linux-kernel - Re: [PATCH] wifi: nl80211: drop impossible negative band 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] [thread-next>] [day] [month] [year] [list]

Message-ID: <074b66ab3d63e6640ecd6962c074702225fba19e.camel@sipsolutions.net>
Date: Wed, 04 Feb 2026 12:11:20 +0100
From: Johannes Berg <johannes@...solutions.net>
To: sun jian <sun.jian.kdev@...il.com>
Cc: linux-wireless@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] wifi: nl80211: drop impossible negative band check

On Wed, 2026-02-04 at 17:13 +0800, sun jian wrote:
> On Wed, Feb 4, 2026 at 4:36 PM Johannes Berg <johannes@...solutions.net> wrote:
> > 
> > On Wed, 2026-02-04 at 16:18 +0800, Sun Jian wrote:
> > > band is derived from nla_type() of a nested netlink attribute, which isAgreed — in general the enum underlying type can be signed.
> > > a masked u16 value and therefore cannot be negative. Drop the dead
> > > "band < 0" checks and keep the upper bound validation.
> > 
> > I've seen this before, but I'm not really convinced it is entirely
> > correct. C says:
> > 
> >    All enumerations have an underlying type. The underlying type can be
> >    explicitly specified using an enum type specifier and is its fixed
> >    underlying type. If it is not explicitly specified, the underlying
> >    type is the enumeration’s compatible type, which is either char or a
> >    standard or extended signed or unsigned integer type.
> > 
> 
> Agreed — in general the enum underlying type can be signed.

But nothing says it cannot be "signed char".

> > It would thus _seem_ to be possible for an enum to generally be a signed
> > type, and therefore a 'signed short', and therefore an nla_type() that's
> > a u16 could end up with a negative value...

I was just using 'signed short' as an example, but your argument:

> The key detail here is that band isn't assigned the raw __u16
> nla->nla_type, but nla_type().
> 
> And nla_type() is effectively:
> nla->nla_type & NLA_TYPE_MASK
> 
> and NLA_TYPE_MASK clears the two high flag bits:
> NLA_F_NESTED (1 << 15)
> NLA_F_NET_BYTEORDER (1 << 14)
> 
> So the result is restricted to the low 14 bits, i.e. 0..0x3fff.
> 
> With that restriction, even if enum nl80211_band ended up with a signed
> 16-bit underlying type, the sign bit (bit 15) can never be set by
> nla_type(), so the value cannot become negative.

applies _only_ to signed short, not to signed char?

Now we can argue a "sane compiler" won't do that, and we can also argue
that "gcc and clang are sane compilers", although sometimes I definitely
have doubts about the latter ;-)

johannes