linux-kernel - Re: [patch] ext4: underflow in alignment check

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] [day] [month] [year] [list]

Message-ID: <20160621130644.GO32247@mwanda>
Date:	Tue, 21 Jun 2016 16:06:44 +0300
From:	Dan Carpenter <dan.carpenter@...cle.com>
To:	Jan Kara <jack@...e.cz>
Cc:	"Theodore Ts'o" <tytso@....edu>,
	Andreas Dilger <adilger.kernel@...ger.ca>,
	linux-ext4@...r.kernel.org, linux-kernel@...r.kernel.org,
	kernel-janitors@...r.kernel.org
Subject: Re: [patch] ext4: underflow in alignment check

On Tue, Jun 21, 2016 at 09:43:53AM +0200, Jan Kara wrote:
> On Mon 20-06-16 22:53:26, Dan Carpenter wrote:
> > On Mon, Jun 20, 2016 at 06:02:04PM +0200, Jan Kara wrote:
> > > On Thu 16-06-16 10:07:09, Dan Carpenter wrote:
> > > > My static checker complains that this can underflow if arg is negative
> > > > which is true.
> > > > 
> > > > Signed-off-by: Dan Carpenter <dan.carpenter@...cle.com>
> > > 
> > > How come? (1 << 30) fits even into 32-bit signed type. So where's the
> > > problem?
> > 
> > Bad changelog...  I was talking about a different issue.  I was casting
> > it to unsigned to take advantage of type promototion.  Assume we have:
> > 
> > int arg = 1 << 31;
> > 
> > (arg > (1 << 30)) // <-- this is false
> > (arg > (1U << 30)) // <-- this is true so there is no underflow.
> 
> I see, but match_int() - or more precisely match_number() returns -ERANGE
> when the number is > INT_MAX, subsequently we check whether the number is <
> 0 (Opt_inode_readahead_blks has flag MOPT_GTE0 set) and bail out if yes. So
> at the place you are modifying we are sure the number is in [0, INT_MAX].
> So the condition (arg > (1 << 30)) is pointless - just defensive
> programming in case we decide e.g. to upgrade the type of 'arg' to long - but
> not wrong...

Ah.  Smatch wasn't able to figure out that MOPT_GTE0 was set.

Thanks for reviewing this.

regards,
dan carpenter