[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CACMJSeu1NYNbQkiZC1BOs4QKhR4+VwYRrpDgq8gczRujt13zEQ@mail.gmail.com>
Date: Fri, 15 Sep 2023 12:09:28 +0200
From: Bartosz Golaszewski <bartosz.golaszewski@...aro.org>
To: Alexey Dobriyan <adobriyan@...il.com>
Cc: Peter Zijlstra <peterz@...radead.org>,
linux-kernel@...r.kernel.org,
Linus Walleij <linus.walleij@...aro.org>,
torvalds@...ux-foundation.org, akpm@...ux-foundation.org
Subject: Re: Buggy __free(kfree) usage pattern already in tree
On Fri, 15 Sept 2023 at 11:56, Alexey Dobriyan <adobriyan@...il.com> wrote:
>
> __free() got some usage and some of the usage is buggy:
>
> 832 static struct fwnode_handle *
> 833 gpio_sim_make_bank_swnode(struct gpio_sim_bank *bank,
> 834 struct fwnode_handle *parent)
> 835 {
> 838 char **line_names __free(kfree) = NULL;
> // returns NULL or ERR_PTR(-E)
> 848 line_names = gpio_sim_make_line_names(bank, &line_names_size);
> 849 if (IS_ERR(line_names))
> 850 return ERR_CAST(line_names);
>
Thanks for the report, I'll send a fix.
Bart
>
> This pattern will result in calling kfree() on error value.
> And there are no compiler or sparse checking these things.
>
> This test module demonstrates the landmine:
>
> [ 812.981089] ------------[ cut here ]------------
> [ 812.981597] WARNING: CPU: 0 PID: 1326 at mm/slab_common.c:991 free_large_kmalloc+0x50/0x80
> [ 813.013266] ---[ end trace 0000000000000000 ]---
> [ 813.013800] object pointer: 0xfffffffffffffff4
>
> #include <linux/module.h>
> #include <linux/slab.h>
> #include <linux/cleanup.h>
>
> struct S {
> int x;
> };
>
> static struct S* f(void)
> {
> struct S* s = kmalloc(sizeof(struct S), GFP_KERNEL);
> s = NULL;
> return s ?: ERR_PTR(-ENOMEM);
> }
>
> static int __init xxx_module_init(void)
> {
> struct S *s __free(kfree) = NULL;
> s = f();
> if (IS_ERR(s)) {
> return PTR_ERR(s);
> }
> return 0;
> }
>
> static void __exit xxx_module_exit(void)
> {
> }
> module_init(xxx_module_init);
> module_exit(xxx_module_exit);
> MODULE_LICENSE("GPL");
Powered by blists - more mailing lists