1// SPDX-License-Identifier: GPL-2.0
2#include <linux/compiler.h>
3#include <linux/rbtree.h>
4#include <inttypes.h>
5#include <string.h>
6#include <ctype.h>
7#include <stdlib.h>
8#include "dso.h"
9#include "map.h"
10#include "symbol.h"
11#include <internal/lib.h> // page_size
12#include "tests.h"
13#include "debug.h"
14#include "machine.h"
15
16#define UM(x) map__unmap_ip(kallsyms_map, (x))
17
18static bool is_ignored_symbol(const char *name, char type)
19{
20	/* Symbol names that exactly match to the following are ignored.*/
21	static const char * const ignored_symbols[] = {
22		/*
23		 * Symbols which vary between passes. Passes 1 and 2 must have
24		 * identical symbol lists. The kallsyms_* symbols below are
25		 * only added after pass 1, they would be included in pass 2
26		 * when --all-symbols is specified so exclude them to get a
27		 * stable symbol list.
28		 */
29		"kallsyms_addresses",
30		"kallsyms_offsets",
31		"kallsyms_relative_base",
32		"kallsyms_num_syms",
33		"kallsyms_names",
34		"kallsyms_markers",
35		"kallsyms_token_table",
36		"kallsyms_token_index",
37		/* Exclude linker generated symbols which vary between passes */
38		"_SDA_BASE_",		/* ppc */
39		"_SDA2_BASE_",		/* ppc */
40		NULL
41	};
42
43	/* Symbol names that begin with the following are ignored.*/
44	static const char * const ignored_prefixes[] = {
45		"$",			/* local symbols for ARM, MIPS, etc. */
46		".L",			/* local labels, .LBB,.Ltmpxxx,.L__unnamed_xx,.LASANPC, etc. */
47		"__crc_",		/* modversions */
48		"__efistub_",		/* arm64 EFI stub namespace */
49		"__kvm_nvhe_$",		/* arm64 local symbols in non-VHE KVM namespace */
50		"__kvm_nvhe_.L",	/* arm64 local symbols in non-VHE KVM namespace */
51		"__AArch64ADRPThunk_",	/* arm64 lld */
52		"__ARMV5PILongThunk_",	/* arm lld */
53		"__ARMV7PILongThunk_",
54		"__ThumbV7PILongThunk_",
55		"__LA25Thunk_",		/* mips lld */
56		"__microLA25Thunk_",
57		NULL
58	};
59
60	/* Symbol names that end with the following are ignored.*/
61	static const char * const ignored_suffixes[] = {
62		"_from_arm",		/* arm */
63		"_from_thumb",		/* arm */
64		"_veneer",		/* arm */
65		NULL
66	};
67
68	/* Symbol names that contain the following are ignored.*/
69	static const char * const ignored_matches[] = {
70		".long_branch.",	/* ppc stub */
71		".plt_branch.",		/* ppc stub */
72		NULL
73	};
74
75	const char * const *p;
76
77	for (p = ignored_symbols; *p; p++)
78		if (!strcmp(name, *p))
79			return true;
80
81	for (p = ignored_prefixes; *p; p++)
82		if (!strncmp(name, *p, strlen(*p)))
83			return true;
84
85	for (p = ignored_suffixes; *p; p++) {
86		int l = strlen(name) - strlen(*p);
87
88		if (l >= 0 && !strcmp(name + l, *p))
89			return true;
90	}
91
92	for (p = ignored_matches; *p; p++) {
93		if (strstr(name, *p))
94			return true;
95	}
96
97	if (type == 'U' || type == 'u')
98		return true;
99	/* exclude debugging symbols */
100	if (type == 'N' || type == 'n')
101		return true;
102
103	if (toupper(type) == 'A') {
104		/* Keep these useful absolute symbols */
105		if (strcmp(name, "__kernel_syscall_via_break") &&
106		    strcmp(name, "__kernel_syscall_via_epc") &&
107		    strcmp(name, "__kernel_sigtramp") &&
108		    strcmp(name, "__gp"))
109			return true;
110	}
111
112	return false;
113}
114
115struct test__vmlinux_matches_kallsyms_cb_args {
116	struct machine kallsyms;
117	struct map *vmlinux_map;
118	bool header_printed;
119};
120
121static int test__vmlinux_matches_kallsyms_cb1(struct map *map, void *data)
122{
123	struct test__vmlinux_matches_kallsyms_cb_args *args = data;
124	struct dso *dso = map__dso(map);
125	/*
126	 * If it is the kernel, kallsyms is always "[kernel.kallsyms]", while
127	 * the kernel will have the path for the vmlinux file being used, so use
128	 * the short name, less descriptive but the same ("[kernel]" in both
129	 * cases.
130	 */
131	struct map *pair = maps__find_by_name(args->kallsyms.kmaps,
132					(dso->kernel ? dso->short_name : dso->name));
133
134	if (pair) {
135		map__set_priv(pair, 1);
136		map__put(pair);
137	} else {
138		if (!args->header_printed) {
139			pr_info("WARN: Maps only in vmlinux:\n");
140			args->header_printed = true;
141		}
142		map__fprintf(map, stderr);
143	}
144	return 0;
145}
146
147static int test__vmlinux_matches_kallsyms_cb2(struct map *map, void *data)
148{
149	struct test__vmlinux_matches_kallsyms_cb_args *args = data;
150	struct map *pair;
151	u64 mem_start = map__unmap_ip(args->vmlinux_map, map__start(map));
152	u64 mem_end = map__unmap_ip(args->vmlinux_map, map__end(map));
153
154	pair = maps__find(args->kallsyms.kmaps, mem_start);
155
156	if (pair != NULL && !map__priv(pair) && map__start(pair) == mem_start) {
157		struct dso *dso = map__dso(map);
158
159		if (!args->header_printed) {
160			pr_info("WARN: Maps in vmlinux with a different name in kallsyms:\n");
161			args->header_printed = true;
162		}
163
164		pr_info("WARN: %" PRIx64 "-%" PRIx64 " %" PRIx64 " %s in kallsyms as",
165			map__start(map), map__end(map), map__pgoff(map), dso->name);
166		if (mem_end != map__end(pair))
167			pr_info(":\nWARN: *%" PRIx64 "-%" PRIx64 " %" PRIx64,
168				map__start(pair), map__end(pair), map__pgoff(pair));
169		pr_info(" %s\n", dso->name);
170		map__set_priv(pair, 1);
171	}
172	map__put(pair);
173	return 0;
174}
175
176static int test__vmlinux_matches_kallsyms_cb3(struct map *map, void *data)
177{
178	struct test__vmlinux_matches_kallsyms_cb_args *args = data;
179
180	if (!map__priv(map)) {
181		if (!args->header_printed) {
182			pr_info("WARN: Maps only in kallsyms:\n");
183			args->header_printed = true;
184		}
185		map__fprintf(map, stderr);
186	}
187	return 0;
188}
189
190static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused,
191					int subtest __maybe_unused)
192{
193	int err = TEST_FAIL;
194	struct rb_node *nd;
195	struct symbol *sym;
196	struct map *kallsyms_map;
197	struct machine vmlinux;
198	struct maps *maps;
199	u64 mem_start, mem_end;
200	struct test__vmlinux_matches_kallsyms_cb_args args;
201
202	/*
203	 * Step 1:
204	 *
205	 * Init the machines that will hold kernel, modules obtained from
206	 * both vmlinux + .ko files and from /proc/kallsyms split by modules.
207	 */
208	machine__init(&args.kallsyms, "", HOST_KERNEL_ID);
209	machine__init(&vmlinux, "", HOST_KERNEL_ID);
210
211	maps = machine__kernel_maps(&vmlinux);
212
213	/*
214	 * Step 2:
215	 *
216	 * Create the kernel maps for kallsyms and the DSO where we will then
217	 * load /proc/kallsyms. Also create the modules maps from /proc/modules
218	 * and find the .ko files that match them in /lib/modules/`uname -r`/.
219	 */
220	if (machine__create_kernel_maps(&args.kallsyms) < 0) {
221		pr_debug("machine__create_kernel_maps failed");
222		err = TEST_SKIP;
223		goto out;
224	}
225
226	/*
227	 * Step 3:
228	 *
229	 * Load and split /proc/kallsyms into multiple maps, one per module.
230	 * Do not use kcore, as this test was designed before kcore support
231	 * and has parts that only make sense if using the non-kcore code.
232	 * XXX: extend it to stress the kcorre code as well, hint: the list
233	 * of modules extracted from /proc/kcore, in its current form, can't
234	 * be compacted against the list of modules found in the "vmlinux"
235	 * code and with the one got from /proc/modules from the "kallsyms" code.
236	 */
237	if (machine__load_kallsyms(&args.kallsyms, "/proc/kallsyms") <= 0) {
238		pr_debug("machine__load_kallsyms failed");
239		err = TEST_SKIP;
240		goto out;
241	}
242
243	/*
244	 * Step 4:
245	 *
246	 * kallsyms will be internally on demand sorted by name so that we can
247	 * find the reference relocation * symbol, i.e. the symbol we will use
248	 * to see if the running kernel was relocated by checking if it has the
249	 * same value in the vmlinux file we load.
250	 */
251	kallsyms_map = machine__kernel_map(&args.kallsyms);
252
253	/*
254	 * Step 5:
255	 *
256	 * Now repeat step 2, this time for the vmlinux file we'll auto-locate.
257	 */
258	if (machine__create_kernel_maps(&vmlinux) < 0) {
259		pr_info("machine__create_kernel_maps failed");
260		goto out;
261	}
262
263	args.vmlinux_map = machine__kernel_map(&vmlinux);
264
265	/*
266	 * Step 6:
267	 *
268	 * Locate a vmlinux file in the vmlinux path that has a buildid that
269	 * matches the one of the running kernel.
270	 *
271	 * While doing that look if we find the ref reloc symbol, if we find it
272	 * we'll have its ref_reloc_symbol.unrelocated_addr and then
273	 * maps__reloc_vmlinux will notice and set proper ->[un]map_ip routines
274	 * to fixup the symbols.
275	 */
276	if (machine__load_vmlinux_path(&vmlinux) <= 0) {
277		pr_info("Couldn't find a vmlinux that matches the kernel running on this machine, skipping test\n");
278		err = TEST_SKIP;
279		goto out;
280	}
281
282	err = 0;
283	/*
284	 * Step 7:
285	 *
286	 * Now look at the symbols in the vmlinux DSO and check if we find all of them
287	 * in the kallsyms dso. For the ones that are in both, check its names and
288	 * end addresses too.
289	 */
290	map__for_each_symbol(args.vmlinux_map, sym, nd) {
291		struct symbol *pair, *first_pair;
292
293		sym  = rb_entry(nd, struct symbol, rb_node);
294
295		if (sym->start == sym->end)
296			continue;
297
298		mem_start = map__unmap_ip(args.vmlinux_map, sym->start);
299		mem_end = map__unmap_ip(args.vmlinux_map, sym->end);
300
301		first_pair = machine__find_kernel_symbol(&args.kallsyms, mem_start, NULL);
302		pair = first_pair;
303
304		if (pair && UM(pair->start) == mem_start) {
305next_pair:
306			if (arch__compare_symbol_names(sym->name, pair->name) == 0) {
307				/*
308				 * kallsyms don't have the symbol end, so we
309				 * set that by using the next symbol start - 1,
310				 * in some cases we get this up to a page
311				 * wrong, trace_kmalloc when I was developing
312				 * this code was one such example, 2106 bytes
313				 * off the real size. More than that and we
314				 * _really_ have a problem.
315				 */
316				s64 skew = mem_end - UM(pair->end);
317				if (llabs(skew) >= page_size)
318					pr_debug("WARN: %#" PRIx64 ": diff end addr for %s v: %#" PRIx64 " k: %#" PRIx64 "\n",
319						 mem_start, sym->name, mem_end,
320						 UM(pair->end));
321
322				/*
323				 * Do not count this as a failure, because we
324				 * could really find a case where it's not
325				 * possible to get proper function end from
326				 * kallsyms.
327				 */
328				continue;
329			} else {
330				pair = machine__find_kernel_symbol_by_name(&args.kallsyms,
331									   sym->name, NULL);
332				if (pair) {
333					if (UM(pair->start) == mem_start)
334						goto next_pair;
335
336					pr_debug("WARN: %#" PRIx64 ": diff name v: %s k: %s\n",
337						 mem_start, sym->name, pair->name);
338				} else {
339					pr_debug("WARN: %#" PRIx64 ": diff name v: %s k: %s\n",
340						 mem_start, sym->name, first_pair->name);
341				}
342
343				continue;
344			}
345		} else if (mem_start == map__end(args.kallsyms.vmlinux_map)) {
346			/*
347			 * Ignore aliases to _etext, i.e. to the end of the kernel text area,
348			 * such as __indirect_thunk_end.
349			 */
350			continue;
351		} else if (is_ignored_symbol(sym->name, sym->type)) {
352			/*
353			 * Ignore hidden symbols, see scripts/kallsyms.c for the details
354			 */
355			continue;
356		} else {
357			pr_debug("ERR : %#" PRIx64 ": %s not on kallsyms\n",
358				 mem_start, sym->name);
359		}
360
361		err = -1;
362	}
363
364	if (verbose <= 0)
365		goto out;
366
367	args.header_printed = false;
368	maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb1, &args);
369
370	args.header_printed = false;
371	maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb2, &args);
372
373	args.header_printed = false;
374	maps = machine__kernel_maps(&args.kallsyms);
375	maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb3, &args);
376
377out:
378	machine__exit(&args.kallsyms);
379	machine__exit(&vmlinux);
380	return err;
381}
382
383DEFINE_SUITE("vmlinux symtab matches kallsyms", vmlinux_matches_kallsyms);
384