linux/drivers/pci/dmar.c

1382 lines
33 KiB
C
Raw Normal View History

Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
/*
* Copyright (c) 2006, Intel Corporation.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program; if not, write to the Free Software Foundation, Inc., 59 Temple
* Place - Suite 330, Boston, MA 02111-1307 USA.
*
* Copyright (C) 2006-2008 Intel Corporation
* Author: Ashok Raj <ashok.raj@intel.com>
* Author: Shaohua Li <shaohua.li@intel.com>
* Author: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
*
* This file implements early detection/parsing of Remapping Devices
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
* reported to OS through BIOS via DMA remapping reporting (DMAR) ACPI
* tables.
*
* These routines are used by both DMA-remapping and Interrupt-remapping
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
*/
#include <linux/pci.h>
#include <linux/dmar.h>
#include <linux/iova.h>
#include <linux/intel-iommu.h>
#include <linux/timer.h>
#include <linux/irq.h>
#include <linux/interrupt.h>
#include <linux/tboot.h>
#include <linux/dmi.h>
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
#define PREFIX "DMAR: "
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
/* No locks are needed as DMA remapping hardware unit
* list is constructed at boot time and hotplug of
* these units are not supported by the architecture.
*/
LIST_HEAD(dmar_drhd_units);
static struct acpi_table_header * __initdata dmar_tbl;
static acpi_size dmar_tbl_size;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
static void __init dmar_register_drhd_unit(struct dmar_drhd_unit *drhd)
{
/*
* add INCLUDE_ALL at the tail, so scan the list will find it at
* the very end.
*/
if (drhd->include_all)
list_add_tail(&drhd->list, &dmar_drhd_units);
else
list_add(&drhd->list, &dmar_drhd_units);
}
static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
struct pci_dev **dev, u16 segment)
{
struct pci_bus *bus;
struct pci_dev *pdev = NULL;
struct acpi_dmar_pci_path *path;
int count;
bus = pci_find_bus(segment, scope->bus);
path = (struct acpi_dmar_pci_path *)(scope + 1);
count = (scope->length - sizeof(struct acpi_dmar_device_scope))
/ sizeof(struct acpi_dmar_pci_path);
while (count) {
if (pdev)
pci_dev_put(pdev);
/*
* Some BIOSes list non-exist devices in DMAR table, just
* ignore it
*/
if (!bus) {
printk(KERN_WARNING
PREFIX "Device scope bus [%d] not found\n",
scope->bus);
break;
}
pdev = pci_get_slot(bus, PCI_DEVFN(path->dev, path->fn));
if (!pdev) {
printk(KERN_WARNING PREFIX
"Device scope device [%04x:%02x:%02x.%02x] not found\n",
segment, bus->number, path->dev, path->fn);
break;
}
path ++;
count --;
bus = pdev->subordinate;
}
if (!pdev) {
printk(KERN_WARNING PREFIX
"Device scope device [%04x:%02x:%02x.%02x] not found\n",
segment, scope->bus, path->dev, path->fn);
*dev = NULL;
return 0;
}
if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT && \
pdev->subordinate) || (scope->entry_type == \
ACPI_DMAR_SCOPE_TYPE_BRIDGE && !pdev->subordinate)) {
pci_dev_put(pdev);
printk(KERN_WARNING PREFIX
"Device scope type does not match for %s\n",
pci_name(pdev));
return -EINVAL;
}
*dev = pdev;
return 0;
}
static int __init dmar_parse_dev_scope(void *start, void *end, int *cnt,
struct pci_dev ***devices, u16 segment)
{
struct acpi_dmar_device_scope *scope;
void * tmp = start;
int index;
int ret;
*cnt = 0;
while (start < end) {
scope = start;
if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE)
(*cnt)++;
else
printk(KERN_WARNING PREFIX
"Unsupported device scope\n");
start += scope->length;
}
if (*cnt == 0)
return 0;
*devices = kcalloc(*cnt, sizeof(struct pci_dev *), GFP_KERNEL);
if (!*devices)
return -ENOMEM;
start = tmp;
index = 0;
while (start < end) {
scope = start;
if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE) {
ret = dmar_parse_one_dev_scope(scope,
&(*devices)[index], segment);
if (ret) {
kfree(*devices);
return ret;
}
index ++;
}
start += scope->length;
}
return 0;
}
/**
* dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition
* structure which uniquely represent one DMA remapping hardware unit
* present in the platform
*/
static int __init
dmar_parse_one_drhd(struct acpi_dmar_header *header)
{
struct acpi_dmar_hardware_unit *drhd;
struct dmar_drhd_unit *dmaru;
int ret = 0;
drhd = (struct acpi_dmar_hardware_unit *)header;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
dmaru = kzalloc(sizeof(*dmaru), GFP_KERNEL);
if (!dmaru)
return -ENOMEM;
dmaru->hdr = header;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
dmaru->reg_base_addr = drhd->address;
dmaru->segment = drhd->segment;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */
ret = alloc_iommu(dmaru);
if (ret) {
kfree(dmaru);
return ret;
}
dmar_register_drhd_unit(dmaru);
return 0;
}
static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
{
struct acpi_dmar_hardware_unit *drhd;
int ret = 0;
drhd = (struct acpi_dmar_hardware_unit *) dmaru->hdr;
if (dmaru->include_all)
return 0;
ret = dmar_parse_dev_scope((void *)(drhd + 1),
((void *)drhd) + drhd->header.length,
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
&dmaru->devices_cnt, &dmaru->devices,
drhd->segment);
if (ret) {
list_del(&dmaru->list);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
kfree(dmaru);
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
return ret;
}
#ifdef CONFIG_DMAR
LIST_HEAD(dmar_rmrr_units);
static void __init dmar_register_rmrr_unit(struct dmar_rmrr_unit *rmrr)
{
list_add(&rmrr->list, &dmar_rmrr_units);
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
static int __init
dmar_parse_one_rmrr(struct acpi_dmar_header *header)
{
struct acpi_dmar_reserved_memory *rmrr;
struct dmar_rmrr_unit *rmrru;
rmrru = kzalloc(sizeof(*rmrru), GFP_KERNEL);
if (!rmrru)
return -ENOMEM;
rmrru->hdr = header;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
rmrr = (struct acpi_dmar_reserved_memory *)header;
rmrru->base_address = rmrr->base_address;
rmrru->end_address = rmrr->end_address;
dmar_register_rmrr_unit(rmrru);
return 0;
}
static int __init
rmrr_parse_dev(struct dmar_rmrr_unit *rmrru)
{
struct acpi_dmar_reserved_memory *rmrr;
int ret;
rmrr = (struct acpi_dmar_reserved_memory *) rmrru->hdr;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
ret = dmar_parse_dev_scope((void *)(rmrr + 1),
((void *)rmrr) + rmrr->header.length,
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
&rmrru->devices_cnt, &rmrru->devices, rmrr->segment);
if (ret || (rmrru->devices_cnt == 0)) {
list_del(&rmrru->list);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
kfree(rmrru);
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
return ret;
}
static LIST_HEAD(dmar_atsr_units);
static int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
{
struct acpi_dmar_atsr *atsr;
struct dmar_atsr_unit *atsru;
atsr = container_of(hdr, struct acpi_dmar_atsr, header);
atsru = kzalloc(sizeof(*atsru), GFP_KERNEL);
if (!atsru)
return -ENOMEM;
atsru->hdr = hdr;
atsru->include_all = atsr->flags & 0x1;
list_add(&atsru->list, &dmar_atsr_units);
return 0;
}
static int __init atsr_parse_dev(struct dmar_atsr_unit *atsru)
{
int rc;
struct acpi_dmar_atsr *atsr;
if (atsru->include_all)
return 0;
atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
rc = dmar_parse_dev_scope((void *)(atsr + 1),
(void *)atsr + atsr->header.length,
&atsru->devices_cnt, &atsru->devices,
atsr->segment);
if (rc || !atsru->devices_cnt) {
list_del(&atsru->list);
kfree(atsru);
}
return rc;
}
int dmar_find_matched_atsr_unit(struct pci_dev *dev)
{
int i;
struct pci_bus *bus;
struct acpi_dmar_atsr *atsr;
struct dmar_atsr_unit *atsru;
list_for_each_entry(atsru, &dmar_atsr_units, list) {
atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
if (atsr->segment == pci_domain_nr(dev->bus))
goto found;
}
return 0;
found:
for (bus = dev->bus; bus; bus = bus->parent) {
struct pci_dev *bridge = bus->self;
if (!bridge || !bridge->is_pcie ||
bridge->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
return 0;
if (bridge->pcie_type == PCI_EXP_TYPE_ROOT_PORT) {
for (i = 0; i < atsru->devices_cnt; i++)
if (atsru->devices[i] == bridge)
return 1;
break;
}
}
if (atsru->include_all)
return 1;
return 0;
}
#endif
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
static void __init
dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
{
struct acpi_dmar_hardware_unit *drhd;
struct acpi_dmar_reserved_memory *rmrr;
struct acpi_dmar_atsr *atsr;
struct acpi_dmar_rhsa *rhsa;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
switch (header->type) {
case ACPI_DMAR_TYPE_HARDWARE_UNIT:
drhd = container_of(header, struct acpi_dmar_hardware_unit,
header);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
printk (KERN_INFO PREFIX
"DRHD base: %#016Lx flags: %#x\n",
(unsigned long long)drhd->address, drhd->flags);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
break;
case ACPI_DMAR_TYPE_RESERVED_MEMORY:
rmrr = container_of(header, struct acpi_dmar_reserved_memory,
header);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
printk (KERN_INFO PREFIX
"RMRR base: %#016Lx end: %#016Lx\n",
(unsigned long long)rmrr->base_address,
(unsigned long long)rmrr->end_address);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
break;
case ACPI_DMAR_TYPE_ATSR:
atsr = container_of(header, struct acpi_dmar_atsr, header);
printk(KERN_INFO PREFIX "ATSR flags: %#x\n", atsr->flags);
break;
case ACPI_DMAR_HARDWARE_AFFINITY:
rhsa = container_of(header, struct acpi_dmar_rhsa, header);
printk(KERN_INFO PREFIX "RHSA base: %#016Lx proximity domain: %#x\n",
(unsigned long long)rhsa->base_address,
rhsa->proximity_domain);
break;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
}
}
/**
* dmar_table_detect - checks to see if the platform supports DMAR devices
*/
static int __init dmar_table_detect(void)
{
acpi_status status = AE_OK;
/* if we could find DMAR table, then there are DMAR devices */
status = acpi_get_table_with_size(ACPI_SIG_DMAR, 0,
(struct acpi_table_header **)&dmar_tbl,
&dmar_tbl_size);
if (ACPI_SUCCESS(status) && !dmar_tbl) {
printk (KERN_WARNING PREFIX "Unable to map DMAR\n");
status = AE_NOT_FOUND;
}
return (ACPI_SUCCESS(status) ? 1 : 0);
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
/**
* parse_dmar_table - parses the DMA reporting table
*/
static int __init
parse_dmar_table(void)
{
struct acpi_table_dmar *dmar;
struct acpi_dmar_header *entry_header;
int ret = 0;
/*
* Do it again, earlier dmar_tbl mapping could be mapped with
* fixed map.
*/
dmar_table_detect();
/*
* ACPI tables may not be DMA protected by tboot, so use DMAR copy
* SINIT saved in SinitMleData in TXT heap (which is DMA protected)
*/
dmar_tbl = tboot_get_dmar_table(dmar_tbl);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
dmar = (struct acpi_table_dmar *)dmar_tbl;
if (!dmar)
return -ENODEV;
if (dmar->width < PAGE_SHIFT - 1) {
printk(KERN_WARNING PREFIX "Invalid DMAR haw\n");
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
return -EINVAL;
}
printk (KERN_INFO PREFIX "Host address width %d\n",
dmar->width + 1);
entry_header = (struct acpi_dmar_header *)(dmar + 1);
while (((unsigned long)entry_header) <
(((unsigned long)dmar) + dmar_tbl->length)) {
intel-iommu: fix endless "Unknown DMAR structure type" loop I have a SuperMicro C2SBX motherboard with BIOS revision 1.0b. With vt-d enabled in the BIOS, Linux gets into an endless loop printing "DMAR:Unknown DMAR structure type" when booting. Here is the DMAR ACPI table: DMAR @ 0x7fe86dec 0000: 44 4d 41 52 98 00 00 00 01 6f 49 6e 74 65 6c 20 DMAR.....oIntel 0010: 4f 45 4d 44 4d 41 52 20 00 00 04 06 4c 4f 48 52 OEMDMAR ....LOHR 0020: 01 00 00 00 23 00 00 00 00 00 00 00 00 00 00 00 ....#........... 0030: 01 00 58 00 00 00 00 00 00 a0 e8 7f 00 00 00 00 ..X............. 0040: ff ff ef 7f 00 00 00 00 01 08 00 00 00 00 1d 00 ................ 0050: 01 08 00 00 00 00 1d 01 01 08 00 00 00 00 1d 02 ................ 0060: 01 08 00 00 00 00 1d 07 01 08 00 00 00 00 1a 00 ................ 0070: 01 08 00 00 00 00 1a 01 01 08 00 00 00 00 1a 02 ................ 0080: 01 08 00 00 00 00 1a 07 01 08 00 00 00 00 1a 07 ................ 0090: c0 00 68 00 04 10 66 60 ..h...f` Here are the messages printed by the kernel: DMAR:Host address width 36 DMAR:RMRR base: 0x000000007fe8a000 end: 0x000000007fefffff DMAR:Unknown DMAR structure type DMAR:Unknown DMAR structure type DMAR:Unknown DMAR structure type ... Although I not very familiar with ACPI, to me it looks like struct acpi_dmar_header::length == 0x0058 is incorrect, causing parse_dmar_table() to look at an invalid offset on the next loop. This offset happens to have struct acpi_dmar_header::length == 0x0000, which prevents the loop from ever terminating. This patch checks for this condition and bails out instead of looping forever. Signed-off-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
2009-02-11 21:24:19 +00:00
/* Avoid looping forever on bad ACPI tables */
if (entry_header->length == 0) {
printk(KERN_WARNING PREFIX
"Invalid 0-length structure\n");
ret = -EINVAL;
break;
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
dmar_table_print_dmar_entry(entry_header);
switch (entry_header->type) {
case ACPI_DMAR_TYPE_HARDWARE_UNIT:
ret = dmar_parse_one_drhd(entry_header);
break;
case ACPI_DMAR_TYPE_RESERVED_MEMORY:
#ifdef CONFIG_DMAR
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
ret = dmar_parse_one_rmrr(entry_header);
#endif
break;
case ACPI_DMAR_TYPE_ATSR:
#ifdef CONFIG_DMAR
ret = dmar_parse_one_atsr(entry_header);
#endif
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
break;
case ACPI_DMAR_HARDWARE_AFFINITY:
/* We don't do anything with RHSA (yet?) */
break;
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
default:
printk(KERN_WARNING PREFIX
"Unknown DMAR structure type %d\n",
entry_header->type);
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
ret = 0; /* for forward compatibility */
break;
}
if (ret)
break;
entry_header = ((void *)entry_header + entry_header->length);
}
return ret;
}
int dmar_pci_device_match(struct pci_dev *devices[], int cnt,
struct pci_dev *dev)
{
int index;
while (dev) {
for (index = 0; index < cnt; index++)
if (dev == devices[index])
return 1;
/* Check our parent */
dev = dev->bus->self;
}
return 0;
}
struct dmar_drhd_unit *
dmar_find_matched_drhd_unit(struct pci_dev *dev)
{
struct dmar_drhd_unit *dmaru = NULL;
struct acpi_dmar_hardware_unit *drhd;
list_for_each_entry(dmaru, &dmar_drhd_units, list) {
drhd = container_of(dmaru->hdr,
struct acpi_dmar_hardware_unit,
header);
if (dmaru->include_all &&
drhd->segment == pci_domain_nr(dev->bus))
return dmaru;
if (dmar_pci_device_match(dmaru->devices,
dmaru->devices_cnt, dev))
return dmaru;
}
return NULL;
}
int __init dmar_dev_scope_init(void)
{
struct dmar_drhd_unit *drhd, *drhd_n;
int ret = -ENODEV;
list_for_each_entry_safe(drhd, drhd_n, &dmar_drhd_units, list) {
ret = dmar_parse_dev(drhd);
if (ret)
return ret;
}
#ifdef CONFIG_DMAR
{
struct dmar_rmrr_unit *rmrr, *rmrr_n;
struct dmar_atsr_unit *atsr, *atsr_n;
list_for_each_entry_safe(rmrr, rmrr_n, &dmar_rmrr_units, list) {
ret = rmrr_parse_dev(rmrr);
if (ret)
return ret;
}
list_for_each_entry_safe(atsr, atsr_n, &dmar_atsr_units, list) {
ret = atsr_parse_dev(atsr);
if (ret)
return ret;
}
}
#endif
return ret;
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
int __init dmar_table_init(void)
{
static int dmar_table_initialized;
int ret;
if (dmar_table_initialized)
return 0;
dmar_table_initialized = 1;
ret = parse_dmar_table();
if (ret) {
if (ret != -ENODEV)
printk(KERN_INFO PREFIX "parse DMAR table failure.\n");
return ret;
}
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
if (list_empty(&dmar_drhd_units)) {
printk(KERN_INFO PREFIX "No DMAR devices found\n");
return -ENODEV;
}
#ifdef CONFIG_DMAR
if (list_empty(&dmar_rmrr_units))
printk(KERN_INFO PREFIX "No RMRR found\n");
if (list_empty(&dmar_atsr_units))
printk(KERN_INFO PREFIX "No ATSR found\n");
#endif
Intel IOMMU: DMAR detection and parsing logic This patch supports the upcomming Intel IOMMU hardware a.k.a. Intel(R) Virtualization Technology for Directed I/O Architecture and the hardware spec for the same can be found here http://www.intel.com/technology/virtualization/index.htm FAQ! (questions from akpm, answers from ak) > So... what's all this code for? > > I assume that the intent here is to speed things up under Xen, etc? Yes in some cases, but not this code. That would be the Xen version of this code that could potentially assign whole devices to guests. I expect this to be only useful in some special cases though because most hardware is not virtualizable and you typically want an own instance for each guest. Ok at some point KVM might implement this too; i likely would use this code for this. > Do we > have any benchmark results to help us to decide whether a merge would be > justified? The main advantage for doing it in the normal kernel is not performance, but more safety. Broken devices won't be able to corrupt memory by doing random DMA. Unfortunately that doesn't work for graphics yet, for that need user space interfaces for the X server are needed. There are some potential performance benefits too: - When you have a device that cannot address the complete address range an IOMMU can remap its memory instead of bounce buffering. Remapping is likely cheaper than copying. - The IOMMU can merge sg lists into a single virtual block. This could potentially speed up SG IO when the device is slow walking SG lists. [I long ago benchmarked 5% on some block benchmark with an old MPT Fusion; but it probably depends a lot on the HBA] And you get better driver debugging because unexpected memory accesses from the devices will cause a trappable event. > > Does it slow anything down? It adds more overhead to each IO so yes. This patch: Add support for early detection and parsing of DMAR's (DMA Remapping) reported to OS via ACPI tables. DMA remapping(DMAR) devices support enables independent address translations for Direct Memory Access(DMA) from Devices. These DMA remapping devices are reported via ACPI tables and includes pci device scope covered by these DMA remapping device. For detailed info on the specification of "Intel(R) Virtualization Technology for Directed I/O Architecture" please see http://www.intel.com/technology/virtualization/index.htm Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: Andi Kleen <ak@suse.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Christoph Lameter <clameter@sgi.com> Cc: Greg KH <greg@kroah.com> Cc: Len Brown <lenb@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-21 23:41:41 +00:00
return 0;
}
int __init check_zero_address(void)
{
struct acpi_table_dmar *dmar;
struct acpi_dmar_header *entry_header;
struct acpi_dmar_hardware_unit *drhd;
dmar = (struct acpi_table_dmar *)dmar_tbl;
entry_header = (struct acpi_dmar_header *)(dmar + 1);
while (((unsigned long)entry_header) <
(((unsigned long)dmar) + dmar_tbl->length)) {
/* Avoid looping forever on bad ACPI tables */
if (entry_header->length == 0) {
printk(KERN_WARNING PREFIX
"Invalid 0-length structure\n");
return 0;
}
if (entry_header->type == ACPI_DMAR_TYPE_HARDWARE_UNIT) {
drhd = (void *)entry_header;
if (!drhd->address) {
/* Promote an attitude of violence to a BIOS engineer today */
WARN(1, "Your BIOS is broken; DMAR reported at address zero!\n"
"BIOS vendor: %s; Ver: %s; Product Version: %s\n",
dmi_get_system_info(DMI_BIOS_VENDOR),
dmi_get_system_info(DMI_BIOS_VERSION),
dmi_get_system_info(DMI_PRODUCT_VERSION));
#ifdef CONFIG_DMAR
dmar_disabled = 1;
#endif
return 0;
}
break;
}
entry_header = ((void *)entry_header + entry_header->length);
}
return 1;
}
void __init detect_intel_iommu(void)
{
int ret;
ret = dmar_table_detect();
if (ret)
ret = check_zero_address();
{
#ifdef CONFIG_INTR_REMAP
struct acpi_table_dmar *dmar;
/*
* for now we will disable dma-remapping when interrupt
* remapping is enabled.
* When support for queued invalidation for IOTLB invalidation
* is added, we will not need this any more.
*/
dmar = (struct acpi_table_dmar *) dmar_tbl;
if (ret && cpu_has_x2apic && dmar->flags & 0x1)
printk(KERN_INFO
"Queued invalidation will be enabled to support "
"x2apic and Intr-remapping.\n");
#endif
#ifdef CONFIG_DMAR
x86: Handle HW IOMMU initialization failure gracefully If HW IOMMU initialization fails (Intel VT-d often does this, typically due to BIOS bugs), we fall back to nommu. It doesn't work for the majority since nowadays we have more than 4GB memory so we must use swiotlb instead of nommu. The problem is that it's too late to initialize swiotlb when HW IOMMU initialization fails. We need to allocate swiotlb memory earlier from bootmem allocator. Chris explained the issue in detail: http://marc.info/?l=linux-kernel&m=125657444317079&w=2 The current x86 IOMMU initialization sequence is too complicated and handling the above issue makes it more hacky. This patch changes x86 IOMMU initialization sequence to handle the above issue cleanly. The new x86 IOMMU initialization sequence are: 1. we initialize the swiotlb (and setting swiotlb to 1) in the case of (max_pfn > MAX_DMA32_PFN && !no_iommu). dma_ops is set to swiotlb_dma_ops or nommu_dma_ops. if swiotlb usage is forced by the boot option, we finish here. 2. we call the detection functions of all the IOMMUs 3. the detection function sets x86_init.iommu.iommu_init to the IOMMU initialization function (so we can avoid calling the initialization functions of all the IOMMUs needlessly). 4. if the IOMMU initialization function doesn't need to swiotlb then sets swiotlb to zero (e.g. the initialization is sucessful). 5. if we find that swiotlb is set to zero, we free swiotlb resource. Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Cc: chrisw@sous-sol.org Cc: dwmw2@infradead.org Cc: joerg.roedel@amd.com Cc: muli@il.ibm.com LKML-Reference: <1257849980-22640-10-git-send-email-fujita.tomonori@lab.ntt.co.jp> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-10 10:46:20 +00:00
if (ret && !no_iommu && !iommu_detected && !dmar_disabled)
iommu_detected = 1;
#endif
#ifdef CONFIG_X86
if (ret)
x86_init.iommu.iommu_init = intel_iommu_init;
#endif
}
early_acpi_os_unmap_memory(dmar_tbl, dmar_tbl_size);
dmar_tbl = NULL;
}
int alloc_iommu(struct dmar_drhd_unit *drhd)
{
struct intel_iommu *iommu;
int map_size;
u32 ver;
static int iommu_allocated = 0;
int agaw = 0;
int msagaw = 0;
iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
if (!iommu)
return -ENOMEM;
iommu->seq_id = iommu_allocated++;
sprintf (iommu->name, "dmar%d", iommu->seq_id);
iommu->reg = ioremap(drhd->reg_base_addr, VTD_PAGE_SIZE);
if (!iommu->reg) {
printk(KERN_ERR "IOMMU: can't map the region\n");
goto error;
}
iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) {
/* Promote an attitude of violence to a BIOS engineer today */
WARN(1, "Your BIOS is broken; DMAR reported at address %llx returns all ones!\n"
"BIOS vendor: %s; Ver: %s; Product Version: %s\n",
drhd->reg_base_addr,
dmi_get_system_info(DMI_BIOS_VENDOR),
dmi_get_system_info(DMI_BIOS_VERSION),
dmi_get_system_info(DMI_PRODUCT_VERSION));
goto err_unmap;
}
#ifdef CONFIG_DMAR
agaw = iommu_calculate_agaw(iommu);
if (agaw < 0) {
printk(KERN_ERR
"Cannot get a valid agaw for iommu (seq_id = %d)\n",
iommu->seq_id);
goto err_unmap;
}
msagaw = iommu_calculate_max_sagaw(iommu);
if (msagaw < 0) {
printk(KERN_ERR
"Cannot get a valid max agaw for iommu (seq_id = %d)\n",
iommu->seq_id);
goto err_unmap;
}
#endif
iommu->agaw = agaw;
iommu->msagaw = msagaw;
/* the registers might be more than one page */
map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
cap_max_fault_reg_offset(iommu->cap));
map_size = VTD_PAGE_ALIGN(map_size);
if (map_size > VTD_PAGE_SIZE) {
iounmap(iommu->reg);
iommu->reg = ioremap(drhd->reg_base_addr, map_size);
if (!iommu->reg) {
printk(KERN_ERR "IOMMU: can't map the region\n");
goto error;
}
}
ver = readl(iommu->reg + DMAR_VER_REG);
pr_info("IOMMU %llx: ver %d:%d cap %llx ecap %llx\n",
(unsigned long long)drhd->reg_base_addr,
DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
(unsigned long long)iommu->cap,
(unsigned long long)iommu->ecap);
spin_lock_init(&iommu->register_lock);
drhd->iommu = iommu;
return 0;
err_unmap:
iounmap(iommu->reg);
error:
kfree(iommu);
return -1;
}
void free_iommu(struct intel_iommu *iommu)
{
if (!iommu)
return;
#ifdef CONFIG_DMAR
free_dmar_iommu(iommu);
#endif
if (iommu->reg)
iounmap(iommu->reg);
kfree(iommu);
}
/*
* Reclaim all the submitted descriptors which have completed its work.
*/
static inline void reclaim_free_desc(struct q_inval *qi)
{
while (qi->desc_status[qi->free_tail] == QI_DONE ||
qi->desc_status[qi->free_tail] == QI_ABORT) {
qi->desc_status[qi->free_tail] = QI_FREE;
qi->free_tail = (qi->free_tail + 1) % QI_LENGTH;
qi->free_cnt++;
}
}
static int qi_check_fault(struct intel_iommu *iommu, int index)
{
u32 fault;
int head, tail;
struct q_inval *qi = iommu->qi;
int wait_index = (index + 1) % QI_LENGTH;
if (qi->desc_status[wait_index] == QI_ABORT)
return -EAGAIN;
fault = readl(iommu->reg + DMAR_FSTS_REG);
/*
* If IQE happens, the head points to the descriptor associated
* with the error. No new descriptors are fetched until the IQE
* is cleared.
*/
if (fault & DMA_FSTS_IQE) {
head = readl(iommu->reg + DMAR_IQH_REG);
if ((head >> DMAR_IQ_SHIFT) == index) {
printk(KERN_ERR "VT-d detected invalid descriptor: "
"low=%llx, high=%llx\n",
(unsigned long long)qi->desc[index].low,
(unsigned long long)qi->desc[index].high);
memcpy(&qi->desc[index], &qi->desc[wait_index],
sizeof(struct qi_desc));
__iommu_flush_cache(iommu, &qi->desc[index],
sizeof(struct qi_desc));
writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG);
return -EINVAL;
}
}
/*
* If ITE happens, all pending wait_desc commands are aborted.
* No new descriptors are fetched until the ITE is cleared.
*/
if (fault & DMA_FSTS_ITE) {
head = readl(iommu->reg + DMAR_IQH_REG);
head = ((head >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
head |= 1;
tail = readl(iommu->reg + DMAR_IQT_REG);
tail = ((tail >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG);
do {
if (qi->desc_status[head] == QI_IN_USE)
qi->desc_status[head] = QI_ABORT;
head = (head - 2 + QI_LENGTH) % QI_LENGTH;
} while (head != tail);
if (qi->desc_status[wait_index] == QI_ABORT)
return -EAGAIN;
}
if (fault & DMA_FSTS_ICE)
writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG);
return 0;
}
/*
* Submit the queued invalidation descriptor to the remapping
* hardware unit and wait for its completion.
*/
int qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
{
int rc;
struct q_inval *qi = iommu->qi;
struct qi_desc *hw, wait_desc;
int wait_index, index;
unsigned long flags;
if (!qi)
return 0;
hw = qi->desc;
restart:
rc = 0;
spin_lock_irqsave(&qi->q_lock, flags);
while (qi->free_cnt < 3) {
spin_unlock_irqrestore(&qi->q_lock, flags);
cpu_relax();
spin_lock_irqsave(&qi->q_lock, flags);
}
index = qi->free_head;
wait_index = (index + 1) % QI_LENGTH;
qi->desc_status[index] = qi->desc_status[wait_index] = QI_IN_USE;
hw[index] = *desc;
wait_desc.low = QI_IWD_STATUS_DATA(QI_DONE) |
QI_IWD_STATUS_WRITE | QI_IWD_TYPE;
wait_desc.high = virt_to_phys(&qi->desc_status[wait_index]);
hw[wait_index] = wait_desc;
__iommu_flush_cache(iommu, &hw[index], sizeof(struct qi_desc));
__iommu_flush_cache(iommu, &hw[wait_index], sizeof(struct qi_desc));
qi->free_head = (qi->free_head + 2) % QI_LENGTH;
qi->free_cnt -= 2;
/*
* update the HW tail register indicating the presence of
* new descriptors.
*/
writel(qi->free_head << DMAR_IQ_SHIFT, iommu->reg + DMAR_IQT_REG);
while (qi->desc_status[wait_index] != QI_DONE) {
/*
* We will leave the interrupts disabled, to prevent interrupt
* context to queue another cmd while a cmd is already submitted
* and waiting for completion on this cpu. This is to avoid
* a deadlock where the interrupt context can wait indefinitely
* for free slots in the queue.
*/
rc = qi_check_fault(iommu, index);
if (rc)
break;
spin_unlock(&qi->q_lock);
cpu_relax();
spin_lock(&qi->q_lock);
}
qi->desc_status[index] = QI_DONE;
reclaim_free_desc(qi);
spin_unlock_irqrestore(&qi->q_lock, flags);
if (rc == -EAGAIN)
goto restart;
return rc;
}
/*
* Flush the global interrupt entry cache.
*/
void qi_global_iec(struct intel_iommu *iommu)
{
struct qi_desc desc;
desc.low = QI_IEC_TYPE;
desc.high = 0;
/* should never fail */
qi_submit_sync(&desc, iommu);
}
void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
u64 type)
{
struct qi_desc desc;
desc.low = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did)
| QI_CC_GRAN(type) | QI_CC_TYPE;
desc.high = 0;
qi_submit_sync(&desc, iommu);
}
void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
unsigned int size_order, u64 type)
{
u8 dw = 0, dr = 0;
struct qi_desc desc;
int ih = 0;
if (cap_write_drain(iommu->cap))
dw = 1;
if (cap_read_drain(iommu->cap))
dr = 1;
desc.low = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw)
| QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE;
desc.high = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih)
| QI_IOTLB_AM(size_order);
qi_submit_sync(&desc, iommu);
}
void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 qdep,
u64 addr, unsigned mask)
{
struct qi_desc desc;
if (mask) {
BUG_ON(addr & ((1 << (VTD_PAGE_SHIFT + mask)) - 1));
addr |= (1 << (VTD_PAGE_SHIFT + mask - 1)) - 1;
desc.high = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE;
} else
desc.high = QI_DEV_IOTLB_ADDR(addr);
if (qdep >= QI_DEV_IOTLB_MAX_INVS)
qdep = 0;
desc.low = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) |
QI_DIOTLB_TYPE;
qi_submit_sync(&desc, iommu);
}
/*
* Disable Queued Invalidation interface.
*/
void dmar_disable_qi(struct intel_iommu *iommu)
{
unsigned long flags;
u32 sts;
cycles_t start_time = get_cycles();
if (!ecap_qis(iommu->ecap))
return;
spin_lock_irqsave(&iommu->register_lock, flags);
sts = dmar_readq(iommu->reg + DMAR_GSTS_REG);
if (!(sts & DMA_GSTS_QIES))
goto end;
/*
* Give a chance to HW to complete the pending invalidation requests.
*/
while ((readl(iommu->reg + DMAR_IQT_REG) !=
readl(iommu->reg + DMAR_IQH_REG)) &&
(DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time)))
cpu_relax();
iommu->gcmd &= ~DMA_GCMD_QIE;
writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl,
!(sts & DMA_GSTS_QIES), sts);
end:
spin_unlock_irqrestore(&iommu->register_lock, flags);
}
/*
* Enable queued invalidation.
*/
static void __dmar_enable_qi(struct intel_iommu *iommu)
{
u32 sts;
unsigned long flags;
struct q_inval *qi = iommu->qi;
qi->free_head = qi->free_tail = 0;
qi->free_cnt = QI_LENGTH;
spin_lock_irqsave(&iommu->register_lock, flags);
/* write zero to the tail reg */
writel(0, iommu->reg + DMAR_IQT_REG);
dmar_writeq(iommu->reg + DMAR_IQA_REG, virt_to_phys(qi->desc));
iommu->gcmd |= DMA_GCMD_QIE;
writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
/* Make sure hardware complete it */
IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts);
spin_unlock_irqrestore(&iommu->register_lock, flags);
}
/*
* Enable Queued Invalidation interface. This is a must to support
* interrupt-remapping. Also used by DMA-remapping, which replaces
* register based IOTLB invalidation.
*/
int dmar_enable_qi(struct intel_iommu *iommu)
{
struct q_inval *qi;
if (!ecap_qis(iommu->ecap))
return -ENOENT;
/*
* queued invalidation is already setup and enabled.
*/
if (iommu->qi)
return 0;
iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC);
if (!iommu->qi)
return -ENOMEM;
qi = iommu->qi;
qi->desc = (void *)(get_zeroed_page(GFP_ATOMIC));
if (!qi->desc) {
kfree(qi);
iommu->qi = 0;
return -ENOMEM;
}
qi->desc_status = kmalloc(QI_LENGTH * sizeof(int), GFP_ATOMIC);
if (!qi->desc_status) {
free_page((unsigned long) qi->desc);
kfree(qi);
iommu->qi = 0;
return -ENOMEM;
}
qi->free_head = qi->free_tail = 0;
qi->free_cnt = QI_LENGTH;
spin_lock_init(&qi->q_lock);
__dmar_enable_qi(iommu);
return 0;
}
/* iommu interrupt handling. Most stuff are MSI-like. */
enum faulttype {
DMA_REMAP,
INTR_REMAP,
UNKNOWN,
};
static const char *dma_remap_fault_reasons[] =
{
"Software",
"Present bit in root entry is clear",
"Present bit in context entry is clear",
"Invalid context entry",
"Access beyond MGAW",
"PTE Write access is not set",
"PTE Read access is not set",
"Next page table ptr is invalid",
"Root table address invalid",
"Context table ptr is invalid",
"non-zero reserved fields in RTP",
"non-zero reserved fields in CTP",
"non-zero reserved fields in PTE",
};
static const char *intr_remap_fault_reasons[] =
{
"Detected reserved fields in the decoded interrupt-remapped request",
"Interrupt index exceeded the interrupt-remapping table size",
"Present field in the IRTE entry is clear",
"Error accessing interrupt-remapping table pointed by IRTA_REG",
"Detected reserved fields in the IRTE entry",
"Blocked a compatibility format interrupt request",
"Blocked an interrupt request due to source-id verification failure",
};
#define MAX_FAULT_REASON_IDX (ARRAY_SIZE(fault_reason_strings) - 1)
const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type)
{
if (fault_reason >= 0x20 && (fault_reason <= 0x20 +
ARRAY_SIZE(intr_remap_fault_reasons))) {
*fault_type = INTR_REMAP;
return intr_remap_fault_reasons[fault_reason - 0x20];
} else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) {
*fault_type = DMA_REMAP;
return dma_remap_fault_reasons[fault_reason];
} else {
*fault_type = UNKNOWN;
return "Unknown";
}
}
void dmar_msi_unmask(unsigned int irq)
{
struct intel_iommu *iommu = get_irq_data(irq);
unsigned long flag;
/* unmask it */
spin_lock_irqsave(&iommu->register_lock, flag);
writel(0, iommu->reg + DMAR_FECTL_REG);
/* Read a reg to force flush the post write */
readl(iommu->reg + DMAR_FECTL_REG);
spin_unlock_irqrestore(&iommu->register_lock, flag);
}
void dmar_msi_mask(unsigned int irq)
{
unsigned long flag;
struct intel_iommu *iommu = get_irq_data(irq);
/* mask it */
spin_lock_irqsave(&iommu->register_lock, flag);
writel(DMA_FECTL_IM, iommu->reg + DMAR_FECTL_REG);
/* Read a reg to force flush the post write */
readl(iommu->reg + DMAR_FECTL_REG);
spin_unlock_irqrestore(&iommu->register_lock, flag);
}
void dmar_msi_write(int irq, struct msi_msg *msg)
{
struct intel_iommu *iommu = get_irq_data(irq);
unsigned long flag;
spin_lock_irqsave(&iommu->register_lock, flag);
writel(msg->data, iommu->reg + DMAR_FEDATA_REG);
writel(msg->address_lo, iommu->reg + DMAR_FEADDR_REG);
writel(msg->address_hi, iommu->reg + DMAR_FEUADDR_REG);
spin_unlock_irqrestore(&iommu->register_lock, flag);
}
void dmar_msi_read(int irq, struct msi_msg *msg)
{
struct intel_iommu *iommu = get_irq_data(irq);
unsigned long flag;
spin_lock_irqsave(&iommu->register_lock, flag);
msg->data = readl(iommu->reg + DMAR_FEDATA_REG);
msg->address_lo = readl(iommu->reg + DMAR_FEADDR_REG);
msg->address_hi = readl(iommu->reg + DMAR_FEUADDR_REG);
spin_unlock_irqrestore(&iommu->register_lock, flag);
}
static int dmar_fault_do_one(struct intel_iommu *iommu, int type,
u8 fault_reason, u16 source_id, unsigned long long addr)
{
const char *reason;
int fault_type;
reason = dmar_get_fault_reason(fault_reason, &fault_type);
if (fault_type == INTR_REMAP)
printk(KERN_ERR "INTR-REMAP: Request device [[%02x:%02x.%d] "
"fault index %llx\n"
"INTR-REMAP:[fault reason %02d] %s\n",
(source_id >> 8), PCI_SLOT(source_id & 0xFF),
PCI_FUNC(source_id & 0xFF), addr >> 48,
fault_reason, reason);
else
printk(KERN_ERR
"DMAR:[%s] Request device [%02x:%02x.%d] "
"fault addr %llx \n"
"DMAR:[fault reason %02d] %s\n",
(type ? "DMA Read" : "DMA Write"),
(source_id >> 8), PCI_SLOT(source_id & 0xFF),
PCI_FUNC(source_id & 0xFF), addr, fault_reason, reason);
return 0;
}
#define PRIMARY_FAULT_REG_LEN (16)
irqreturn_t dmar_fault(int irq, void *dev_id)
{
struct intel_iommu *iommu = dev_id;
int reg, fault_index;
u32 fault_status;
unsigned long flag;
spin_lock_irqsave(&iommu->register_lock, flag);
fault_status = readl(iommu->reg + DMAR_FSTS_REG);
if (fault_status)
printk(KERN_ERR "DRHD: handling fault status reg %x\n",
fault_status);
/* TBD: ignore advanced fault log currently */
if (!(fault_status & DMA_FSTS_PPF))
goto clear_rest;
fault_index = dma_fsts_fault_record_index(fault_status);
reg = cap_fault_reg_offset(iommu->cap);
while (1) {
u8 fault_reason;
u16 source_id;
u64 guest_addr;
int type;
u32 data;
/* highest 32 bits */
data = readl(iommu->reg + reg +
fault_index * PRIMARY_FAULT_REG_LEN + 12);
if (!(data & DMA_FRCD_F))
break;
fault_reason = dma_frcd_fault_reason(data);
type = dma_frcd_type(data);
data = readl(iommu->reg + reg +
fault_index * PRIMARY_FAULT_REG_LEN + 8);
source_id = dma_frcd_source_id(data);
guest_addr = dmar_readq(iommu->reg + reg +
fault_index * PRIMARY_FAULT_REG_LEN);
guest_addr = dma_frcd_page_addr(guest_addr);
/* clear the fault */
writel(DMA_FRCD_F, iommu->reg + reg +
fault_index * PRIMARY_FAULT_REG_LEN + 12);
spin_unlock_irqrestore(&iommu->register_lock, flag);
dmar_fault_do_one(iommu, type, fault_reason,
source_id, guest_addr);
fault_index++;
if (fault_index >= cap_num_fault_regs(iommu->cap))
fault_index = 0;
spin_lock_irqsave(&iommu->register_lock, flag);
}
clear_rest:
/* clear all the other faults */
fault_status = readl(iommu->reg + DMAR_FSTS_REG);
writel(fault_status, iommu->reg + DMAR_FSTS_REG);
spin_unlock_irqrestore(&iommu->register_lock, flag);
return IRQ_HANDLED;
}
int dmar_set_interrupt(struct intel_iommu *iommu)
{
int irq, ret;
/*
* Check if the fault interrupt is already initialized.
*/
if (iommu->irq)
return 0;
irq = create_irq();
if (!irq) {
printk(KERN_ERR "IOMMU: no free vectors\n");
return -EINVAL;
}
set_irq_data(irq, iommu);
iommu->irq = irq;
ret = arch_setup_dmar_msi(irq);
if (ret) {
set_irq_data(irq, NULL);
iommu->irq = 0;
destroy_irq(irq);
return ret;
}
ret = request_irq(irq, dmar_fault, 0, iommu->name, iommu);
if (ret)
printk(KERN_ERR "IOMMU: can't request irq\n");
return ret;
}
int __init enable_drhd_fault_handling(void)
{
struct dmar_drhd_unit *drhd;
/*
* Enable fault control interrupt.
*/
for_each_drhd_unit(drhd) {
int ret;
struct intel_iommu *iommu = drhd->iommu;
ret = dmar_set_interrupt(iommu);
if (ret) {
printk(KERN_ERR "DRHD %Lx: failed to enable fault, "
" interrupt, ret %d\n",
(unsigned long long)drhd->reg_base_addr, ret);
return -1;
}
}
return 0;
}
/*
* Re-enable Queued Invalidation interface.
*/
int dmar_reenable_qi(struct intel_iommu *iommu)
{
if (!ecap_qis(iommu->ecap))
return -ENOENT;
if (!iommu->qi)
return -ENOENT;
/*
* First disable queued invalidation.
*/
dmar_disable_qi(iommu);
/*
* Then enable queued invalidation again. Since there is no pending
* invalidation requests now, it's safe to re-enable queued
* invalidation.
*/
__dmar_enable_qi(iommu);
return 0;
}
/*
* Check interrupt remapping support in DMAR table description.
*/
int dmar_ir_support(void)
{
struct acpi_table_dmar *dmar;
dmar = (struct acpi_table_dmar *)dmar_tbl;
return dmar->flags & 0x1;
}