linux-kernel - Re: [PATCH] coccinelle: semantic patch to check for inappropriate do

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: <alpine.DEB.2.21.2001041454020.6944@hadrien>
Date:   Sat, 4 Jan 2020 14:54:32 +0100 (CET)
From:   Julia Lawall <julia.lawall@...ia.fr>
To:     Wen Yang <wenyang@...ux.alibaba.com>
cc:     Gilles Muller <Gilles.Muller@...6.fr>,
        Nicolas Palix <nicolas.palix@...g.fr>,
        Michal Marek <michal.lkml@...kovi.net>,
        Matthias Maennich <maennich@...gle.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Thomas Gleixner <tglx@...utronix.de>, cocci@...teme.lip6.fr,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] coccinelle: semantic patch to check for inappropriate
 do_div() calls



--- Please note the new email address ---


On Sat, 4 Jan 2020, Wen Yang wrote:

>
>
> On 2020/1/4 4:55 下午, Julia Lawall wrote:
> > On Sat, 4 Jan 2020, Wen Yang wrote:
> >
> > >
> > >
> > > On 2020/1/4 3:00 下午, Julia Lawall wrote:
> > > > On Sat, 4 Jan 2020, Wen Yang wrote:
> > > >
> > > > > do_div() does a 64-by-32 division.
> > > > > When the divisor is unsigned long, u64, or s64,
> > > > > do_div() truncates it to 32 bits, this means it
> > > > > can test non-zero and be truncated to zero for division.
> > > > > This semantic patch is inspired by Mateusz Guzik's patch:
> > > > > commit b0ab99e7736a ("sched: Fix possible divide by zero in avg_atom()
> > > > > calculation")
> > > > >
> > > > > Signed-off-by: Wen Yang <wenyang@...ux.alibaba.com>
> > > > > Cc: Julia Lawall <Julia.Lawall@...6.fr>
> > > > > Cc: Gilles Muller <Gilles.Muller@...6.fr>
> > > > > Cc: Nicolas Palix <nicolas.palix@...g.fr>
> > > > > Cc: Michal Marek <michal.lkml@...kovi.net>
> > > > > Cc: Matthias Maennich <maennich@...gle.com>
> > > > > Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
> > > > > Cc: Masahiro Yamada <yamada.masahiro@...ionext.com>
> > > > > Cc: Thomas Gleixner <tglx@...utronix.de>
> > > > > Cc: cocci@...teme.lip6.fr
> > > > > Cc: linux-kernel@...r.kernel.org
> > > > > ---
> > > > >    scripts/coccinelle/misc/do_div.cocci | 66
> > > > > ++++++++++++++++++++++++++++++++++++
> > > > >    1 file changed, 66 insertions(+)
> > > > >    create mode 100644 scripts/coccinelle/misc/do_div.cocci
> > > > >
> > > > > diff --git a/scripts/coccinelle/misc/do_div.cocci
> > > > > b/scripts/coccinelle/misc/do_div.cocci
> > > > > new file mode 100644
> > > > > index 0000000..f1b72d1
> > > > > --- /dev/null
> > > > > +++ b/scripts/coccinelle/misc/do_div.cocci
> > > > > @@ -0,0 +1,66 @@
> > > > > +// SPDX-License-Identifier: GPL-2.0-only
> > > > > +/// do_div() does a 64-by-32 division.
> > > > > +/// When the divisor is unsigned long, u64, or s64,
> > > > > +/// do_div() truncates it to 32 bits, this means it
> > > > > +/// can test non-zero and be truncated to zero for division.
> > > > > +///
> > > > > +//# This makes an effort to find those inappropriate do_div () calls.
> > > > > +//
> > > > > +// Confidence: Moderate
> > > > > +// Copyright: (C) 2020 Wen Yang, Alibaba.
> > > > > +// Comments:
> > > > > +// Options: --no-includes --include-headers
> > > > > +
> > > > > +virtual context
> > > > > +virtual org
> > > > > +virtual report
> > > > > +
> > > > > +@...ends on context@
> > > > > +expression f;
> > > > > +long l;
> > > > > +unsigned long ul;
> > > > > +u64 ul64;
> > > > > +s64 sl64;
> > > > > +
> > > > > +@@
> > > > > +(
> > > > > +* do_div(f, l);
> > > > > +|
> > > > > +* do_div(f, ul);
> > > > > +|
> > > > > +* do_div(f, ul64);
> > > > > +|
> > > > > +* do_div(f, sl64);
> > > > > +)
> > > > > +
> > > > > +@r depends on (org || report)@
> > > > > +expression f;
> > > > > +long l;
> > > > > +unsigned long ul;
> > > > > +position p;
> > > > > +u64 ul64;
> > > > > +s64 sl64;
> > > > > +@@
> > > > > +(
> > > > > +do_div@p(f, l);
> > > > > +|
> > > > > +do_div@p(f, ul);
> > > > > +|
> > > > > +do_div@p(f, ul64);
> > > > > +|
> > > > > +do_div@p(f, sl64);
> > > > > +)
> > > > > +
> > > > > +@...ipt:python depends on org@
> > > > > +p << r.p;
> > > > > +@@
> > > > > +
> > > > > +msg="WARNING: WARNING: do_div() does a 64-by-32 division, which may
> > > > > truncation the divisor to 32-bit"
> > > > > +coccilib.org.print_todo(p[0], msg)
> > > > > +
> > > > > +@...ipt:python depends on report@
> > > > > +p << r.p;
> > > > > +@@
> > > > > +
> > > > > +msg="WARNING: WARNING: do_div() does a 64-by-32 division, which may
> > > > > truncation the divisor to 32-bit"
> > > > > +coccilib.report.print_report(p[0], msg)
> > > >
> > > > A few small issues: You have WARNING: twice in each case, and truncation
> > > > should be truncate.
> > > >
> > >
> > > Thanks for your comments, we will fix it soon.
> > >
> > > > Is there any generic strategy for fixing these issues?
> > > >
> > >
> > > We have done some experiments, such as:
> > > https://lkml.org/lkml/2020/1/2/1354
> >
> > Thanks.  Actually, I would appreciate knowing about such experiments when
> > the semantic patch is submitted, since eg in this case I am not really an
> > expert in this issue.
> >
> > >
> > > -	avg = rec->time;
> > > -	do_div(avg, rec->counter);
> > > +	avg = div64_ul(rec->time, rec->counter);
> > >
> > > --> Function replacement was performed here,
> > >      and simple code cleanup was also performed.
> > >
> > >
> > > -		do_div(stddev, rec->counter * (rec->counter - 1) * 1000);
> > > +		stddev = div64_ul(stddev,
> > > +				  rec->counter * (rec->counter - 1) * 1000);
> > >
> > > --> Only the function replacement is performed here (because the variable
> > > ‘stddev’ corresponds to a more complicated equation, cleaning it will
> > > reduce
> > > readability).
> >
> > Would it be reasonable to extend the warning to say "consider using
> > div64_ul instead"?  Or do you think it is obvious to everyone?
> >
>
> Thank you for your comments.
> We plan to modify it as follows:
> msg="WARNING: do_div() does a 64-by-32 division, please consider using
> div64_ul, div64_long, div64_u64 or div64_s64 instead."

This message seems helpful, thanks.

julia

>
> > > In addition, there are some codes that do not need to be modified:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/net/can/dev.c#n263
> >
> > Would it be worth having a special case for constants and checking whether
> > the value is obviously safe and no warning is needed?
> >
> Thanks.
> This is very valuable in reducing false positives, and we'll try to implement
> it.
>
> --
> Best Wishes,
> Wen
>
> > > So we just print a warning.
> > > As for how to fix it, we need to analyze the code carefully.
> > >
> > > --
> > > Best Wishes,
> > > Wen
> > >
> > >
> > >
> >
>