diff options
Diffstat (limited to 'drivers/gpu/nvgpu/gv11b/subctx_gv11b.c')
-rw-r--r-- | drivers/gpu/nvgpu/gv11b/subctx_gv11b.c | 147 |
1 files changed, 147 insertions, 0 deletions
diff --git a/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c b/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c new file mode 100644 index 00000000..3acc53f6 --- /dev/null +++ b/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c | |||
@@ -0,0 +1,147 @@ | |||
1 | /* | ||
2 | * Volta GPU series Subcontext | ||
3 | * | ||
4 | * Copyright (c) 2016, NVIDIA CORPORATION. All rights reserved. | ||
5 | * | ||
6 | * This program is free software; you can redistribute it and/or modify it | ||
7 | * under the terms and conditions of the GNU General Public License, | ||
8 | * version 2, as published by the Free Software Foundation. | ||
9 | * | ||
10 | * This program is distributed in the hope it will be useful, but WITHOUT | ||
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | ||
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | ||
13 | * more details. | ||
14 | * | ||
15 | * You should have received a copy of the GNU General Public License along with | ||
16 | * this program. | ||
17 | */ | ||
18 | |||
19 | #include "gk20a/gk20a.h" | ||
20 | #include "gk20a/semaphore_gk20a.h" | ||
21 | #include "gv11b/subctx_gv11b.h" | ||
22 | #include "gv11b/hw_ram_gv11b.h" | ||
23 | #include "gv11b/hw_ctxsw_prog_gv11b.h" | ||
24 | |||
25 | static void gv11b_init_subcontext_pdb(struct channel_gk20a *c, | ||
26 | struct mem_desc *inst_block); | ||
27 | |||
28 | void gv11b_free_subctx_header(struct channel_gk20a *c) | ||
29 | { | ||
30 | struct ctx_header_desc *ctx = &c->ch_ctx.ctx_header; | ||
31 | struct gk20a *g = c->g; | ||
32 | |||
33 | gk20a_dbg_fn(""); | ||
34 | |||
35 | if (ctx->mem.gpu_va) { | ||
36 | gk20a_gmmu_unmap(c->vm, ctx->mem.gpu_va, | ||
37 | ctx->mem.size, gk20a_mem_flag_none); | ||
38 | |||
39 | gk20a_gmmu_free_attr(g, DMA_ATTR_NO_KERNEL_MAPPING, &ctx->mem); | ||
40 | } | ||
41 | } | ||
42 | |||
43 | int gv11b_alloc_subctx_header(struct channel_gk20a *c) | ||
44 | { | ||
45 | struct ctx_header_desc *ctx = &c->ch_ctx.ctx_header; | ||
46 | struct gk20a *g = c->g; | ||
47 | int ret = 0; | ||
48 | |||
49 | gk20a_dbg_fn(""); | ||
50 | |||
51 | if (ctx->mem.gpu_va == 0) { | ||
52 | ret = gk20a_gmmu_alloc_attr_sys(g, | ||
53 | DMA_ATTR_NO_KERNEL_MAPPING, | ||
54 | ctxsw_prog_fecs_header_v(), | ||
55 | &ctx->mem); | ||
56 | if (ret) { | ||
57 | gk20a_err(dev_from_gk20a(g), | ||
58 | "failed to allocate sub ctx header"); | ||
59 | return ret; | ||
60 | } | ||
61 | ctx->mem.gpu_va = gk20a_gmmu_map(c->vm, | ||
62 | &ctx->mem.sgt, | ||
63 | ctx->mem.size, | ||
64 | NVGPU_MAP_BUFFER_FLAGS_CACHEABLE_TRUE, | ||
65 | gk20a_mem_flag_none, true, | ||
66 | ctx->mem.aperture); | ||
67 | if (!ctx->mem.gpu_va) { | ||
68 | gk20a_err(dev_from_gk20a(g), | ||
69 | "failed to map ctx header"); | ||
70 | gk20a_gmmu_free_attr(g, DMA_ATTR_NO_KERNEL_MAPPING, | ||
71 | &ctx->mem); | ||
72 | return -ENOMEM; | ||
73 | } | ||
74 | /* Now clear the buffer */ | ||
75 | if (gk20a_mem_begin(g, &ctx->mem)) | ||
76 | return -ENOMEM; | ||
77 | |||
78 | gk20a_memset(g, &ctx->mem, 0, 0, ctx->mem.size); | ||
79 | gk20a_mem_end(g, &ctx->mem); | ||
80 | |||
81 | gv11b_init_subcontext_pdb(c, &c->inst_block); | ||
82 | |||
83 | } | ||
84 | return ret; | ||
85 | } | ||
86 | |||
87 | static void gv11b_init_subcontext_pdb(struct channel_gk20a *c, | ||
88 | struct mem_desc *inst_block) | ||
89 | { | ||
90 | struct gk20a *g = c->g; | ||
91 | struct vm_gk20a *vm; | ||
92 | u64 pdb_addr, pdb_addr_lo, pdb_addr_hi; | ||
93 | u32 format_word; | ||
94 | u32 lo, hi; | ||
95 | |||
96 | gk20a_dbg_fn(""); | ||
97 | /* load main pdb as veid0 pdb also */ | ||
98 | vm = c->vm; | ||
99 | pdb_addr = g->ops.mm.get_iova_addr(g, vm->pdb.mem.sgt->sgl, 0); | ||
100 | pdb_addr_lo = u64_lo32(pdb_addr >> ram_in_base_shift_v()); | ||
101 | pdb_addr_hi = u64_hi32(pdb_addr); | ||
102 | format_word = ram_in_sc_page_dir_base_target_f( | ||
103 | ram_in_sc_page_dir_base_target_sys_mem_ncoh_v(), 0) | | ||
104 | ram_in_sc_page_dir_base_vol_f( | ||
105 | ram_in_sc_page_dir_base_vol_true_v(), 0) | | ||
106 | ram_in_sc_page_dir_base_fault_replay_tex_f(0, 0) | | ||
107 | ram_in_sc_page_dir_base_fault_replay_gcc_f(0, 0) | | ||
108 | ram_in_sc_use_ver2_pt_format_f(1, 0) | | ||
109 | ram_in_sc_big_page_size_f(1, 0) | | ||
110 | ram_in_sc_page_dir_base_lo_0_f(pdb_addr_lo); | ||
111 | lo = ram_in_sc_page_dir_base_vol_0_w(); | ||
112 | hi = ram_in_sc_page_dir_base_hi_0_w(); | ||
113 | gk20a_mem_wr32(g, inst_block, lo, format_word); | ||
114 | gk20a_mem_wr32(g, inst_block, hi, pdb_addr_hi); | ||
115 | |||
116 | /* make subcontext0 address space to valid */ | ||
117 | /* TODO fix proper hw register definations */ | ||
118 | gk20a_mem_wr32(g, inst_block, 166, 0x1); | ||
119 | gk20a_mem_wr32(g, inst_block, 167, 0); | ||
120 | gk20a_mem_wr32(g, inst_block, ram_in_engine_wfi_veid_w(), | ||
121 | ram_in_engine_wfi_veid_f(0)); | ||
122 | |||
123 | } | ||
124 | |||
125 | int gv11b_update_subctx_header(struct channel_gk20a *c, u64 gpu_va) | ||
126 | { | ||
127 | struct ctx_header_desc *ctx = &c->ch_ctx.ctx_header; | ||
128 | struct mem_desc *gr_mem; | ||
129 | struct gk20a *g = c->g; | ||
130 | int ret = 0; | ||
131 | u32 addr_lo, addr_hi; | ||
132 | |||
133 | addr_lo = u64_lo32(gpu_va); | ||
134 | addr_hi = u64_hi32(gpu_va); | ||
135 | |||
136 | gr_mem = &ctx->mem; | ||
137 | g->ops.mm.l2_flush(g, true); | ||
138 | if (gk20a_mem_begin(g, gr_mem)) | ||
139 | return -ENOMEM; | ||
140 | |||
141 | gk20a_mem_wr(g, gr_mem, | ||
142 | ctxsw_prog_main_image_context_buffer_ptr_hi_o(), addr_hi); | ||
143 | gk20a_mem_wr(g, gr_mem, | ||
144 | ctxsw_prog_main_image_context_buffer_ptr_o(), addr_lo); | ||
145 | gk20a_mem_end(g, gr_mem); | ||
146 | return ret; | ||
147 | } | ||