nv50_vm.c 4.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182
  1. /*
  2. * Copyright 2010 Red Hat 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. * Authors: Ben Skeggs
  23. */
  24. #include "drmP.h"
  25. #include "nouveau_drv.h"
  26. #include "nouveau_vm.h"
  27. void
  28. nv50_vm_map_pgt(struct nouveau_gpuobj *pgd, u32 pde,
  29. struct nouveau_gpuobj *pgt[2])
  30. {
  31. u64 phys = 0xdeadcafe00000000ULL;
  32. u32 coverage = 0;
  33. if (pgt[0]) {
  34. phys = 0x00000003 | pgt[0]->vinst; /* present, 4KiB pages */
  35. coverage = (pgt[0]->size >> 3) << 12;
  36. } else
  37. if (pgt[1]) {
  38. phys = 0x00000001 | pgt[1]->vinst; /* present */
  39. coverage = (pgt[1]->size >> 3) << 16;
  40. }
  41. if (phys & 1) {
  42. if (coverage <= 32 * 1024 * 1024)
  43. phys |= 0x60;
  44. else if (coverage <= 64 * 1024 * 1024)
  45. phys |= 0x40;
  46. else if (coverage < 128 * 1024 * 1024)
  47. phys |= 0x20;
  48. }
  49. nv_wo32(pgd, (pde * 8) + 0, lower_32_bits(phys));
  50. nv_wo32(pgd, (pde * 8) + 4, upper_32_bits(phys));
  51. }
  52. static inline u64
  53. nv50_vm_addr(struct nouveau_vma *vma, u64 phys, u32 memtype, u32 target)
  54. {
  55. struct drm_nouveau_private *dev_priv = vma->vm->dev->dev_private;
  56. phys |= 1; /* present */
  57. phys |= (u64)memtype << 40;
  58. /* IGPs don't have real VRAM, re-target to stolen system memory */
  59. if (target == 0 && dev_priv->vram_sys_base) {
  60. phys += dev_priv->vram_sys_base;
  61. target = 3;
  62. }
  63. phys |= target << 4;
  64. if (vma->access & NV_MEM_ACCESS_SYS)
  65. phys |= (1 << 6);
  66. if (!(vma->access & NV_MEM_ACCESS_WO))
  67. phys |= (1 << 3);
  68. return phys;
  69. }
  70. void
  71. nv50_vm_map(struct nouveau_vma *vma, struct nouveau_gpuobj *pgt,
  72. struct nouveau_mem *mem, u32 pte, u32 cnt, u64 phys, u64 delta)
  73. {
  74. u32 comp = (mem->memtype & 0x180) >> 7;
  75. u32 block;
  76. int i;
  77. phys = nv50_vm_addr(vma, phys, mem->memtype, 0);
  78. pte <<= 3;
  79. cnt <<= 3;
  80. while (cnt) {
  81. u32 offset_h = upper_32_bits(phys);
  82. u32 offset_l = lower_32_bits(phys);
  83. for (i = 7; i >= 0; i--) {
  84. block = 1 << (i + 3);
  85. if (cnt >= block && !(pte & (block - 1)))
  86. break;
  87. }
  88. offset_l |= (i << 7);
  89. phys += block << (vma->node->type - 3);
  90. cnt -= block;
  91. if (comp) {
  92. u32 tag = mem->tag->start + ((delta >> 16) * comp);
  93. offset_h |= (tag << 17);
  94. delta += block << (vma->node->type - 3);
  95. }
  96. while (block) {
  97. nv_wo32(pgt, pte + 0, offset_l);
  98. nv_wo32(pgt, pte + 4, offset_h);
  99. pte += 8;
  100. block -= 8;
  101. }
  102. }
  103. }
  104. void
  105. nv50_vm_map_sg(struct nouveau_vma *vma, struct nouveau_gpuobj *pgt,
  106. struct nouveau_mem *mem, u32 pte, u32 cnt, dma_addr_t *list)
  107. {
  108. pte <<= 3;
  109. while (cnt--) {
  110. u64 phys = nv50_vm_addr(vma, (u64)*list++, mem->memtype, 2);
  111. nv_wo32(pgt, pte + 0, lower_32_bits(phys));
  112. nv_wo32(pgt, pte + 4, upper_32_bits(phys));
  113. pte += 8;
  114. }
  115. }
  116. void
  117. nv50_vm_unmap(struct nouveau_gpuobj *pgt, u32 pte, u32 cnt)
  118. {
  119. pte <<= 3;
  120. while (cnt--) {
  121. nv_wo32(pgt, pte + 0, 0x00000000);
  122. nv_wo32(pgt, pte + 4, 0x00000000);
  123. pte += 8;
  124. }
  125. }
  126. void
  127. nv50_vm_flush(struct nouveau_vm *vm)
  128. {
  129. struct drm_nouveau_private *dev_priv = vm->dev->dev_private;
  130. struct nouveau_instmem_engine *pinstmem = &dev_priv->engine.instmem;
  131. struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
  132. int i;
  133. pinstmem->flush(vm->dev);
  134. /* BAR */
  135. if (vm == dev_priv->bar1_vm || vm == dev_priv->bar3_vm) {
  136. nv50_vm_flush_engine(vm->dev, 6);
  137. return;
  138. }
  139. pfifo->tlb_flush(vm->dev);
  140. for (i = 0; i < NVOBJ_ENGINE_NR; i++) {
  141. if (atomic_read(&vm->engref[i]))
  142. dev_priv->eng[i]->tlb_flush(vm->dev, i);
  143. }
  144. }
  145. void
  146. nv50_vm_flush_engine(struct drm_device *dev, int engine)
  147. {
  148. struct drm_nouveau_private *dev_priv = dev->dev_private;
  149. unsigned long flags;
  150. spin_lock_irqsave(&dev_priv->vm_lock, flags);
  151. nv_wr32(dev, 0x100c80, (engine << 16) | 1);
  152. if (!nv_wait(dev, 0x100c80, 0x00000001, 0x00000000))
  153. NV_ERROR(dev, "vm flush timeout: engine %d\n", engine);
  154. spin_unlock_irqrestore(&dev_priv->vm_lock, flags);
  155. }