Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 10 Oct 2022 03:35:03 +0900 (JST)
Message-Id: <20221010.033503.2099269269123366991.syoshida@redhat.com>
To:     maarten.lankhorst@linux.intel.com, mripard@kernel.org,
        tzimmermann@suse.de, airlied@linux.ie, daniel@ffwll.ch
Cc:     dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org,
        syoshida@redhat.com,
        syzbot+c512687fff9d22327436@syzkaller.appspotmail.com
Subject: Re: [PATCH] drm/gem: Avoid use-after-free on drm_gem_mmap_obj()
 failure
From:   Shigeru Yoshida <syoshida@redhat.com>
In-Reply-To: <20220920072408.387105-1-syoshida@redhat.com>
References: <20220920072408.387105-1-syoshida@redhat.com>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

ping?

On Tue, 20 Sep 2022 16:24:08 +0900, Shigeru Yoshida wrote:
> syzbot reported use-after-free for drm_gem_object [1].  This causes
> the call trace like below:
> 
> [   75.327400][ T5723] Call Trace:
> [   75.327611][ T5723]  <TASK>
> [   75.327803][ T5723]  drm_gem_object_handle_put_unlocked+0x11e/0x1a0
> [   75.328209][ T5723]  drm_gem_object_release_handle+0x5d/0x70
> [   75.328568][ T5723]  ? drm_gem_object_handle_put_unlocked+0x1a0/0x1a0
> [   75.328965][ T5723]  idr_for_each+0x99/0x160
> [   75.329253][ T5723]  drm_gem_release+0x20/0x30
> [   75.329544][ T5723]  drm_file_free.part.0+0x269/0x310
> [   75.329867][ T5723]  drm_close_helper.isra.0+0x88/0xa0
> [   75.330305][ T5723]  drm_release+0x8e/0x1a0
> [   75.330674][ T5723]  ? drm_release_noglobal+0xc0/0xc0
> [   75.331138][ T5723]  __fput+0x10e/0x440
> [   75.331503][ T5723]  task_work_run+0x73/0xd0
> [   75.331895][ T5723]  do_exit+0x535/0x1200
> [   75.332280][ T5723]  ? ktime_get_coarse_real_ts64+0x13b/0x170
> [   75.332810][ T5723]  do_group_exit+0x51/0x100
> [   75.333215][ T5723]  __x64_sys_exit_group+0x18/0x20
> [   75.333654][ T5723]  do_syscall_64+0x37/0x90
> [   75.334058][ T5723]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> If drm_gem_mmap_obj() failed, it drops the reference count by calling
> drm_gem_object_put().  However, drm_gem_mmap() drops the reference
> count after calling drm_gem_mmap_obj() even if it failed, so it breaks
> the balance of the reference count.
> 
> This patch fixes this issue by calling drm_gem_object_put() only if
> drm_gem_mmap_obj() succeeds, and returns immediately if
> drm_gem_mmap_obj() failed without calling drm_gem_object_put().
> 
> Link: https://syzkaller.appspot.com/bug?id=c42a72b0b3bcedd95e5f132a4ccd7cd550334160 [1]
> Reported-by: syzbot+c512687fff9d22327436@syzkaller.appspotmail.com
> Signed-off-by: Shigeru Yoshida <syoshida@redhat.com>
> ---
>  drivers/gpu/drm/drm_gem.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
> index ad068865ba20..f345d38df50a 100644
> --- a/drivers/gpu/drm/drm_gem.c
> +++ b/drivers/gpu/drm/drm_gem.c
> @@ -1115,10 +1115,12 @@ int drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
>  
>  	ret = drm_gem_mmap_obj(obj, drm_vma_node_size(node) << PAGE_SHIFT,
>  			       vma);
> +	if (ret)
> +		return ret;
>  
>  	drm_gem_object_put(obj);
>  
> -	return ret;
> +	return 0;
>  }
>  EXPORT_SYMBOL(drm_gem_mmap);
>  
> -- 
> 2.37.3
>