Re: [Linux-ia64] Re: [PATCH] head.S fix for unusual load addrs

From: Jesse Barnes <jbarnes_at_sgi.com>
Date: 2003-05-09 03:20:05
On Thu, May 08, 2003 at 10:07:49AM -0700, David Mosberger wrote:
> Ah, yes, I had forgotten about -q.  It probably would be sufficient.
> And the backwards-compatibility it would achieve would definitely be
> a plus.

I thought so too.

> A new concern though: I don't think text-replication will work well
> with this scheme.  If each replica is linked for a different address,
> function pointers become a performance problem: whenever you call
> through a function pointer, you'll end up executing on whatever
> replica initialized the function pointer.  Not good.

Right, I thought about that.  I don't think we need text replication
yet though anyway (at least our platform doesn't).

> Unless I'm missing something, kernel relocation is therefore a bit of
> a stillborn idea (unless text-replication really isn't all that
> important because we all have humongous caches between nodes...).
> 
> If we do have to go the virtual remapping route, my preference would
> be to stick the kernel somewhere in region 5 (0xa..).  Has anyone
> tried that?  It should work fine in principle (modules already live in
> that space).

I guess we can revisit that if we need/want to implement text
replication at some point in the future.  It might not be that bad
though--I don't envision relocating each replica, but rather just
seperating the kernel text and data with the right translation
registers that point to local text and global data, but I haven't
thought about it much (this is what Tony's original patch did, and it
seemed to work well).

Jesse
Received on Thu May 08 10:20:11 2003

This archive was generated by hypermail 2.1.8 : 2005-08-02 09:20:14 EST