linux-kernel - Re: [PATCH] checkpatch: check for nested (un)?likely calls

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: <c93541d7-9e0f-75e6-56c4-597ba440fb88@linux.com>
Date:   Tue, 27 Aug 2019 20:32:43 +0300
From:   Denis Efremov <efremov@...ux.com>
To:     Joe Perches <joe@...ches.com>, Andy Whitcroft <apw@...onical.com>
Cc:     linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH] checkpatch: check for nested (un)?likely calls

On 8/27/19 8:21 PM, Joe Perches wrote:
> On Tue, 2019-08-27 at 19:55 +0300, Denis Efremov wrote:
>> IS_ERR, IS_ERR_OR_NULL, IS_ERR_VALUE already contain unlikely optimization
>> internally. Thus, there is no point in calling these functions under
>> likely/unlikely.
>>
>> This check is based on the coccinelle rule developed by Enrico Weigelt
>> https://lore.kernel.org/lkml/1559767582-11081-1-git-send-email-info@metux.net/
> []
>> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> []
>> @@ -6480,6 +6480,13 @@ sub process {
>>  			     "Using $1 should generally have parentheses around the comparison\n" . $herecurr);
>>  		}
>>  
>> +# nested likely/unlikely calls
>> +		if ($perl_version_ok &&
>> +		    $line =~ /\b(?:(?:un)?likely)\s*\(!?\s*(IS_ERR(?:_OR_NULL|_VALUE)?)\s*${balanced_parens}\s*\)/) {
>> +			WARN("LIKELY_MISUSE",
>> +			     "nested (un)?likely calls, unlikely already used in $1 internally\n" . $herecurr);
>> +		}
>> +
> 
> Couple things:
> 
> 1:
> 
> Are you going to submit patches for the just 10 instances that exist?
> 
> $ git grep -P -n '\b(?:(?:un)?likely)\s*\(!?\s*(IS_ERR(?:_OR_NULL|_VALUE)?)\s*\([^\)]+\)\s*\)'
> arch/x86/kernel/irq.c:246:      if (likely(!IS_ERR_OR_NULL(desc))) {
> drivers/infiniband/hw/hfi1/verbs.c:1044:        if (unlikely(IS_ERR_OR_NULL(pbuf))) {
> drivers/input/mouse/alps.c:1479:        } else if (unlikely(IS_ERR_OR_NULL(priv->dev3))) {
> fs/ntfs/mft.c:74:       if (likely(!IS_ERR(page))) {
> fs/ntfs/mft.c:157:      if (likely(!IS_ERR(m)))
> fs/ntfs/mft.c:274:              if (likely(!IS_ERR(m))) {
> fs/ntfs/mft.c:1779:             if (likely(!IS_ERR(rl2)))
> fs/ntfs/namei.c:118:            if (likely(!IS_ERR(dent_inode))) {
> fs/ntfs/runlist.c:954:  if (likely(!IS_ERR(old_rl)))
> include/net/udp.h:483:  if (unlikely(IS_ERR_OR_NULL(segs))) {

Yes, I will do it in days.

> 
> 2:
> 
> This will not warn about code like:
> 
> fs/ntfs/mft.c:  if (unlikely(IS_ERR(rl) || !rl->length || rl->lcn < 0)) {
> 
> that could probably be better written as:
> 
> 		if (IS_ERR(rl) || unlikely(!rl->length || rl->lcn < 0)) {
> 
> 

Ok, I will change the regex in v2 to:
$line =~ /\b(?:(?:un)?likely)\s*\(!?\s*(IS_ERR(?:_OR_NULL|_VALUE)?)/

Should I skip $perl_version_ok check then?

If there are no other suggestions about, for example, warn message or commit
description I will send v2.

Thanks,
Denis