From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail-qk1-x743.google.com ([2607:f8b0:4864:20::743]) by bombadil.infradead.org with esmtps (Exim 4.92.3 #3 (Red Hat Linux)) id 1jbbUL-0008Tz-VH for barebox@lists.infradead.org; Thu, 21 May 2020 03:03:20 +0000 Received: by mail-qk1-x743.google.com with SMTP id m11so5959677qka.4 for ; Wed, 20 May 2020 20:03:17 -0700 (PDT) Date: Wed, 20 May 2020 23:04:56 -0400 From: David Dgien Message-ID: <20200521030456.GB111592@fizzbox.localdomain> References: <20200520035555.84422-1-dgienda125@gmail.com> <20200520062632.hfbmedgtofv6y665@pengutronix.de> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20200520062632.hfbmedgtofv6y665@pengutronix.de> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Sender: "barebox" Errors-To: barebox-bounces+u.kleine-koenig=pengutronix.de@lists.infradead.org Subject: Re: [PATCH] password: Fix warning with empty default password To: Uwe =?iso-8859-1?Q?Kleine-K=F6nig?= Cc: barebox@lists.infradead.org Hello On Wed, May 20, 2020 at 08:26:32AM +0200, Uwe Kleine-K=F6nig wrote: > On Tue, May 19, 2020 at 11:55:55PM -0400, David Dgien wrote: > > When CONFIG_PASSWORD_DEFAULT is unset, the default_passwd buffer is set > = > I assume you mean "If CONFIG_PASSWORD_DEFAULT is set to an empty > string". Yes. > = > > to the empty string. The read_default_passwd() function wants to read at > > least two characters from that buffer, causing GCC to generate an array > > bounds warning. > = > I cannot reproduce that warning. Which gcc version do you use and on > which platform? Mentioning the exact warning in the commit log helps > finding the resulting commit when searching for a fix. arm-none-eabi-gcc --version prints "arm-none-eabi-gcc (Arch Repository) 10.1.0" I found the issue when building for rpi_defconfig and vexpress_defconfig. The warning I get when building from master (commit c10b20dc83ac): barebox/common/password.c: In function 'login': barebox/common/password.c:173:5: warning: array subscript [1, 2147483647] i= s outside array bounds of 'const char[1]' [-Warray-bounds] 173 | c =3D buf[i]; | ~~^~~~~~~~ In file included from barebox/common/password.c:30: include/generated/passwd.h:1:19: note: while referencing 'default_passwd' 1 | static const char default_passwd[] =3D ""; | ^~~~~~~~~~~~~~ I guess the compiler doesn't know that strlen(default_passwd) =3D 0, just that length > 0 so the most it can assume is that the loop has to consume at least two chars, and the empty string only contains one. > = > > Make the default_passwd buffer have at least 2 bytes so > > this warning is not generated. > > = > > Since the read_default_passwd() function is only called when > > default_passwd is not the empty string, this is not a functional change. > = > I don't understand the problem for the empty password. With > default_passwd =3D "" we have strlen(default_passwd) =3D 0 so the for loop > doesn't run at all. Yes, that's correct, which is one reason why this is not functionally different. But the compiler doesn't seem to be smart enough to know that. > = > As I understand the code (at commit c10b20dc83ac) for uneven lengths of > default_passwd the last accessed byte is the trailing '\0' and for even > length it's the byte before the trailing '\0'. This should be ok?! > = > Am I missing something? When working on this reply, I realized there was another solution I missed when I was trying to find ways to short-circut the compiler previously. If I add: if (ARRAY_SIZE(default_passwd) =3D=3D 1) return -ENOSYS; in the read_default_passwd() function, that would short-circut the compiler preventing the warning message, and is less hacky. I can resubmit with that change instead. > = > Best regards > Uwe > = > -- = > Pengutronix e.K. | Uwe Kleine-K=F6nig = | > Industrial Linux Solutions | https://www.pengutronix.de/ | Thanks, David Dgien _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox