linux-kernel - Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer initialization

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: <aPkii_HgX1f0MUVc@smile.fi.intel.com>
Date: Wed, 22 Oct 2025 21:29:31 +0300
From: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To: Thorsten Blum <thorsten.blum@...ux.dev>
Cc: Giovanni Cabiddu <giovanni.cabiddu@...el.com>,
	Herbert Xu <herbert@...dor.apana.org.au>,
	"David S. Miller" <davem@...emloft.net>,
	Jack Xu <jack.xu@...el.com>,
	Suman Kumar Chakraborty <suman.kumar.chakraborty@...el.com>,
	Qianfeng Rong <rongqianfeng@...o.com>, qat-linux@...el.com,
	linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] crypto: qat - use strscpy_pad to simplify buffer
 initialization

On Wed, Oct 22, 2025 at 09:17:22PM +0300, Andy Shevchenko wrote:
> On Wed, Oct 22, 2025 at 02:36:19PM +0200, Thorsten Blum wrote:
> > Use strscpy_pad() to copy the string and zero-pad the destination buffer
> > in a single step instead of zero-initializing the buffer first and then
> > immediately overwriting it using strscpy().
> > 
> > Replace the magic number 16 with sizeof(buf) and remove the redundant
> > parentheses around kstrtoul() while we're at it.
> 
> I understand that you focused on strscpy*() conversions, but the below I think
> needs a bigger refactoring, see my remarks.

...

> > -	char buf[16] = {0};
> > +	char buf[16] = {};
> >  	unsigned long ae = 0;
> >  	int i;
> >  
> > -	strscpy(buf, str, sizeof(buf));
> > -	for (i = 0; i < 16; i++) {
> > +	strscpy_pad(buf, str);
> 
> First of all, why do we need a _pad() version here? Is the data somehow being
> used as a whole?
> 
> > +	for (i = 0; i < sizeof(buf); i++) {
> >  		if (!isdigit(buf[i])) {
> >  			buf[i] = '\0';
> >  			break;
> >  		}
> >  	}
> > -	if ((kstrtoul(buf, 10, &ae)))
> > +	if (kstrtoul(buf, 10, &ae))
> >  		return -EFAULT;

On top of that the function is called only from one place and returns different
error code, instead it would have returned what kstrtoul() gives...

> Looking at this, it tries to work around the kstrtoul() inability to perform
> partial parses. Instead, this should do something like
> 
> 	unsigned long long x;
> 	const char *end;
> 
> 	simple_strtoull(...);
> 	if (x > UINT_MAX || end == buf)
> 		return $ERR; // wrong input / overflow

Yeah, the overflow check here is not comprehensive, it won't catch the overflow
(wrap around) of 64-bit value. But we can add a check for the end not to be
farther than ~19 characters from the start, which would correspond the initial
copy of 16 characters.

-- 
With Best Regards,
Andy Shevchenko