Jerome Glisse | 771fe6b | 2009-06-05 14:42:42 +0200 | [diff] [blame] | 1 | /* |
| 2 | * Copyright 2008 Advanced Micro Devices, Inc. |
| 3 | * Copyright 2008 Red Hat Inc. |
| 4 | * Copyright 2009 Jerome Glisse. |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a |
| 7 | * copy of this software and associated documentation files (the "Software"), |
| 8 | * to deal in the Software without restriction, including without limitation |
| 9 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
| 10 | * and/or sell copies of the Software, and to permit persons to whom the |
| 11 | * Software is furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in |
| 14 | * all copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
| 19 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
| 20 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
| 21 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| 22 | * OTHER DEALINGS IN THE SOFTWARE. |
| 23 | * |
| 24 | * Authors: Dave Airlie |
| 25 | * Alex Deucher |
| 26 | * Jerome Glisse |
| 27 | */ |
| 28 | #include "drmP.h" |
| 29 | #include "radeon_drm.h" |
| 30 | #include "radeon.h" |
| 31 | #include "radeon_reg.h" |
| 32 | |
| 33 | /* |
| 34 | * Common GART table functions. |
| 35 | */ |
| 36 | int radeon_gart_table_ram_alloc(struct radeon_device *rdev) |
| 37 | { |
| 38 | void *ptr; |
| 39 | |
| 40 | ptr = pci_alloc_consistent(rdev->pdev, rdev->gart.table_size, |
| 41 | &rdev->gart.table_addr); |
| 42 | if (ptr == NULL) { |
| 43 | return -ENOMEM; |
| 44 | } |
| 45 | #ifdef CONFIG_X86 |
| 46 | if (rdev->family == CHIP_RS400 || rdev->family == CHIP_RS480 || |
| 47 | rdev->family == CHIP_RS690 || rdev->family == CHIP_RS740) { |
| 48 | set_memory_uc((unsigned long)ptr, |
| 49 | rdev->gart.table_size >> PAGE_SHIFT); |
| 50 | } |
| 51 | #endif |
| 52 | rdev->gart.table.ram.ptr = ptr; |
| 53 | memset((void *)rdev->gart.table.ram.ptr, 0, rdev->gart.table_size); |
| 54 | return 0; |
| 55 | } |
| 56 | |
| 57 | void radeon_gart_table_ram_free(struct radeon_device *rdev) |
| 58 | { |
| 59 | if (rdev->gart.table.ram.ptr == NULL) { |
| 60 | return; |
| 61 | } |
| 62 | #ifdef CONFIG_X86 |
| 63 | if (rdev->family == CHIP_RS400 || rdev->family == CHIP_RS480 || |
| 64 | rdev->family == CHIP_RS690 || rdev->family == CHIP_RS740) { |
| 65 | set_memory_wb((unsigned long)rdev->gart.table.ram.ptr, |
| 66 | rdev->gart.table_size >> PAGE_SHIFT); |
| 67 | } |
| 68 | #endif |
| 69 | pci_free_consistent(rdev->pdev, rdev->gart.table_size, |
| 70 | (void *)rdev->gart.table.ram.ptr, |
| 71 | rdev->gart.table_addr); |
| 72 | rdev->gart.table.ram.ptr = NULL; |
| 73 | rdev->gart.table_addr = 0; |
| 74 | } |
| 75 | |
| 76 | int radeon_gart_table_vram_alloc(struct radeon_device *rdev) |
| 77 | { |
| 78 | uint64_t gpu_addr; |
| 79 | int r; |
| 80 | |
| 81 | if (rdev->gart.table.vram.robj == NULL) { |
| 82 | r = radeon_object_create(rdev, NULL, |
| 83 | rdev->gart.table_size, |
| 84 | true, |
| 85 | RADEON_GEM_DOMAIN_VRAM, |
| 86 | false, &rdev->gart.table.vram.robj); |
| 87 | if (r) { |
| 88 | return r; |
| 89 | } |
| 90 | } |
| 91 | r = radeon_object_pin(rdev->gart.table.vram.robj, |
| 92 | RADEON_GEM_DOMAIN_VRAM, &gpu_addr); |
| 93 | if (r) { |
| 94 | radeon_object_unref(&rdev->gart.table.vram.robj); |
| 95 | return r; |
| 96 | } |
| 97 | r = radeon_object_kmap(rdev->gart.table.vram.robj, |
| 98 | (void **)&rdev->gart.table.vram.ptr); |
| 99 | if (r) { |
| 100 | radeon_object_unpin(rdev->gart.table.vram.robj); |
| 101 | radeon_object_unref(&rdev->gart.table.vram.robj); |
| 102 | DRM_ERROR("radeon: failed to map gart vram table.\n"); |
| 103 | return r; |
| 104 | } |
| 105 | rdev->gart.table_addr = gpu_addr; |
| 106 | return 0; |
| 107 | } |
| 108 | |
| 109 | void radeon_gart_table_vram_free(struct radeon_device *rdev) |
| 110 | { |
| 111 | if (rdev->gart.table.vram.robj == NULL) { |
| 112 | return; |
| 113 | } |
| 114 | radeon_object_kunmap(rdev->gart.table.vram.robj); |
| 115 | radeon_object_unpin(rdev->gart.table.vram.robj); |
| 116 | radeon_object_unref(&rdev->gart.table.vram.robj); |
| 117 | } |
| 118 | |
| 119 | |
| 120 | |
| 121 | |
| 122 | /* |
| 123 | * Common gart functions. |
| 124 | */ |
| 125 | void radeon_gart_unbind(struct radeon_device *rdev, unsigned offset, |
| 126 | int pages) |
| 127 | { |
| 128 | unsigned t; |
| 129 | unsigned p; |
| 130 | int i, j; |
| 131 | |
| 132 | if (!rdev->gart.ready) { |
| 133 | WARN(1, "trying to unbind memory to unitialized GART !\n"); |
| 134 | return; |
| 135 | } |
| 136 | t = offset / 4096; |
| 137 | p = t / (PAGE_SIZE / 4096); |
| 138 | for (i = 0; i < pages; i++, p++) { |
| 139 | if (rdev->gart.pages[p]) { |
| 140 | pci_unmap_page(rdev->pdev, rdev->gart.pages_addr[p], |
| 141 | PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); |
| 142 | rdev->gart.pages[p] = NULL; |
| 143 | rdev->gart.pages_addr[p] = 0; |
| 144 | for (j = 0; j < (PAGE_SIZE / 4096); j++, t++) { |
| 145 | radeon_gart_set_page(rdev, t, 0); |
| 146 | } |
| 147 | } |
| 148 | } |
| 149 | mb(); |
| 150 | radeon_gart_tlb_flush(rdev); |
| 151 | } |
| 152 | |
| 153 | int radeon_gart_bind(struct radeon_device *rdev, unsigned offset, |
| 154 | int pages, struct page **pagelist) |
| 155 | { |
| 156 | unsigned t; |
| 157 | unsigned p; |
| 158 | uint64_t page_base; |
| 159 | int i, j; |
| 160 | |
| 161 | if (!rdev->gart.ready) { |
| 162 | DRM_ERROR("trying to bind memory to unitialized GART !\n"); |
| 163 | return -EINVAL; |
| 164 | } |
| 165 | t = offset / 4096; |
| 166 | p = t / (PAGE_SIZE / 4096); |
| 167 | |
| 168 | for (i = 0; i < pages; i++, p++) { |
| 169 | /* we need to support large memory configurations */ |
| 170 | /* assume that unbind have already been call on the range */ |
| 171 | rdev->gart.pages_addr[p] = pci_map_page(rdev->pdev, pagelist[i], |
| 172 | 0, PAGE_SIZE, |
| 173 | PCI_DMA_BIDIRECTIONAL); |
| 174 | if (pci_dma_mapping_error(rdev->pdev, rdev->gart.pages_addr[p])) { |
| 175 | /* FIXME: failed to map page (return -ENOMEM?) */ |
| 176 | radeon_gart_unbind(rdev, offset, pages); |
| 177 | return -ENOMEM; |
| 178 | } |
| 179 | rdev->gart.pages[p] = pagelist[i]; |
| 180 | page_base = (uint32_t)rdev->gart.pages_addr[p]; |
| 181 | for (j = 0; j < (PAGE_SIZE / 4096); j++, t++) { |
| 182 | radeon_gart_set_page(rdev, t, page_base); |
| 183 | page_base += 4096; |
| 184 | } |
| 185 | } |
| 186 | mb(); |
| 187 | radeon_gart_tlb_flush(rdev); |
| 188 | return 0; |
| 189 | } |
| 190 | |
| 191 | int radeon_gart_init(struct radeon_device *rdev) |
| 192 | { |
| 193 | if (rdev->gart.pages) { |
| 194 | return 0; |
| 195 | } |
| 196 | /* We need PAGE_SIZE >= 4096 */ |
| 197 | if (PAGE_SIZE < 4096) { |
| 198 | DRM_ERROR("Page size is smaller than GPU page size!\n"); |
| 199 | return -EINVAL; |
| 200 | } |
| 201 | /* Compute table size */ |
| 202 | rdev->gart.num_cpu_pages = rdev->mc.gtt_size / PAGE_SIZE; |
| 203 | rdev->gart.num_gpu_pages = rdev->mc.gtt_size / 4096; |
| 204 | DRM_INFO("GART: num cpu pages %u, num gpu pages %u\n", |
| 205 | rdev->gart.num_cpu_pages, rdev->gart.num_gpu_pages); |
| 206 | /* Allocate pages table */ |
| 207 | rdev->gart.pages = kzalloc(sizeof(void *) * rdev->gart.num_cpu_pages, |
| 208 | GFP_KERNEL); |
| 209 | if (rdev->gart.pages == NULL) { |
| 210 | radeon_gart_fini(rdev); |
| 211 | return -ENOMEM; |
| 212 | } |
| 213 | rdev->gart.pages_addr = kzalloc(sizeof(dma_addr_t) * |
| 214 | rdev->gart.num_cpu_pages, GFP_KERNEL); |
| 215 | if (rdev->gart.pages_addr == NULL) { |
| 216 | radeon_gart_fini(rdev); |
| 217 | return -ENOMEM; |
| 218 | } |
| 219 | return 0; |
| 220 | } |
| 221 | |
| 222 | void radeon_gart_fini(struct radeon_device *rdev) |
| 223 | { |
| 224 | if (rdev->gart.pages && rdev->gart.pages_addr && rdev->gart.ready) { |
| 225 | /* unbind pages */ |
| 226 | radeon_gart_unbind(rdev, 0, rdev->gart.num_cpu_pages); |
| 227 | } |
| 228 | rdev->gart.ready = false; |
| 229 | kfree(rdev->gart.pages); |
| 230 | kfree(rdev->gart.pages_addr); |
| 231 | rdev->gart.pages = NULL; |
| 232 | rdev->gart.pages_addr = NULL; |
| 233 | } |