netdev - Re: [PATCH] nfp: remove h from printk format specifier

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 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]

Date:   Fri, 25 Dec 2020 15:00:52 -0800
From:   Joe Perches <joe@...ches.com>
To:     Tom Rix <trix@...hat.com>,
        Simon Horman <simon.horman@...ronome.com>
Cc:     kuba@...nel.org, davem@...emloft.net, ast@...nel.org,
        daniel@...earbox.net, andrii@...nel.org, kafai@...com,
        songliubraving@...com, yhs@...com, john.fastabend@...il.com,
        kpsingh@...nel.org, gustavoars@...nel.org,
        louis.peens@...ronome.com, netdev@...r.kernel.org,
        bpf@...r.kernel.org, oss-drivers@...ronome.com,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] nfp: remove h from printk format specifier

On Fri, 2020-12-25 at 14:13 -0800, Tom Rix wrote:
> On 12/25/20 9:06 AM, Joe Perches wrote:
> > On Fri, 2020-12-25 at 06:56 -0800, Tom Rix wrote:
> > > On 12/24/20 2:39 PM, Joe Perches wrote:
> > []
> > > > Kernel code doesn't use a signed char or short with %hx or %hu very often
> > > > but in case you didn't already know, any signed char/short emitted with
> > > > anything like %hx or %hu needs to be left alone as sign extension occurs so:
> > > Yes, this would also effect checkpatch.
> > Of course but checkpatch is stupid and doesn't know types
> > so it just assumes that the type argument is not signed.
> > 
> > In general, that's a reasonable but imperfect assumption.
> > 
> > coccinelle could probably do this properly as it's a much
> > better parser.  clang-tidy should be able to as well.
> > 
> Ok.
> 
> But types not matching the format string is a larger problem.
> 
> Has there been an effort to clean these up ?

Not really no.  __printf already does a reasonable job for that.

The biggest issue for format type mismatches is the %p<foo> extensions.

__printf can only verify that the argument is a pointer, not
necessarily the 'right' type of pointed to object.

There are overflow possibilities like '"%*ph", len, pointer'
where pointer may not have len bytes available and, for instance,
mismatched uses of %pI4 and %pI6 where %pI4 expects a pointer to
4 bytes and %pI6 expects a pointer to 16 bytes.

Anyway it's not that easy a problem to analyze.