linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: David Woodhouse <dwmw2@infradead.org>
To: Mark Fortescue <mark@mtfhpc.demon.co.uk>
Cc: linux-mm@kvack.org, Andrew Morton <akpm@linux-foundation.org>,
	LKML <linux-kernel@vger.kernel.org>,
	sparclinux@vger.kernel.org, David Miller <davem@davemloft.net>,
	Christoph Lameter <clameter@engr.sgi.com>,
	William Lee Irwin III <wli@holomorphy.com>
Subject: Re: [PATCH] Re: Sparc32: random invalid instruction occourances on sparc32 (sun4c)
Date: Tue, 03 Jul 2007 19:36:27 -0400	[thread overview]
Message-ID: <1183505787.29081.62.camel@shinybook.infradead.org> (raw)
In-Reply-To: <Pine.LNX.4.61.0707032317590.30376@mtfhpc.demon.co.uk>

On Tue, 2007-07-03 at 23:47 +0100, Mark Fortescue wrote:
> Hi David,
> 
> I will try out your patch shortly.

Thanks.

> I may be wrong about the size calculations but if you take a look at lines 
> 2174 to 2188 and 2207 to 2203, reading the comments suggest to me that 
> these need to be changed to match the changes to the RedZone words. 
> Failing to change these means that 32bit aligned access of the 64bit 
> RedZone words is still posible and this will kill sun4c.

Why do we need more than the existing:

	if (flags & SLAB_RED_ZONE || flags & SLAB_STORE_USER)
		ralign = __alignof__(unsigned long long);

> For the 64bit RedZone word to be 64bit aligned (required by sun4c), the 
> User word must be 64bit aligned. I don't see where in your patch, this is 
> enforced.

Where __alignof__(long long) > BYTES_PER_WORD my patch should lead to
this layout (32-bit words):

    [ redzone1 bits 63-32 ]
    [ redzone1 bits 31-0  ]
    [    ... object ...   ]
    [    ... object ...   ]
    [ redzone2 bits 63-32 ]
    [ redzone2 bits 31-0  ]
    [        unused       ]
    [      user word      ]

The user word is a 32-bit value; there's no requirement for _it_ to be
aligned.

Hm, actually I think my patch may be incomplete -- I need to adjust the
size of the actual object too. This patch should be better...

diff --git a/mm/slab.c b/mm/slab.c
index a9c4472..8081c07 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -547,7 +547,7 @@ static unsigned long long *dbg_redzone2(struct kmem_cache *cachep, void *objp)
 	if (cachep->flags & SLAB_STORE_USER)
 		return (unsigned long long *)(objp + cachep->buffer_size -
 					      sizeof(unsigned long long) -
-					      BYTES_PER_WORD);
+					      max(BYTES_PER_WORD, __alignof__(unsigned long long)));
 	return (unsigned long long *) (objp + cachep->buffer_size -
 				       sizeof(unsigned long long));
 }
@@ -2223,8 +2223,11 @@ kmem_cache_create (const char *name, size_t size, size_t align,
 	 * overridden by architecture or caller mandated alignment if either
 	 * is greater than BYTES_PER_WORD.
 	 */
-	if (flags & SLAB_RED_ZONE || flags & SLAB_STORE_USER)
+	if (flags & SLAB_RED_ZONE || flags & SLAB_STORE_USER) {
 		ralign = __alignof__(unsigned long long);
+		size += (__alignof__(unsigned long long) - 1);
+		size &= ~(__alignof__(unsigned long long) - 1);
+	}
 
 	/* 2) arch mandated alignment */
 	if (ralign < ARCH_SLAB_MINALIGN) {
@@ -2261,9 +2264,14 @@ kmem_cache_create (const char *name, size_t size, size_t align,
 	}
 	if (flags & SLAB_STORE_USER) {
 		/* user store requires one word storage behind the end of
-		 * the real object.
+		 * the real object. But if the second red zone must be
+		 * aligned 'better' than that, allow for it.
 		 */
-		size += BYTES_PER_WORD;
+		if (flags & SLAB_RED_ZONE
+		    && BYTES_PER_WORD < __alignof__(unsigned long long))
+			size += __alignof__(unsigned long long);
+		else
+			size += BYTES_PER_WORD;
 	}
 #if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
 	if (size >= malloc_sizes[INDEX_L3 + 1].cs_size


-- 
dwmw2

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2007-07-03 23:36 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <468A7D14.1050505@googlemail.com>
2007-07-03 17:29 ` Mark Fortescue
2007-07-03 18:57   ` [PATCH] " Mark Fortescue
2007-07-03 19:26     ` David Woodhouse
2007-07-03 21:25       ` Mark Fortescue
2007-07-03 21:56         ` David Woodhouse
2007-07-03 22:47           ` Mark Fortescue
2007-07-03 23:36             ` David Woodhouse [this message]
2007-07-04  3:27               ` Mark Fortescue
2007-07-04  3:33                 ` David Woodhouse
2007-07-04 10:27                   ` Mark Fortescue
2007-07-04 14:46                     ` David Woodhouse
2007-07-04 18:38                       ` Mark Fortescue
2007-07-03 21:41       ` David Miller, David Woodhouse
2007-07-03 22:01         ` David Woodhouse

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1183505787.29081.62.camel@shinybook.infradead.org \
    --to=dwmw2@infradead.org \
    --cc=akpm@linux-foundation.org \
    --cc=clameter@engr.sgi.com \
    --cc=davem@davemloft.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mark@mtfhpc.demon.co.uk \
    --cc=sparclinux@vger.kernel.org \
    --cc=wli@holomorphy.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox