]>
Commit | Line | Data |
---|---|---|
1 | // Copyright (C) Andrew Tridgell 2002 (original file) | |
2 | // Copyright (C) 2006-2011 Red Hat Inc. (systemtap changes) | |
3 | // | |
4 | // This program is free software; you can redistribute it and/or modify | |
5 | // it under the terms of the GNU General Public License as published by | |
6 | // the Free Software Foundation; either version 2 of the License, or | |
7 | // (at your option) any later version. | |
8 | // | |
9 | // This program is distributed in the hope that it will be useful, | |
10 | // but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | // GNU General Public License for more details. | |
13 | // | |
14 | // You should have received a copy of the GNU General Public License | |
15 | // along with this program. If not, see <http://www.gnu.org/licenses/>. | |
16 | ||
17 | #include "config.h" | |
18 | #include "session.h" | |
19 | #include "hash.h" | |
20 | #include "util.h" | |
21 | ||
22 | #include <cstdlib> | |
23 | #include <cstring> | |
24 | #include <fstream> | |
25 | #include <sstream> | |
26 | #include <iomanip> | |
27 | #include <cerrno> | |
28 | ||
29 | extern "C" { | |
30 | #include <sys/types.h> | |
31 | #include <sys/stat.h> | |
32 | #include <unistd.h> | |
33 | #include "mdfour.h" | |
34 | } | |
35 | ||
36 | using namespace std; | |
37 | ||
38 | ||
39 | class hash | |
40 | { | |
41 | private: | |
42 | struct mdfour md4; | |
43 | std::ostringstream parm_stream; | |
44 | ||
45 | public: | |
46 | hash() { start(); } | |
47 | hash(const hash &base) { md4 = base.md4; parm_stream << base.parm_stream.str(); } | |
48 | ||
49 | void start(); | |
50 | ||
51 | void add(const std::string& description, const unsigned char *buffer, size_t size); | |
52 | template<typename T> void add(const std::string& d, const T& x); | |
53 | void add(const std::string& d, const char *s) { add((const std::string&)d, (const unsigned char *)s, strlen(s)); } | |
54 | void add(const std:: string& d, const std::string& s) { add(d, (const unsigned char *)s.c_str(), s.length()); } | |
55 | ||
56 | void add_path(const std::string& description, const std::string& path); | |
57 | ||
58 | void result(std::string& r); | |
59 | std::string get_parms() { return parm_stream.str(); } | |
60 | }; | |
61 | ||
62 | ||
63 | void | |
64 | hash::start() | |
65 | { | |
66 | mdfour_begin(&md4); | |
67 | } | |
68 | ||
69 | ||
70 | void | |
71 | hash::add(const std::string& description, const unsigned char *buffer, size_t size) | |
72 | { | |
73 | // If the string is longer than 80 and it is not a path, only | |
74 | // put the last 80 characters from the string into the hash log. | |
75 | // string stripped_buffer = lex_cast_qstring(buffer); | |
76 | if(size > 80 && buffer[0] != '/') | |
77 | parm_stream << description << "..." << lex_cast_qstring(buffer+(size-80)) << endl; | |
78 | else | |
79 | parm_stream << description << lex_cast_qstring(buffer) << endl; | |
80 | ||
81 | mdfour_update(&md4, buffer, size); | |
82 | } | |
83 | ||
84 | ||
85 | template <typename T> void | |
86 | hash::add(const std::string& d, const T& x) | |
87 | { | |
88 | parm_stream << d << x << endl; | |
89 | mdfour_update(&md4, (const unsigned char *)&x, sizeof(x)); | |
90 | } | |
91 | ||
92 | ||
93 | void | |
94 | hash::add_path(const std::string& description, const std::string& path) | |
95 | { | |
96 | struct stat st; | |
97 | memset (&st, 0, sizeof(st)); | |
98 | ||
99 | if (stat(path.c_str(), &st) != 0) | |
100 | st.st_size = st.st_mtime = -1; | |
101 | ||
102 | add(description + "Path: ", path); | |
103 | add(description + "Size: ", st.st_size); | |
104 | add(description + "Timestamp: ", st.st_mtime); | |
105 | } | |
106 | ||
107 | ||
108 | void | |
109 | hash::result(string& r) | |
110 | { | |
111 | ostringstream rstream; | |
112 | unsigned char sum[16]; | |
113 | ||
114 | mdfour_update(&md4, NULL, 0); | |
115 | mdfour_result(&md4, sum); | |
116 | ||
117 | for (int i=0; i<16; i++) | |
118 | { | |
119 | rstream << hex << setfill('0') << setw(2) << (unsigned)sum[i]; | |
120 | } | |
121 | rstream << "_" << setw(0) << dec << (unsigned)md4.totalN; | |
122 | r = rstream.str(); | |
123 | } | |
124 | ||
125 | void create_hash_log(const string &type_str, const string &parms, const string &result, const string &hash_log_path) | |
126 | { | |
127 | ofstream log_file; | |
128 | time_t rawtime; | |
129 | time ( &rawtime ); | |
130 | string time_str(ctime (&rawtime)); | |
131 | ||
132 | log_file.open(hash_log_path.c_str()); | |
133 | log_file << "[" << time_str.substr(0,time_str.length()-1); // erase terminated '\n' | |
134 | log_file << "] " << type_str << ":" << endl; | |
135 | log_file << parms << endl; | |
136 | log_file << _("result:") << result << endl; | |
137 | log_file.close(); | |
138 | } | |
139 | ||
140 | static const hash& | |
141 | get_base_hash (systemtap_session& s) | |
142 | { | |
143 | if (s.base_hash) | |
144 | return *s.base_hash; | |
145 | ||
146 | s.base_hash = new hash(); | |
147 | hash& h = *s.base_hash; | |
148 | ||
149 | // Hash kernel release and arch. | |
150 | h.add("Kernel Release: ", s.kernel_release); | |
151 | h.add_path("Kernel Build Tree ", s.kernel_build_tree); | |
152 | h.add("Architecture: ", s.architecture); | |
153 | ||
154 | // Hash a few kernel version/build-id files too | |
155 | // (useful for kernel developers reusing a single source tree) | |
156 | h.add_path("Kernel Build Tree .config ", s.kernel_build_tree + "/.config"); | |
157 | h.add_path("Kernel Build Tree .version ", s.kernel_build_tree + "/.version"); | |
158 | h.add_path("Kernel Build Tree compile.h ", s.kernel_build_tree + "/include/linux/compile.h"); | |
159 | h.add_path("Kernel Build Tree version.h ", s.kernel_build_tree + "/include/linux/version.h"); | |
160 | h.add_path("Kernel Build Tree utsrelease.h ", s.kernel_build_tree + "/include/linux/utsrelease.h"); | |
161 | ||
162 | // Hash runtime path (that gets added in as "-R path"). | |
163 | h.add_path("Runtime ", s.runtime_path); | |
164 | ||
165 | // Hash compiler path, size, and mtime. We're just going to assume | |
166 | // we'll be using gcc. XXX: getting kbuild to spit out out would be | |
167 | // better, especially since this is fooled by ccache. | |
168 | h.add_path("Compiler ", find_executable("gcc")); | |
169 | ||
170 | // Hash the systemtap size and mtime. We could use VERSION/DATE, | |
171 | // but when developing systemtap that doesn't work well (since you | |
172 | // can compile systemtap multiple times in 1 day). Since we don't | |
173 | // know exactly where we're getting run from, we'll use | |
174 | // /proc/self/exe. | |
175 | // XXX well almost exactly -- valgrind throws this off | |
176 | h.add_path("Systemtap ", "/proc/self/exe"); | |
177 | ||
178 | return h; | |
179 | } | |
180 | ||
181 | ||
182 | static bool | |
183 | create_hashdir (systemtap_session& s, const string& result, string& hashdir) | |
184 | { | |
185 | int nlevels = 1; | |
186 | ||
187 | // Use a N level subdir for the cache path to reduce the impact on | |
188 | // filesystems which are slow for large directories. Let N be adjustable. | |
189 | const char *s_n = getenv("SYSTEMTAP_NLEVELS"); | |
190 | if (s_n) | |
191 | { | |
192 | nlevels = atoi(s_n); | |
193 | if (nlevels < 1) nlevels = 1; | |
194 | if (nlevels > 8) nlevels = 8; | |
195 | } | |
196 | ||
197 | hashdir = s.cache_path; | |
198 | ||
199 | for (int i = 0; i < nlevels; i++) | |
200 | { | |
201 | hashdir += string("/") + result[i*2] + result[i*2 + 1]; | |
202 | if (create_dir(hashdir.c_str()) != 0) | |
203 | { | |
204 | if (! s.suppress_warnings) | |
205 | cerr << _F("Warning: failed to create cache directory (\"%s\") %s, disabling cache support", | |
206 | hashdir.c_str(), strerror(errno)) << endl; | |
207 | s.use_cache = s.use_script_cache = false; | |
208 | return false; | |
209 | } | |
210 | } | |
211 | return true; | |
212 | } | |
213 | ||
214 | ||
215 | void | |
216 | find_script_hash (systemtap_session& s, const string& script) | |
217 | { | |
218 | hash h(get_base_hash(s)); | |
219 | struct stat st; | |
220 | ||
221 | // Hash getuid. This really shouldn't be necessary (since who you | |
222 | // are doesn't change the generated output), but the hash gets used | |
223 | // as the module name. If two different users try to run the same | |
224 | // script at the same time, we need something to differentiate the | |
225 | // module name. | |
226 | h.add("UID: ", getuid()); | |
227 | ||
228 | // Hash user-specified arguments (that change the generated module). | |
229 | h.add("Bulk Mode (-b): ", s.bulk_mode); // '-b' | |
230 | h.add("Timing (-t): ", s.timing); // '-t' | |
231 | h.add("Prologue Searching (-P): ", s.prologue_searching); // '-P' | |
232 | h.add("Ignore Vmlinux (--ignore-vmlinux): ", s.ignore_vmlinux); // --ignore-vmlinux | |
233 | h.add("Ignore Dwarf (--ignore-dwarf): ", s.ignore_dwarf); // --ignore-dwarf | |
234 | h.add("Consult Symtab (--kelf, --kmap): ", s.consult_symtab); // --kelf, --kmap | |
235 | h.add("Skip Badvars (--skip-badvars): ", s.skip_badvars); // --skip-badvars | |
236 | h.add("Unprivileged (--unprivileged): ", s.unprivileged); // --unprivileged | |
237 | h.add("Compatible (--compatible): ", s.compatible); // --compatible | |
238 | h.add("Omit Werror (undocumented): ", s.omit_werror); // undocumented, evil | |
239 | if (!s.kernel_symtab_path.empty()) // --kmap | |
240 | { | |
241 | h.add("Kernel Symtab Path: ", s.kernel_symtab_path); | |
242 | if (stat(s.kernel_symtab_path.c_str(), &st) == 0) | |
243 | { | |
244 | // NB: stat of /proc/kallsyms always returns size=0, mtime=now... | |
245 | // which is a good reason to use the default /boot/System.map-2.6.xx | |
246 | // instead. | |
247 | h.add("Kernel Symtab Size: ", st.st_size); | |
248 | h.add("Kernel Symtab Timestamp: ", st.st_mtime); | |
249 | } | |
250 | } | |
251 | for (unsigned i = 0; i < s.macros.size(); i++) | |
252 | h.add("Macros: ", s.macros[i]); | |
253 | ||
254 | // Add any custom kbuild flags (-B) | |
255 | for (unsigned i = 0; i < s.kbuildflags.size(); i++) | |
256 | h.add("Kbuildflags: ", s.kbuildflags[i]); | |
257 | ||
258 | // -d MODULE | |
259 | for (set<string>::iterator it = s.unwindsym_modules.begin(); | |
260 | it != s.unwindsym_modules.end(); | |
261 | it++) | |
262 | h.add_path("Unwindsym Modules ", *it); | |
263 | ||
264 | // Add the build id of each module | |
265 | for(vector<string>::iterator it = s.build_ids.begin(); | |
266 | it != s.build_ids.end(); | |
267 | it++) | |
268 | h.add("Build ID: ", *it); | |
269 | ||
270 | // Add in pass 2 script output. | |
271 | h.add("Script: ", script); | |
272 | ||
273 | // Get the directory path to store our cached script | |
274 | string result, hashdir; | |
275 | h.result(result); | |
276 | if (!create_hashdir(s, result, hashdir)) | |
277 | return; | |
278 | ||
279 | // Update module name to be 'stap_{hash start}'. '{hash start}' | |
280 | // must not be too long. This shouldn't happen, since the maximum | |
281 | // size of a hash is 32 fixed chars + 1 (for the '_') + a max of 11. | |
282 | s.module_name = "stap_" + result; | |
283 | if (s.module_name.size() >= (MODULE_NAME_LEN - 1)) | |
284 | s.module_name.resize(MODULE_NAME_LEN - 1); | |
285 | ||
286 | // 'ccache' would use a hash path of something like: | |
287 | // s.hash_path = hashdir + "/" + result.substr(nlevels); | |
288 | // which would look like: | |
289 | // ~/.stap_cache/A/B/CDEFGHIJKLMNOPQRSTUVWXYZABCDEF_XXX | |
290 | // | |
291 | // We're using the following so that the module can be used straight | |
292 | // from the cache if desired. This ends up looking like this: | |
293 | // ~/.stap_cache/A/B/stap_ABCDEFGHIJKLMNOPQRSTUVWXYZABCDEF_XXX.ko | |
294 | s.hash_path = hashdir + "/" + s.module_name + ".ko"; | |
295 | ||
296 | // Update C source name with new module_name. | |
297 | s.translated_source = string(s.tmpdir) + "/" + s.module_name + ".c"; | |
298 | create_hash_log(string("script_hash"), h.get_parms(), result, | |
299 | hashdir + "/" + s.module_name + "_hash.log"); | |
300 | } | |
301 | ||
302 | ||
303 | void | |
304 | find_stapconf_hash (systemtap_session& s) | |
305 | { | |
306 | hash h(get_base_hash(s)); | |
307 | ||
308 | // Add any custom kbuild flags | |
309 | for (unsigned i = 0; i < s.kbuildflags.size(); i++) | |
310 | h.add("Kbuildflags: ", s.kbuildflags[i]); | |
311 | ||
312 | // Get the directory path to store our cached stapconf parameters | |
313 | string result, hashdir; | |
314 | h.result(result); | |
315 | if (!create_hashdir(s, result, hashdir)) | |
316 | return; | |
317 | ||
318 | s.stapconf_name = "stapconf_" + result + ".h"; | |
319 | s.stapconf_path = hashdir + "/" + s.stapconf_name; | |
320 | create_hash_log(string("stapconf_hash"), h.get_parms(), result, | |
321 | hashdir + "/stapconf_" + result + "_hash.log"); | |
322 | } | |
323 | ||
324 | ||
325 | string | |
326 | find_tracequery_hash (systemtap_session& s, const vector<string>& headers) | |
327 | { | |
328 | hash h(get_base_hash(s)); | |
329 | ||
330 | // Add the tracepoint headers to the computed hash | |
331 | for (size_t i = 0; i < headers.size(); ++i) | |
332 | h.add_path("Headers ", headers[i]); | |
333 | ||
334 | // Add any custom kbuild flags | |
335 | for (unsigned i = 0; i < s.kbuildflags.size(); i++) | |
336 | h.add("Kbuildflags: ", s.kbuildflags[i]); | |
337 | ||
338 | // Get the directory path to store our cached module | |
339 | string result, hashdir; | |
340 | h.result(result); | |
341 | if (!create_hashdir(s, result, hashdir)) | |
342 | return ""; | |
343 | ||
344 | create_hash_log(string("tracequery_hash"), h.get_parms(), result, | |
345 | hashdir + "/tracequery_" + result + "_hash.log"); | |
346 | return hashdir + "/tracequery_" + result + ".ko"; | |
347 | } | |
348 | ||
349 | ||
350 | string | |
351 | find_typequery_hash (systemtap_session& s, const string& name) | |
352 | { | |
353 | hash h(get_base_hash(s)); | |
354 | ||
355 | // Add the typequery name to distinguish the hash | |
356 | h.add("Typequery Name: ", name); | |
357 | ||
358 | if (name[0] == 'k') | |
359 | // Add any custom kbuild flags | |
360 | for (unsigned i = 0; i < s.kbuildflags.size(); i++) | |
361 | h.add("Kbuildflags: ", s.kbuildflags[i]); | |
362 | ||
363 | // Get the directory path to store our cached module | |
364 | string result, hashdir; | |
365 | h.result(result); | |
366 | if (!create_hashdir(s, result, hashdir)) | |
367 | return ""; | |
368 | ||
369 | create_hash_log(string("typequery_hash"), h.get_parms(), result, | |
370 | hashdir + "/typequery_" + result + "_hash.log"); | |
371 | return hashdir + "/typequery_" + result | |
372 | + (name[0] == 'k' ? ".ko" : ".so"); | |
373 | } | |
374 | ||
375 | ||
376 | string | |
377 | find_uprobes_hash (systemtap_session& s) | |
378 | { | |
379 | hash h(get_base_hash(s)); | |
380 | ||
381 | // Hash runtime uprobes paths | |
382 | h.add_path("Uprobes Runtime Path /uprobes ", s.runtime_path + "/uprobes"); | |
383 | h.add_path("Uprobes Runtime Path /uprobes2 ", s.runtime_path + "/uprobes2"); | |
384 | ||
385 | // Add any custom kbuild flags | |
386 | for (unsigned i = 0; i < s.kbuildflags.size(); i++) | |
387 | h.add("Kbuildflags: ", s.kbuildflags[i]); | |
388 | ||
389 | // Get the directory path to store our cached module | |
390 | string result, hashdir; | |
391 | h.result(result); | |
392 | if (!create_hashdir(s, result, hashdir)) | |
393 | return ""; | |
394 | ||
395 | create_hash_log(string("uprobes_hash"), h.get_parms(), result, | |
396 | hashdir + "/uprobes_" + result + "_hash.log"); | |
397 | return hashdir + "/uprobes_" + result; | |
398 | } | |
399 | ||
400 | /* vim: set sw=2 ts=8 cino=>4,n-2,{2,^-2,t0,(0,u0,w1,M1 : */ |