linux-kernel - Re: [PATCH v2] ubifs: Fix read out-of-bounds in ubifs_jnl_write

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: <244303467.160590.1608764840819.JavaMail.zimbra@nod.at>
Date:   Thu, 24 Dec 2020 00:07:20 +0100 (CET)
From:   Richard Weinberger <richard@....at>
To:     Chengsong Ke <kechengsong@...wei.com>
Cc:     Sascha Hauer <s.hauer@...gutronix.de>,
        linux-mtd <linux-mtd@...ts.infradead.org>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        wangfangpeng1 <wangfangpeng1@...wei.com>,
        chengzhihao1 <chengzhihao1@...wei.com>
Subject: Re: [PATCH v2] ubifs: Fix read out-of-bounds in
 ubifs_jnl_write_inode()

Chengsong Ke,

----- Ursprüngliche Mail -----
> The memory area allocated in ubifs_jnl_write_inode() is not aligned with 8
> bytes:
> ino_start = ino = kmalloc(write_len, GFP_NOFS);
> 
> When ino_start passed into write_head -> ubifs_wbuf_write_nolock:
>    n = aligned_len >> c->max_write_shift;
>    if (n) {
>      n <<= c->max_write_shift;
>      err = ubifs_leb_write(c, wbuf->lnum, buf + written, wbuf->offs, n);
>      // Read oob occurs here, read n bytes from buf, and buf is passed from
>      @ino_start which is
>      // not 8 bytes aligned(write_len < n). Program read (n - write_len) more bytes.
>    }
> 
> Reproducer:
> 0. config KASAN && apply print.patch
> 1. mount ubifs on /root/temp
> 2. run test.sh

What does test.sh do?

> 3. cd /root/temp && ls // change atime for link_file
> 4. wait 1~2 minutes
> 
> In order to solve the read oob problem in ubifs_wbuf_write_nolock, just align
> the write_len to
> 8 bytes when alloc the memory. So that this patch will not affect the use of
> write_len in other
> functions, such as ubifs_jnl_write_inode->make_reservation and
> ubifs_jnl_write_inode->ubifs_node_calc_hash.

I gave this a second thought and I'm not so sure anymore what exactly is going on.
The problem is real, I fully agree with you but I need to dig deeper into
the journal and wbuf code to double check that we really fix the right thing
and not just paper other something.

Thanks,
//richard