linux-kernel - Re: seq_putc is preferable to seq

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: <33bbe7e9c0f41ee0ef73ffd5bbca1e9c84c259ec.camel@perches.com>
Date:   Fri, 18 May 2018 10:50:49 -0700
From:   Joe Perches <joe@...ches.com>
To:     Matthew Wilcox <willy@...radead.org>
Cc:     linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: seq_putc is preferable to seq_puts

On Fri, 2018-05-18 at 07:09 -0700, Matthew Wilcox wrote:
> Hi Joe,
> 
> A nice checkpatch addition would be to look for seq_puts(x, "s")
> and recommend the author use seq_putc(x, 's') instead.  The primary
> offender is seq_puts(x, "\n") (a hundred or so, scattered throughout
> the kernel), but seq_puts(x, " ") has its adherents too.

Firstly, has Markus Elfring left the building?

Here's a grep to find the existing uses:

$ git grep -P '\b(?:seq_puts|seq_printf)\s*\(\s*[^,]+,\s*"(?:\\.|.)"\s*\)\s*;' | \
  wc -l
326

and a trivial script to convert them all treewide

$ git grep -P --name-only '\b(?:seq_puts|seq_printf)\s*\(\s*[^,]+,\s*"(?:\\.|.)"\s*\)\s*;' | \
  xargs perl -i -e 'local $/; while (<>) { s/\b(?:seq_puts|seq_printf)\s*\(\s*([^,]+),\s*"(\\.|.)"\s*\)\s*;/seq_putc(\1, '"'\2'"')/g; print;}'

Anyway, I still like a variant of Steven Rostedt's
trace_puts macro silliness hack that avoids all
this by allowing seq_printf without format arguments
to call the appropriate seq_puts/seq_putc invisibly.

https://groups.google.com/forum/#!topic/linux.kernel/YGZhTD-n4UA

It's relatively easy to expand this for seq_putc, but here's
an expanded checkpatch rule anyway that seems to work.
---
diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index baddac9379f0..62b36d7bb3af 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -5791,11 +5791,23 @@ sub process {
 			     "struct spinlock should be spinlock_t\n" . $herecurr);
 		}
 
-# check for seq_printf uses that could be seq_puts
-		if ($sline =~ /\bseq_printf\s*\(.*"\s*\)\s*;\s*$/) {
+# check for seq_printf/seq_puts uses that could be seq_puts or seq_putc
+		if ($sline =~ /\b(seq_(?:printf|puts))\s*\(.*"\s*\)\s*;\s*$/) {
+			my $call = $1;
 			my $fmt = get_quoted_string($line, $rawline);
+			my $fmt_nq = substr($fmt, 1, length($fmt) - 2); #strip outer quotes
 			$fmt =~ s/%%//g;
-			if ($fmt !~ /%/) {
+
+			if (length($fmt_nq) == 1 ||
+			    (length($fmt_nq) == 2 && ($fmt_nq =~ /^\\/ || $fmt_nq =~ /\%\%/))) {
+				my $fmt_sub = $fmt_nq;
+				$fmt_sub =~ s/\%\%/%/;
+				if (WARN("PREFER_SEQ_PUTC",
+					 "Prefer seq_putc to $call\n" . $herecurr) &&
+				    $fix) {
+					$fixed[$fixlinenr] =~ s/\b$call\s*\(\s*($FuncArg)\s*,\s*"\Q$fmt_nq\E"\s*\)/seq_putc($1, '$fmt_sub')/x;
+				}
+			} elsif ($fmt !~ /%/) {
 				if (WARN("PREFER_SEQ_PUTS",
 					 "Prefer seq_puts to seq_printf\n" . $herecurr) &&
 				    $fix) {