Re: [PATCH] fs: pstore: fix double-free on ramoops_init_przs

[Kees Cook <keescook@chromium.org>]

On Tue, Jan 07, 2020 at 02:04:46PM +0300, Cengiz Can wrote:
> According to Coverity scanner (CID 1457526) kfree on ram.c:591 frees
> label which has already been freed.
> 
> Here's the flow as I have understood (this is my first time reading
> pstore's files):
> 
> Whenever `persistent_ram_new` fails, it implicitly calls
> `persistent_ram_free(prz)` which already does `kfree(prz->label)` and a
> `kfree(prz)` consequently.
> 
> Removed `kfree(label)` to prevent double-free.

I think this is a false positive (have you actually hit the
double-free?). The logic in this area is:

label = kmalloc(...)
prz[i] = persistent_ram_new(..., label, ...)
if (IS_ERR(prz[i])) {
    kfree(label)
    while (i > 0) {
        i--;
        persistent_ram_free(prz[i]);
    }
}

nothing was freeing the label on the failed prz, but all the other prz
labels were free (i.e. there is a "i--" that skips the failed prz
alloc).

-Kees

> 
> Signed-off-by: Cengiz Can <cengiz@kernel.wtf>
> ---
>  fs/pstore/ram.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/fs/pstore/ram.c b/fs/pstore/ram.c
> index 487ee39b4..e196aa08f 100644
> --- a/fs/pstore/ram.c
> +++ b/fs/pstore/ram.c
> @@ -588,7 +588,6 @@ static int ramoops_init_przs(const char *name,
>  			dev_err(dev, "failed to request %s mem region (0x%zx@0x%llx): %d\n",
>  				name, record_size,
>  				(unsigned long long)*paddr, err);
> -			kfree(label);
>  
>  			while (i > 0) {
>  				i--;
> -- 
> 2.24.1
> 

-- 
Kees Cook