From patchwork Fri Jun 12 19:31:11 2015 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Oded Gabbay X-Patchwork-Id: 6600501 Return-Path: X-Original-To: patchwork-dri-devel@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork1.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.136]) by patchwork1.web.kernel.org (Postfix) with ESMTP id 11E7F9F326 for ; Fri, 12 Jun 2015 19:31:37 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 35F7A205F9 for ; Fri, 12 Jun 2015 19:31:35 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by mail.kernel.org (Postfix) with ESMTP id 22CD12060E for ; Fri, 12 Jun 2015 19:31:32 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id B9BC66E456; Fri, 12 Jun 2015 12:31:29 -0700 (PDT) X-Original-To: dri-devel@lists.freedesktop.org Delivered-To: dri-devel@lists.freedesktop.org Received: from mail-wg0-f51.google.com (mail-wg0-f51.google.com [74.125.82.51]) by gabe.freedesktop.org (Postfix) with ESMTP id A73B66E456 for ; Fri, 12 Jun 2015 12:31:28 -0700 (PDT) Received: by wgv5 with SMTP id 5so30575455wgv.1 for ; Fri, 12 Jun 2015 12:31:28 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=nNVA0Ru/P8tq9u2g/reJwHdvD2QSph8A9YAWd550564=; b=XPc2D8aa/RrZ+y0psUM0yfo/DE8Vf7ats+TUcW51XiwDSBJxyf6349viiaBGCdmC2j y0eC9O0T7yvl4ELuPk6/nSrQxxweKnOfi3RyhUM8jYNCzt9l5ITqhWhzQZPDlvQt/Tt1 eHs8Mdc4zFtsevvD2ltVI/wRjaLbmMR1ZBHapCbXILbJ5BnHVh9f5ZWXuJEKI68aq9uX /RpsYdZYStnaln6XIKVeNRAzegMPayA7ijOsVdwJ3RJTEJ8IUIT4Up6TZWCz2eZikLnf 2eooBoQ/FjtKwQfGOviOZMRoXLnUuS3XOM6lxJO7XYD5bskHp9C0ogWZU6j3xlBhO4Sm smeQ== X-Received: by 10.194.184.14 with SMTP id eq14mr29413195wjc.58.1434137487839; Fri, 12 Jun 2015 12:31:27 -0700 (PDT) Received: from odedg-home.redhat.com ([77.126.137.55]) by mx.google.com with ESMTPSA id v3sm4089284wiz.14.2015.06.12.12.31.26 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Fri, 12 Jun 2015 12:31:27 -0700 (PDT) From: Oded Gabbay To: dri-devel@lists.freedesktop.org, alexdeucher@gmail.com Subject: [PATCH 03/13] drm/amdgpu: add amdgpu <--> amdkfd gfx7 interface Date: Fri, 12 Jun 2015 22:31:11 +0300 Message-Id: <1434137481-13315-4-git-send-email-oded.gabbay@gmail.com> X-Mailer: git-send-email 2.4.3 In-Reply-To: <1434137481-13315-1-git-send-email-oded.gabbay@gmail.com> References: <1434137481-13315-1-git-send-email-oded.gabbay@gmail.com> X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.18 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" X-Spam-Status: No, score=-4.1 required=5.0 tests=BAYES_00, DKIM_ADSP_CUSTOM_MED, DKIM_SIGNED, FREEMAIL_FROM, RCVD_IN_DNSWL_MED, T_DKIM_INVALID, T_RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=unavailable version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP This patch adds the gfx7 interface file between amdgpu and amdkfd. This interface file mirrors (some) of the functions in radeon_kfd.c (the interface file between radeon and amdkfd). The gfx7 interface is used when it is run on a Kaveri-based system. This interface file was used for bring-up of amdkfd on amdgpu and for debugging purposes. For users who would like to run HSA on Kaveri, please use the radeon graphic driver. Note: CONFIG_DRM_AMDGPU_CIK must be selected for amdgpu to handle Kaveri. Signed-off-by: Oded Gabbay --- MAINTAINERS | 1 + drivers/gpu/drm/amd/amdgpu/Makefile | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 670 ++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/cikd.h | 12 + 5 files changed, 687 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c diff --git a/MAINTAINERS b/MAINTAINERS index 581c8d4..324bc2d 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -637,6 +637,7 @@ T: git git://people.freedesktop.org/~gabbayo/linux.git S: Supported F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c F: drivers/gpu/drm/amd/amdkfd/ F: drivers/gpu/drm/amd/include/cik_structs.h F: drivers/gpu/drm/amd/include/kgd_kfd_interface.h diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 891f02c..af5397c 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -73,7 +73,8 @@ amdgpu-y += \ # add amdkfd interfaces amdgpu-y += \ - amdgpu_amdkfd.o + amdgpu_amdkfd.o \ + amdgpu_amdkfd_gfx_v7.o amdgpu-$(CONFIG_COMPAT) += amdgpu_ioc32.o amdgpu-$(CONFIG_VGA_SWITCHEROO) += amdgpu_atpx_handler.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 2071a9b..7aa5ab0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -51,6 +51,8 @@ bool amdgpu_amdkfd_load_interface(struct amdgpu_device *rdev) switch (rdev->asic_type) { case CHIP_KAVERI: + kfd2kgd = amdgpu_amdkfd_gfx_7_get_functions(); + break; default: return false; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c new file mode 100644 index 0000000..2daad33 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -0,0 +1,670 @@ +/* + * 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 +#include +#include +#include +#include "amdgpu.h" +#include "amdgpu_amdkfd.h" +#include "cikd.h" +#include "cik_sdma.h" +#include "amdgpu_ucode.h" +#include "gca/gfx_7_2_d.h" +#include "gca/gfx_7_2_enum.h" +#include "gca/gfx_7_2_sh_mask.h" +#include "oss/oss_2_0_d.h" +#include "oss/oss_2_0_sh_mask.h" +#include "gmc/gmc_7_1_d.h" +#include "gmc/gmc_7_1_sh_mask.h" +#include "cik_structs.h" + +#define CIK_PIPE_PER_MEC (4) + +enum { + MAX_TRAPID = 8, /* 3 bits in the bitfield. */ + MAX_WATCH_ADDRESSES = 4 +}; + +enum { + ADDRESS_WATCH_REG_ADDR_HI = 0, + ADDRESS_WATCH_REG_ADDR_LO, + ADDRESS_WATCH_REG_CNTL, + ADDRESS_WATCH_REG_MAX +}; + +/* not defined in the CI/KV reg file */ +enum { + ADDRESS_WATCH_REG_CNTL_ATC_BIT = 0x10000000UL, + ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK = 0x00FFFFFF, + ADDRESS_WATCH_REG_ADDLOW_MASK_EXTENSION = 0x03000000, + /* extend the mask to 26 bits to match the low address field */ + ADDRESS_WATCH_REG_ADDLOW_SHIFT = 6, + ADDRESS_WATCH_REG_ADDHIGH_MASK = 0xFFFF +}; + +static const uint32_t watchRegs[MAX_WATCH_ADDRESSES * ADDRESS_WATCH_REG_MAX] = { + mmTCP_WATCH0_ADDR_H, mmTCP_WATCH0_ADDR_L, mmTCP_WATCH0_CNTL, + mmTCP_WATCH1_ADDR_H, mmTCP_WATCH1_ADDR_L, mmTCP_WATCH1_CNTL, + mmTCP_WATCH2_ADDR_H, mmTCP_WATCH2_ADDR_L, mmTCP_WATCH2_CNTL, + mmTCP_WATCH3_ADDR_H, mmTCP_WATCH3_ADDR_L, mmTCP_WATCH3_CNTL +}; + +union TCP_WATCH_CNTL_BITS { + struct { + uint32_t mask:24; + uint32_t vmid:4; + uint32_t atc:1; + uint32_t mode:2; + uint32_t valid:1; + } bitfields, bits; + uint32_t u32All; + signed int i32All; + float f32All; +}; + +/* + * 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); +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, 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 uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); + +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, + .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 +}; + +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions() +{ + return (struct kfd2kgd_calls *)&kfd2kgd; +} + +static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) +{ + return (struct amdgpu_device *)kgd; +} + +static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, + uint32_t queue, uint32_t vmid) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue); + + mutex_lock(&adev->srbm_mutex); + WREG32(mmSRBM_GFX_CNTL, value); +} + +static void unlock_srbm(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + WREG32(mmSRBM_GFX_CNTL, 0); + mutex_unlock(&adev->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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + lock_srbm(kgd, 0, 0, 0, vmid); + + WREG32(mmSH_MEM_CONFIG, sh_mem_config); + WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base); + WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit); + WREG32(mmSH_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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + /* + * 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_VMID0_PASID_MAPPING__VALID_MASK; + + WREG32(mmATC_VMID0_PASID_MAPPING + vmid, pasid_mapping); + + while (!(RREG32(mmATC_VMID_PASID_MAPPING_UPDATE_STATUS) & (1U << vmid))) + cpu_relax(); + WREG32(mmATC_VMID_PASID_MAPPING_UPDATE_STATUS, 1U << vmid); + + /* Mapping vmid to pasid also for IH block */ + WREG32(mmIH_VMID_0_LUT + vmid, 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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1; + uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC); + + lock_srbm(kgd, mec, pipe, 0, 0); + WREG32(mmCP_HPD_EOP_BASE_ADDR, lower_32_bits(hpd_gpu_addr >> 8)); + WREG32(mmCP_HPD_EOP_BASE_ADDR_HI, upper_32_bits(hpd_gpu_addr >> 8)); + WREG32(mmCP_HPD_EOP_VMID, 0); + WREG32(mmCP_HPD_EOP_CONTROL, hpd_size); + unlock_srbm(kgd); + + return 0; +} + +static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + 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); + + WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | + CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); + + 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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + 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); + WREG32(mmCP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo); + WREG32(mmCP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi); + WREG32(mmCP_MQD_CONTROL, m->cp_mqd_control); + + WREG32(mmCP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo); + WREG32(mmCP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi); + WREG32(mmCP_HQD_PQ_CONTROL, m->cp_hqd_pq_control); + + WREG32(mmCP_HQD_IB_CONTROL, m->cp_hqd_ib_control); + WREG32(mmCP_HQD_IB_BASE_ADDR, m->cp_hqd_ib_base_addr_lo); + WREG32(mmCP_HQD_IB_BASE_ADDR_HI, m->cp_hqd_ib_base_addr_hi); + + WREG32(mmCP_HQD_IB_RPTR, m->cp_hqd_ib_rptr); + + WREG32(mmCP_HQD_PERSISTENT_STATE, m->cp_hqd_persistent_state); + WREG32(mmCP_HQD_SEMA_CMD, m->cp_hqd_sema_cmd); + WREG32(mmCP_HQD_MSG_TYPE, m->cp_hqd_msg_type); + + WREG32(mmCP_HQD_ATOMIC0_PREOP_LO, m->cp_hqd_atomic0_preop_lo); + WREG32(mmCP_HQD_ATOMIC0_PREOP_HI, m->cp_hqd_atomic0_preop_hi); + WREG32(mmCP_HQD_ATOMIC1_PREOP_LO, m->cp_hqd_atomic1_preop_lo); + WREG32(mmCP_HQD_ATOMIC1_PREOP_HI, m->cp_hqd_atomic1_preop_hi); + + WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR, m->cp_hqd_pq_rptr_report_addr_lo); + WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI, + m->cp_hqd_pq_rptr_report_addr_hi); + + WREG32(mmCP_HQD_PQ_RPTR, m->cp_hqd_pq_rptr); + + WREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR, m->cp_hqd_pq_wptr_poll_addr_lo); + WREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR_HI, m->cp_hqd_pq_wptr_poll_addr_hi); + + WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, m->cp_hqd_pq_doorbell_control); + + WREG32(mmCP_HQD_VMID, m->cp_hqd_vmid); + + WREG32(mmCP_HQD_QUANTUM, m->cp_hqd_quantum); + + WREG32(mmCP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority); + WREG32(mmCP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority); + + WREG32(mmCP_HQD_IQ_RPTR, m->cp_hqd_iq_rptr); + + if (is_wptr_shadow_valid) + WREG32(mmCP_HQD_PQ_WPTR, wptr_shadow); + + WREG32(mmCP_HQD_ACTIVE, m->cp_hqd_active); + release_queue(kgd); + + return 0; +} + +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct cik_sdma_rlc_registers *m; + uint32_t sdma_base_addr; + + m = get_sdma_mqd(mqd); + sdma_base_addr = get_sdma_base_addr(m); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_VIRTUAL_ADDR, + m->sdma_rlc_virtual_addr); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, + m->sdma_rlc_rb_base); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE_HI, + m->sdma_rlc_rb_base_hi); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_LO, + m->sdma_rlc_rb_rptr_addr_lo); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_HI, + m->sdma_rlc_rb_rptr_addr_hi); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, + m->sdma_rlc_doorbell); + + WREG32(sdma_base_addr + mmSDMA0_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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t act; + bool retval = false; + uint32_t low, high; + + acquire_queue(kgd, pipe_id, queue_id); + act = RREG32(mmCP_HQD_ACTIVE); + if (act) { + low = lower_32_bits(queue_address >> 8); + high = upper_32_bits(queue_address >> 8); + + if (low == RREG32(mmCP_HQD_PQ_BASE) && + high == RREG32(mmCP_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 amdgpu_device *adev = get_amdgpu_device(kgd); + 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 = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL); + + if (sdma_rlc_rb_cntl & SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK) + return true; + + return false; +} + +static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t temp; + + acquire_queue(kgd, pipe_id, queue_id); + WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0); + + WREG32(mmCP_HQD_DEQUEUE_REQUEST, reset_type); + + while (true) { + temp = RREG32(mmCP_HQD_ACTIVE); + if (temp & CP_HQD_ACTIVE__ACTIVE__SHIFT) + 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 amdgpu_device *adev = get_amdgpu_device(kgd); + 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 = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL); + temp = temp & ~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK; + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, temp); + + while (true) { + temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); + if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT) + break; + if (timeout == 0) + return -ETIME; + msleep(20); + timeout -= 20; + } + + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, 0); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, 0); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, 0); + + return 0; +} + +static int kgd_address_watch_disable(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = get_amdgpu_device(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++) + WREG32(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) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + 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; + WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + + ADDRESS_WATCH_REG_CNTL], cntl.u32All); + + WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + + ADDRESS_WATCH_REG_ADDR_HI], addr_hi); + + WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + + ADDRESS_WATCH_REG_ADDR_LO], addr_lo); + + /* Enable the watch point */ + cntl.bitfields.valid = 1; + + WREG32(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 amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t data; + + mutex_lock(&adev->grbm_idx_mutex); + + WREG32(mmGRBM_GFX_INDEX, gfx_index_val); + WREG32(mmSQ_CMD, sq_cmd); + + /* Restore the GRBM_GFX_INDEX register */ + + data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK | + GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | + GRBM_GFX_INDEX__SE_BROADCAST_WRITES_MASK; + + WREG32(mmGRBM_GFX_INDEX, data); + + mutex_unlock(&adev->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]; +} + +static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, + uint8_t vmid) +{ + uint32_t reg; + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + reg = RREG32(mmATC_VMID0_PASID_MAPPING + vmid); + return reg & ATC_VMID0_PASID_MAPPING__VALID_MASK; +} + +static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, + uint8_t vmid) +{ + uint32_t reg; + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + reg = RREG32(mmATC_VMID0_PASID_MAPPING + vmid); + return reg & ATC_VMID0_PASID_MAPPING__VALID_MASK; +} + +static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid); +} + +static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + const union amdgpu_firmware_header *hdr; + + BUG_ON(kgd == NULL); + + switch (type) { + case KGD_ENGINE_PFP: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.pfp_fw->data; + break; + + case KGD_ENGINE_ME: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.me_fw->data; + break; + + case KGD_ENGINE_CE: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.ce_fw->data; + break; + + case KGD_ENGINE_MEC1: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.mec_fw->data; + break; + + case KGD_ENGINE_MEC2: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.mec2_fw->data; + break; + + case KGD_ENGINE_RLC: + hdr = (const union amdgpu_firmware_header *) + adev->gfx.rlc_fw->data; + break; + + case KGD_ENGINE_SDMA1: + hdr = (const union amdgpu_firmware_header *) + adev->sdma[0].fw->data; + break; + + case KGD_ENGINE_SDMA2: + hdr = (const union amdgpu_firmware_header *) + adev->sdma[1].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/amd/amdgpu/cikd.h b/drivers/gpu/drm/amd/amdgpu/cikd.h index 220865a..a3e3dfa 100644 --- a/drivers/gpu/drm/amd/amdgpu/cikd.h +++ b/drivers/gpu/drm/amd/amdgpu/cikd.h @@ -552,4 +552,16 @@ #define VCE_CMD_IB_AUTO 0x00000005 #define VCE_CMD_SEMAPHORE 0x00000006 +/* if PTR32, these are the bases for scratch and lds */ +#define PRIVATE_BASE(x) ((x) << 0) /* scratch */ +#define SHARED_BASE(x) ((x) << 16) /* LDS */ + +#define KFD_CIK_SDMA_QUEUE_OFFSET 0x200 + +/* valid for both DEFAULT_MTYPE and APE1_MTYPE */ +enum { + MTYPE_CACHED = 0, + MTYPE_NONCACHED = 3 +}; + #endif