From mboxrd@z Thu Jan  1 00:00:00 1970
From: Bob Picco <bpicco@meloft.net>
Date: Sun, 05 Oct 2014 13:51:30 +0000
Subject: Re: [PATCH v2 0/8] sparc64: MM/IRQ patch queue.
Message-Id: <20141005135130.GF22365@zareason>
List-Id: <sparclinux.vger.kernel.org>
References: <20140927.142812.2031647355756795530.davem@davemloft.net>
In-Reply-To: <20140927.142812.2031647355756795530.davem@davemloft.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: sparclinux@vger.kernel.org

David Miller wrote:	[Sat Oct 04 2014, 03:12:37PM EDT]
> From: Bob Picco <bpicco@meloft.net>
> Date: Thu, 2 Oct 2014 10:24:35 -0400
> 
> > At this commit:
> > 59a35b1 sparc64: Use kernel page tables for vmemmap.
> 
> I think I figured out what the bug is.
> 
> It has to do with how KERN_PGTABLE_WALK masks in the sub-pagesize
> bits for huge pages.
> 
> That assembler works fine for PAGE_OFFSET mappings where the
> sub-pagesize bits always have a direct correspondance to the
> physical bits.  But for something like vmemmap that doesn't
> work.
> 
> I'll verify this and work on a fix.
This is exactly where I was on Friday afternoon. Unfortunately I was too
exhausted to contribute.