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;
Message-ID: <038f8365-b23b-9d81-f7b2-8f8c6eb3a065@redhat.com>
Date:   Wed, 4 May 2022 11:28:07 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.0
Subject: Re: [PATCH v2] fbdev: Use helper to get fb_info in all file
 operations
Content-Language: en-US
To:     linux-kernel@vger.kernel.org, linux-fbdev@vger.kernel.org,
        Junxiao Chang <junxiao.chang@intel.com>,
        dri-devel@lists.freedesktop.org, Maxime Ripard <maxime@cerno.tech>,
        Thomas Zimmermann <tzimmermann@suse.de>,
        Sam Ravnborg <sam@ravnborg.org>
References: <20220503201934.681276-1-javierm@redhat.com>
 <YnJBGpvlViLV+0/a@phenom.ffwll.local>
From:   Javier Martinez Canillas <javierm@redhat.com>
In-Reply-To: <YnJBGpvlViLV+0/a@phenom.ffwll.local>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hello Daniel,

On 5/4/22 11:02, Daniel Vetter wrote:
> On Tue, May 03, 2022 at 10:19:34PM +0200, Javier Martinez Canillas wrote:
>> A reference to the framebuffer device struct fb_info is stored in the file
>> private data, but this reference could no longer be valid and must not be
>> accessed directly. Instead, the file_fb_info() accessor function must be
>> used since it does sanity checking to make sure that the fb_info is valid.
>>
>> This can happen for example if the registered framebuffer device is for a
>> driver that just uses a framebuffer provided by the system firmware. In
>> that case, the fbdev core would unregister the framebuffer device when a
>> real video driver is probed and ask to remove conflicting framebuffers.
>>
>> Most fbdev file operations already use the helper to get the fb_info but
>> get_fb_unmapped_area() and fb_deferred_io_fsync() don't. Fix those two.
>>
>> Since fb_deferred_io_fsync() is not in fbmem.o, the helper has to be
>> exported. Rename it and add a fb_ prefix to denote that is public now.
>>
>> Reported-by: Junxiao Chang <junxiao.chang@intel.com>
>> Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
> 
> Note that fb_file_info is hilariously racy since there's nothing
> preventing a concurrenct framebuffer_unregister. Or at least I'm not
> seeing anything. See cf4a3ae4ef33 ("fbdev: lock_fb_info cannot fail") for
> context, maybe reference that commit here in your patch.
>
> Either way this doesn't really make anything worse, so
> Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
>

Yes, I noticed is racy but at least checking this makes less likely to
occur. And thanks, I'll reference that commit in the description of v3.

BTW, I also noticed that the same race that happens with open(),read(),
close(), etc happens with the VM operations:

int fb_deferred_io_mmap(struct fb_info *info, struct vm_area_struct *vma)
{
...
	vma->vm_private_data = info;
...
}

static vm_fault_t fb_deferred_io_fault(struct vm_fault *vmf)
{
...
	struct fb_info *info = vmf->vma->vm_private_data;
...
}

static vm_fault_t fb_deferred_io_mkwrite(struct vm_fault *vmf)
{
...
	struct fb_info *info = vmf->vma->vm_private_data;
...
}

So something similar to fb_file_fb_info() is needed to check if
the vm_private_data is still valid. I guess that could be done
by using the vmf->vma->vm_file and attempting the same trick that
fb_file_fb_info() does ?

-- 
Best regards,

Javier Martinez Canillas
Linux Engineering
Red Hat