2013-04-06 13:01:26

by Sebastian Wankerl

Subject: [PATCH] Make information about modules available to kgdb.

From: Philip Kranz <[email protected]>

To be able to properly debug kernel modules kgbd needs to know all SHF_ALLOC
sections of the module. This patch add an array of those sections to struct
module. One cannot use sysfs since it does not contain all sections needed.

Signed-off-by: Philip Kranz <[email protected]>
Signed-off-by: Sebastian Wankerl <[email protected]>

---
include/linux/module.h | 5 +++++
kernel/module.c | 42 +++++++++++++++++++++++++++++++++++++++---
2 files changed, 44 insertions(+), 3 deletions(-)

diff --git a/include/linux/module.h b/include/linux/module.h
index 46f1ea0..0209f02 100644
--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -375,6 +375,11 @@ struct module
ctor_fn_t *ctors;
unsigned int num_ctors;
#endif
+
+#ifdef CONFIG_KGDB
+ unsigned int num_alloc_sections;
+ unsigned long *alloc_sections;
+#endif
};
#ifndef MODULE_ARCH_INIT
#define MODULE_ARCH_INIT {}
diff --git a/kernel/module.c b/kernel/module.c
index 3c2c72d..a959b2f 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -103,9 +103,10 @@
DEFINE_MUTEX(module_mutex);
EXPORT_SYMBOL_GPL(module_mutex);
static LIST_HEAD(modules);
-#ifdef CONFIG_KGDB_KDB
-struct list_head *kdb_modules = &modules; /* kdb needs the list of modules */
-#endif /* CONFIG_KGDB_KDB */
+#ifdef CONFIG_KGDB
+/* k(g)db needs the list of modules */
+struct list_head *kdb_modules = &modules;
+#endif /* CONFIG_KGDB */

#ifdef CONFIG_MODULE_SIG
#ifdef CONFIG_MODULE_SIG_FORCE
@@ -1868,6 +1869,10 @@ static void free_module(struct module *mod)
mutex_unlock(&module_mutex);
mod_sysfs_teardown(mod);

+#ifdef CONFIG_KGDB
+ kfree(mod->alloc_sections);
+#endif
+
/* Remove dynamic debug info */
ddebug_remove_module(mod->name);

@@ -3206,6 +3211,33 @@ out:
return err;
}

+#ifdef CONFIG_KGDB
+static void save_alloc_sections(struct module *mod, struct load_info *info)
+{
+ int i, j;
+
+ mod->num_alloc_sections = 0;
+
+ for (i = 0; i < info->hdr->e_shnum; i++)
+ if (info->sechdrs[i].sh_flags & SHF_ALLOC)
+ mod->num_alloc_sections++;
+
+ mod->alloc_sections =
+ kzalloc(sizeof(unsigned long) * mod->num_alloc_sections,
+ GFP_KERNEL);
+ if (mod->alloc_sections == NULL)
+ return;
+
+ j = 0;
+ for (i = 0; i < info->hdr->e_shnum; i++) {
+ if (info->sechdrs[i].sh_flags & SHF_ALLOC) {
+ mod->alloc_sections[j] = info->sechdrs[i].sh_addr;
+ j++;
+ }
+ }
+}
+#endif
+
/* Allocate and load the module: note that size of section 0 is always
zero, and we rely on this for optional sections. */
static int load_module(struct load_info *info, const char __user *uargs,
@@ -3301,6 +3333,10 @@ static int load_module(struct load_info *info, const char __user *uargs,
if (err < 0)
goto bug_cleanup;

+#ifdef CONFIG_KGDB
+ save_alloc_sections(mod, info);
+#endif
+
/* Get rid of temporary copy. */
free_copy(info);

--
1.7.10.4

2013-04-06 16:57:37

by Greg Kroah-Hartman

Subject: Re: [PATCH] Make information about modules available to kgdb.

On Sat, Apr 06, 2013 at 03:00:55PM +0200, Sebastian Wankerl wrote:
> From: Philip Kranz <[email protected]>
>
> To be able to properly debug kernel modules kgbd needs to know all SHF_ALLOC
> sections of the module. This patch add an array of those sections to struct
> module. One cannot use sysfs since it does not contain all sections needed.

Why not just add those sections to sysfs so that everyone can see them?

thanks,

greg k-h

2013-04-08 04:33:39

by Rusty Russell

Subject: Re: [PATCH] Make information about modules available to kgdb.

Greg KH <[email protected]> writes:
> On Sat, Apr 06, 2013 at 03:00:55PM +0200, Sebastian Wankerl wrote:
>> From: Philip Kranz <[email protected]>
>>
>> To be able to properly debug kernel modules kgbd needs to know all SHF_ALLOC
>> sections of the module. This patch add an array of those sections to struct
>> module. One cannot use sysfs since it does not contain all sections needed.
>
> Why not just add those sections to sysfs so that everyone can see them?

That's what we're going to do, but it breaks PARISC. So we're debating
the exact method with James.

Cheers,
Rusty.