From: Jarod Wilson <jwilson@redhat.com>
Date: Tue, 8 Jan 2008 12:27:09 -0500
Subject: [xen] ia64: access extended I/O spaces from dom0
Message-id: 4783B26D.8030005@redhat.com
O-Subject: [RHEL5.2 PATCH][XEN] ia64: access extended I/O spaces from dom0
Bugzilla: 249629

Jarod Wilson wrote:
> Don Zickus wrote:
>> On Thu, Aug 09, 2007 at 04:54:30PM -0400, Aron Griffis wrote:
>>> Bugzilla Bug 249629: dom0 can't access extended I/O spaces
>>> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=249629
>>>
>>> Summary
>>> -------
>>> These two patches were added to xen-unstable after 3.1 was released.
>>> Without these, extended I/O port spaces will not be registered with
>>> Xen. It they're not registered, then Xen is not able to grant dom0
>>> access to poke at them.  If dom0 can't poke at them, then I/O port
>>> spaces effectively don't work.
>>>
>>> See the first changeset below for a longer description.
>>>
>>> http://xenbits.xensource.com/xen-unstable.hg?rev/601509daabfc
>>> http://xenbits.xensource.com/linux-2.6.18-xen.hg?rev/a62cfa35d3b5
>>>
>>> Testing
>>> -------
>>> 2.6.18-36.el5xen boots with these patches on rx6600.  I haven't yet
>>> understood how to demonstrate it's actually accomplishing the goal, so
>>> I'm still working on that.  I'll follow up with that info ASAP but
>>> wanted to post this for review immediately considering where we are in
>>> the 5.1 schedule.
>>
>> Is this patch still needed?  If so, can it be reposted into chunks, one
>> for the linux kernel, the other for the hypervisor.
>
> I believe the hypervisor chunk is included via the xen 3.1.2 rebase, but
> we do still need the kernel portion for this.

This is a repost on Aron's behalf of the kernel-only portions of this
patch, as the hypervisor parts are unnecessary with the xen 3.1.2
rebase. I don't have access to any hardware with extended I/O spaces to
verify full functionality of this patch, but I've been running my
private branch ia64 xen kernels carrying this patch for several months
now, never had an issue with it, and its been in upstream xen for quite
some time now.

Please ACK

--
Jarod Wilson
jwilson@redhat.com

http://xenbits.xensource.com/linux-2.6.18-xen.hg?rev/a62cfa35d3b5
ported to kernel-2.6.18-36.el5

Acked-by: Bill Burns <bburns@redhat.com>
Acked-by: "Stephen C. Tweedie" <sct@redhat.com>
Acked-by: Prarit Bhargava <prarit@redhat.com>

diff --git a/arch/ia64/pci/pci.c b/arch/ia64/pci/pci.c
index bdb3bfa..d65c412 100644
--- a/arch/ia64/pci/pci.c
+++ b/arch/ia64/pci/pci.c
@@ -165,6 +165,11 @@ new_space (u64 phys_base, int sparse)
 	io_space[i].mmio_base = mmio_base;
 	io_space[i].sparse = sparse;
 
+#ifdef CONFIG_XEN
+	if (is_initial_xendomain())
+		HYPERVISOR_add_io_space(phys_base, sparse, i);
+#endif
+
 	return i;
 }
 
diff --git a/include/asm-ia64/hypercall.h b/include/asm-ia64/hypercall.h
index 77716a7..2c4ef37 100644
--- a/include/asm-ia64/hypercall.h
+++ b/include/asm-ia64/hypercall.h
@@ -364,6 +364,15 @@ HYPERVISOR_expose_p2m(unsigned long conv_start_gpfn,
 }
 #endif
 
+static inline int
+HYPERVISOR_add_io_space(unsigned long phys_base,
+			unsigned long sparse,
+			unsigned long space_number)
+{
+	return _hypercall4(int, ia64_dom0vp_op, IA64_DOM0VP_add_io_space,
+			   phys_base, sparse, space_number);
+}
+
 // for balloon driver
 #define HYPERVISOR_update_va_mapping(va, new_val, flags) (0)
 
diff --git a/include/xen/interface/arch-ia64.h b/include/xen/interface/arch-ia64.h
index fdb45a7..a15ba1f 100644
--- a/include/xen/interface/arch-ia64.h
+++ b/include/xen/interface/arch-ia64.h
@@ -358,6 +358,9 @@ DEFINE_XEN_GUEST_HANDLE(vcpu_guest_context_t);
 /* gmfn version of IA64_DOM0VP_add_physmap */
 #define IA64_DOM0VP_add_physmap_with_gmfn       9
 
+/* Add an I/O port space range */
+#define IA64_DOM0VP_add_io_space        11
+
 // flags for page assignement to pseudo physical address space
 #define _ASSIGN_readonly                0
 #define ASSIGN_readonly                 (1UL << _ASSIGN_readonly)