linux-kernel - Re: [PATCH] jffs2:freely allocate memory when parameters are invalid

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <20190920125442.GA20754@ZenIV.linux.org.uk>
Date:   Fri, 20 Sep 2019 13:54:42 +0100
From:   Al Viro <viro@...iv.linux.org.uk>
To:     Xiaoming Ni <nixiaoming@...wei.com>
Cc:     dwmw2@...radead.org, dilinger@...ued.net, richard@....at,
        houtao1@...wei.com, bbrezillon@...nel.org, daniel.santos@...ox.com,
        linux-mtd@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] jffs2:freely allocate memory when parameters are invalid

On Fri, Sep 20, 2019 at 01:45:33PM +0100, Al Viro wrote:
> On Fri, Sep 20, 2019 at 08:21:53PM +0800, Xiaoming Ni wrote:
> > 
> > 
> > On 2019/9/20 19:43, Al Viro wrote:
> > > On Fri, Sep 20, 2019 at 02:54:38PM +0800, Xiaoming Ni wrote:
> > >> Use kzalloc() to allocate memory in jffs2_fill_super().
> > >> Freeing memory when jffs2_parse_options() fails will cause
> > >> use-after-free and double-free in jffs2_kill_sb()
> > > 
> > > ... so we are not freeing it there.  What's the problem?
> > 
> > No code logic issues, no memory leaks
> > 
> > But there is too much code logic between memory allocation and free,
> > which is difficult to understand.
> 
> Er?  An instance of jffs2 superblock might have a related object
> attached to it; it is created in jffs2 superblock constructor and
> freed in destructor.
> 
> > The modified code is easier to understand.
> 
> You are making the cleanup logics harder to follow.

PS: the whole point of ->kill_sb() is that it's always called on
superblock destruction, whether that instance had been fully set
up of failed halfway through.

In particular, anything like foofs_fill_super() *will* be followed
by ->kill_sb().  Always.  Which allows for simpler logics in
failure exits.  And the main thing about those is that they are
always the bitrot hot spots - they are systematically undertested,
so that's the last place where you want something non-trivial.

As for "too much code between"...  Huh?  We fail jffs2_fill_super()
immediately, which has get_tree_mtd() (or mount_mtd() in slightly
earlier kernels) destroy the superblock there and then...