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;
From:   Jim Cromie <jim.cromie@gmail.com>
To:     daniel.vetter@ffwll.ch, daniel@ffwll.ch, jbaron@akamai.com,
        gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org,
        dri-devel@lists.freedesktop.org, amd-gfx@lists.freedesktop.org,
        intel-gvt-dev@lists.freedesktop.org,
        intel-gfx@lists.freedesktop.org
Cc:     linux@rasmusvillemoes.dk, joe@perches.com, jani.nikula@intel.com,
        ville.syrjala@linux.intel.com, seanpaul@chromium.org,
        robdclark@gmail.com, Jim Cromie <jim.cromie@gmail.com>,
        apw@canonical.com
Subject: [PATCH v5 13/22] checkpatch: special case for file-scoped extern linker-symbol
Date:   Tue,  1 Aug 2023 11:02:46 -0600
Message-ID: <20230801170255.163237-15-jim.cromie@gmail.com>
In-Reply-To: <20230801170255.163237-1-jim.cromie@gmail.com>
References: <20230801170255.163237-1-jim.cromie@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

"externs should be avoided in .c files" needs an exception for linker
symbols, like those that mark the start, stop of many kernel sections.

Since code already checks REALNAME to avoid linker-scripts entirely,
add a new else-if block to look at them instead.  As a simple
heuristic, treat all words (in the patch-line) as possible symbols,
and save them to screen the WARN quoted above.

For my test case, this included BOUNDED_BY (a macro), which is extra,
but not troublesome - any extra words collected would have to also be
symbols the script would otherwise complain about.

Where the WARN is issued, precede it with an else-if block to catch
one common extern-in-c use case: "extern struct foo bar[]".  Here we
can at least issue a softer warning, after checking for a match with a
maybe-linker-symbol parsed earlier from the patch.  Though heuristic,
it worked for my test-case, allowing both (start|stop)$symbol's,
matched by $symbol, it wasn't thrown by noise (BOUNDED_BY) in the
maybe-linker-symbols.

NB: git diff ordering dependence on vmlinux.lds.h before c files ?

Cc: apw@canonical.com
Cc: joe@perches.com
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
---
 scripts/checkpatch.pl | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index b30114d637c4..f6b6b2a50dfe 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -74,6 +74,8 @@ my $git_command ='export LANGUAGE=en_US.UTF-8; git';
 my $tabsize = 8;
 my ${CONFIG_} = "CONFIG_";
 
+my %maybe_linker_symbol; # for externs in c exceptions, when seen in *vmlinux.lds.h
+
 sub help {
 	my ($exitcode) = @_;
 
@@ -6051,6 +6053,9 @@ sub process {
 
 # check for line continuations outside of #defines, preprocessor #, and asm
 
+		} elsif ($realfile =~ m@/vmlinux.lds.h$@) {
+		    $line =~ s/(\w+)/$maybe_linker_symbol{$1}++/ge;
+		    #print "REAL: $realfile\nln: $line\nkeys:", sort keys %maybe_linker_symbol;
 		} else {
 			if ($prevline !~ /^..*\\$/ &&
 			    $line !~ /^\+\s*\#.*\\$/ &&		# preprocessor
@@ -7107,6 +7112,21 @@ sub process {
 				     "arguments for function declarations should follow identifier\n" . $herecurr);
 			}
 
+		} elsif ($realfile =~ /\.c$/ && defined $stat &&
+		    $stat =~ /^\+extern struct\s+(\w+)\s+(\w+)\[\];/)
+		{
+			my ($st_type, $st_name) = ($1, $2);
+
+			for my $s (keys %maybe_linker_symbol) {
+			    #print "Linker symbol? $st_name : $s\n";
+			    goto LIKELY_LINKER_SYMBOL
+				if $st_name =~ /$s/;
+			}
+			WARN("AVOID_EXTERNS",
+			     "found a file-scoped extern type:$st_type name:$st_name in .c file\n"
+			     . "is this a linker symbol ?\n" . $herecurr);
+		  LIKELY_LINKER_SYMBOL:
+
 		} elsif ($realfile =~ /\.c$/ && defined $stat &&
 		    $stat =~ /^.\s*extern\s+/)
 		{
-- 
2.41.0