linux-kernel - Re: [PATCH] scsi: aic94xx: Fix fall-through warning for Clang

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] [day] [month] [year] [list]

Message-ID: <6ddc71efad6cd45e45991028a7656eab203df12d.camel@perches.com>
Date:   Thu, 01 Jul 2021 09:40:56 -0700
From:   Joe Perches <joe@...ches.com>
To:     "Gustavo A. R. Silva" <gustavoars@...nel.org>,
        "James E.J. Bottomley" <jejb@...ux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@...cle.com>
Cc:     linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-hardening@...r.kernel.org, Kees Cook <keescook@...omium.org>
Subject: Re: [PATCH] scsi: aic94xx: Fix fall-through warning for Clang

On Wed, 2021-06-30 at 20:04 -0500, Gustavo A. R. Silva wrote:
> In preparation to enable -Wimplicit-fallthrough for Clang, fix a
> warning by explicitly adding a fallthrough; statement.
> 
> Notice that this seems to be a Duff device for performance[1]. So,
> although the code looks a bit _funny_, I didn't want to refactor
> or modify it beyond merely adding a fallthrough marking, which
> might be the least disruptive way to fix this issue.

If you read the surrounding calling code, it appears it's
not performant.

 * @start has to be the _base_ element start, since the
 * linked list entries's offset is from this pointer.
 * Some linked list entries use only the first id, in which case
 * you can pass 0xFF for the second.

> diff --git a/drivers/scsi/aic94xx/aic94xx_sds.c b/drivers/scsi/aic94xx/aic94xx_sds.c
[]
> @@ -718,10 +718,12 @@ static void *asd_find_ll_by_id(void * const start, const u8 id0, const u8 id1)
>  	do {
>  		switch (id1) {
>  		default:
> -			if (el->id1 == id1)
> +			if (el->id1 == id1) {
> +			fallthrough;
>  		case 0xFF:
>  				if (el->id0 == id0)
>  					return el;
> +			}
>  		}
>  		el = start + le16_to_cpu(el->next);
>  	} while (el != start);

And this is still horrible code to read.

I think the below is rather more sensible and it doesn't change
the defconfig with aic94xx object size.

$ size drivers/scsi/aic94xx/aic94xx_sds.o*
   text	   data	    bss	    dec	    hex	filename
  10010	    192	      0	  10202	   27da	drivers/scsi/aic94xx/aic94xx_sds.o.new
  10010	    192	      0	  10202	   27da	drivers/scsi/aic94xx/aic94xx_sds.o.old
---
 drivers/scsi/aic94xx/aic94xx_sds.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/drivers/scsi/aic94xx/aic94xx_sds.c b/drivers/scsi/aic94xx/aic94xx_sds.c
index 297a66770260c..069278f6d6cea 100644
--- a/drivers/scsi/aic94xx/aic94xx_sds.c
+++ b/drivers/scsi/aic94xx/aic94xx_sds.c
@@ -717,11 +717,14 @@ static void *asd_find_ll_by_id(void * const start, const u8 id0, const u8 id1)
 
 	do {
 		switch (id1) {
-		default:
-			if (el->id1 == id1)
 		case 0xFF:
-				if (el->id0 == id0)
-					return el;
+			if (el->id0 == id0)
+				return el;
+			break;
+		default:
+			if (el->id0 == id0 && el->id1 == id1)
+				return el;
+			break;
 		}
 		el = start + le16_to_cpu(el->next);
 	} while (el != start);