Message-ID: <alpine.LNX.2.00.1412121737280.6587@pobox.suse.cz>
Date:	Fri, 12 Dec 2014 17:58:19 +0100 (CET)
From:	Miroslav Benes <mbenes@...e.cz>
To:	Josh Poimboeuf <jpoimboe@...hat.com>
cc:	Seth Jennings <sjenning@...hat.com>, Jiri Kosina <jkosina@...e.cz>,
	Vojtech Pavlik <vojtech@...e.cz>,
	Steven Rostedt <rostedt@...dmis.org>,
	Petr Mladek <pmladek@...e.cz>,
	Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>,
	Jiri Slaby <jslaby@...e.cz>,
	Christoph Hellwig <hch@...radead.org>,
	Greg KH <gregkh@...uxfoundation.org>,
	Andy Lutomirski <luto@...capital.net>,
	live-patching@...r.kernel.org, x86@...nel.org, kpatch@...hat.com,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCHv6 2/3] kernel: add support for live patching


Hi, 

I think we are really close (or I hope so). I found few suspicious things 
or nitpicks though. They might have applied also to v5, but I didn't 
manage to look at that. Sorry about that.

On Wed, 10 Dec 2014, Josh Poimboeuf wrote:

> +/* klp_mutex must be held by caller */
> +static bool klp_patch_is_registered(struct klp_patch *patch)

Maybe klp_is_patch_registered is more appropriate name (consistent with 
other predicates in the file).

> +{
> +	struct klp_patch *mypatch;
> +
> +	list_for_each_entry(mypatch, &klp_patches, list)
> +		if (mypatch == patch)
> +			return true;
> +
> +	return false;
> +}

[...]

> +static int klp_disable_func(struct klp_func *func)
> +{
> +	int ret;
> +
> +	if (WARN_ON(func->state != KLP_ENABLED))
> +		return -EINVAL;
> +
> +	if (WARN_ON(!func->old_addr))
> +		return -EINVAL;
> +
> +	ret = unregister_ftrace_function(func->fops);
> +	if (ret) {
> +		pr_err("failed to unregister ftrace handler for function '%s' (%d)\n",
> +		       func->old_name, ret);
> +		return ret;
> +	}
> +
> +	ret = ftrace_set_filter_ip(func->fops, func->old_addr, 1, 0);
> +	if (ret)
> +		pr_warn("function unregister succeeded but failed to clear the filter\n");
> +
> +	func->state = KLP_DISABLED;
> +
> +	return 0;
> +}
> +
> +static int klp_enable_func(struct klp_func *func)
> +{
> +	int ret;
> +
> +	if (WARN_ON(!func->old_addr))
> +		return -EINVAL;
> +
> +	if (WARN_ON(func->state != KLP_DISABLED))
> +		return -EINVAL;
> +
> +	ret = ftrace_set_filter_ip(func->fops, func->old_addr, 0, 0);
> +	if (ret) {
> +		pr_err("failed to set ftrace filter for function '%s' (%d)\n",
> +		       func->old_name, ret);
> +		return ret;
> +	}
> +
> +	ret = register_ftrace_function(func->fops);
> +	if (ret) {
> +		pr_err("failed to register ftrace handler for function '%s' (%d)\n",
> +		       func->old_name, ret);
> +		ftrace_set_filter_ip(func->fops, func->old_addr, 1, 0);
> +	} else {
> +		func->state = KLP_ENABLED;
> +	}
> +
> +	return ret;
> +}

Just to be sure about our policy. We want to be stricter during enabling 
than in disabling process. Is that correct? Otherwise there is 
inconsistency in pr_* macros and return values. Also fops could be 
hypothetically registered back when ftrace_set_filter_ip fails in 
klp_disable_func. I just want to be sure that we didn't overlook 
anything...

[...]

> +static int klp_init_func(struct klp_object *obj, struct klp_func *func)
> +{
> +	struct ftrace_ops *ops;
> +	int ret;
> +
> +	func->state = KLP_DISABLED;
> +
> +	ops = kzalloc(sizeof(*ops), GFP_KERNEL);
> +	if (!ops)
> +		ret = -ENOMEM;

There should be return -ENOMEM.

> +
> +	ops->private = func;
> +	ops->func = klp_ftrace_handler;
> +	ops->flags = FTRACE_OPS_FL_SAVE_REGS | FTRACE_OPS_FL_DYNAMIC;
> +	func->fops = ops;
> +
> +	ret = kobject_init_and_add(&func->kobj, &klp_ktype_func,
> +				   obj->kobj, func->old_name);
> +	if (ret) {
> +		kfree(func->fops);
> +		return ret;
> +	}
> +
> +	return 0;
> +}

[...]

> +static int klp_init_object(struct klp_patch *patch, struct klp_object *obj)
> +{
> +	struct klp_func *func;
> +	int ret;
> +	const char *name;
> +
> +	if (!obj->funcs)
> +		return -EINVAL;
> +
> +	obj->state = KLP_DISABLED;
> +
> +	klp_find_object_module(obj);
> +
> +	name = klp_is_module(obj) ? obj->name : "vmlinux";
> +	obj->kobj = kobject_create_and_add(name, &patch->kobj);
> +	if (!obj->kobj)
> +		return -ENOMEM;
> +
> +	for (func = obj->funcs; func->old_name; func++) {
> +		ret = klp_init_func(obj, func);
> +		if (ret)
> +			goto free;
> +	}
> +
> +	if (klp_is_object_loaded(obj)) {
> +		ret = klp_init_object_loaded(patch, obj);
> +		if (ret)
> +			goto free;
> +	}
> +
> +	return 0;
> +
> +free:
> +	klp_free_funcs_limited(obj, func);
> +	return ret;
> +}

Shouldn't we call kobject_put(obj->kobj) in free branch? If I am not wrong 
it is not freed anywhere else. We free only already initialized functions 
and already initialized objects later in klp_init_patch, but not the 
kobject of the currently failing object.

> +static int klp_init_patch(struct klp_patch *patch)
> +{
> +	struct klp_object *obj;
> +	int ret;
> +
> +	if (!patch->objs)
> +		return -EINVAL;
> +
> +	mutex_lock(&klp_mutex);
> +
> +	patch->state = KLP_DISABLED;
> +
> +	ret = kobject_init_and_add(&patch->kobj, &klp_ktype_patch,
> +				   klp_root_kobj, patch->mod->name);
> +	if (ret)
> +		goto unlock;
> +
> +	for (obj = patch->objs; obj->funcs; obj++) {
> +		ret = klp_init_object(patch, obj);
> +		if (ret)
> +			goto free;
> +	}
> +
> +	list_add(&patch->list, &klp_patches);
> +
> +	mutex_unlock(&klp_mutex);
> +
> +	return 0;
> +
> +free:
> +	klp_free_objects_limited(patch, obj);
> +	kobject_put(&patch->kobj);
> +unlock:
> +	mutex_unlock(&klp_mutex);
> +	return ret;
> +}

And that is everything. I like it, it has improved a lot. I hope that 
there are no other problems. I am getting blind looking at it all the 
time :)

Thank you
Mira
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists