Re: [PATCH 2/2] module: fix bne2 "gave up waiting for init of module libcrc32c"

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Linus Torvalds
Date: Tuesday, June 1, 2010 - 9:44 pm

On Tue, 1 Jun 2010, Linus Torvalds wrote:

In fact, I didn't nest it right.

The "free_modinfo()" pairs with the "setup_modinfo()" call, and should go 
into the "cleanup" error case, not the "sysfs_uninit" error case. IOW, I 
moved one too many error case cleanup lines.

So in that patch, the "free_modinfo()" call should move back to the 
cleanup case. Like the appended (still untested - I just stared at the 
code some more, rather than do anything as mundane as _test_ it) patch.

It may still not be right, of course. But it might be closer.

(That function _really_ should be peeled like an onion, and split into 
many smaller functions, so that we don't have ten error cases needing 
unwinding. I like "goto error", but at some point you can't see the 
unwinding any more, and that function has passed that point a long time 
ago, I think)

		Linus

---
 kernel/module.c |   33 ++++++++++++++++++---------------
 1 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/kernel/module.c b/kernel/module.c
index a1f46a5..135577c 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2198,11 +2198,6 @@ static noinline struct module *load_module(void __user *umod,
 		goto free_mod;
 	}
 
-	if (find_module(mod->name)) {
-		err = -EEXIST;
-		goto free_mod;
-	}
-
 	mod->state = MODULE_STATE_COMING;
 
 	/* Allow arches to frob section contents and sizes.  */
@@ -2293,11 +2288,6 @@ static noinline struct module *load_module(void __user *umod,
 	/* Now we've moved module, initialize linked lists, etc. */
 	module_unload_init(mod);
 
-	/* add kobject, so we can reference it. */
-	err = mod_sysfs_init(mod);
-	if (err)
-		goto free_unload;
-
 	/* Set up license info based on the info section */
 	set_license(mod, get_modinfo(sechdrs, infoindex, "license"));
 
@@ -2486,16 +2476,28 @@ static noinline struct module *load_module(void __user *umod,
 	 * The mutex protects against concurrent writers.
 	 */
 	mutex_lock(&module_mutex);
+
+	if (find_module(mod->name)) {
+		err = -EEXIST;
+		/* This will also unlock the mutex */
+		goto already_exists;
+	}
+
 	list_add_rcu(&mod->list, &modules);
 	mutex_unlock(&module_mutex);
 
+	/* add kobject, so we can reference it. */
+	err = mod_sysfs_init(mod);
+	if (err)
+		goto unlink;
+
 	err = parse_args(mod->name, mod->args, mod->kp, mod->num_kp, NULL);
 	if (err < 0)
-		goto unlink;
+		goto sysfs_uninit;
 
 	err = mod_sysfs_setup(mod, mod->kp, mod->num_kp);
 	if (err < 0)
-		goto unlink;
+		goto sysfs_uninit;
 	add_sect_attrs(mod, hdr->e_shnum, secstrings, sechdrs);
 	add_notes_attrs(mod, hdr->e_shnum, secstrings, sechdrs);
 
@@ -2507,18 +2509,19 @@ static noinline struct module *load_module(void __user *umod,
 	/* Done! */
 	return mod;
 
+ sysfs_uninit:
+	kobject_del(&mod->mkobj.kobj);
+	kobject_put(&mod->mkobj.kobj);
  unlink:
 	mutex_lock(&module_mutex);
 	/* Unlink carefully: kallsyms could be walking list. */
 	list_del_rcu(&mod->list);
+ already_exists:
 	mutex_unlock(&module_mutex);
 	synchronize_sched();
 	module_arch_cleanup(mod);
  cleanup:
 	free_modinfo(mod);
-	kobject_del(&mod->mkobj.kobj);
-	kobject_put(&mod->mkobj.kobj);
- free_unload:
 	module_unload_free(mod);
 #if defined(CONFIG_MODULE_UNLOAD)
 	free_percpu(mod->refptr);
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[Regression] Crash in load_module() while freeing args, Rafael J. Wysocki, (Tue May 25, 2:00 pm)
Re: [Regression] Crash in load_module() while freeing args, Rafael J. Wysocki, (Tue May 25, 3:54 pm)
Re: [Regression] Crash in load_module() while freeing args, Linus Torvalds, (Tue May 25, 4:47 pm)
Re: [Regression] Crash in load_module() while freeing args, Linus Torvalds, (Wed May 26, 8:41 am)
Re: [Regression] Crash in load_module() while freeing args, Rafael J. Wysocki, (Wed May 26, 3:56 pm)
Re: [Regression] Crash in load_module() while freeing args, Linus Torvalds, (Wed May 26, 4:07 pm)
Re: [Regression] Crash in load_module() while freeing args, Rusty Russell, (Wed May 26, 10:26 pm)
Re: [Regression] Crash in load_module() while freeing args, Brandon Philips, (Thu May 27, 11:46 am)
Re: [Regression] Crash in load_module() while freeing args, Rafael J. Wysocki, (Thu May 27, 2:57 pm)
Re: [Regression] Crash in load_module() while freeing args, Rusty Russell, (Mon May 31, 12:54 am)
[PATCH 0/2] kernel/module.c locking changes, Rusty Russell, (Mon May 31, 5:00 am)
[PATCH 1/2] module: make locking more fine-grained., Rusty Russell, (Mon May 31, 5:01 am)
[PATCH 1/2] Make the module 'usage' lists be two-way, Linus Torvalds, (Mon May 31, 1:16 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Rusty Russell, (Mon May 31, 6:37 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Américo Wang, (Mon May 31, 7:44 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Rusty Russell, (Mon May 31, 8:42 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Linus Torvalds, (Mon May 31, 8:51 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Linus Torvalds, (Mon May 31, 9:00 pm)
Re: [PATCH 1/2] Make the module 'usage' lists be two-way, Linus Torvalds, (Mon May 31, 9:05 pm)
Re: [PATCH 1/2] module: make locking more fine-grained., Américo Wang, (Mon May 31, 10:38 pm)
Re: [PATCH 1/2] module: make locking more fine-grained., Rusty Russell, (Mon May 31, 10:55 pm)
Re: [PATCH 2/2] module: fix bne2 "gave up waiting for init ..., Linus Torvalds, (Tue Jun 1, 9:44 pm)