RepoMirrors/kpatch
1
0
Fork 0
mirror of https://github.com/dynup/kpatch synced 2024-12-25 14:52:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

create-diff-object: simplify mangled function correlation

Browse Source 
The RHEL powerpc kernel is compiled with -O3, which triggers some
"interesting" new optimizations.  One of them, which seems to be
relatively common, is the replacing of a function with two separate
"constprop" functions.

Previously we only ever saw a single constprop clone, so we just renamed
the patched version of the function to match the original version.  Now
that we can have two such clones, that no longer makes sense.

Instead of renaming functions, just improve the correlation logic such
that they can be correlated despite having slightly different symbol
names.  The first clone in the original object is correlated with the
first clone in the patched object; the second clone is correlated with
the second clone; and so on.

This assumes that the order of the symbols and sections doesn't change,
which seems to be a reasonable assumption based on past experience with
the compiler.  Otherwise it will just unnecessarily mark the cloned
constprop functions as changed, which is annoying but harmless, and
noticeable by a human anyway.

Fixes #935.

Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
This commit is contained in:
Josh Poimboeuf 2019-10-02 18:34:48 -05:00
parent 683289206b
commit 935f199875
2 changed files with 14 additions and 82 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

94
kpatch-build/create-diff-object.c
View File

@ -297,6 +297,13 @@ static int is_special_static(struct symbol *sym)
*/
static int kpatch_mangled_strcmp(char *s1, char *s2)
{
/*
* ELF string sections aren't mangled, though they look that way. Just
* compare them normally.
*/
if (strstr(s1, ".str1."))
return strcmp(s1, s2);
while (*s1 == *s2) {
if (!*s1)
return 0;
@ -389,11 +396,7 @@ static int rela_equal(struct rela *rela1, struct rela *rela2)
if (rela_toc1->addend != rela_toc2->addend)
return 0;
if (is_special_static(rela_toc1->sym))
return !kpatch_mangled_strcmp(rela_toc1->sym->name,
rela_toc2->sym->name);
return !strcmp(rela_toc1->sym->name, rela_toc2->sym->name);
return !kpatch_mangled_strcmp(rela_toc1->sym->name, rela_toc2->sym->name);
}
static void kpatch_compare_correlated_rela_section(struct section *sec)
@ -443,7 +446,7 @@ static void kpatch_compare_correlated_section(struct section *sec)
sec1->sh.sh_entsize != sec2->sh.sh_entsize ||
(sec1->sh.sh_addralign != sec2->sh.sh_addralign &&
!is_text_section(sec1)))
DIFF_FATAL("%s section header details differ", sec1->name);
DIFF_FATAL("%s section header details differ from %s", sec1->name, sec2->name);
/* Short circuit for mcount sections, we rebuild regardless */
if (!strcmp(sec->name, ".rela__mcount_loc") ||
@ -763,7 +766,8 @@ static void kpatch_correlate_sections(struct list_head *seclist1, struct list_he
list_for_each_entry(sec1, seclist1, list) {
list_for_each_entry(sec2, seclist2, list) {
if (strcmp(sec1->name, sec2->name))
if (kpatch_mangled_strcmp(sec1->name, sec2->name) ||
sec2->twin)
continue;
if (is_special_static(is_rela_section(sec1) ?
@ -797,8 +801,8 @@ static void kpatch_correlate_symbols(struct list_head *symlist1, struct list_hea
list_for_each_entry(sym1, symlist1, list) {
list_for_each_entry(sym2, symlist2, list) {
if (strcmp(sym1->name, sym2->name) ||
sym1->type != sym2->type)
if (kpatch_mangled_strcmp(sym1->name, sym2->name) ||
sym1->type != sym2->type || sym2->twin)
continue;
if (is_special_static(sym1))
@ -894,77 +898,6 @@ static void kpatch_mark_grouped_sections(struct kpatch_elf *kelf)
}
}
/*
* When gcc makes compiler optimizations which affect a function's calling
* interface, it mangles the function's name. For example, sysctl_print_dir is
* renamed to sysctl_print_dir.isra.2. The problem is that the trailing number
* is chosen arbitrarily, and the patched version of the function may end up
* with a different trailing number. Rename any mangled patched functions to
* match their base counterparts.
*/
static void kpatch_rename_mangled_functions(struct kpatch_elf *base,
struct kpatch_elf *patched)
{
struct symbol *sym, *basesym;
char name[256], *origname;
struct section *sec, *basesec;
int found;
list_for_each_entry(sym, &patched->symbols, list) {
if (sym->type != STT_FUNC)
continue;
if (!strstr(sym->name, ".isra.") &&
!strstr(sym->name, ".constprop.") &&
!strstr(sym->name, ".cold.") &&
!strstr(sym->name, ".part."))
continue;
found = 0;
list_for_each_entry(basesym, &base->symbols, list) {
if (!kpatch_mangled_strcmp(basesym->name, sym->name)) {
found = 1;
break;
}
}
if (!found)
continue;
if (!strcmp(sym->name, basesym->name))
continue;
log_debug("renaming %s to %s\n", sym->name, basesym->name);
origname = sym->name;
sym->name = strdup(basesym->name);
if (sym != sym->sec->sym)
continue;
sym->sec->name = strdup(basesym->sec->name);
if (sym->sec->rela)
sym->sec->rela->name = strdup(basesym->sec->rela->name);
/*
* When function foo.isra.1 has a switch statement, it might
* have a corresponding bundled .rodata.foo.isra.1 section (in
* addition to .text.foo.isra.1 which we renamed above).
*/
sprintf(name, ".rodata.%s", origname);
sec = find_section_by_name(&patched->sections, name);
if (!sec)
continue;
sprintf(name, ".rodata.%s", basesym->name);
basesec = find_section_by_name(&base->sections, name);
if (!basesec)
continue;
sec->name = strdup(basesec->name);
sec->secsym->name = sec->name;
if (sec->rela)
sec->rela->name = strdup(basesec->rela->name);
}
}
static char *kpatch_section_function_name(struct section *sec)
{
if (is_rela_section(sec))
@ -3429,7 +3362,6 @@ int main(int argc, char *argv[])
kpatch_mark_grouped_sections(kelf_patched);
kpatch_replace_sections_syms(kelf_base);
kpatch_replace_sections_syms(kelf_patched);
kpatch_rename_mangled_functions(kelf_base, kelf_patched);
kpatch_correlate_elfs(kelf_base, kelf_patched);
kpatch_correlate_static_local_variables(kelf_base, kelf_patched);

2
test/unit/objs

@ -1 +1 @@
Subproject commit 510a90ca8d35c09574d449ea0f171a7a9baaf087
Subproject commit a5fd47f575552ef989b6b14de84460de5ce0b4ea