Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <ca6e40fff619aa16cb178382f5404d047ed08f3c.camel@HansenPartnership.com>
Subject: Re: [PATCH linux-next] parisc: use strscpy() to instead of strncpy()
From:   James Bottomley <James.Bottomley@HansenPartnership.com>
To:     Helge Deller <deller@gmx.de>, yang.yang29@zte.com.cn
Cc:     linux-parisc@vger.kernel.org, linux-kernel@vger.kernel.org,
        xu.panda@zte.com.cn
Date:   Tue, 27 Dec 2022 17:43:56 -0500
In-Reply-To: <279ea596-6a20-0bb8-39c3-67b45d7425a6@gmx.de>
References: <202212231040562072342@zte.com.cn>
         <0fb8a86a-ca92-8d5f-99da-6815b2d5ec3e@gmx.de>
         <eaad272203baa65ad65ac2206e5197595c88508e.camel@HansenPartnership.com>
         <279ea596-6a20-0bb8-39c3-67b45d7425a6@gmx.de>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.42.4 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Tue, 2022-12-27 at 22:38 +0100, Helge Deller wrote:
> Hi James,
> 
> On 12/27/22 13:38, James Bottomley wrote:
> > On Fri, 2022-12-23 at 08:55 +0100, Helge Deller wrote:
> > > On 12/23/22 03:40, yang.yang29@zte.com.cn wrote:
> > > > From: Xu Panda <xu.panda@zte.com.cn>
> > > > 
> > > > The implementation of strscpy() is more robust and safer.
> > > > That's now the recommended way to copy NUL-terminated strings.
> > > 
> > > Thanks for your patch, but....
> > > 
> > > > Signed-off-by: Xu Panda <xu.panda@zte.com.cn>
> > > > Signed-off-by: Yang Yang <yang.yang29@zte.com>
> > > > ---
> > > >    drivers/parisc/pdc_stable.c | 9 +++------
> > > >    1 file changed, 3 insertions(+), 6 deletions(-)
> > > > 
> > > > diff --git a/drivers/parisc/pdc_stable.c
> > > > b/drivers/parisc/pdc_stable.c
> > > > index d6af5726ddf3..403bca0021c5 100644
> > > > --- a/drivers/parisc/pdc_stable.c
> > > > +++ b/drivers/parisc/pdc_stable.c
> > > > @@ -274,8 +274,7 @@ pdcspath_hwpath_write(struct pdcspath_entry
> > > > *entry, const char *buf, size_t coun
> > > > 
> > > >          /* We'll use a local copy of buf */
> > > >          count = min_t(size_t, count, sizeof(in)-1);
> > > > -       strncpy(in, buf, count);
> > > > -       in[count] = '\0';
> > > > +       strscpy(in, buf, count + 1);
> > > 
> > > could you resend it somewhat simplified, e.g.
> > > strscpy(in, buf, sizeof(in));
> > 
> > I don't think you can: count is the size of buf, if that's <
> > sizeof(in) you've introduced a write beyond end of buffer.  In fact
> > sysfs tends to pass pages as buffers, so there's no actual problem,
> > but if that ever changed ...
> 
> Huh?... he doesn't change "count", so what's wrong with the latest
> patch?

the array buf[] is actually buf[count], so if count < 64 then
sizeof(buf) < sizeof(in) and you're copying whatever is after buf on
the stack or wherever it comes from. The amount you copy into in[]
truly has to be the smaller of count and sizeof(in).  These are file
operations, so you shouldn't rely on buf[] being null terminated
(kernfs ensures it is, but it's a dangerous thing to rely on in the
face of someone trying to exploit a stack smashing attack).

James