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

   1 /*
   2  * Copyright 2014 Advanced Micro Devices, Inc.
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice shall be included in
  12  * all copies or substantial portions of the Software.
  13  *
  14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  15  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  17  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
  18  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
  19  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  20  * OTHER DEALINGS IN THE SOFTWARE.
  21  *
  22  */
  23
  24 #include "kfd_device_queue_manager.h"
  25 #include "cik_regs.h"
  26
  27 static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm,
  28                                    struct qcm_process_device *qpd,
  29                                    enum cache_policy default_policy,
  30                                    enum cache_policy alternate_policy,
  31                                    void __user *alternate_aperture_base,
  32                                    uint64_t alternate_aperture_size);
  33 static int register_process_cik(struct device_queue_manager *dqm,
  34                                         struct qcm_process_device *qpd);
  35 static int initialize_cpsch_cik(struct device_queue_manager *dqm);
  36 static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
  37                                 struct qcm_process_device *qpd);
  38
  39 void device_queue_manager_init_cik(struct device_queue_manager_asic_ops *ops)
  40 {
  41         ops->set_cache_memory_policy = set_cache_memory_policy_cik;
  42         ops->register_process = register_process_cik;
  43         ops->initialize = initialize_cpsch_cik;
  44         ops->init_sdma_vm = init_sdma_vm;
  45 }
  46
  47 static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble)
  48 {
  49         /* In 64-bit mode, we can only control the top 3 bits of the LDS,
  50          * scratch and GPUVM apertures.
  51          * The hardware fills in the remaining 59 bits according to the
  52          * following pattern:
  53          * LDS:         X0000000'00000000 - X0000001'00000000 (4GB)
  54          * Scratch:     X0000001'00000000 - X0000002'00000000 (4GB)
  55          * GPUVM:       Y0010000'00000000 - Y0020000'00000000 (1TB)
  56          *
  57          * (where X/Y is the configurable nybble with the low-bit 0)
  58          *
  59          * LDS and scratch will have the same top nybble programmed in the
  60          * top 3 bits of SH_MEM_BASES.PRIVATE_BASE.
  61          * GPUVM can have a different top nybble programmed in the
  62          * top 3 bits of SH_MEM_BASES.SHARED_BASE.
  63          * We don't bother to support different top nybbles
  64          * for LDS/Scratch and GPUVM.
  65          */
  66
  67         BUG_ON((top_address_nybble & 1) || top_address_nybble > 0xE ||
  68                 top_address_nybble == 0);
  69
  70         return PRIVATE_BASE(top_address_nybble << 12) |
  71                         SHARED_BASE(top_address_nybble << 12);
  72 }
  73
  74 static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm,
  75                                    struct qcm_process_device *qpd,
  76                                    enum cache_policy default_policy,
  77                                    enum cache_policy alternate_policy,
  78                                    void __user *alternate_aperture_base,
  79                                    uint64_t alternate_aperture_size)
  80 {
  81         uint32_t default_mtype;
  82         uint32_t ape1_mtype;
  83
  84         default_mtype = (default_policy == cache_policy_coherent) ?
  85                         MTYPE_NONCACHED :
  86                         MTYPE_CACHED;
  87
  88         ape1_mtype = (alternate_policy == cache_policy_coherent) ?
  89                         MTYPE_NONCACHED :
  90                         MTYPE_CACHED;
  91
  92         qpd->sh_mem_config = (qpd->sh_mem_config & PTR32)
  93                         | ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
  94                         | DEFAULT_MTYPE(default_mtype)
  95                         | APE1_MTYPE(ape1_mtype);
  96
  97         return true;
  98 }
  99
 100 static int register_process_cik(struct device_queue_manager *dqm,
 101                 struct qcm_process_device *qpd)
 102 {
 103         struct kfd_process_device *pdd;
 104         unsigned int temp;
 105
 106         BUG_ON(!dqm || !qpd);
 107
 108         pdd = qpd_to_pdd(qpd);
 109
 110         /* check if sh_mem_config register already configured */
 111         if (qpd->sh_mem_config == 0) {
 112                 qpd->sh_mem_config =
 113                         ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) |
 114                         DEFAULT_MTYPE(MTYPE_NONCACHED) |
 115                         APE1_MTYPE(MTYPE_NONCACHED);
 116                 qpd->sh_mem_ape1_limit = 0;
 117                 qpd->sh_mem_ape1_base = 0;
 118         }
 119
 120         if (qpd->pqm->process->is_32bit_user_mode) {
 121                 temp = get_sh_mem_bases_32(pdd);
 122                 qpd->sh_mem_bases = SHARED_BASE(temp);
 123                 qpd->sh_mem_config |= PTR32;
 124         } else {
 125                 temp = get_sh_mem_bases_nybble_64(pdd);
 126                 qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp);
 127         }
 128
 129         pr_debug("kfd: is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n",
 130                 qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases);
 131
 132         return 0;
 133 }
 134
 135 static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
 136                                 struct qcm_process_device *qpd)
 137 {
 138         uint32_t value = SDMA_ATC;
 139
 140         if (q->process->is_32bit_user_mode)
 141                 value |= SDMA_VA_PTR32 | get_sh_mem_bases_32(qpd_to_pdd(qpd));
 142         else
 143                 value |= SDMA_VA_SHARED_BASE(get_sh_mem_bases_nybble_64(
 144                                                         qpd_to_pdd(qpd)));
 145         q->properties.sdma_vm_addr = value;
 146 }
 147
 148 static int initialize_cpsch_cik(struct device_queue_manager *dqm)
 149 {
 150         return init_pipelines(dqm, get_pipes_num(dqm), get_first_pipe(dqm));
 151 }