Skip to contentSkip to conversation

module: treat exit sections the same as init sections when !CONFIG_MODULE_UNLOAD

PATCH In review Jessica Yu 2 participants 2 messages Mar 23, 2021

lore
Share & tools
lore

Jessica Yu

author
#0 lore raw

PATCH 1/1 module: treat exit sections the same as init sections when !CONFIG_MODULE_UNLOAD

Dynamic code patching (alternatives, jump_label and static_call) can have sites in __exit code, even if __exit is never executed. Therefore __exit must be present at runtime, at least for as long as __init code is.

Additionally, for jump_label and static_call, the __exit sites must also identify as within_module_init(), such that the infrastructure is aware to never touch them after module init -- alternatives are only ran once at init and hence don't have this particular constraint.

By making __exit identify as __init for !MODULE_UNLOAD, the above is satisfied.

So the section ordering should look like the following when !CONFIG_MODULE_UNLOAD, with the .exit sections moved to the init region of the module.

Core section allocation order:
.text
.rodata
__ksymtab_gpl
__ksymtab_strings
.note.* sections
.bss
.data
.gnu.linkonce.this_module
Init section allocation order:
.init.text
.exit.text
.symtab
.strtab

[jeyu: thanks to Peter Zijlstra for most of the changelog]

---

Do you want to take this patch with the other static_call patches? Or should I take this through modules-next?

 kernel/module.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
kernel/module.c +4 -5
index 30479355ab85..173a09175511 100644
@@ -2802,7 +2802,11 @@ void * __weak module_alloc(unsigned long size)
 
 bool __weak module_init_section(const char *name)
 {
+#ifndef CONFIG_MODULE_UNLOAD
+	return strstarts(name, ".init") || module_exit_section(name);
+#else
 	return strstarts(name, ".init");
+#endif
 }
 
 bool __weak module_exit_section(const char *name)
@@ -3116,11 +3120,6 @@ static int rewrite_section_headers(struct load_info *info, int flags)
 		 */
 		shdr->sh_addr = (size_t)info->hdr + shdr->sh_offset;
 
-#ifndef CONFIG_MODULE_UNLOAD
-		/* Don't load .exit sections */
-		if (module_exit_section(info->secstrings+shdr->sh_name))
-			shdr->sh_flags &= ~(unsigned long)SHF_ALLOC;
-#endif
 	}
 
 	/* Track but don't keep modinfo and version sections. */
Signature · 1 line
2.30.1
Acked-by on v1 1/1
#1 lore raw
Quoted 32 lines · Jessica Yu

Dynamic code patching (alternatives, jump_label and static_call) can have sites in __exit code, even if __exit is never executed. Therefore __exit must be present at runtime, at least for as long as __init code is.

Additionally, for jump_label and static_call, the __exit sites must also identify as within_module_init(), such that the infrastructure is aware to never touch them after module init -- alternatives are only ran once at init and hence don't have this particular constraint.

By making __exit identify as __init for !MODULE_UNLOAD, the above is satisfied.

So the section ordering should look like the following when !CONFIG_MODULE_UNLOAD, with the .exit sections moved to the init region of the module.

Core section allocation order:

(quoted diff) +0 -0
 	.text
 	.rodata
 	__ksymtab_gpl
 	__ksymtab_strings
 	.note.* sections
 	.bss
 	.data
 	.gnu.linkonce.this_module
 Init section allocation order:
 	.init.text
 	.exit.text
 	.symtab
 	.strtab
Quoted 4 lines

[jeyu: thanks to Peter Zijlstra for most of the changelog]

Quoted 4 lines

---

Do you want to take this patch with the other static_call patches? Or should I take this through modules-next?

Take it through modules-next, I haven't seen an actual report on this and it's been broken forever afaict. I only found it by accident while chasing this other problem.