_H, TCP_WATCH0_ADDR_L, TCP_WATCH0_CNTL,
>> -       TCP_WATCH1_ADDR_H, TCP_WATCH1_ADDR_L, TCP_WATCH1_CNTL,
>> -       TCP_WATCH2_ADDR_H, TCP_WATCH2_ADDR_L, TCP_WATCH2_CNTL,
>> -       TCP_WATCH3_ADDR_H, TCP_WATCH3_ADDR_L, TCP_WATCH3_CNTL
>> -};
>> -
>> -struct kgd_mem {
>> -       struct radeon_bo *bo;
>> -       uint64_t gpu_addr;
>> -       void *cpu_ptr;
>> -};
>> -
>> -
>> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
>> -                       void **mem_obj, uint64_t *gpu_addr,
>> -                       void **cpu_ptr);
>> -
>> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
>> -
>> -static uint64_t get_vmem_size(struct kgd_dev *kgd);
>> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd);
>> -
>> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
>> -
>> -static int alloc_pasid(unsigned int bits);
>> -static void free_pasid(unsigned int pasid);
>> -
>> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
>> -
>> -/*
>> - * Register access functions
>> - */
>> -
>> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
>> -               uint32_t sh_mem_config, uint32_t sh_mem_ape1_base,
>> -               uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases);
>> -
>> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
>> -                                       unsigned int vmid);
>> -
>> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
>> -                               uint32_t hpd_size, uint64_t hpd_gpu_addr);
>> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
>> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
>> -                       uint32_t queue_id, uint32_t __user *wptr,
>> -                       uint32_t wptr_shift, uint32_t wptr_mask,
>> -                       struct mm_struct *mm);
>> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
>> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
>> -                               uint32_t pipe_id, uint32_t queue_id);
>> -
>> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
>> -                               unsigned int timeout, uint32_t pipe_id,
>> -                               uint32_t queue_id);
>> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
>> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
>> -                               unsigned int timeout);
>> -static int kgd_address_watch_disable(struct kgd_dev *kgd);
>> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
>> -                                       unsigned int watch_point_id,
>> -                                       uint32_t cntl_val,
>> -                                       uint32_t addr_hi,
>> -                                       uint32_t addr_lo);
>> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
>> -                                       uint32_t gfx_index_val,
>> -                                       uint32_t sq_cmd);
>> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
>> -                                       unsigned int watch_point_id,
>> -                                       unsigned int reg_offset);
>> -
>> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid);
>> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
>> -                                                       uint8_t vmid);
>> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid);
>> -
>> -static const struct kfd2kgd_calls kfd2kgd = {
>> -       .init_gtt_mem_allocation = alloc_gtt_mem,
>> -       .free_gtt_mem = free_gtt_mem,
>> -       .get_vmem_size = get_vmem_size,
>> -       .get_gpu_clock_counter = get_gpu_clock_counter,
>> -       .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
>> -       .alloc_pasid = alloc_pasid,
>> -       .free_pasid = free_pasid,
>> -       .program_sh_mem_settings = kgd_program_sh_mem_settings,
>> -       .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
>> -       .init_pipeline = kgd_init_pipeline,
>> -       .init_interrupts = kgd_init_interrupts,
>> -       .hqd_load = kgd_hqd_load,
>> -       .hqd_sdma_load = kgd_hqd_sdma_load,
>> -       .hqd_is_occupied = kgd_hqd_is_occupied,
>> -       .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
>> -       .hqd_destroy = kgd_hqd_destroy,
>> -       .hqd_sdma_destroy = kgd_hqd_sdma_destroy,
>> -       .address_watch_disable = kgd_address_watch_disable,
>> -       .address_watch_execute = kgd_address_watch_execute,
>> -       .wave_control_execute = kgd_wave_control_execute,
>> -       .address_watch_get_offset = kgd_address_watch_get_offset,
>> -       .get_atc_vmid_pasid_mapping_pasid = get_atc_vmid_pasid_mapping_pasid,
>> -       .get_atc_vmid_pasid_mapping_valid = get_atc_vmid_pasid_mapping_valid,
>> -       .write_vmid_invalidate_request = write_vmid_invalidate_request,
>> -       .get_fw_version = get_fw_version
>> -};
>> -
>> -static const struct kgd2kfd_calls *kgd2kfd;
>> -
>> -int radeon_kfd_init(void)
>> -{
>> -       int ret;
>> -
>> -#if defined(CONFIG_HSA_AMD_MODULE)
>> -       int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
>> -
>> -       kgd2kfd_init_p = symbol_request(kgd2kfd_init);
>> -
>> -       if (kgd2kfd_init_p == NULL)
>> -               return -ENOENT;
>> -
>> -       ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd);
>> -       if (ret) {
>> -               symbol_put(kgd2kfd_init);
>> -               kgd2kfd = NULL;
>> -       }
>> -
>> -#elif defined(CONFIG_HSA_AMD)
>> -       ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
>> -       if (ret)
>> -               kgd2kfd = NULL;
>> -
>> -#else
>> -       ret = -ENOENT;
>> -#endif
>> -
>> -       return ret;
>> -}
>> -
>> -void radeon_kfd_fini(void)
>> -{
>> -       if (kgd2kfd) {
>> -               kgd2kfd->exit();
>> -               symbol_put(kgd2kfd_init);
>> -       }
>> -}
>> -
>> -void radeon_kfd_device_probe(struct radeon_device *rdev)
>> -{
>> -       if (kgd2kfd)
>> -               rdev->kfd = kgd2kfd->probe((struct kgd_dev *)rdev,
>> -                       rdev->pdev, &kfd2kgd);
>> -}
>> -
>> -void radeon_kfd_device_init(struct radeon_device *rdev)
>> -{
>> -       int i, queue, pipe, mec;
>> -
>> -       if (rdev->kfd) {
>> -               struct kgd2kfd_shared_resources gpu_resources = {
>> -                       .compute_vmid_bitmap = 0xFF00,
>> -                       .num_pipe_per_mec = 4,
>> -                       .num_queue_per_pipe = 8
>> -               };
>> -
>> -               bitmap_zero(gpu_resources.queue_bitmap, KGD_MAX_QUEUES);
>> -
>> -               for (i = 0; i < KGD_MAX_QUEUES; ++i) {
>> -                       queue = i % gpu_resources.num_queue_per_pipe;
>> -                       pipe = (i / gpu_resources.num_queue_per_pipe)
>> -                               % gpu_resources.num_pipe_per_mec;
>> -                       mec = (i / gpu_resources.num_queue_per_pipe)
>> -                               / gpu_resources.num_pipe_per_mec;
>> -
>> -                       if (mec == 0 && pipe > 0)
>> -                               set_bit(i, gpu_resources.queue_bitmap);
>> -               }
>> -
>> -               radeon_doorbell_get_kfd_info(rdev,
>> -                               &gpu_resources.doorbell_physical_address,
>> -                               &gpu_resources.doorbell_aperture_size,
>> -                               &gpu_resources.doorbell_start_offset);
>> -
>> -               kgd2kfd->device_init(rdev->kfd, &gpu_resources);
>> -       }
>> -}
>> -
>> -void radeon_kfd_device_fini(struct radeon_device *rdev)
>> -{
>> -       if (rdev->kfd) {
>> -               kgd2kfd->device_exit(rdev->kfd);
>> -               rdev->kfd = NULL;
>> -       }
>> -}
>> -
>> -void radeon_kfd_interrupt(struct radeon_device *rdev, const void *ih_ring_entry)
>> -{
>> -       if (rdev->kfd)
>> -               kgd2kfd->interrupt(rdev->kfd, ih_ring_entry);
>> -}
>> -
>> -void radeon_kfd_suspend(struct radeon_device *rdev)
>> -{
>> -       if (rdev->kfd)
>> -               kgd2kfd->suspend(rdev->kfd);
>> -}
>> -
>> -int radeon_kfd_resume(struct radeon_device *rdev)
>> -{
>> -       int r = 0;
>> -
>> -       if (rdev->kfd)
>> -               r = kgd2kfd->resume(rdev->kfd);
>> -
>> -       return r;
>> -}
>> -
>> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
>> -                       void **mem_obj, uint64_t *gpu_addr,
>> -                       void **cpu_ptr)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>> -       struct kgd_mem **mem = (struct kgd_mem **) mem_obj;
>> -       int r;
>> -
>> -       BUG_ON(kgd == NULL);
>> -       BUG_ON(gpu_addr == NULL);
>> -       BUG_ON(cpu_ptr == NULL);
>> -
>> -       *mem = kmalloc(sizeof(struct kgd_mem), GFP_KERNEL);
>> -       if ((*mem) == NULL)
>> -               return -ENOMEM;
>> -
>> -       r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_GTT,
>> -                               RADEON_GEM_GTT_WC, NULL, NULL, &(*mem)->bo);
>> -       if (r) {
>> -               dev_err(rdev->dev,
>> -                       "failed to allocate BO for amdkfd (%d)\n", r);
>> -               return r;
>> -       }
>> -
>> -       /* map the buffer */
>> -       r = radeon_bo_reserve((*mem)->bo, true);
>> -       if (r) {
>> -               dev_err(rdev->dev, "(%d) failed to reserve bo for amdkfd\n", r);
>> -               goto allocate_mem_reserve_bo_failed;
>> -       }
>> -
>> -       r = radeon_bo_pin((*mem)->bo, RADEON_GEM_DOMAIN_GTT,
>> -                               &(*mem)->gpu_addr);
>> -       if (r) {
>> -               dev_err(rdev->dev, "(%d) failed to pin bo for amdkfd\n", r);
>> -               goto allocate_mem_pin_bo_failed;
>> -       }
>> -       *gpu_addr = (*mem)->gpu_addr;
>> -
>> -       r = radeon_bo_kmap((*mem)->bo, &(*mem)->cpu_ptr);
>> -       if (r) {
>> -               dev_err(rdev->dev,
>> -                       "(%d) failed to map bo to kernel for amdkfd\n", r);
>> -               goto allocate_mem_kmap_bo_failed;
>> -       }
>> -       *cpu_ptr = (*mem)->cpu_ptr;
>> -
>> -       radeon_bo_unreserve((*mem)->bo);
>> -
>> -       return 0;
>> -
>> -allocate_mem_kmap_bo_failed:
>> -       radeon_bo_unpin((*mem)->bo);
>> -allocate_mem_pin_bo_failed:
>> -       radeon_bo_unreserve((*mem)->bo);
>> -allocate_mem_reserve_bo_failed:
>> -       radeon_bo_unref(&(*mem)->bo);
>> -
>> -       return r;
>> -}
>> -
>> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
>> -{
>> -       struct kgd_mem *mem = (struct kgd_mem *) mem_obj;
>> -
>> -       BUG_ON(mem == NULL);
>> -
>> -       radeon_bo_reserve(mem->bo, true);
>> -       radeon_bo_kunmap(mem->bo);
>> -       radeon_bo_unpin(mem->bo);
>> -       radeon_bo_unreserve(mem->bo);
>> -       radeon_bo_unref(&(mem->bo));
>> -       kfree(mem);
>> -}
>> -
>> -static uint64_t get_vmem_size(struct kgd_dev *kgd)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>> -
>> -       BUG_ON(kgd == NULL);
>> -
>> -       return rdev->mc.real_vram_size;
>> -}
>> -
>> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>> -
>> -       return rdev->asic->get_gpu_clock_counter(rdev);
>> -}
>> -
>> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>> -
>> -       /* The sclk is in quantas of 10kHz */
>> -       return rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk / 100;
>> -}
>> -
>> -/*
>> - * PASID manager
>> - */
>> -static DEFINE_IDA(pasid_ida);
>> -
>> -static int alloc_pasid(unsigned int bits)
>> -{
>> -       int pasid = -EINVAL;
>> -
>> -       for (bits = min(bits, 31U); bits > 0; bits--) {
>> -               pasid = ida_simple_get(&pasid_ida,
>> -                                      1U << (bits - 1), 1U << bits,
>> -                                      GFP_KERNEL);
>> -               if (pasid != -ENOSPC)
>> -                       break;
>> -       }
>> -
>> -       return pasid;
>> -}
>> -
>> -static void free_pasid(unsigned int pasid)
>> -{
>> -       ida_simple_remove(&pasid_ida, pasid);
>> -}
>> -
>> -static inline struct radeon_device *get_radeon_device(struct kgd_dev *kgd)
>> -{
>> -       return (struct radeon_device *)kgd;
>> -}
>> -
>> -static void write_register(struct kgd_dev *kgd, uint32_t offset, uint32_t value)
>> -{
>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>> -
>> -       writel(value, (void __iomem *)(rdev->rmmio + offset));
>> -}
>> -
>> -static uint32_t read_register(struct kgd_dev *kgd, uint32_t offset)
>> -{
>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>> -
>> -       return readl((void __iomem *)(rdev->rmmio + offset));
>> -}
>> -
>> -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
>> -                       uint32_t queue, uint32_t vmid)
>> -{
>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>> -       uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
>> -
>> -       mutex_lock(&rdev->srbm_mutex);
>> -       write_register(kgd, SRBM_GFX_CNTL, value);
>> -}
>> -
>> -static void unlock_srbm(struct kgd_dev *kgd)
>> -{
>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>> -
>> -       write_register(kgd, SRBM_GFX_CNTL, 0);
>> -       mutex_unlock(&rdev->srbm_mutex);
>> -}
>> -
>> -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
>> -                               uint32_t queue_id)
>> -{
>> -       uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1;
>> -       uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC);
>> -
>> -       lock_srbm(kgd, mec, pipe, queue_id, 0);
>> -}
>> -
>> -static void release_queue(struct kgd_dev *kgd)
>> -{
>> -       unlock_srbm(kgd);
>> -}
>> -
>> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
>> -                                       uint32_t sh_mem_config,
>> -                                       uint32_t sh_mem_ape1_base,
>> -                                       uint32_t sh_mem_ape1_limit,
>> -                                       uint32_t sh_mem_bases)
>> -{
>> -       lock_srbm(kgd, 0, 0, 0, vmid);
>> -
>> -       write_register(kgd, SH_MEM_CONFIG, sh_mem_config);
>> -       write_register(kgd, SH_MEM_APE1_BASE, sh_mem_ape1_base);
>> -       write_register(kgd, SH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
>> -       write_register(kgd, SH_MEM_BASES, sh_mem_bases);
>> -
>> -       unlock_srbm(kgd);
>> -}
>> -
>> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
>> -                                       unsigned int vmid)
>> -{
>> -       /*
>> -        * We have to assume that there is no outstanding mapping.
>> -        * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0
>> -        * because a mapping is in progress or because a mapping finished and
>> -        * the SW cleared it.
>> -        * So the protocol is to always wait & clear.
>> -        */
>> -       uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
>> -                                       ATC_VMID_PASID_MAPPING_VALID_MASK;
>> -
>> -       write_register(kgd, ATC_VMID0_PASID_MAPPING + vmid*sizeof(uint32_t),
>> -                       pasid_mapping);
>> -
>> -       while (!(read_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS) &
>> -                                                               (1U << vmid)))
>> -               cpu_relax();
>> -       write_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U << vmid);
>> -
>> -       /* Mapping vmid to pasid also for IH block */
>> -       write_register(kgd, IH_VMID_0_LUT + vmid * sizeof(uint32_t),
>> -                       pasid_mapping);
>> -
>> -       return 0;
>> -}
>> -
>> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
>> -                               uint32_t hpd_size, uint64_t hpd_gpu_addr)
>> -{
>> -       /* nothing to do here */
>> -       return 0;
>> -}
>> -
>> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
>> -{
>> -       uint32_t mec;
>> -       uint32_t pipe;
>> -
>> -       mec = (pipe_id / CIK_PIPE_PER_MEC) + 1;
>> -       pipe = (pipe_id % CIK_PIPE_PER_MEC);
>> -
>> -       lock_srbm(kgd, mec, pipe, 0, 0);
>> -
>> -       write_register(kgd, CPC_INT_CNTL,
>> -                       TIME_STAMP_INT_ENABLE | OPCODE_ERROR_INT_ENABLE);
>> -
>> -       unlock_srbm(kgd);
>> -
>> -       return 0;
>> -}
>> -
>> -static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers *m)
>> -{
>> -       uint32_t retval;
>> -
>> -       retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET +
>> -                       m->sdma_queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
>> -
>> -       pr_debug("kfd: sdma base address: 0x%x\n", retval);
>> -
>> -       return retval;
>> -}
>> -
>> -static inline struct cik_mqd *get_mqd(void *mqd)
>> -{
>> -       return (struct cik_mqd *)mqd;
>> -}
>> -
>> -static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
>> -{
>> -       return (struct cik_sdma_rlc_registers *)mqd;
>> -}
>> -
>> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
>> -                       uint32_t queue_id, uint32_t __user *wptr,
>> -                       uint32_t wptr_shift, uint32_t wptr_mask,
>> -                       struct mm_struct *mm)
>> -{
>> -       uint32_t wptr_shadow, is_wptr_shadow_valid;
>> -       struct cik_mqd *m;
>> -
>> -       m = get_mqd(mqd);
>> -
>> -       is_wptr_shadow_valid = !get_user(wptr_shadow, wptr);
>> -
>> -       acquire_queue(kgd, pipe_id, queue_id);
>> -       write_register(kgd, CP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo);
>> -       write_register(kgd, CP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi);
>> -       write_register(kgd, CP_MQD_CONTROL, m->cp_mqd_control);
>> -
>> -       write_register(kgd, CP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo);
>> -       write_register(kgd, CP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi);
>> -       write_register(kgd, CP_HQD_PQ_CONTROL, m->cp_hqd_pq_control);
>> -
>> -       write_register(kgd, CP_HQD_IB_CONTROL, m->cp_hqd_ib_control);
>> -       write_register(kgd, CP_HQD_IB_BASE_ADDR, m->cp_hqd_ib_base_addr_lo);
>> -       write_register(kgd, CP_HQD_IB_BASE_ADDR_HI, m->cp_hqd_ib_base_addr_hi);
>> -
>> -       write_register(kgd, CP_HQD_IB_RPTR, m->cp_hqd_ib_rptr);
>> -
>> -       write_register(kgd, CP_HQD_PERSISTENT_STATE,
>> -                       m->cp_hqd_persistent_state);
>> -       write_register(kgd, CP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd);
>> -       write_register(kgd, CP_HQD_MSG_TYPE, m->cp_hqd_msg_type);
>> -
>> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_LO,
>> -                       m->cp_hqd_atomic0_preop_lo);
>> -
>> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_HI,
>> -                       m->cp_hqd_atomic0_preop_hi);
>> -
>> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_LO,
>> -                       m->cp_hqd_atomic1_preop_lo);
>> -
>> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_HI,
>> -                       m->cp_hqd_atomic1_preop_hi);
>> -
>> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR,
>> -                       m->cp_hqd_pq_rptr_report_addr_lo);
>> -
>> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR_HI,
>> -                       m->cp_hqd_pq_rptr_report_addr_hi);
>> -
>> -       write_register(kgd, CP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr);
>> -
>> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR,
>> -                       m->cp_hqd_pq_wptr_poll_addr_lo);
>> -
>> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR_HI,
>> -                       m->cp_hqd_pq_wptr_poll_addr_hi);
>> -
>> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL,
>> -                       m->cp_hqd_pq_doorbell_control);
>> -
>> -       write_register(kgd, CP_HQD_VMID, m->cp_hqd_vmid);
>> -
>> -       write_register(kgd, CP_HQD_QUANTUM, m->cp_hqd_quantum);
>> -
>> -       write_register(kgd, CP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority);
>> -       write_register(kgd, CP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority);
>> -
>> -       write_register(kgd, CP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr);
>> -
>> -       if (is_wptr_shadow_valid)
>> -               write_register(kgd, CP_HQD_PQ_WPTR, wptr_shadow);
>> -
>> -       write_register(kgd, CP_HQD_ACTIVE, m->cp_hqd_active);
>> -       release_queue(kgd);
>> -
>> -       return 0;
>> -}
>> -
>> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
>> -{
>> -       struct cik_sdma_rlc_registers *m;
>> -       uint32_t sdma_base_addr;
>> -
>> -       m = get_sdma_mqd(mqd);
>> -       sdma_base_addr = get_sdma_base_addr(m);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_VIRTUAL_ADDR,
>> -                       m->sdma_rlc_virtual_addr);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE,
>> -                       m->sdma_rlc_rb_base);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE_HI,
>> -                       m->sdma_rlc_rb_base_hi);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_LO,
>> -                       m->sdma_rlc_rb_rptr_addr_lo);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_HI,
>> -                       m->sdma_rlc_rb_rptr_addr_hi);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_DOORBELL,
>> -                       m->sdma_rlc_doorbell);
>> -
>> -       write_register(kgd,
>> -                       sdma_base_addr + SDMA0_RLC0_RB_CNTL,
>> -                       m->sdma_rlc_rb_cntl);
>> -
>> -       return 0;
>> -}
>> -
>> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
>> -                               uint32_t pipe_id, uint32_t queue_id)
>> -{
>> -       uint32_t act;
>> -       bool retval = false;
>> -       uint32_t low, high;
>> -
>> -       acquire_queue(kgd, pipe_id, queue_id);
>> -       act = read_register(kgd, CP_HQD_ACTIVE);
>> -       if (act) {
>> -               low = lower_32_bits(queue_address >> 8);
>> -               high = upper_32_bits(queue_address >> 8);
>> -
>> -               if (low == read_register(kgd, CP_HQD_PQ_BASE) &&
>> -                               high == read_register(kgd, CP_HQD_PQ_BASE_HI))
>> -                       retval = true;
>> -       }
>> -       release_queue(kgd);
>> -       return retval;
>> -}
>> -
>> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
>> -{
>> -       struct cik_sdma_rlc_registers *m;
>> -       uint32_t sdma_base_addr;
>> -       uint32_t sdma_rlc_rb_cntl;
>> -
>> -       m = get_sdma_mqd(mqd);
>> -       sdma_base_addr = get_sdma_base_addr(m);
>> -
>> -       sdma_rlc_rb_cntl = read_register(kgd,
>> -                                       sdma_base_addr + SDMA0_RLC0_RB_CNTL);
>> -
>> -       if (sdma_rlc_rb_cntl & SDMA_RB_ENABLE)
>> -               return true;
>> -
>> -       return false;
>> -}
>> -
>> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
>> -                               unsigned int timeout, uint32_t pipe_id,
>> -                               uint32_t queue_id)
>> -{
>> -       uint32_t temp;
>> -
>> -       acquire_queue(kgd, pipe_id, queue_id);
>> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, 0);
>> -
>> -       write_register(kgd, CP_HQD_DEQUEUE_REQUEST, reset_type);
>> -
>> -       while (true) {
>> -               temp = read_register(kgd, CP_HQD_ACTIVE);
>> -               if (temp & 0x1)
>> -                       break;
>> -               if (timeout == 0) {
>> -                       pr_err("kfd: cp queue preemption time out (%dms)\n",
>> -                               temp);
>> -                       release_queue(kgd);
>> -                       return -ETIME;
>> -               }
>> -               msleep(20);
>> -               timeout -= 20;
>> -       }
>> -
>> -       release_queue(kgd);
>> -       return 0;
>> -}
>> -
>> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
>> -                               unsigned int timeout)
>> -{
>> -       struct cik_sdma_rlc_registers *m;
>> -       uint32_t sdma_base_addr;
>> -       uint32_t temp;
>> -
>> -       m = get_sdma_mqd(mqd);
>> -       sdma_base_addr = get_sdma_base_addr(m);
>> -
>> -       temp = read_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL);
>> -       temp = temp & ~SDMA_RB_ENABLE;
>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL, temp);
>> -
>> -       while (true) {
>> -               temp = read_register(kgd, sdma_base_addr +
>> -                                               SDMA0_RLC0_CONTEXT_STATUS);
>> -               if (temp & SDMA_RLC_IDLE)
>> -                       break;
>> -               if (timeout == 0)
>> -                       return -ETIME;
>> -               msleep(20);
>> -               timeout -= 20;
>> -       }
>> -
>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_DOORBELL, 0);
>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_RPTR, 0);
>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_WPTR, 0);
>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_BASE, 0);
>> -
>> -       return 0;
>> -}
>> -
>> -static int kgd_address_watch_disable(struct kgd_dev *kgd)
>> -{
>> -       union TCP_WATCH_CNTL_BITS cntl;
>> -       unsigned int i;
>> -
>> -       cntl.u32All = 0;
>> -
>> -       cntl.bitfields.valid = 0;
>> -       cntl.bitfields.mask = ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK;
>> -       cntl.bitfields.atc = 1;
>> -
>> -       /* Turning off this address until we set all the registers */
>> -       for (i = 0; i < MAX_WATCH_ADDRESSES; i++)
>> -               write_register(kgd,
>> -                               watchRegs[i * ADDRESS_WATCH_REG_MAX +
>> -                                       ADDRESS_WATCH_REG_CNTL],
>> -                               cntl.u32All);
>> -
>> -       return 0;
>> -}
>> -
>> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
>> -                                       unsigned int watch_point_id,
>> -                                       uint32_t cntl_val,
>> -                                       uint32_t addr_hi,
>> -                                       uint32_t addr_lo)
>> -{
>> -       union TCP_WATCH_CNTL_BITS cntl;
>> -
>> -       cntl.u32All = cntl_val;
>> -
>> -       /* Turning off this watch point until we set all the registers */
>> -       cntl.bitfields.valid = 0;
>> -       write_register(kgd,
>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
>> -                               ADDRESS_WATCH_REG_CNTL],
>> -                       cntl.u32All);
>> -
>> -       write_register(kgd,
>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
>> -                               ADDRESS_WATCH_REG_ADDR_HI],
>> -                       addr_hi);
>> -
>> -       write_register(kgd,
>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
>> -                               ADDRESS_WATCH_REG_ADDR_LO],
>> -                       addr_lo);
>> -
>> -       /* Enable the watch point */
>> -       cntl.bitfields.valid = 1;
>> -
>> -       write_register(kgd,
>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
>> -                               ADDRESS_WATCH_REG_CNTL],
>> -                       cntl.u32All);
>> -
>> -       return 0;
>> -}
>> -
>> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
>> -                                       uint32_t gfx_index_val,
>> -                                       uint32_t sq_cmd)
>> -{
>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>> -       uint32_t data;
>> -
>> -       mutex_lock(&rdev->grbm_idx_mutex);
>> -
>> -       write_register(kgd, GRBM_GFX_INDEX, gfx_index_val);
>> -       write_register(kgd, SQ_CMD, sq_cmd);
>> -
>> -       /*  Restore the GRBM_GFX_INDEX register  */
>> -
>> -       data = INSTANCE_BROADCAST_WRITES | SH_BROADCAST_WRITES |
>> -               SE_BROADCAST_WRITES;
>> -
>> -       write_register(kgd, GRBM_GFX_INDEX, data);
>> -
>> -       mutex_unlock(&rdev->grbm_idx_mutex);
>> -
>> -       return 0;
>> -}
>> -
>> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
>> -                                       unsigned int watch_point_id,
>> -                                       unsigned int reg_offset)
>> -{
>> -       return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset]
>> -               / 4;
>> -}
>> -
>> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, uint8_t vmid)
>> -{
>> -       uint32_t reg;
>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>> -
>> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
>> -       return reg & ATC_VMID_PASID_MAPPING_VALID_MASK;
>> -}
>> -
>> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
>> -                                                       uint8_t vmid)
>> -{
>> -       uint32_t reg;
>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>> -
>> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
>> -       return reg & ATC_VMID_PASID_MAPPING_PASID_MASK;
>> -}
>> -
>> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>> -
>> -       return WREG32(VM_INVALIDATE_REQUEST, 1 << vmid);
>> -}
>> -
>> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
>> -{
>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>> -       const union radeon_firmware_header *hdr;
>> -
>> -       BUG_ON(kgd == NULL || rdev->mec_fw == NULL);
>> -
>> -       switch (type) {
>> -       case KGD_ENGINE_PFP:
>> -               hdr = (const union radeon_firmware_header *) rdev->pfp_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_ME:
>> -               hdr = (const union radeon_firmware_header *) rdev->me_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_CE:
>> -               hdr = (const union radeon_firmware_header *) rdev->ce_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_MEC1:
>> -               hdr = (const union radeon_firmware_header *) rdev->mec_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_MEC2:
>> -               hdr = (const union radeon_firmware_header *)
>> -                                                       rdev->mec2_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_RLC:
>> -               hdr = (const union radeon_firmware_header *) rdev->rlc_fw->data;
>> -               break;
>> -
>> -       case KGD_ENGINE_SDMA1:
>> -       case KGD_ENGINE_SDMA2:
>> -               hdr = (const union radeon_firmware_header *)
>> -                                                       rdev->sdma_fw->data;
>> -               break;
>> -
>> -       default:
>> -               return 0;
>> -       }
>> -
>> -       if (hdr == NULL)
>> -               return 0;
>> -
>> -       /* Only 12 bit in use*/
>> -       return hdr->common.ucode_version;
>> -}
>> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.h b/drivers/gpu/drm/radeon/radeon_kfd.h
>> deleted file mode 100644
>> index 9df1fea..0000000
>> --- a/drivers/gpu/drm/radeon/radeon_kfd.h
>> +++ /dev/null
>> @@ -1,47 +0,0 @@
>> -/*
>> - * Copyright 2014 Advanced Micro Devices, Inc.
>> - *
>> - * Permission is hereby granted, free of charge, to any person obtaining a
>> - * copy of this software and associated documentation files (the "Software"),
>> - * to deal in the Software without restriction, including without limitation
>> - * the rights to use, copy, modify, merge, publish, distribute, sublicense,
>> - * and/or sell copies of the Software, and to permit persons to whom the
>> - * Software is furnished to do so, subject to the following conditions:
>> - *
>> - * The above copyright notice and this permission notice shall be included in
>> - * all copies or substantial portions of the Software.
>> - *
>> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
>> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
>> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
>> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>> - * OTHER DEALINGS IN THE SOFTWARE.
>> - */
>> -
>> -/*
>> - * radeon_kfd.h defines the private interface between the
>> - * AMD kernel graphics drivers and the AMD KFD.
>> - */
>> -
>> -#ifndef RADEON_KFD_H_INCLUDED
>> -#define RADEON_KFD_H_INCLUDED
>> -
>> -#include <linux/types.h>
>> -#include "kgd_kfd_interface.h"
>> -
>> -struct radeon_device;
>> -
>> -int radeon_kfd_init(void);
>> -void radeon_kfd_fini(void);
>> -
>> -void radeon_kfd_suspend(struct radeon_device *rdev);
>> -int radeon_kfd_resume(struct radeon_device *rdev);
>> -void radeon_kfd_interrupt(struct radeon_device *rdev,
>> -                       const void *ih_ring_entry);
>> -void radeon_kfd_device_probe(struct radeon_device *rdev);
>> -void radeon_kfd_device_init(struct radeon_device *rdev);
>> -void radeon_kfd_device_fini(struct radeon_device *rdev);
>> -
>> -#endif /* RADEON_KFD_H_INCLUDED */
>> diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
>> index dfee8f7..cde037f 100644
>> --- a/drivers/gpu/drm/radeon/radeon_kms.c
>> +++ b/drivers/gpu/drm/radeon/radeon_kms.c
>> @@ -34,8 +34,6 @@
>>   #include <linux/slab.h>
>>   #include <linux/pm_runtime.h>
>>
>> -#include "radeon_kfd.h"
>> -
>>   #if defined(CONFIG_VGA_SWITCHEROO)
>>   bool radeon_has_atpx(void);
>>   #else
>> @@ -68,8 +66,6 @@ void radeon_driver_unload_kms(struct drm_device *dev)
>>                  pm_runtime_forbid(dev->dev);
>>          }
>>
>> -       radeon_kfd_device_fini(rdev);
>> -
>>          radeon_acpi_fini(rdev);
>>
>>          radeon_modeset_fini(rdev);
>> @@ -174,9 +170,6 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
>>                                  "Error during ACPI methods call\n");
>>          }
>>
>> -       radeon_kfd_device_probe(rdev);
>> -       radeon_kfd_device_init(rdev);
>> -
>>          if (radeon_is_px(dev)) {
>>                  pm_runtime_use_autosuspend(dev->dev);
>>                  pm_runtime_set_autosuspend_delay(dev->dev, 5000);
>> --
>> 2.7.4
>>
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
ok, taken to -next.

On Tue, Oct 31, 2017 at 4:56 PM, Christian König <
ckoenig.leichtzumerken@gmail.com> wrote:

> Am 31.10.2017 um 11:44 schrieb Oded Gabbay:
>
>> Don't have any strong objection, but I just want to ask if current
>> users can just move to using amdgpu on KV and their current usermode
>> stack will work as usual.
>>
>
> Yes, I think so.
>
> btw, are there any "current users" that you are aware of ?
>>
>
> Not the slightest idea, but from Felix comments at least the current user
> mode stack + old kernel won't work.
>
> Regards,
> Christian.
>
>
>
>>
>> On Mon, Oct 30, 2017 at 3:16 PM, Christian König
>> <deathsimple@vodafone.de> wrote:
>>
>>> From: Christian König <christian.koenig@amd.com>
>>>
>>> To quote Felix: "For testing KV with current user mode stack, please use
>>> amdgpu. I don't expect this to work with radeon and I'm not planning to
>>> spend
>>> any effort on making radeon work with a current user mode stack."
>>>
>>> Only compile tested, but should be straight forward.
>>>
>>> Signed-off-by: Christian König <christian.koenig@amd.com>
>>> ---
>>>   drivers/gpu/drm/amd/amdkfd/Kconfig  |   2 +-
>>>   drivers/gpu/drm/radeon/Makefile     |   3 +-
>>>   drivers/gpu/drm/radeon/cik.c        |  14 +-
>>>   drivers/gpu/drm/radeon/cikd.h       |   2 -
>>>   drivers/gpu/drm/radeon/radeon.h     |   3 -
>>>   drivers/gpu/drm/radeon/radeon_drv.c |  10 -
>>>   drivers/gpu/drm/radeon/radeon_kfd.c | 901
>>> ------------------------------------
>>>   drivers/gpu/drm/radeon/radeon_kfd.h |  47 --
>>>   drivers/gpu/drm/radeon/radeon_kms.c |   7 -
>>>   9 files changed, 4 insertions(+), 985 deletions(-)
>>>   delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.c
>>>   delete mode 100644 drivers/gpu/drm/radeon/radeon_kfd.h
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig
>>> b/drivers/gpu/drm/amd/amdkfd/Kconfig
>>> index e13c67c..bc5a294 100644
>>> --- a/drivers/gpu/drm/amd/amdkfd/Kconfig
>>> +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
>>> @@ -4,6 +4,6 @@
>>>
>>>   config HSA_AMD
>>>          tristate "HSA kernel driver for AMD GPU devices"
>>> -       depends on (DRM_RADEON || DRM_AMDGPU) && AMD_IOMMU_V2 && X86_64
>>> +       depends on DRM_AMDGPU && AMD_IOMMU_V2 && X86_64
>>>          help
>>>            Enable this if you want to use HSA features on AMD GPU
>>> devices.
>>> diff --git a/drivers/gpu/drm/radeon/Makefile
>>> b/drivers/gpu/drm/radeon/Makefile
>>> index be16c63..cf3e598 100644
>>> --- a/drivers/gpu/drm/radeon/Makefile
>>> +++ b/drivers/gpu/drm/radeon/Makefile
>>> @@ -102,8 +102,7 @@ radeon-y += \
>>>   radeon-y += \
>>>          radeon_vce.o \
>>>          vce_v1_0.o \
>>> -       vce_v2_0.o \
>>> -       radeon_kfd.o
>>> +       vce_v2_0.o
>>>
>>>   radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
>>>   radeon-$(CONFIG_ACPI) += radeon_acpi.o
>>> diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c
>>> index 3cb6c55..898f9a0 100644
>>> --- a/drivers/gpu/drm/radeon/cik.c
>>> +++ b/drivers/gpu/drm/radeon/cik.c
>>> @@ -33,7 +33,6 @@
>>>   #include "cik_blit_shaders.h"
>>>   #include "radeon_ucode.h"
>>>   #include "clearstate_ci.h"
>>> -#include "radeon_kfd.h"
>>>
>>>   #define SH_MEM_CONFIG_GFX_DEFAULT \
>>>          ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
>>> @@ -5684,10 +5683,9 @@ int cik_vm_init(struct radeon_device *rdev)
>>>          /*
>>>           * number of VMs
>>>           * VMID 0 is reserved for System
>>> -        * radeon graphics/compute will use VMIDs 1-7
>>> -        * amdkfd will use VMIDs 8-15
>>> +        * radeon graphics/compute will use VMIDs 1-15
>>>           */
>>> -       rdev->vm_manager.nvm = RADEON_NUM_OF_VMIDS;
>>> +       rdev->vm_manager.nvm = 16;
>>>          /* base offset of vram pages */
>>>          if (rdev->flags & RADEON_IS_IGP) {
>>>                  u64 tmp = RREG32(MC_VM_FB_OFFSET);
>>> @@ -7589,9 +7587,6 @@ int cik_irq_process(struct radeon_device *rdev)
>>>                  /* wptr/rptr are in bytes! */
>>>                  ring_index = rptr / 4;
>>>
>>> -               radeon_kfd_interrupt(rdev,
>>> -                               (const void *)
>>> &rdev->ih.ring[ring_index]);
>>> -
>>>                  src_id =  le32_to_cpu(rdev->ih.ring[ring_index]) &
>>> 0xff;
>>>                  src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1])
>>> & 0xfffffff;
>>>                  ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) &
>>> 0xff;
>>> @@ -8486,10 +8481,6 @@ static int cik_startup(struct radeon_device *rdev)
>>>          if (r)
>>>                  return r;
>>>
>>> -       r = radeon_kfd_resume(rdev);
>>> -       if (r)
>>> -               return r;
>>> -
>>>          return 0;
>>>   }
>>>
>>> @@ -8538,7 +8529,6 @@ int cik_resume(struct radeon_device *rdev)
>>>    */
>>>   int cik_suspend(struct radeon_device *rdev)
>>>   {
>>> -       radeon_kfd_suspend(rdev);
>>>          radeon_pm_suspend(rdev);
>>>          radeon_audio_fini(rdev);
>>>          radeon_vm_manager_fini(rdev);
>>> diff --git a/drivers/gpu/drm/radeon/cikd.h
>>> b/drivers/gpu/drm/radeon/cikd.h
>>> index e210154..cda16fc 100644
>>> --- a/drivers/gpu/drm/radeon/cikd.h
>>> +++ b/drivers/gpu/drm/radeon/cikd.h
>>> @@ -30,8 +30,6 @@
>>>   #define CIK_RB_BITMAP_WIDTH_PER_SH     2
>>>   #define HAWAII_RB_BITMAP_WIDTH_PER_SH  4
>>>
>>> -#define RADEON_NUM_OF_VMIDS    8
>>> -
>>>   /* DIDT IND registers */
>>>   #define DIDT_SQ_CTRL0                                     0x0
>>>   #       define DIDT_CTRL_EN                               (1 << 0)
>>> diff --git a/drivers/gpu/drm/radeon/radeon.h
>>> b/drivers/gpu/drm/radeon/radeon.h
>>> index ec63bc5..d94741b 100644
>>> --- a/drivers/gpu/drm/radeon/radeon.h
>>> +++ b/drivers/gpu/drm/radeon/radeon.h
>>> @@ -2456,9 +2456,6 @@ struct radeon_device {
>>>          u64 vram_pin_size;
>>>          u64 gart_pin_size;
>>>
>>> -       /* amdkfd interface */
>>> -       struct kfd_dev          *kfd;
>>> -
>>>          struct mutex    mn_lock;
>>>          DECLARE_HASHTABLE(mn_hash, 7);
>>>   };
>>> diff --git a/drivers/gpu/drm/radeon/radeon_drv.c
>>> b/drivers/gpu/drm/radeon/radeon_drv.c
>>> index f4becad..31dd04f 100644
>>> --- a/drivers/gpu/drm/radeon/radeon_drv.c
>>> +++ b/drivers/gpu/drm/radeon/radeon_drv.c
>>> @@ -43,7 +43,6 @@
>>>   #include <drm/drm_fb_helper.h>
>>>
>>>   #include <drm/drm_crtc_helper.h>
>>> -#include "radeon_kfd.h"
>>>
>>>   /*
>>>    * KMS wrapper.
>>> @@ -338,14 +337,6 @@ static int radeon_pci_probe(struct pci_dev *pdev,
>>>   {
>>>          int ret;
>>>
>>> -       /*
>>> -        * Initialize amdkfd before starting radeon. If it was not
>>> loaded yet,
>>> -        * defer radeon probing
>>> -        */
>>> -       ret = radeon_kfd_init();
>>> -       if (ret == -EPROBE_DEFER)
>>> -               return ret;
>>> -
>>>          if (vga_switcheroo_client_probe_defer(pdev))
>>>                  return -EPROBE_DEFER;
>>>
>>> @@ -645,7 +636,6 @@ static int __init radeon_init(void)
>>>
>>>   static void __exit radeon_exit(void)
>>>   {
>>> -       radeon_kfd_fini();
>>>          pci_unregister_driver(pdriver);
>>>          radeon_unregister_atpx_handler();
>>>   }
>>> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c
>>> b/drivers/gpu/drm/radeon/radeon_kfd.c
>>> deleted file mode 100644
>>> index 385b4d7..0000000
>>> --- a/drivers/gpu/drm/radeon/radeon_kfd.c
>>> +++ /dev/null
>>> @@ -1,901 +0,0 @@
>>> -/*
>>> - * Copyright 2014 Advanced Micro Devices, Inc.
>>> - *
>>> - * Permission is hereby granted, free of charge, to any person
>>> obtaining a
>>> - * copy of this software and associated documentation files (the
>>> "Software"),
>>> - * to deal in the Software without restriction, including without
>>> limitation
>>> - * the rights to use, copy, modify, merge, publish, distribute,
>>> sublicense,
>>> - * and/or sell copies of the Software, and to permit persons to whom the
>>> - * Software is furnished to do so, subject to the following conditions:
>>> - *
>>> - * The above copyright notice and this permission notice shall be
>>> included in
>>> - * all copies or substantial portions of the Software.
>>> - *
>>> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>>> EXPRESS OR
>>> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>>> MERCHANTABILITY,
>>> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT
>>> SHALL
>>> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM,
>>> DAMAGES OR
>>> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
>>> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>>> - * OTHER DEALINGS IN THE SOFTWARE.
>>> - */
>>> -
>>> -#include <linux/module.h>
>>> -#include <linux/fdtable.h>
>>> -#include <linux/uaccess.h>
>>> -#include <drm/drmP.h>
>>> -#include "radeon.h"
>>> -#include "cikd.h"
>>> -#include "cik_reg.h"
>>> -#include "radeon_kfd.h"
>>> -#include "radeon_ucode.h"
>>> -#include <linux/firmware.h>
>>> -#include "cik_structs.h"
>>> -
>>> -#define CIK_PIPE_PER_MEC       (4)
>>> -
>>> -static const uint32_t watchRegs[MAX_WATCH_ADDRESSES *
>>> ADDRESS_WATCH_REG_MAX] = {
>>> -       TCP_WATCH0_ADDR_H, TCP_WATCH0_ADDR_L, TCP_WATCH0_CNTL,
>>> -       TCP_WATCH1_ADDR_H, TCP_WATCH1_ADDR_L, TCP_WATCH1_CNTL,
>>> -       TCP_WATCH2_ADDR_H, TCP_WATCH2_ADDR_L, TCP_WATCH2_CNTL,
>>> -       TCP_WATCH3_ADDR_H, TCP_WATCH3_ADDR_L, TCP_WATCH3_CNTL
>>> -};
>>> -
>>> -struct kgd_mem {
>>> -       struct radeon_bo *bo;
>>> -       uint64_t gpu_addr;
>>> -       void *cpu_ptr;
>>> -};
>>> -
>>> -
>>> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
>>> -                       void **mem_obj, uint64_t *gpu_addr,
>>> -                       void **cpu_ptr);
>>> -
>>> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
>>> -
>>> -static uint64_t get_vmem_size(struct kgd_dev *kgd);
>>> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd);
>>> -
>>> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
>>> -
>>> -static int alloc_pasid(unsigned int bits);
>>> -static void free_pasid(unsigned int pasid);
>>> -
>>> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum
>>> kgd_engine_type type);
>>> -
>>> -/*
>>> - * Register access functions
>>> - */
>>> -
>>> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t
>>> vmid,
>>> -               uint32_t sh_mem_config, uint32_t sh_mem_ape1_base,
>>> -               uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases);
>>> -
>>> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned
>>> int pasid,
>>> -                                       unsigned int vmid);
>>> -
>>> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
>>> -                               uint32_t hpd_size, uint64_t
>>> hpd_gpu_addr);
>>> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
>>> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t
>>> pipe_id,
>>> -                       uint32_t queue_id, uint32_t __user *wptr,
>>> -                       uint32_t wptr_shift, uint32_t wptr_mask,
>>> -                       struct mm_struct *mm);
>>> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
>>> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t
>>> queue_address,
>>> -                               uint32_t pipe_id, uint32_t queue_id);
>>> -
>>> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t
>>> reset_type,
>>> -                               unsigned int timeout, uint32_t pipe_id,
>>> -                               uint32_t queue_id);
>>> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
>>> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
>>> -                               unsigned int timeout);
>>> -static int kgd_address_watch_disable(struct kgd_dev *kgd);
>>> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
>>> -                                       unsigned int watch_point_id,
>>> -                                       uint32_t cntl_val,
>>> -                                       uint32_t addr_hi,
>>> -                                       uint32_t addr_lo);
>>> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
>>> -                                       uint32_t gfx_index_val,
>>> -                                       uint32_t sq_cmd);
>>> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
>>> -                                       unsigned int watch_point_id,
>>> -                                       unsigned int reg_offset);
>>> -
>>> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd,
>>> uint8_t vmid);
>>> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
>>> -                                                       uint8_t vmid);
>>> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t
>>> vmid);
>>> -
>>> -static const struct kfd2kgd_calls kfd2kgd = {
>>> -       .init_gtt_mem_allocation = alloc_gtt_mem,
>>> -       .free_gtt_mem = free_gtt_mem,
>>> -       .get_vmem_size = get_vmem_size,
>>> -       .get_gpu_clock_counter = get_gpu_clock_counter,
>>> -       .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
>>> -       .alloc_pasid = alloc_pasid,
>>> -       .free_pasid = free_pasid,
>>> -       .program_sh_mem_settings = kgd_program_sh_mem_settings,
>>> -       .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
>>> -       .init_pipeline = kgd_init_pipeline,
>>> -       .init_interrupts = kgd_init_interrupts,
>>> -       .hqd_load = kgd_hqd_load,
>>> -       .hqd_sdma_load = kgd_hqd_sdma_load,
>>> -       .hqd_is_occupied = kgd_hqd_is_occupied,
>>> -       .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
>>> -       .hqd_destroy = kgd_hqd_destroy,
>>> -       .hqd_sdma_destroy = kgd_hqd_sdma_destroy,
>>> -       .address_watch_disable = kgd_address_watch_disable,
>>> -       .address_watch_execute = kgd_address_watch_execute,
>>> -       .wave_control_execute = kgd_wave_control_execute,
>>> -       .address_watch_get_offset = kgd_address_watch_get_offset,
>>> -       .get_atc_vmid_pasid_mapping_pasid =
>>> get_atc_vmid_pasid_mapping_pasid,
>>> -       .get_atc_vmid_pasid_mapping_valid =
>>> get_atc_vmid_pasid_mapping_valid,
>>> -       .write_vmid_invalidate_request = write_vmid_invalidate_request,
>>> -       .get_fw_version = get_fw_version
>>> -};
>>> -
>>> -static const struct kgd2kfd_calls *kgd2kfd;
>>> -
>>> -int radeon_kfd_init(void)
>>> -{
>>> -       int ret;
>>> -
>>> -#if defined(CONFIG_HSA_AMD_MODULE)
>>> -       int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
>>> -
>>> -       kgd2kfd_init_p = symbol_request(kgd2kfd_init);
>>> -
>>> -       if (kgd2kfd_init_p == NULL)
>>> -               return -ENOENT;
>>> -
>>> -       ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd);
>>> -       if (ret) {
>>> -               symbol_put(kgd2kfd_init);
>>> -               kgd2kfd = NULL;
>>> -       }
>>> -
>>> -#elif defined(CONFIG_HSA_AMD)
>>> -       ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
>>> -       if (ret)
>>> -               kgd2kfd = NULL;
>>> -
>>> -#else
>>> -       ret = -ENOENT;
>>> -#endif
>>> -
>>> -       return ret;
>>> -}
>>> -
>>> -void radeon_kfd_fini(void)
>>> -{
>>> -       if (kgd2kfd) {
>>> -               kgd2kfd->exit();
>>> -               symbol_put(kgd2kfd_init);
>>> -       }
>>> -}
>>> -
>>> -void radeon_kfd_device_probe(struct radeon_device *rdev)
>>> -{
>>> -       if (kgd2kfd)
>>> -               rdev->kfd = kgd2kfd->probe((struct kgd_dev *)rdev,
>>> -                       rdev->pdev, &kfd2kgd);
>>> -}
>>> -
>>> -void radeon_kfd_device_init(struct radeon_device *rdev)
>>> -{
>>> -       int i, queue, pipe, mec;
>>> -
>>> -       if (rdev->kfd) {
>>> -               struct kgd2kfd_shared_resources gpu_resources = {
>>> -                       .compute_vmid_bitmap = 0xFF00,
>>> -                       .num_pipe_per_mec = 4,
>>> -                       .num_queue_per_pipe = 8
>>> -               };
>>> -
>>> -               bitmap_zero(gpu_resources.queue_bitmap, KGD_MAX_QUEUES);
>>> -
>>> -               for (i = 0; i < KGD_MAX_QUEUES; ++i) {
>>> -                       queue = i % gpu_resources.num_queue_per_pipe;
>>> -                       pipe = (i / gpu_resources.num_queue_per_pipe)
>>> -                               % gpu_resources.num_pipe_per_mec;
>>> -                       mec = (i / gpu_resources.num_queue_per_pipe)
>>> -                               / gpu_resources.num_pipe_per_mec;
>>> -
>>> -                       if (mec == 0 && pipe > 0)
>>> -                               set_bit(i, gpu_resources.queue_bitmap);
>>> -               }
>>> -
>>> -               radeon_doorbell_get_kfd_info(rdev,
>>> -                               &gpu_resources.doorbell_physi
>>> cal_address,
>>> -                               &gpu_resources.doorbell_aperture_size,
>>> -                               &gpu_resources.doorbell_start_offset);
>>> -
>>> -               kgd2kfd->device_init(rdev->kfd, &gpu_resources);
>>> -       }
>>> -}
>>> -
>>> -void radeon_kfd_device_fini(struct radeon_device *rdev)
>>> -{
>>> -       if (rdev->kfd) {
>>> -               kgd2kfd->device_exit(rdev->kfd);
>>> -               rdev->kfd = NULL;
>>> -       }
>>> -}
>>> -
>>> -void radeon_kfd_interrupt(struct radeon_device *rdev, const void
>>> *ih_ring_entry)
>>> -{
>>> -       if (rdev->kfd)
>>> -               kgd2kfd->interrupt(rdev->kfd, ih_ring_entry);
>>> -}
>>> -
>>> -void radeon_kfd_suspend(struct radeon_device *rdev)
>>> -{
>>> -       if (rdev->kfd)
>>> -               kgd2kfd->suspend(rdev->kfd);
>>> -}
>>> -
>>> -int radeon_kfd_resume(struct radeon_device *rdev)
>>> -{
>>> -       int r = 0;
>>> -
>>> -       if (rdev->kfd)
>>> -               r = kgd2kfd->resume(rdev->kfd);
>>> -
>>> -       return r;
>>> -}
>>> -
>>> -static int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
>>> -                       void **mem_obj, uint64_t *gpu_addr,
>>> -                       void **cpu_ptr)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>>> -       struct kgd_mem **mem = (struct kgd_mem **) mem_obj;
>>> -       int r;
>>> -
>>> -       BUG_ON(kgd == NULL);
>>> -       BUG_ON(gpu_addr == NULL);
>>> -       BUG_ON(cpu_ptr == NULL);
>>> -
>>> -       *mem = kmalloc(sizeof(struct kgd_mem), GFP_KERNEL);
>>> -       if ((*mem) == NULL)
>>> -               return -ENOMEM;
>>> -
>>> -       r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
>>> RADEON_GEM_DOMAIN_GTT,
>>> -                               RADEON_GEM_GTT_WC, NULL, NULL,
>>> &(*mem)->bo);
>>> -       if (r) {
>>> -               dev_err(rdev->dev,
>>> -                       "failed to allocate BO for amdkfd (%d)\n", r);
>>> -               return r;
>>> -       }
>>> -
>>> -       /* map the buffer */
>>> -       r = radeon_bo_reserve((*mem)->bo, true);
>>> -       if (r) {
>>> -               dev_err(rdev->dev, "(%d) failed to reserve bo for
>>> amdkfd\n", r);
>>> -               goto allocate_mem_reserve_bo_failed;
>>> -       }
>>> -
>>> -       r = radeon_bo_pin((*mem)->bo, RADEON_GEM_DOMAIN_GTT,
>>> -                               &(*mem)->gpu_addr);
>>> -       if (r) {
>>> -               dev_err(rdev->dev, "(%d) failed to pin bo for amdkfd\n",
>>> r);
>>> -               goto allocate_mem_pin_bo_failed;
>>> -       }
>>> -       *gpu_addr = (*mem)->gpu_addr;
>>> -
>>> -       r = radeon_bo_kmap((*mem)->bo, &(*mem)->cpu_ptr);
>>> -       if (r) {
>>> -               dev_err(rdev->dev,
>>> -                       "(%d) failed to map bo to kernel for amdkfd\n",
>>> r);
>>> -               goto allocate_mem_kmap_bo_failed;
>>> -       }
>>> -       *cpu_ptr = (*mem)->cpu_ptr;
>>> -
>>> -       radeon_bo_unreserve((*mem)->bo);
>>> -
>>> -       return 0;
>>> -
>>> -allocate_mem_kmap_bo_failed:
>>> -       radeon_bo_unpin((*mem)->bo);
>>> -allocate_mem_pin_bo_failed:
>>> -       radeon_bo_unreserve((*mem)->bo);
>>> -allocate_mem_reserve_bo_failed:
>>> -       radeon_bo_unref(&(*mem)->bo);
>>> -
>>> -       return r;
>>> -}
>>> -
>>> -static void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
>>> -{
>>> -       struct kgd_mem *mem = (struct kgd_mem *) mem_obj;
>>> -
>>> -       BUG_ON(mem == NULL);
>>> -
>>> -       radeon_bo_reserve(mem->bo, true);
>>> -       radeon_bo_kunmap(mem->bo);
>>> -       radeon_bo_unpin(mem->bo);
>>> -       radeon_bo_unreserve(mem->bo);
>>> -       radeon_bo_unref(&(mem->bo));
>>> -       kfree(mem);
>>> -}
>>> -
>>> -static uint64_t get_vmem_size(struct kgd_dev *kgd)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>>> -
>>> -       BUG_ON(kgd == NULL);
>>> -
>>> -       return rdev->mc.real_vram_size;
>>> -}
>>> -
>>> -static uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>>> -
>>> -       return rdev->asic->get_gpu_clock_counter(rdev);
>>> -}
>>> -
>>> -static uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *)kgd;
>>> -
>>> -       /* The sclk is in quantas of 10kHz */
>>> -       return rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk /
>>> 100;
>>> -}
>>> -
>>> -/*
>>> - * PASID manager
>>> - */
>>> -static DEFINE_IDA(pasid_ida);
>>> -
>>> -static int alloc_pasid(unsigned int bits)
>>> -{
>>> -       int pasid = -EINVAL;
>>> -
>>> -       for (bits = min(bits, 31U); bits > 0; bits--) {
>>> -               pasid = ida_simple_get(&pasid_ida,
>>> -                                      1U << (bits - 1), 1U << bits,
>>> -                                      GFP_KERNEL);
>>> -               if (pasid != -ENOSPC)
>>> -                       break;
>>> -       }
>>> -
>>> -       return pasid;
>>> -}
>>> -
>>> -static void free_pasid(unsigned int pasid)
>>> -{
>>> -       ida_simple_remove(&pasid_ida, pasid);
>>> -}
>>> -
>>> -static inline struct radeon_device *get_radeon_device(struct kgd_dev
>>> *kgd)
>>> -{
>>> -       return (struct radeon_device *)kgd;
>>> -}
>>> -
>>> -static void write_register(struct kgd_dev *kgd, uint32_t offset,
>>> uint32_t value)
>>> -{
>>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>>> -
>>> -       writel(value, (void __iomem *)(rdev->rmmio + offset));
>>> -}
>>> -
>>> -static uint32_t read_register(struct kgd_dev *kgd, uint32_t offset)
>>> -{
>>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>>> -
>>> -       return readl((void __iomem *)(rdev->rmmio + offset));
>>> -}
>>> -
>>> -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
>>> -                       uint32_t queue, uint32_t vmid)
>>> -{
>>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>>> -       uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) |
>>> QUEUEID(queue);
>>> -
>>> -       mutex_lock(&rdev->srbm_mutex);
>>> -       write_register(kgd, SRBM_GFX_CNTL, value);
>>> -}
>>> -
>>> -static void unlock_srbm(struct kgd_dev *kgd)
>>> -{
>>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>>> -
>>> -       write_register(kgd, SRBM_GFX_CNTL, 0);
>>> -       mutex_unlock(&rdev->srbm_mutex);
>>> -}
>>> -
>>> -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
>>> -                               uint32_t queue_id)
>>> -{
>>> -       uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1;
>>> -       uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC);
>>> -
>>> -       lock_srbm(kgd, mec, pipe, queue_id, 0);
>>> -}
>>> -
>>> -static void release_queue(struct kgd_dev *kgd)
>>> -{
>>> -       unlock_srbm(kgd);
>>> -}
>>> -
>>> -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t
>>> vmid,
>>> -                                       uint32_t sh_mem_config,
>>> -                                       uint32_t sh_mem_ape1_base,
>>> -                                       uint32_t sh_mem_ape1_limit,
>>> -                                       uint32_t sh_mem_bases)
>>> -{
>>> -       lock_srbm(kgd, 0, 0, 0, vmid);
>>> -
>>> -       write_register(kgd, SH_MEM_CONFIG, sh_mem_config);
>>> -       write_register(kgd, SH_MEM_APE1_BASE, sh_mem_ape1_base);
>>> -       write_register(kgd, SH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
>>> -       write_register(kgd, SH_MEM_BASES, sh_mem_bases);
>>> -
>>> -       unlock_srbm(kgd);
>>> -}
>>> -
>>> -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned
>>> int pasid,
>>> -                                       unsigned int vmid)
>>> -{
>>> -       /*
>>> -        * We have to assume that there is no outstanding mapping.
>>> -        * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0
>>> -        * because a mapping is in progress or because a mapping
>>> finished and
>>> -        * the SW cleared it.
>>> -        * So the protocol is to always wait & clear.
>>> -        */
>>> -       uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
>>> -                                       ATC_VMID_PASID_MAPPING_VALID_
>>> MASK;
>>> -
>>> -       write_register(kgd, ATC_VMID0_PASID_MAPPING +
>>> vmid*sizeof(uint32_t),
>>> -                       pasid_mapping);
>>> -
>>> -       while (!(read_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS)
>>> &
>>> -                                                               (1U <<
>>> vmid)))
>>> -               cpu_relax();
>>> -       write_register(kgd, ATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U <<
>>> vmid);
>>> -
>>> -       /* Mapping vmid to pasid also for IH block */
>>> -       write_register(kgd, IH_VMID_0_LUT + vmid * sizeof(uint32_t),
>>> -                       pasid_mapping);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
>>> -                               uint32_t hpd_size, uint64_t hpd_gpu_addr)
>>> -{
>>> -       /* nothing to do here */
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
>>> -{
>>> -       uint32_t mec;
>>> -       uint32_t pipe;
>>> -
>>> -       mec = (pipe_id / CIK_PIPE_PER_MEC) + 1;
>>> -       pipe = (pipe_id % CIK_PIPE_PER_MEC);
>>> -
>>> -       lock_srbm(kgd, mec, pipe, 0, 0);
>>> -
>>> -       write_register(kgd, CPC_INT_CNTL,
>>> -                       TIME_STAMP_INT_ENABLE | OPCODE_ERROR_INT_ENABLE);
>>> -
>>> -       unlock_srbm(kgd);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers
>>> *m)
>>> -{
>>> -       uint32_t retval;
>>> -
>>> -       retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET +
>>> -                       m->sdma_queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
>>> -
>>> -       pr_debug("kfd: sdma base address: 0x%x\n", retval);
>>> -
>>> -       return retval;
>>> -}
>>> -
>>> -static inline struct cik_mqd *get_mqd(void *mqd)
>>> -{
>>> -       return (struct cik_mqd *)mqd;
>>> -}
>>> -
>>> -static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
>>> -{
>>> -       return (struct cik_sdma_rlc_registers *)mqd;
>>> -}
>>> -
>>> -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t
>>> pipe_id,
>>> -                       uint32_t queue_id, uint32_t __user *wptr,
>>> -                       uint32_t wptr_shift, uint32_t wptr_mask,
>>> -                       struct mm_struct *mm)
>>> -{
>>> -       uint32_t wptr_shadow, is_wptr_shadow_valid;
>>> -       struct cik_mqd *m;
>>> -
>>> -       m = get_mqd(mqd);
>>> -
>>> -       is_wptr_shadow_valid = !get_user(wptr_shadow, wptr);
>>> -
>>> -       acquire_queue(kgd, pipe_id, queue_id);
>>> -       write_register(kgd, CP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo);
>>> -       write_register(kgd, CP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi);
>>> -       write_register(kgd, CP_MQD_CONTROL, m->cp_mqd_control);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo);
>>> -       write_register(kgd, CP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi);
>>> -       write_register(kgd, CP_HQD_PQ_CONTROL, m->cp_hqd_pq_control);
>>> -
>>> -       write_register(kgd, CP_HQD_IB_CONTROL, m->cp_hqd_ib_control);
>>> -       write_register(kgd, CP_HQD_IB_BASE_ADDR,
>>> m->cp_hqd_ib_base_addr_lo);
>>> -       write_register(kgd, CP_HQD_IB_BASE_ADDR_HI,
>>> m->cp_hqd_ib_base_addr_hi);
>>> -
>>> -       write_register(kgd, CP_HQD_IB_RPTR, m->cp_hqd_ib_rptr);
>>> -
>>> -       write_register(kgd, CP_HQD_PERSISTENT_STATE,
>>> -                       m->cp_hqd_persistent_state);
>>> -       write_register(kgd, CP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd);
>>> -       write_register(kgd, CP_HQD_MSG_TYPE, m->cp_hqd_msg_type);
>>> -
>>> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_LO,
>>> -                       m->cp_hqd_atomic0_preop_lo);
>>> -
>>> -       write_register(kgd, CP_HQD_ATOMIC0_PREOP_HI,
>>> -                       m->cp_hqd_atomic0_preop_hi);
>>> -
>>> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_LO,
>>> -                       m->cp_hqd_atomic1_preop_lo);
>>> -
>>> -       write_register(kgd, CP_HQD_ATOMIC1_PREOP_HI,
>>> -                       m->cp_hqd_atomic1_preop_hi);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR,
>>> -                       m->cp_hqd_pq_rptr_report_addr_lo);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_RPTR_REPORT_ADDR_HI,
>>> -                       m->cp_hqd_pq_rptr_report_addr_hi);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR,
>>> -                       m->cp_hqd_pq_wptr_poll_addr_lo);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_WPTR_POLL_ADDR_HI,
>>> -                       m->cp_hqd_pq_wptr_poll_addr_hi);
>>> -
>>> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL,
>>> -                       m->cp_hqd_pq_doorbell_control);
>>> -
>>> -       write_register(kgd, CP_HQD_VMID, m->cp_hqd_vmid);
>>> -
>>> -       write_register(kgd, CP_HQD_QUANTUM, m->cp_hqd_quantum);
>>> -
>>> -       write_register(kgd, CP_HQD_PIPE_PRIORITY,
>>> m->cp_hqd_pipe_priority);
>>> -       write_register(kgd, CP_HQD_QUEUE_PRIORITY,
>>> m->cp_hqd_queue_priority);
>>> -
>>> -       write_register(kgd, CP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr);
>>> -
>>> -       if (is_wptr_shadow_valid)
>>> -               write_register(kgd, CP_HQD_PQ_WPTR, wptr_shadow);
>>> -
>>> -       write_register(kgd, CP_HQD_ACTIVE, m->cp_hqd_active);
>>> -       release_queue(kgd);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
>>> -{
>>> -       struct cik_sdma_rlc_registers *m;
>>> -       uint32_t sdma_base_addr;
>>> -
>>> -       m = get_sdma_mqd(mqd);
>>> -       sdma_base_addr = get_sdma_base_addr(m);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_VIRTUAL_ADDR,
>>> -                       m->sdma_rlc_virtual_addr);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE,
>>> -                       m->sdma_rlc_rb_base);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_RB_BASE_HI,
>>> -                       m->sdma_rlc_rb_base_hi);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_LO,
>>> -                       m->sdma_rlc_rb_rptr_addr_lo);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_RB_RPTR_ADDR_HI,
>>> -                       m->sdma_rlc_rb_rptr_addr_hi);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_DOORBELL,
>>> -                       m->sdma_rlc_doorbell);
>>> -
>>> -       write_register(kgd,
>>> -                       sdma_base_addr + SDMA0_RLC0_RB_CNTL,
>>> -                       m->sdma_rlc_rb_cntl);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t
>>> queue_address,
>>> -                               uint32_t pipe_id, uint32_t queue_id)
>>> -{
>>> -       uint32_t act;
>>> -       bool retval = false;
>>> -       uint32_t low, high;
>>> -
>>> -       acquire_queue(kgd, pipe_id, queue_id);
>>> -       act = read_register(kgd, CP_HQD_ACTIVE);
>>> -       if (act) {
>>> -               low = lower_32_bits(queue_address >> 8);
>>> -               high = upper_32_bits(queue_address >> 8);
>>> -
>>> -               if (low == read_register(kgd, CP_HQD_PQ_BASE) &&
>>> -                               high == read_register(kgd,
>>> CP_HQD_PQ_BASE_HI))
>>> -                       retval = true;
>>> -       }
>>> -       release_queue(kgd);
>>> -       return retval;
>>> -}
>>> -
>>> -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
>>> -{
>>> -       struct cik_sdma_rlc_registers *m;
>>> -       uint32_t sdma_base_addr;
>>> -       uint32_t sdma_rlc_rb_cntl;
>>> -
>>> -       m = get_sdma_mqd(mqd);
>>> -       sdma_base_addr = get_sdma_base_addr(m);
>>> -
>>> -       sdma_rlc_rb_cntl = read_register(kgd,
>>> -                                       sdma_base_addr +
>>> SDMA0_RLC0_RB_CNTL);
>>> -
>>> -       if (sdma_rlc_rb_cntl & SDMA_RB_ENABLE)
>>> -               return true;
>>> -
>>> -       return false;
>>> -}
>>> -
>>> -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t
>>> reset_type,
>>> -                               unsigned int timeout, uint32_t pipe_id,
>>> -                               uint32_t queue_id)
>>> -{
>>> -       uint32_t temp;
>>> -
>>> -       acquire_queue(kgd, pipe_id, queue_id);
>>> -       write_register(kgd, CP_HQD_PQ_DOORBELL_CONTROL, 0);
>>> -
>>> -       write_register(kgd, CP_HQD_DEQUEUE_REQUEST, reset_type);
>>> -
>>> -       while (true) {
>>> -               temp = read_register(kgd, CP_HQD_ACTIVE);
>>> -               if (temp & 0x1)
>>> -                       break;
>>> -               if (timeout == 0) {
>>> -                       pr_err("kfd: cp queue preemption time out
>>> (%dms)\n",
>>> -                               temp);
>>> -                       release_queue(kgd);
>>> -                       return -ETIME;
>>> -               }
>>> -               msleep(20);
>>> -               timeout -= 20;
>>> -       }
>>> -
>>> -       release_queue(kgd);
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
>>> -                               unsigned int timeout)
>>> -{
>>> -       struct cik_sdma_rlc_registers *m;
>>> -       uint32_t sdma_base_addr;
>>> -       uint32_t temp;
>>> -
>>> -       m = get_sdma_mqd(mqd);
>>> -       sdma_base_addr = get_sdma_base_addr(m);
>>> -
>>> -       temp = read_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL);
>>> -       temp = temp & ~SDMA_RB_ENABLE;
>>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_CNTL, temp);
>>> -
>>> -       while (true) {
>>> -               temp = read_register(kgd, sdma_base_addr +
>>> -
>>>  SDMA0_RLC0_CONTEXT_STATUS);
>>> -               if (temp & SDMA_RLC_IDLE)
>>> -                       break;
>>> -               if (timeout == 0)
>>> -                       return -ETIME;
>>> -               msleep(20);
>>> -               timeout -= 20;
>>> -       }
>>> -
>>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_DOORBELL, 0);
>>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_RPTR, 0);
>>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_WPTR, 0);
>>> -       write_register(kgd, sdma_base_addr + SDMA0_RLC0_RB_BASE, 0);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_address_watch_disable(struct kgd_dev *kgd)
>>> -{
>>> -       union TCP_WATCH_CNTL_BITS cntl;
>>> -       unsigned int i;
>>> -
>>> -       cntl.u32All = 0;
>>> -
>>> -       cntl.bitfields.valid = 0;
>>> -       cntl.bitfields.mask = ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK;
>>> -       cntl.bitfields.atc = 1;
>>> -
>>> -       /* Turning off this address until we set all the registers */
>>> -       for (i = 0; i < MAX_WATCH_ADDRESSES; i++)
>>> -               write_register(kgd,
>>> -                               watchRegs[i * ADDRESS_WATCH_REG_MAX +
>>> -                                       ADDRESS_WATCH_REG_CNTL],
>>> -                               cntl.u32All);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_address_watch_execute(struct kgd_dev *kgd,
>>> -                                       unsigned int watch_point_id,
>>> -                                       uint32_t cntl_val,
>>> -                                       uint32_t addr_hi,
>>> -                                       uint32_t addr_lo)
>>> -{
>>> -       union TCP_WATCH_CNTL_BITS cntl;
>>> -
>>> -       cntl.u32All = cntl_val;
>>> -
>>> -       /* Turning off this watch point until we set all the registers */
>>> -       cntl.bitfields.valid = 0;
>>> -       write_register(kgd,
>>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX
>>> +
>>> -                               ADDRESS_WATCH_REG_CNTL],
>>> -                       cntl.u32All);
>>> -
>>> -       write_register(kgd,
>>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX
>>> +
>>> -                               ADDRESS_WATCH_REG_ADDR_HI],
>>> -                       addr_hi);
>>> -
>>> -       write_register(kgd,
>>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX
>>> +
>>> -                               ADDRESS_WATCH_REG_ADDR_LO],
>>> -                       addr_lo);
>>> -
>>> -       /* Enable the watch point */
>>> -       cntl.bitfields.valid = 1;
>>> -
>>> -       write_register(kgd,
>>> -                       watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX
>>> +
>>> -                               ADDRESS_WATCH_REG_CNTL],
>>> -                       cntl.u32All);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static int kgd_wave_control_execute(struct kgd_dev *kgd,
>>> -                                       uint32_t gfx_index_val,
>>> -                                       uint32_t sq_cmd)
>>> -{
>>> -       struct radeon_device *rdev = get_radeon_device(kgd);
>>> -       uint32_t data;
>>> -
>>> -       mutex_lock(&rdev->grbm_idx_mutex);
>>> -
>>> -       write_register(kgd, GRBM_GFX_INDEX, gfx_index_val);
>>> -       write_register(kgd, SQ_CMD, sq_cmd);
>>> -
>>> -       /*  Restore the GRBM_GFX_INDEX register  */
>>> -
>>> -       data = INSTANCE_BROADCAST_WRITES | SH_BROADCAST_WRITES |
>>> -               SE_BROADCAST_WRITES;
>>> -
>>> -       write_register(kgd, GRBM_GFX_INDEX, data);
>>> -
>>> -       mutex_unlock(&rdev->grbm_idx_mutex);
>>> -
>>> -       return 0;
>>> -}
>>> -
>>> -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
>>> -                                       unsigned int watch_point_id,
>>> -                                       unsigned int reg_offset)
>>> -{
>>> -       return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX +
>>> reg_offset]
>>> -               / 4;
>>> -}
>>> -
>>> -static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd,
>>> uint8_t vmid)
>>> -{
>>> -       uint32_t reg;
>>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>>> -
>>> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
>>> -       return reg & ATC_VMID_PASID_MAPPING_VALID_MASK;
>>> -}
>>> -
>>> -static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
>>> -                                                       uint8_t vmid)
>>> -{
>>> -       uint32_t reg;
>>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>>> -
>>> -       reg = RREG32(ATC_VMID0_PASID_MAPPING + vmid*4);
>>> -       return reg & ATC_VMID_PASID_MAPPING_PASID_MASK;
>>> -}
>>> -
>>> -static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t
>>> vmid)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>>> -
>>> -       return WREG32(VM_INVALIDATE_REQUEST, 1 << vmid);
>>> -}
>>> -
>>> -static uint16_t get_fw_version(struct kgd_dev *kgd, enum
>>> kgd_engine_type type)
>>> -{
>>> -       struct radeon_device *rdev = (struct radeon_device *) kgd;
>>> -       const union radeon_firmware_header *hdr;
>>> -
>>> -       BUG_ON(kgd == NULL || rdev->mec_fw == NULL);
>>> -
>>> -       switch (type) {
>>> -       case KGD_ENGINE_PFP:
>>> -               hdr = (const union radeon_firmware_header *)
>>> rdev->pfp_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_ME:
>>> -               hdr = (const union radeon_firmware_header *)
>>> rdev->me_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_CE:
>>> -               hdr = (const union radeon_firmware_header *)
>>> rdev->ce_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_MEC1:
>>> -               hdr = (const union radeon_firmware_header *)
>>> rdev->mec_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_MEC2:
>>> -               hdr = (const union radeon_firmware_header *)
>>> -
>>>  rdev->mec2_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_RLC:
>>> -               hdr = (const union radeon_firmware_header *)
>>> rdev->rlc_fw->data;
>>> -               break;
>>> -
>>> -       case KGD_ENGINE_SDMA1:
>>> -       case KGD_ENGINE_SDMA2:
>>> -               hdr = (const union radeon_firmware_header *)
>>> -
>>>  rdev->sdma_fw->data;
>>> -               break;
>>> -
>>> -       default:
>>> -               return 0;
>>> -       }
>>> -
>>> -       if (hdr == NULL)
>>> -               return 0;
>>> -
>>> -       /* Only 12 bit in use*/
>>> -       return hdr->common.ucode_version;
>>> -}
>>> diff --git a/drivers/gpu/drm/radeon/radeon_kfd.h
>>> b/drivers/gpu/drm/radeon/radeon_kfd.h
>>> deleted file mode 100644
>>> index 9df1fea..0000000
>>> --- a/drivers/gpu/drm/radeon/radeon_kfd.h
>>> +++ /dev/null
>>> @@ -1,47 +0,0 @@
>>> -/*
>>> - * Copyright 2014 Advanced Micro Devices, Inc.
>>> - *
>>> - * Permission is hereby granted, free of charge, to any person
>>> obtaining a
>>> - * copy of this software and associated documentation files (the
>>> "Software"),
>>> - * to deal in the Software without restriction, including without
>>> limitation
>>> - * the rights to use, copy, modify, merge, publish, distribute,
>>> sublicense,
>>> - * and/or sell copies of the Software, and to permit persons to whom the
>>> - * Software is furnished to do so, subject to the following conditions:
>>> - *
>>> - * The above copyright notice and this permission notice shall be
>>> included in
>>> - * all copies or substantial portions of the Software.
>>> - *
>>> - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>>> EXPRESS OR
>>> - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>>> MERCHANTABILITY,
>>> - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT
>>> SHALL
>>> - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM,
>>> DAMAGES OR
>>> - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
>>> - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>>> - * OTHER DEALINGS IN THE SOFTWARE.
>>> - */
>>> -
>>> -/*
>>> - * radeon_kfd.h defines the private interface between the
>>> - * AMD kernel graphics drivers and the AMD KFD.
>>> - */
>>> -
>>> -#ifndef RADEON_KFD_H_INCLUDED
>>> -#define RADEON_KFD_H_INCLUDED
>>> -
>>> -#include <linux/types.h>
>>> -#include "kgd_kfd_interface.h"
>>> -
>>> -struct radeon_device;
>>> -
>>> -int radeon_kfd_init(void);
>>> -void radeon_kfd_fini(void);
>>> -
>>> -void radeon_kfd_suspend(struct radeon_device *rdev);
>>> -int radeon_kfd_resume(struct radeon_device *rdev);
>>> -void radeon_kfd_interrupt(struct radeon_device *rdev,
>>> -                       const void *ih_ring_entry);
>>> -void radeon_kfd_device_probe(struct radeon_device *rdev);
>>> -void radeon_kfd_device_init(struct radeon_device *rdev);
>>> -void radeon_kfd_device_fini(struct radeon_device *rdev);
>>> -
>>> -#endif /* RADEON_KFD_H_INCLUDED */
>>> diff --git a/drivers/gpu/drm/radeon/radeon_kms.c
>>> b/drivers/gpu/drm/radeon/radeon_kms.c
>>> index dfee8f7..cde037f 100644
>>> --- a/drivers/gpu/drm/radeon/radeon_kms.c
>>> +++ b/drivers/gpu/drm/radeon/radeon_kms.c
>>> @@ -34,8 +34,6 @@
>>>   #include <linux/slab.h>
>>>   #include <linux/pm_runtime.h>
>>>
>>> -#include "radeon_kfd.h"
>>> -
>>>   #if defined(CONFIG_VGA_SWITCHEROO)
>>>   bool radeon_has_atpx(void);
>>>   #else
>>> @@ -68,8 +66,6 @@ void radeon_driver_unload_kms(struct drm_device *dev)
>>>                  pm_runtime_forbid(dev->dev);
>>>          }
>>>
>>> -       radeon_kfd_device_fini(rdev);
>>> -
>>>          radeon_acpi_fini(rdev);
>>>
>>>          radeon_modeset_fini(rdev);
>>> @@ -174,9 +170,6 @@ int radeon_driver_load_kms(struct drm_device *dev,
>>> unsigned long flags)
>>>                                  "Error during ACPI methods call\n");
>>>          }
>>>
>>> -       radeon_kfd_device_probe(rdev);
>>> -       radeon_kfd_device_init(rdev);
>>> -
>>>          if (radeon_is_px(dev)) {
>>>                  pm_runtime_use_autosuspend(dev->dev);
>>>                  pm_runtime_set_autosuspend_delay(dev->dev, 5000);
>>> --
>>> 2.7.4
>>>
>>> _______________________________________________
>>> amd-gfx mailing list
>>> amd-gfx@lists.freedesktop.org
>>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
>>>
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
>>
>
>
>
On 2017-11-01 09:31 AM, Oded Gabbay wrote:
> ok, taken to -next.

This change broke the radeon driver on my Kaveri laptop. The gdm login
screen works, but logging into the GNOME on Xorg session quickly results
in a GPU hang and associated badness, see the attached dmesg.

Reverting this change on top of drm-next makes it work again.

On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
reduce number of free VMIDs and pipes in KV") and 28b57b856b63
("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.

Any ideas for what else is missing?

Note that the amdkfd driver isn't actually active anyway, because I'm
disabling the IOMMU. Is it possible that it's still doing or triggering
some needed HW setup before it bails in that case?


P.S. Assuming we can fix this without reverting, maybe we could also
remove rdev->grbm_idx_mutex again?
On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>> ok, taken to -next.
>
> This change broke the radeon driver on my Kaveri laptop. The gdm login
> screen works, but logging into the GNOME on Xorg session quickly results
> in a GPU hang and associated badness, see the attached dmesg.
>
> Reverting this change on top of drm-next makes it work again.
>
> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>
> Any ideas for what else is missing?
>
> Note that the amdkfd driver isn't actually active anyway, because I'm
> disabling the IOMMU. Is it possible that it's still doing or triggering
> some needed HW setup before it bails in that case?
>
>
> P.S. Assuming we can fix this without reverting, maybe we could also
> remove rdev->grbm_idx_mutex again?
>
> --
> Earthling Michel Dänzer               |               http://www.amd.com
> Libre software enthusiast             |             Mesa and X developer

Hi Michel,
Even without IOMMU, amdkfd will initialize the module and internal
structures per device, up to the point where it tries to register a
callback with the iommu driver.
If IOMMU is disabled, it will fail then with the following error
message (in dmesg): "error getting iommu info. is the iommu enabled?"

Having said that, it doesn't initialize anything in the device H/W
itself, so I find this very weird.

I looked at the patch itself again and I don't see anything suspicious.

I'll try to resurrect my Kaveri machine to check this, but it will
take some time.

Oded
On Wed, Nov 29, 2017 at 2:31 PM, Oded Gabbay <oded.gabbay@gmail.com> wrote:
> On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
>> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>>> ok, taken to -next.
>>
>> This change broke the radeon driver on my Kaveri laptop. The gdm login
>> screen works, but logging into the GNOME on Xorg session quickly results
>> in a GPU hang and associated badness, see the attached dmesg.
>>
>> Reverting this change on top of drm-next makes it work again.
>>
>> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
>> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
>> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>>
>> Any ideas for what else is missing?
>>
>> Note that the amdkfd driver isn't actually active anyway, because I'm
>> disabling the IOMMU. Is it possible that it's still doing or triggering
>> some needed HW setup before it bails in that case?
>>
>>
>> P.S. Assuming we can fix this without reverting, maybe we could also
>> remove rdev->grbm_idx_mutex again?
>>
>> --
>> Earthling Michel Dänzer               |               http://www.amd.com
>> Libre software enthusiast             |             Mesa and X developer
>
> Hi Michel,
> Even without IOMMU, amdkfd will initialize the module and internal
> structures per device, up to the point where it tries to register a
> callback with the iommu driver.
> If IOMMU is disabled, it will fail then with the following error
> message (in dmesg): "error getting iommu info. is the iommu enabled?"
>
> Having said that, it doesn't initialize anything in the device H/W
> itself, so I find this very weird.
>
> I looked at the patch itself again and I don't see anything suspicious.
>
> I'll try to resurrect my Kaveri machine to check this, but it will
> take some time.
>
> Oded

Any chance that the increase of VMIDs from 8 to 16 somehow (although I
don't know how) caused this problem ?
The desktop gui also didn't work for me, but when I changed the VMID
number back to 8 (in cik.c) the gui worked again.

Michel, could you try this as well ?

Thanks,
Oded
Am 29.11.2017 um 13:31 schrieb Oded Gabbay:
> On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
>> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>>> ok, taken to -next.
>> This change broke the radeon driver on my Kaveri laptop. The gdm login
>> screen works, but logging into the GNOME on Xorg session quickly results
>> in a GPU hang and associated badness, see the attached dmesg.
>>
>> Reverting this change on top of drm-next makes it work again.
>>
>> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
>> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
>> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>>
>> Any ideas for what else is missing?
>>
>> Note that the amdkfd driver isn't actually active anyway, because I'm
>> disabling the IOMMU. Is it possible that it's still doing or triggering
>> some needed HW setup before it bails in that case?
>>
>>
>> P.S. Assuming we can fix this without reverting, maybe we could also
>> remove rdev->grbm_idx_mutex again?
>>
>> --
>> Earthling Michel Dänzer               |               http://www.amd.com
>> Libre software enthusiast             |             Mesa and X developer
> Hi Michel,
> Even without IOMMU, amdkfd will initialize the module and internal
> structures per device, up to the point where it tries to register a
> callback with the iommu driver.
> If IOMMU is disabled, it will fail then with the following error
> message (in dmesg): "error getting iommu info. is the iommu enabled?"
>
> Having said that, it doesn't initialize anything in the device H/W
> itself, so I find this very weird.
>
> I looked at the patch itself again and I don't see anything suspicious.
>
> I'll try to resurrect my Kaveri machine to check this, but it will
> take some time.

My best guess is that there is something broken with the VMID handling 
on Radeon.

I have a Kaveri running amdgpu as one of my test systems here and 
switching to radeon for a test should be trivial.

Going to give that a try later today if I have time.

Christian.

>
> Oded
Am 29.11.2017 um 15:40 schrieb Oded Gabbay:
> On Wed, Nov 29, 2017 at 2:31 PM, Oded Gabbay <oded.gabbay@gmail.com> wrote:
>> On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
>>> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>>>> ok, taken to -next.
>>> This change broke the radeon driver on my Kaveri laptop. The gdm login
>>> screen works, but logging into the GNOME on Xorg session quickly results
>>> in a GPU hang and associated badness, see the attached dmesg.
>>>
>>> Reverting this change on top of drm-next makes it work again.
>>>
>>> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
>>> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
>>> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>>>
>>> Any ideas for what else is missing?
>>>
>>> Note that the amdkfd driver isn't actually active anyway, because I'm
>>> disabling the IOMMU. Is it possible that it's still doing or triggering
>>> some needed HW setup before it bails in that case?
>>>
>>>
>>> P.S. Assuming we can fix this without reverting, maybe we could also
>>> remove rdev->grbm_idx_mutex again?
>>>
>>> --
>>> Earthling Michel Dänzer               |               http://www.amd.com
>>> Libre software enthusiast             |             Mesa and X developer
>> Hi Michel,
>> Even without IOMMU, amdkfd will initialize the module and internal
>> structures per device, up to the point where it tries to register a
>> callback with the iommu driver.
>> If IOMMU is disabled, it will fail then with the following error
>> message (in dmesg): "error getting iommu info. is the iommu enabled?"
>>
>> Having said that, it doesn't initialize anything in the device H/W
>> itself, so I find this very weird.
>>
>> I looked at the patch itself again and I don't see anything suspicious.
>>
>> I'll try to resurrect my Kaveri machine to check this, but it will
>> take some time.
>>
>> Oded
> Any chance that the increase of VMIDs from 8 to 16 somehow (although I
> don't know how) caused this problem ?

Yeah, that's my first suspicion as well. We haven't used the higher 
VMIDs in a while on radeon and so there could be bugs in the code.

> The desktop gui also didn't work for me, but when I changed the VMID
> number back to 8 (in cik.c) the gui worked again.

Good to know, going to test this as well later today.

Regards,
Christian.

>
> Michel, could you try this as well ?
>
> Thanks,
> Oded
On 2017-11-29 03:40 PM, Oded Gabbay wrote:
> On Wed, Nov 29, 2017 at 2:31 PM, Oded Gabbay <oded.gabbay@gmail.com> wrote:
>> On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
>>> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>>>> ok, taken to -next.
>>>
>>> This change broke the radeon driver on my Kaveri laptop. The gdm login
>>> screen works, but logging into the GNOME on Xorg session quickly results
>>> in a GPU hang and associated badness, see the attached dmesg.
>>>
>>> Reverting this change on top of drm-next makes it work again.
>>>
>>> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
>>> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
>>> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>>>
>>> Any ideas for what else is missing?
>>>
>>> Note that the amdkfd driver isn't actually active anyway, because I'm
>>> disabling the IOMMU. Is it possible that it's still doing or triggering
>>> some needed HW setup before it bails in that case?
>>>
>>>
>>> P.S. Assuming we can fix this without reverting, maybe we could also
>>> remove rdev->grbm_idx_mutex again?
>>>
>>> --
>>> Earthling Michel Dänzer               |               http://www.amd.com
>>> Libre software enthusiast             |             Mesa and X developer
>>
>> Hi Michel,
>> Even without IOMMU, amdkfd will initialize the module and internal
>> structures per device, up to the point where it tries to register a
>> callback with the iommu driver.
>> If IOMMU is disabled, it will fail then with the following error
>> message (in dmesg): "error getting iommu info. is the iommu enabled?"
>>
>> Having said that, it doesn't initialize anything in the device H/W
>> itself, so I find this very weird.
>>
>> I looked at the patch itself again and I don't see anything suspicious.
>>
>> I'll try to resurrect my Kaveri machine to check this, but it will
>> take some time.
>>
>> Oded
> 
> Any chance that the increase of VMIDs from 8 to 16 somehow (although I
> don't know how) caused this problem ?
> The desktop gui also didn't work for me, but when I changed the VMID
> number back to 8 (in cik.c) the gui worked again.
> 
> Michel, could you try this as well ?

Yeah, that also occurred to me in the meantime, and I can confirm your
findings.

My guess right now is that it's related to cik_pcie_init_compute_vmid.
On Wed, Nov 29, 2017 at 4:54 PM, Michel Dänzer <michel@daenzer.net> wrote:
> On 2017-11-29 03:40 PM, Oded Gabbay wrote:
>> On Wed, Nov 29, 2017 at 2:31 PM, Oded Gabbay <oded.gabbay@gmail.com> wrote:
>>> On Wed, Nov 29, 2017 at 1:16 PM, Michel Dänzer <michel@daenzer.net> wrote:
>>>> On 2017-11-01 09:31 AM, Oded Gabbay wrote:
>>>>> ok, taken to -next.
>>>>
>>>> This change broke the radeon driver on my Kaveri laptop. The gdm login
>>>> screen works, but logging into the GNOME on Xorg session quickly results
>>>> in a GPU hang and associated badness, see the attached dmesg.
>>>>
>>>> Reverting this change on top of drm-next makes it work again.
>>>>
>>>> On a hunch, I've tried reverting commits 62a7b7fbd08e ("drm/radeon:
>>>> reduce number of free VMIDs and pipes in KV") and 28b57b856b63
>>>> ("drm/radeon/cik: Don't touch int of pipes 1-7"), but no luck.
>>>>
>>>> Any ideas for what else is missing?
>>>>
>>>> Note that the amdkfd driver isn't actually active anyway, because I'm
>>>> disabling the IOMMU. Is it possible that it's still doing or triggering
>>>> some needed HW setup before it bails in that case?
>>>>
>>>>
>>>> P.S. Assuming we can fix this without reverting, maybe we could also
>>>> remove rdev->grbm_idx_mutex again?
>>>>
>>>> --
>>>> Earthling Michel Dänzer               |               http://www.amd.com
>>>> Libre software enthusiast             |             Mesa and X developer
>>>
>>> Hi Michel,
>>> Even without IOMMU, amdkfd will initialize the module and internal
>>> structures per device, up to the point where it tries to register a
>>> callback with the iommu driver.
>>> If IOMMU is disabled, it will fail then with the following error
>>> message (in dmesg): "error getting iommu info. is the iommu enabled?"
>>>
>>> Having said that, it doesn't initialize anything in the device H/W
>>> itself, so I find this very weird.
>>>
>>> I looked at the patch itself again and I don't see anything suspicious.
>>>
>>> I'll try to resurrect my Kaveri machine to check this, but it will
>>> take some time.
>>>
>>> Oded
>>
>> Any chance that the increase of VMIDs from 8 to 16 somehow (although I
>> don't know how) caused this problem ?
>> The desktop gui also didn't work for me, but when I changed the VMID
>> number back to 8 (in cik.c) the gui worked again.
>>
>> Michel, could you try this as well ?
>
> Yeah, that also occurred to me in the meantime, and I can confirm your
> findings.
>
> My guess right now is that it's related to cik_pcie_init_compute_vmid.
>
>
> --
> Earthling Michel Dänzer               |               http://www.amd.com
> Libre software enthusiast             |             Mesa and X developer

Yeah, that seems reasonable.
That initialization was part of kfd but then we moved it to radeon.
Now it collides with radeon's initialization.
I removed it completely and returned the number of VMIDs to 16 and the
GUI is working.
I'll send a patch.

Oded