Go to most recent revision | Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
5564 | serge | 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 (including the next |
||
12 | * paragraph) shall be included in all copies or substantial portions of the |
||
13 | * Software. |
||
14 | * |
||
15 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
||
16 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
||
17 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
||
18 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
||
19 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
||
20 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
||
21 | * SOFTWARE. |
||
22 | * |
||
23 | * Authors: Tom Stellard |
||
24 | * |
||
25 | */ |
||
26 | |||
27 | #include "radeon_elf_util.h" |
||
28 | #include "r600_pipe_common.h" |
||
29 | |||
30 | #include "util/u_memory.h" |
||
31 | |||
32 | #include |
||
33 | #include |
||
34 | #include |
||
35 | |||
36 | static void parse_symbol_table(Elf_Data *symbol_table_data, |
||
37 | const GElf_Shdr *symbol_table_header, |
||
38 | struct radeon_shader_binary *binary) |
||
39 | { |
||
40 | GElf_Sym symbol; |
||
41 | unsigned i = 0; |
||
42 | unsigned symbol_count = |
||
43 | symbol_table_header->sh_size / symbol_table_header->sh_entsize; |
||
44 | |||
45 | /* We are over allocating this list, because symbol_count gives the |
||
46 | * total number of symbols, and we will only be filling the list |
||
47 | * with offsets of global symbols. The memory savings from |
||
48 | * allocating the correct size of this list will be small, and |
||
49 | * I don't think it is worth the cost of pre-computing the number |
||
50 | * of global symbols. |
||
51 | */ |
||
52 | binary->global_symbol_offsets = CALLOC(symbol_count, sizeof(uint64_t)); |
||
53 | |||
54 | while (gelf_getsym(symbol_table_data, i++, &symbol)) { |
||
55 | unsigned i; |
||
56 | if (GELF_ST_BIND(symbol.st_info) != STB_GLOBAL || |
||
57 | symbol.st_shndx == 0 /* Undefined symbol */) { |
||
58 | continue; |
||
59 | } |
||
60 | |||
61 | binary->global_symbol_offsets[binary->global_symbol_count] = |
||
62 | symbol.st_value; |
||
63 | |||
64 | /* Sort the list using bubble sort. This list will usually |
||
65 | * be small. */ |
||
66 | for (i = binary->global_symbol_count; i > 0; --i) { |
||
67 | uint64_t lhs = binary->global_symbol_offsets[i - 1]; |
||
68 | uint64_t rhs = binary->global_symbol_offsets[i]; |
||
69 | if (lhs < rhs) { |
||
70 | break; |
||
71 | } |
||
72 | binary->global_symbol_offsets[i] = lhs; |
||
73 | binary->global_symbol_offsets[i - 1] = rhs; |
||
74 | } |
||
75 | ++binary->global_symbol_count; |
||
76 | } |
||
77 | } |
||
78 | |||
79 | static void parse_relocs(Elf *elf, Elf_Data *relocs, Elf_Data *symbols, |
||
80 | unsigned symbol_sh_link, |
||
81 | struct radeon_shader_binary *binary) |
||
82 | { |
||
83 | unsigned i; |
||
84 | |||
85 | if (!relocs || !symbols || !binary->reloc_count) { |
||
86 | return; |
||
87 | } |
||
88 | binary->relocs = CALLOC(binary->reloc_count, |
||
89 | sizeof(struct radeon_shader_reloc)); |
||
90 | for (i = 0; i < binary->reloc_count; i++) { |
||
91 | GElf_Sym symbol; |
||
92 | GElf_Rel rel; |
||
93 | char *symbol_name; |
||
94 | struct radeon_shader_reloc *reloc = &binary->relocs[i]; |
||
95 | |||
96 | gelf_getrel(relocs, i, &rel); |
||
97 | gelf_getsym(symbols, GELF_R_SYM(rel.r_info), &symbol); |
||
98 | symbol_name = elf_strptr(elf, symbol_sh_link, symbol.st_name); |
||
99 | |||
100 | reloc->offset = rel.r_offset; |
||
101 | reloc->name = strdup(symbol_name); |
||
102 | } |
||
103 | } |
||
104 | |||
105 | void radeon_elf_read(const char *elf_data, unsigned elf_size, |
||
106 | struct radeon_shader_binary *binary, |
||
107 | unsigned debug) |
||
108 | { |
||
109 | char *elf_buffer; |
||
110 | Elf *elf; |
||
111 | Elf_Scn *section = NULL; |
||
112 | Elf_Data *symbols = NULL, *relocs = NULL; |
||
113 | size_t section_str_index; |
||
114 | unsigned symbol_sh_link = 0; |
||
115 | |||
116 | /* One of the libelf implementations |
||
117 | * (http://www.mr511.de/software/english.htm) requires calling |
||
118 | * elf_version() before elf_memory(). |
||
119 | */ |
||
120 | elf_version(EV_CURRENT); |
||
121 | elf_buffer = MALLOC(elf_size); |
||
122 | memcpy(elf_buffer, elf_data, elf_size); |
||
123 | |||
124 | elf = elf_memory(elf_buffer, elf_size); |
||
125 | |||
126 | elf_getshdrstrndx(elf, §ion_str_index); |
||
127 | binary->disassembled = 0; |
||
128 | |||
129 | while ((section = elf_nextscn(elf, section))) { |
||
130 | const char *name; |
||
131 | Elf_Data *section_data = NULL; |
||
132 | GElf_Shdr section_header; |
||
133 | if (gelf_getshdr(section, §ion_header) != §ion_header) { |
||
134 | fprintf(stderr, "Failed to read ELF section header\n"); |
||
135 | return; |
||
136 | } |
||
137 | name = elf_strptr(elf, section_str_index, section_header.sh_name); |
||
138 | if (!strcmp(name, ".text")) { |
||
139 | section_data = elf_getdata(section, section_data); |
||
140 | binary->code_size = section_data->d_size; |
||
141 | binary->code = MALLOC(binary->code_size * sizeof(unsigned char)); |
||
142 | memcpy(binary->code, section_data->d_buf, binary->code_size); |
||
143 | } else if (!strcmp(name, ".AMDGPU.config")) { |
||
144 | section_data = elf_getdata(section, section_data); |
||
145 | binary->config_size = section_data->d_size; |
||
146 | binary->config = MALLOC(binary->config_size * sizeof(unsigned char)); |
||
147 | memcpy(binary->config, section_data->d_buf, binary->config_size); |
||
148 | } else if (debug && !strcmp(name, ".AMDGPU.disasm")) { |
||
149 | binary->disassembled = 1; |
||
150 | section_data = elf_getdata(section, section_data); |
||
151 | fprintf(stderr, "\nShader Disassembly:\n\n"); |
||
152 | fprintf(stderr, "%.*s\n", (int)section_data->d_size, |
||
153 | (char *)section_data->d_buf); |
||
154 | } else if (!strncmp(name, ".rodata", 7)) { |
||
155 | section_data = elf_getdata(section, section_data); |
||
156 | binary->rodata_size = section_data->d_size; |
||
157 | binary->rodata = MALLOC(binary->rodata_size * sizeof(unsigned char)); |
||
158 | memcpy(binary->rodata, section_data->d_buf, binary->rodata_size); |
||
159 | } else if (!strncmp(name, ".symtab", 7)) { |
||
160 | symbols = elf_getdata(section, section_data); |
||
161 | symbol_sh_link = section_header.sh_link; |
||
162 | parse_symbol_table(symbols, §ion_header, binary); |
||
163 | } else if (!strcmp(name, ".rel.text")) { |
||
164 | relocs = elf_getdata(section, section_data); |
||
165 | binary->reloc_count = section_header.sh_size / |
||
166 | section_header.sh_entsize; |
||
167 | } |
||
168 | } |
||
169 | |||
170 | parse_relocs(elf, relocs, symbols, symbol_sh_link, binary); |
||
171 | |||
172 | if (elf){ |
||
173 | elf_end(elf); |
||
174 | } |
||
175 | FREE(elf_buffer); |
||
176 | |||
177 | /* Cache the config size per symbol */ |
||
178 | if (binary->global_symbol_count) { |
||
179 | binary->config_size_per_symbol = |
||
180 | binary->config_size / binary->global_symbol_count; |
||
181 | } else { |
||
182 | binary->global_symbol_count = 1; |
||
183 | binary->config_size_per_symbol = binary->config_size; |
||
184 | } |
||
185 | } |
||
186 | |||
187 | const unsigned char *radeon_shader_binary_config_start( |
||
188 | const struct radeon_shader_binary *binary, |
||
189 | uint64_t symbol_offset) |
||
190 | { |
||
191 | unsigned i; |
||
192 | for (i = 0; i < binary->global_symbol_count; ++i) { |
||
193 | if (binary->global_symbol_offsets[i] == symbol_offset) { |
||
194 | unsigned offset = i * binary->config_size_per_symbol; |
||
195 | return binary->config + offset; |
||
196 | } |
||
197 | } |
||
198 | return binary->config; |
||
199 | } |
||
200 | |||
201 | void radeon_shader_binary_free_relocs(struct radeon_shader_reloc *relocs, |
||
202 | unsigned reloc_count) |
||
203 | { |
||
204 | unsigned i; |
||
205 | for (i = 0; i < reloc_count; i++) { |
||
206 | FREE(relocs[i].name); |
||
207 | } |
||
208 | FREE(relocs); |
||
209 | } |
||
210 | |||
211 | void radeon_shader_binary_free_members(struct radeon_shader_binary *binary, |
||
212 | unsigned free_relocs) |
||
213 | { |
||
214 | FREE(binary->code); |
||
215 | FREE(binary->config); |
||
216 | FREE(binary->rodata); |
||
217 | |||
218 | if (free_relocs) { |
||
219 | radeon_shader_binary_free_relocs(binary->relocs, |
||
220 | binary->reloc_count); |
||
221 | } |
||
222 | }>>>> |