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 
25 #include "drmP.h"
26 
27 #include "nouveau_drv.h"
28 #include "nouveau_vm.h"
29 
30 void
nvc0_vm_map_pgt(struct nouveau_gpuobj * pgd,u32 index,struct nouveau_gpuobj * pgt[2])31 nvc0_vm_map_pgt(struct nouveau_gpuobj *pgd, u32 index,
32 		struct nouveau_gpuobj *pgt[2])
33 {
34 	u32 pde[2] = { 0, 0 };
35 
36 	if (pgt[0])
37 		pde[1] = 0x00000001 | (pgt[0]->vinst >> 8);
38 	if (pgt[1])
39 		pde[0] = 0x00000001 | (pgt[1]->vinst >> 8);
40 
41 	nv_wo32(pgd, (index * 8) + 0, pde[0]);
42 	nv_wo32(pgd, (index * 8) + 4, pde[1]);
43 }
44 
45 static inline u64
nvc0_vm_addr(struct nouveau_vma * vma,u64 phys,u32 memtype,u32 target)46 nvc0_vm_addr(struct nouveau_vma *vma, u64 phys, u32 memtype, u32 target)
47 {
48 	phys >>= 8;
49 
50 	phys |= 0x00000001; /* present */
51 	if (vma->access & NV_MEM_ACCESS_SYS)
52 		phys |= 0x00000002;
53 
54 	phys |= ((u64)target  << 32);
55 	phys |= ((u64)memtype << 36);
56 
57 	return phys;
58 }
59 
60 void
nvc0_vm_map(struct nouveau_vma * vma,struct nouveau_gpuobj * pgt,struct nouveau_mem * mem,u32 pte,u32 cnt,u64 phys,u64 delta)61 nvc0_vm_map(struct nouveau_vma *vma, struct nouveau_gpuobj *pgt,
62 	    struct nouveau_mem *mem, u32 pte, u32 cnt, u64 phys, u64 delta)
63 {
64 	u32 next = 1 << (vma->node->type - 8);
65 
66 	phys  = nvc0_vm_addr(vma, phys, mem->memtype, 0);
67 	pte <<= 3;
68 	while (cnt--) {
69 		nv_wo32(pgt, pte + 0, lower_32_bits(phys));
70 		nv_wo32(pgt, pte + 4, upper_32_bits(phys));
71 		phys += next;
72 		pte  += 8;
73 	}
74 }
75 
76 void
nvc0_vm_map_sg(struct nouveau_vma * vma,struct nouveau_gpuobj * pgt,struct nouveau_mem * mem,u32 pte,u32 cnt,dma_addr_t * list)77 nvc0_vm_map_sg(struct nouveau_vma *vma, struct nouveau_gpuobj *pgt,
78 	       struct nouveau_mem *mem, u32 pte, u32 cnt, dma_addr_t *list)
79 {
80 	pte <<= 3;
81 	while (cnt--) {
82 		u64 phys = nvc0_vm_addr(vma, *list++, mem->memtype, 5);
83 		nv_wo32(pgt, pte + 0, lower_32_bits(phys));
84 		nv_wo32(pgt, pte + 4, upper_32_bits(phys));
85 		pte += 8;
86 	}
87 }
88 
89 void
nvc0_vm_unmap(struct nouveau_gpuobj * pgt,u32 pte,u32 cnt)90 nvc0_vm_unmap(struct nouveau_gpuobj *pgt, u32 pte, u32 cnt)
91 {
92 	pte <<= 3;
93 	while (cnt--) {
94 		nv_wo32(pgt, pte + 0, 0x00000000);
95 		nv_wo32(pgt, pte + 4, 0x00000000);
96 		pte += 8;
97 	}
98 }
99 
100 void
nvc0_vm_flush(struct nouveau_vm * vm)101 nvc0_vm_flush(struct nouveau_vm *vm)
102 {
103 	struct drm_nouveau_private *dev_priv = vm->dev->dev_private;
104 	struct nouveau_instmem_engine *pinstmem = &dev_priv->engine.instmem;
105 	struct drm_device *dev = vm->dev;
106 	struct nouveau_vm_pgd *vpgd;
107 	unsigned long flags;
108 	u32 engine;
109 
110 	engine = 1;
111 	if (vm == dev_priv->bar1_vm || vm == dev_priv->bar3_vm)
112 		engine |= 4;
113 
114 	pinstmem->flush(vm->dev);
115 
116 	spin_lock_irqsave(&dev_priv->vm_lock, flags);
117 	list_for_each_entry(vpgd, &vm->pgd_list, head) {
118 		/* looks like maybe a "free flush slots" counter, the
119 		 * faster you write to 0x100cbc to more it decreases
120 		 */
121 		if (!nv_wait_ne(dev, 0x100c80, 0x00ff0000, 0x00000000)) {
122 			NV_ERROR(dev, "vm timeout 0: 0x%08x %d\n",
123 				 nv_rd32(dev, 0x100c80), engine);
124 		}
125 		nv_wr32(dev, 0x100cb8, vpgd->obj->vinst >> 8);
126 		nv_wr32(dev, 0x100cbc, 0x80000000 | engine);
127 		/* wait for flush to be queued? */
128 		if (!nv_wait(dev, 0x100c80, 0x00008000, 0x00008000)) {
129 			NV_ERROR(dev, "vm timeout 1: 0x%08x %d\n",
130 				 nv_rd32(dev, 0x100c80), engine);
131 		}
132 	}
133 	spin_unlock_irqrestore(&dev_priv->vm_lock, flags);
134 }
135