Re: [RFC PATCH] xen/x86: Increase xen_e820_map to E820_X_MAX possible entries

From: Juergen Gross <jgross@suse.com>
To: Alex Thorlton <athorlton@sgi.com>
Cc: Jan Beulich <JBeulich@suse.com>,
	David Vrabel <david.vrabel@citrix.com>,
	x86@kernel.org, Thomas Gleixner <tglx@linutronix.de>,
	xen-devel@lists.xenproject.org, Ingo Molnar <mingo@redhat.com>,
	Russ Anderson <rja@sgi.com>,
	linux-kernel@vger.kernel.org, "H. Peter Anvin" <hpa@zytor.com>
Subject: Re: [RFC PATCH] xen/x86: Increase xen_e820_map to E820_X_MAX possible entries
Date: Wed, 16 Nov 2016 07:06:31 +0100	[thread overview]
Message-ID: <633e8146-d181-a599-498b-abae4af11687@suse.com> (raw)
In-Reply-To: <20161115152209.GA67776@stormcage.americas.sgi.com>

On 15/11/16 16:22, Alex Thorlton wrote:
> On Tue, Nov 15, 2016 at 10:55:49AM +0100, Juergen Gross wrote:
>> I'd go with the new error code. What about E2BIG or ENOSPC?
>>
>> I think the hypervisor should fill in the number of entries required
>> in this case.
>>
>> In case nobody objects I can post patches for this purpose (both Xen
>> and Linux).
> 
> This sounds like a good solution to me.  I think it's definitely more
> appropriate than simply bumping up the size of xen_e820_map, especially
> considering the fact that it's theoretically possible for the e820 map
> generated by the hypercall to grow too large, even on a non-EFI machine,
> where my change would have no effect.

Well, it won't help with the current hypervisor, so bumping up the size
of xen_e820_map will still be a good idea. I think using E820_X_MAX is
okay since in the end xen_e820_map will be transferred into a struct
e820map which can't hold more than E820_X_MAX entries (additional
entries are ignored here, so this won't let the boot fail).

Juergen