[Top] [All Lists]

Re: [PATCH v2] MIPS: Alchemy: fix dbdma ring destruction memory debugche

To: Manuel Lauss <>
Subject: Re: [PATCH v2] MIPS: Alchemy: fix dbdma ring destruction memory debugcheck.
From: Ralf Baechle <>
Date: Thu, 28 Jan 2010 12:22:52 +0100
Cc: Linux-MIPS <>, Manuel Lauss <>
In-reply-to: <>
References: <> <> <>
User-agent: Mutt/1.5.20 (2009-08-17)
On Thu, Jan 28, 2010 at 09:23:56AM +0100, Manuel Lauss wrote:

> >> This fix is only necessary with the SLAB allocator and CONFIG_DEBUG_SLAB
> >> enabled;  non-debug SLAB, SLUB do return nicely aligned addresses,
> >> debug-enabled SLUB currently panics early in the boot process.
> >
> > Queued for 2.6.34 - should this also go into 2.6.33?
> 2.6.33 and .32 at least.  That code has been in there since the dawn of 2.6;
> the fact that nobody has tripped over this so far means that noone build 
> kernels
> with slab debugging (I tripped over this while looking for something else) or
> the slab allocator behaviour changed recently.

The behaviour of the slab has been like this as long as I can remember.
Cherrypicking into the -stable branches as I'm writing this.

> > Have you considered increasing the value ARCH_KMALLOC_MINALIGN which
> > defaults to 8?  Or your own slab cache of suitable alignment?  The latter
> > is more something for frequent allocations.
> I have to admit I know nothing about Linux' memory management.  Are slab
> caches not affected by the what I presume are "guard bytes" inserted into
> the memory areas when slab debug is on?

When creating a slab cache an alignment for objects allocated from this
cache can be specified.


<Prev in Thread] Current Thread [Next in Thread>