linux-kernel - Re: [PATCH v1 1/1] s390: Use string_upper() instead of open coded variant

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: <YVwSxGyx45gs2+ZW@osiris>
Date:   Tue, 5 Oct 2021 10:54:28 +0200
From:   Heiko Carstens <hca@...ux.ibm.com>
To:     Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Cc:     linux-s390@...r.kernel.org, linux-kernel@...r.kernel.org,
        Vasily Gorbik <gor@...ux.ibm.com>,
        Christian Borntraeger <borntraeger@...ibm.com>
Subject: Re: [PATCH v1 1/1] s390: Use string_upper() instead of open coded
 variant

On Tue, Oct 05, 2021 at 11:18:38AM +0300, Andy Shevchenko wrote:
> On Mon, Oct 04, 2021 at 10:31:46PM +0200, Heiko Carstens wrote:
> > On Fri, Oct 01, 2021 at 04:02:01PM +0300, Andy Shevchenko wrote:
> > > +	/* Segment name is limited by 8 characters + NUL */
> > > +	char tmp[8 + 1];
> > >  	int i;
> > >  
> > > -	for (i = 0; i < 8; i++) {
> > > -		if (name[i] == '\0')
> > > -			break;
> > > -		dcss_name[i] = toupper(name[i]);
> > > -	}
> > > -	for (; i < 8; i++)
> > > -		dcss_name[i] = ' ';
> > > +	/*
> > > +	 * This snprintf() call does two things:
> > > +	 * - makes a NUL-terminated copy of the input string
> > > +	 * - pads it with spaces
> > > +	 */
> > > +	snprintf(tmp, sizeof(tmp), "%s        ", name);
> > 
> > I can't say I like code where I have to count spaces in order to
> > verify if the code is actually correct.
> 
> I understand your point, but have any idea how to make it differently
> and not ugly at the same time?

Don't know. You could use strncopy+strlen+memset (with space
character). After all I'm not very convinced that the resulting code
buys us anything compared to the current variant.

> > > +	string_upper(dcss_name, tmp);
> 
> ...
> 
> > >  static struct dcss_segment *
> > >  segment_by_name (char *name)
> > >  {
> > > -	char dcss_name[9];
> > > +	char dcss_name[8];
> > 
> > string_upper will copy the terminating NUL-byte. By reducing the size
> > of dcss_name to 8 bytes this will result in stack corruption.
> 
> Nope. Even in the original code this additional byte is left unused.

I'm talking about the new code, not the old code: If "name" points to
a NUL terminated eight chararacter string, then the new code will use
snprintf to copy it 1:1 to tmp, and the subsequent string_upper() will
copy the string (upper cased) to dcss_name, now including the NUL
terminating byte, which won't fit into dcss_name.
Am I missing something here?