linux-kernel - Re: [PATCH] tty_io: Fix a missing-check bug in drivers/tty/tty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20190522102900.GC2200@localhost>
Date:   Wed, 22 May 2019 12:29:00 +0200
From:   Johan Hovold <johan@...nel.org>
To:     Jiri Slaby <jslaby@...e.cz>
Cc:     Gen Zhang <blackgod016574@...il.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] tty_io: Fix a missing-check bug in drivers/tty/tty_io.c

On Wed, May 22, 2019 at 10:15:56AM +0200, Jiri Slaby wrote:
> On 22. 05. 19, 10:06, Gen Zhang wrote:
> > On Wed, May 22, 2019 at 06:25:36AM +0200, Jiri Slaby wrote:
> >> On 22. 05. 19, 3:40, Gen Zhang wrote:
> >>> In alloc_tty_struct(), tty->dev is assigned by tty_get_device(). And it
> >>> calls class_find_device(). And class_find_device() may return NULL.
> >>> And tty->dev is dereferenced in the following codes. When 
> >>> tty_get_device() returns NULL, dereferencing this tty->dev null pointer
> >>> may cause the kernel go wrong. Thus we should check tty->dev.

Where do you see that the kernel is dereferencing tty->dev without
checking for NULL first? If you can find that, then that would indeed be
a bug that needs fixing.

> >>> Further, if tty_get_device() returns NULL, we should free tty and 
> >>> return NULL.
> >>>
> >>> Signed-off-by: Gen Zhang <blackgod016574@...il.com>
> >>>
> >>> ---
> >>> diff --git a/drivers/tty/tty_io.c b/drivers/tty/tty_io.c
> >>> index 033ac7e..1444b59 100644
> >>> --- a/drivers/tty/tty_io.c
> >>> +++ b/drivers/tty/tty_io.c
> >>> @@ -3008,6 +3008,10 @@ struct tty_struct *alloc_tty_struct(struct tty_driver *driver, int idx)
> >>>  	tty->index = idx;
> >>>  	tty_line_name(driver, idx, tty->name);
> >>>  	tty->dev = tty_get_device(tty);
> >>> +	if (!tty->dev) {
> >>> +		kfree(tty);
> >>> +		return NULL;
> >>> +	}
> >>
> >> This is incorrect, you introduced an ldisc reference leak.
> > Thanks for your reply, Jiri!
> > And what do you mean by an ldisc reference leak? I did't get the reason
> > of introducing it.
> 
> Look at the top of alloc_tty_struct: there is tty_ldisc_init. If
> tty_get_device fails here, you have to call tty_ldisc_deinit. Better,
> you should add a failure-handling tail to this function and "goto" there.
> 
> >> And can this happen at all?
> > I think tty_get_device() may happen to return NULL. Because it calls 
> > class_find_device() and there's a chance class_find_device() returns
> > NULL.
> 
> Sure, but can class_find_device return NULL in this tty case here?

Yes, it can and will and that's fine, not all ttys have a struct device
(e.g. ptys). 

This patch is broken and should be dropped.

Johan