linux-kernel - Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

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: <CAK8P3a2nTi6AOf-4xJN3wurCXC52mLPBVbjgvcv6Cz=8A05X0A@mail.gmail.com>
Date:   Fri, 1 May 2020 19:36:55 +0200
From:   Arnd Bergmann <arnd@...db.de>
To:     James Bottomley <jejb@...ux.ibm.com>
Cc:     John Garry <john.garry@...wei.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "Martin K. Petersen" <martin.petersen@...cle.com>,
        James Bottomley <James.Bottomley@...eleye.com>,
        Hannes Reinecke <hare@...e.com>,
        linux-scsi <linux-scsi@...r.kernel.org>
Subject: Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

On Fri, May 1, 2020 at 4:53 PM James Bottomley <jejb@...ux.ibm.com> wrote:
> On Fri, 2020-05-01 at 09:54 +0200, Arnd Bergmann wrote:
> > On Fri, May 1, 2020 at 9:48 AM John Garry <john.garry@...wei.com>
> > wrote:
> > I found one hack that would work, but I think it's too ugly and
> > likely not well-defined either:
> >
> > struct ssp_response_iu {
> > ...
> >         struct {
> >                 u8      dummy[0]; /* a struct must have at least one
> > non-flexible member */
>
> If gcc is now warning about zero length members, why isn't it warning
> about this one ... are unions temporarily excluded?

It does not warn about all zero-length arrays, but it does warn when you
try to access an array with an out-of-range index, and this apparently
got extended in gcc-10 to any index for zero-length arrays.

> >                 u8      resp_data[]; /* allowed here because it's at
> > the one of a struct */
> >         };
> >         u8     sense_data[];
> > } __attribute__ ((packed));
>
> Let's go back to what the standard says:  we want the data beyond the
> ssp_response_iu to be addressable either as sense_data if it's an error
> return or resp_data if it's a real response.  What about trying to use
> an alias attribute inside the structure ... will that work on gcc-10?

I think alias attributes only work for functions and variables, but not
for struct members.

A "#define sense_data resp_data" would obviously work, but it's
rather error-prone when other code uses the same identifiers.

Another option would be an inline helper like

static inline u8 *ssp_response_data(struct ssp_response_iu *iu)
{
      return iu.resp_data;
}

      Arnd