Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 23 Apr 2021 18:40:33 +0000
From:   Luis Chamberlain <mcgrof@kernel.org>
To:     Shuah Khan <skhan@linuxfoundation.org>
Cc:     Anirudh Rayabharam <mail@anirudhrb.com>,
        Shuah Khan <shuah@kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Junyong Sun <sunjy516@gmail.com>,
        syzbot+de271708674e2093097b@syzkaller.appspotmail.com,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3] firmware_loader: fix use-after-free in
 firmware_fallback_sysfs
Message-ID: <20210423184033.GQ4332@42.do-not-panic.com>
References: <20210414085406.1842-1-mail@anirudhrb.com>
 <20210414125540.GJ4332@42.do-not-panic.com>
 <1e12bb24-9bb8-dad7-d852-02173ce41b6a@linuxfoundation.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1e12bb24-9bb8-dad7-d852-02173ce41b6a@linuxfoundation.org>
Precedence: bulk

On Wed, Apr 14, 2021 at 09:26:55AM -0600, Shuah Khan wrote:
> On 4/14/21 6:55 AM, Luis Chamberlain wrote:
> > In either case I documented well *why* we do these error checks
> > before sending a code to userspace on fw_sysfs_wait_timeout() since
> > otherwise it would be easy to regress that code, so please also
> > document that as I did.
> > 
> > I'll re-iterate again also:
> > 
> > 	Shuah's commit 0542ad88fbdd81bb ("firmware loader: Fix
> > 	_request_firmware_load() return val for fw load abort") also wanted to
> > 	distinguish the timeout vs -ENOMEM, but for some reason in the timeout
> > 	case -EAGAIN was being sent back to userspace. I am no longer sure if
> > 	that is a good idea, but since we started doing that at some point I
> > 	guess we want to keep that behaviour.
> > 
> > Shuah, can you think of any reason to retain -EAGAIN other than you
> > introduced it here? If there's no real good reason I think it can
> > simplify the error handling here. But, we *would* change what we do
> > to userspace... and for that reason we may have to live with it.
> > 
> 
> As I recall the reason for this patch was to be able to differentiate
> between timing out vs no memory case when driver was attempting to
> load firmware. I wish I added why to the change log.
> 
> The code seems to have changed a lot since my commit. I will take a look
> at the closely and let you know if this is still necessary late on
> today.

Shuah, *poke*

  Luis