linux-kernel - Re: [PATCH RFC] checkpatch: fix multi-statement macro checks

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: <dc985938aa3986702815a0bd68dfca8a03c85447.camel@perches.com>
Date:   Thu, 01 Oct 2020 08:26:26 -0700
From:   Joe Perches <joe@...ches.com>
To:     Dwaipayan Ray <dwaipayanray1@...il.com>
Cc:     linux-kernel@...r.kernel.org,
        linux-kernel-mentees@...ts.linuxfoundation.org,
        Lukas Bulwahn <lukas.bulwahn@...il.com>
Subject: Re: [PATCH RFC] checkpatch: fix multi-statement macro checks

On Thu, 2020-10-01 at 07:38 -0700, Joe Perches wrote:
> On Thu, 2020-10-01 at 19:44 +0530, Dwaipayan Ray wrote:
> > On Thu, Oct 1, 2020 at 7:12 PM Joe Perches <joe@...ches.com> wrote:
> > > On Thu, 2020-10-01 at 18:57 +0530, Dwaipayan Ray wrote:
> > > > On Thu, Oct 1, 2020 at 6:47 PM Joe Perches <joe@...ches.com> wrote:
> > > > > On Thu, 2020-10-01 at 16:03 +0530, Dwaipayan Ray wrote:
> > > > > > Checkpatch.pl doesn't have a check for excluding while (...) {...}
> > > > > > blocks from MULTISTATEMENT_MACRO_USE_DO_WHILE error.
> > > > > > 
> > > > > > For example, running checkpatch.pl on the file mm/access.c in the
> > > > > > kernel generates the following error:
> > > > > > 
> > > > > > ERROR: Macros with complex values should be enclosed in parentheses
> > > > > > +#define copy_from_kernel_nofault_loop(dst, src, len, type, err_label)        \
> > > > > > +     while (len >= sizeof(type)) {                                   \
> > > > > > +             __get_kernel_nofault(dst, src, type, err_label);        \
> > > > > > +             dst += sizeof(type);                                    \
> > > > > > +             src += sizeof(type);                                    \
> > > > > > +             len -= sizeof(type);                                    \
> > > > > > +     }
> > > > > > 
> > > > > > The error is misleading for this case. Enclosing it in parantheses
> > > > > > doesn't make any sense.
> > > > > 
> > > > > OK
> > > > > 
> > > > > > Checkpatch already has an exception list for such common macro types.
> > > > > > Added a new exception for while (...) {...} style blocks to the same.
> > > > > > This effectively fixed the wrong error message.
> > > > > []
> > > > > > diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> > > > > []
> > > > > > @@ -5342,6 +5342,7 @@ sub process {
> > > > > >                           $dstat !~ /^\.$Ident\s*=/ &&                                # .foo =
> > > > > >                           $dstat !~ /^(?:\#\s*$Ident|\#\s*$Constant)\s*$/ &&          # stringification #foo
> > > > > >                           $dstat !~ /^do\s*$Constant\s*while\s*$Constant;?$/ &&       # do {...} while (...); // do {...} while (...)
> > > > > > +                         $dstat !~ /^while\s*$Constant\s*$Constant\s*$/ &&           # while (...) {...}
> > > 
> > > Note the \s*
> > >                                                               ^
> > > 
> > > > > Did you try to output $dstat for some matching cases?
> > > > > What was the $dstat value for the cases you tried?
> > > > > 
> > > > > 
> > > > Hi,
> > > > I did check $dstat values.
> > > > 
> > > > For example on file mm/maccess.c, there were two such macros:
> > > > 
> > > > Case 1:
> > > > 
> > > > $ctx:
> > > > +#define copy_from_kernel_nofault_loop(dst, src, len, type, err_label) \
> > > > + while (len >= sizeof(type)) { \
> > > > + __get_kernel_nofault(dst, src, type, err_label); \
> > > > + dst += sizeof(type); \
> > > > + src += sizeof(type); \
> > > > + len -= sizeof(type); \
> > > > + }
> > > > 
> > > > $dstat:
> > > > while 1 1
> > > 
> > > And perhaps this test should use \s+ instead.
> > > What is $dstat with a #define like:
> > > 
> > > #define foo(bar,baz)while(bar){bar--;baz++;}
> > > 
> > > (no spaces anywhere bot the required one after define
> > > 
> > 
> > In this case, $dstat is: while11
> > 
> > So, if \s+ is used, it won't match with this. I ran checkpatch
> > on it and some other condition seems to match, so it is
> > excluded from the error.
> > 
> > However, if the macro is like:
> > 
> > #define foo(bar,baz)while(bar) {bar--;baz++;}
> > (one space after condition)
> > 
> > $dstat is: while1 1
> > (space after first 1)
> > and the same error is again emitted.
> > 
> > So I think \s* works better since there can be
> > 0 or more whitespaces between them.
> 
> All I'm trying to point out to you is that $Constant\s*$Constant
> isn't a proper test as the first $Constant will pull the test
> entire sequence of digits and the second $Constant will not be
> met.
> 
> It may take some conversion of the collapsing of the dstat
> block to work appropriately
> 
> 
> 			# Flatten any parentheses and braces
> 			while ($dstat =~ s/\([^\(\)]*\)/1/ ||
> 			       $dstat =~ s/\{[^\{\}]*\}/1/ ||
> 			       $dstat =~ s/.\[[^\[\]]*\]/1/)
> 			{
> 			}
> 
> Maybe the /1/ should be / 1 / but I didn't look to see what
> happens to the exclusion tests below that.

I think your patch would work well enough if the /1/ bits
here were simply changed to /1u/.

1 is a $Constant as it's just a number.
11 though is also a $Constant.
1u is also a $Constant but it stops the acquisition of
digits that 11 would not and the sequence of
"while1u1u" should match your newly introduced test
of $Constant\s*$Constant as "while11" would not match.







>