xfs
[Top] [All Lists]

Re: XFS unhappy with large holey loopback and syncs -- [OT] vmalloc 45-b

To: "Bryan J. Smith" <b.j.smith@xxxxxxxx>
Subject: Re: XFS unhappy with large holey loopback and syncs -- [OT] vmalloc 45-bit/32PiB address limit
From: Andi Kleen <ak@xxxxxxx>
Date: Tue, 29 Nov 2005 16:41:39 +0100
Cc: Steve Lord <lord@xxxxxxx>, Andi Kleen <ak@xxxxxxx>, David Chinner <dgc@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <20051129153133.42577.qmail@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <20051129152148.GH19515@xxxxxxxxxxxxx> <20051129153133.42577.qmail@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Tue, Nov 29, 2005 at 07:31:33AM -0800, Bryan J. Smith wrote:
> Andi Kleen <ak@xxxxxxx> wrote:
> > For that it should use vmalloc - also no problem because it
> > is backed in distributed order 0 4k pages. On 32bit the
> vmalloc
> > space is a bit limited, but 8MB is still only a small part
> of
> > it On x86-64 it's VmallocTotal: 34359738367 kB
> 
> If I'm reading that correctly, that's 45-bit/32PiB
> addressing.  Where does that limit come from?

x86-64 supports 48bits of address space. Half of it is for 
user space, the other half for the kernel.  Of the kernel
half half is for physical memory, and half of the other half 
is for kernel text and holes. This leaves 45bits for vmalloc.

-Andi


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