Skip to contentSkip to conversation

x86/mcfg: relax checking of MCFG position

PATCH In review Roger Pau Monne 3 participants 7 messages Jun 4 – Jun 5

lore
Share & tools
lore
#0 lore raw

v2 cover x86/mcfg: relax checking of MCFG position

Hello,

First patch is a trivial cleanup of header sorting, second patch is the real fix. Possibly something to consider for inclusion in 4.22.

Thanks, Roger.

Roger Pau Monne (2):
x86/mcfg: sort header includes
x86/mcfg: relax memory map checks on newer firmware

 xen/arch/x86/x86_64/mmconfig-shared.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)
Signature · 1 line
2.53.0
#1 lore raw

PATCH v2 1/2 x86/mcfg: sort header includes

No functional change intended.

---
 xen/arch/x86/x86_64/mmconfig-shared.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
xen/arch/x86/x86_64/mmconfig-shared.c +5 -3
index ab082b5f5b37..d0cbc151705d 100644
@@ -12,17 +12,19 @@
  * Author: Allen Kay <[email protected]> -  adapted to xen from Linux
  */
 
+#include <xen/acpi.h>
 #include <xen/init.h>
 #include <xen/mm.h>
 #include <xen/param.h>
-#include <xen/acpi.h>
-#include <xen/xmalloc.h>
 #include <xen/pci.h>
-#include <xen/pci_regs.h>
 #include <xen/pci_ids.h>
+#include <xen/pci_regs.h>
+#include <xen/xmalloc.h>
+
 #include <asm/e820.h>
 #include <asm/msr.h>
 #include <asm/msr-index.h>
+
 #include <public/physdev.h>
 
 #include "mmconfig.h"
Signature · 1 line
2.53.0
#3 lore raw

PATCH v2 2/2 x86/mcfg: relax memory map checks on newer firmware

In reply to Roger Pau Monne

Per PCI Firmware 3.3 specification, section 4.1.2, ECAM space must be reserved by declaring a motherboard resource, but there's no requirement to mention it in E820. The specification additionally states that: the resources can optionally be returned in Int15 E820h or EFIGetMemoryMap as reserved memory.

This fixes an issue with detection of extended capabilities when running Xen on a Lenovo system that doesn't list the MCFG area as an EfiMemoryMappedIO region (or as any other type) in the EFI memory map.

The more strict logic was introduced in Linux in 2006 as 946f2ee5c731 ("[PATCH] i386/x86-64: Check that MCFG points to an e820 reserved area"). This was picked up by Xen when MCFG support was added in 3b35911d709e ("Enable pci mmcfg and ATS for x86_64"). Apply an approach similar to what Linux has done in 199f968f1484 ("x86/pci: Skip early E820 check for ECAM region") and relax the strict reserved region checking so it's only done for firmware manufactured prior to 2016. For firmware from 2016 and newer allow MCFG region to reside in holes on the memory map. Note Xen is still more strict than Linux however, as it will refuse to use MCFG regions that overlap with memory map regions different than reserved.

When dom0 boots it can prevent access to misconfigured MCFG regions by using the PHYSDEVOP_pci_mmcfg_reserved hypercall. This brings Xen's early usage of MCFG (prior to ACPI AML parsing) more in line with the implementation in Linux.

---
Changes since v1:
- Shift/adjust commit message.
- Reword in-code comment.

---
 xen/arch/x86/x86_64/mmconfig-shared.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
xen/arch/x86/x86_64/mmconfig-shared.c +11 -1
index d0cbc151705d..b33e2f56e6e3 100644
@@ -13,6 +13,7 @@
  */
 
 #include <xen/acpi.h>
+#include <xen/dmi.h>
 #include <xen/init.h>
 #include <xen/mm.h>
 #include <xen/param.h>
@@ -369,12 +370,15 @@ static bool __init pci_mmcfg_reject_broken(void)
     typeof(pci_mmcfg_config[0]) *cfg;
     int i;
     bool valid = true;
+    int year;
 
     if ((pci_mmcfg_config_num == 0) ||
         (pci_mmcfg_config == NULL) ||
         (pci_mmcfg_config[0].address == 0))
         return 0;
 
+    dmi_get_date(DMI_BIOS_DATE, &year, NULL, NULL);
+
     for (i = 0; i < pci_mmcfg_config_num; i++) {
         u64 addr, size;
 
@@ -390,7 +394,13 @@ static bool __init pci_mmcfg_reject_broken(void)
                (unsigned int)cfg->start_bus_number,
                (unsigned int)cfg->end_bus_number);
 
-        if ( !is_mmconf_reserved(addr, size, i, cfg) ||
+        /*
+         * For firmwares prior to 2016, confirm that MMCFG is marked as
+         * reserved.  For 2016 and later, also allow MMCFG being in a hole.
+         */
+        if ( ((year < 2016 || !is_memory_hole(maddr_to_mfn(addr),
+                                              maddr_to_mfn(addr + size - 1))) &&
+              !is_mmconf_reserved(addr, size, i, cfg)) ||
              pci_mmcfg_arch_enable(i) < 0 )
         {
             pci_mmcfg_arch_disable(i);
Signature · 1 line
2.53.0
#4 lore raw

Re: [PATCH for-4.22 v2 2/2] x86/mcfg: relax memory map checks on newer firmware

Quoted 21 lines · Roger Pau Monne

Per PCI Firmware 3.3 specification, section 4.1.2, ECAM space must be reserved by declaring a motherboard resource, but there's no requirement to mention it in E820. The specification additionally states that: the resources can optionally be returned in Int15 E820h or EFIGetMemoryMap as reserved memory.

This fixes an issue with detection of extended capabilities when running Xen on a Lenovo system that doesn't list the MCFG area as an EfiMemoryMappedIO region (or as any other type) in the EFI memory map.

The more strict logic was introduced in Linux in 2006 as 946f2ee5c731 ("[PATCH] i386/x86-64: Check that MCFG points to an e820 reserved area"). This was picked up by Xen when MCFG support was added in 3b35911d709e ("Enable pci mmcfg and ATS for x86_64"). Apply an approach similar to what Linux has done in 199f968f1484 ("x86/pci: Skip early E820 check for ECAM region") and relax the strict reserved region checking so it's only done for firmware manufactured prior to 2016. For firmware from 2016 and newer allow MCFG region to reside in holes on the memory map. Note Xen is still more strict than Linux however, as it will refuse to use MCFG regions that overlap with memory map regions different than reserved.

Quoted 4 lines

When dom0 boots it can prevent access to misconfigured MCFG regions by using the PHYSDEVOP_pci_mmcfg_reserved hypercall. This brings Xen's early usage of MCFG (prior to ACPI AML parsing) more in line with the implementation in Linux.

Don't we want to add Fixes: or Ammends: here?

Quoted 1 line

Generally it looks low-risk patch so:

Thanks.

~ Oleksii

#5 lore raw

Re: [PATCH for-4.22 v2 2/2] x86/mcfg: relax memory map checks on newer firmware

Quoted 24 lines · Oleksii Kurochko
Quoted 21 lines · Roger Pau Monne

Per PCI Firmware 3.3 specification, section 4.1.2, ECAM space must be reserved by declaring a motherboard resource, but there's no requirement to mention it in E820. The specification additionally states that: the resources can optionally be returned in Int15 E820h or EFIGetMemoryMap as reserved memory.

This fixes an issue with detection of extended capabilities when running Xen on a Lenovo system that doesn't list the MCFG area as an EfiMemoryMappedIO region (or as any other type) in the EFI memory map.

The more strict logic was introduced in Linux in 2006 as 946f2ee5c731 ("[PATCH] i386/x86-64: Check that MCFG points to an e820 reserved area"). This was picked up by Xen when MCFG support was added in 3b35911d709e ("Enable pci mmcfg and ATS for x86_64"). Apply an approach similar to what Linux has done in 199f968f1484 ("x86/pci: Skip early E820 check for ECAM region") and relax the strict reserved region checking so it's only done for firmware manufactured prior to 2016. For firmware from 2016 and newer allow MCFG region to reside in holes on the memory map. Note Xen is still more strict than Linux however, as it will refuse to use MCFG regions that overlap with memory map regions different than reserved.

Quoted 6 lines
Quoted 4 lines

When dom0 boots it can prevent access to misconfigured MCFG regions by using the PHYSDEVOP_pci_mmcfg_reserved hypercall. This brings Xen's early usage of MCFG (prior to ACPI AML parsing) more in line with the implementation in Linux.

Don't we want to add Fixes: or Ammends: here?

Hm, when the code was added to Xen the checks were fine, that's why we keep the existing behavior for firmware < 2016. The change here is to cope with new firmware behavior, so I'm not sure we can reference a commit from 2009 with a fixes tag when the new behavior only applies to firmware > 2015.

Quoted 4 lines
Quoted 1 line

Generally it looks low-risk patch so:

Thanks, Roger.

Reviewed-by on v2 2/2
#6 lore raw

Re: [PATCH for-4.22 v2 2/2] x86/mcfg: relax memory map checks on newer firmware

In reply to Roger Pau Monne
Quoted 9 lines · Roger Pau Monne

Per PCI Firmware 3.3 specification, section 4.1.2, ECAM space must be reserved by declaring a motherboard resource, but there's no requirement to mention it in E820. The specification additionally states that: the resources can optionally be returned in Int15 E820h or EFIGetMemoryMap as reserved memory.

This fixes an issue with detection of extended capabilities when running Xen on a Lenovo system that doesn't list the MCFG area as an EfiMemoryMappedIO region (or as any other type) in the EFI memory map.

This still doesn't quite track properly.

"Recent Lenovo systems have been found to have the MMCFG region in an E820 hole, which looks to be technically spec compliant, but is rejected by Xen."

That then bridges nicely into the subsequent paragraph, perhaps dropping "more strict" too.

Quoted 11 lines

The more strict logic was introduced in Linux in 2006 as 946f2ee5c731 ("[PATCH] i386/x86-64: Check that MCFG points to an e820 reserved area"). This was picked up by Xen when MCFG support was added in 3b35911d709e ("Enable pci mmcfg and ATS for x86_64"). Apply an approach similar to what Linux has done in 199f968f1484 ("x86/pci: Skip early E820 check for ECAM region") and relax the strict reserved region checking so it's only done for firmware manufactured prior to 2016. For firmware from 2016 and newer allow MCFG region to reside in holes on the memory map. Note Xen is still more strict than Linux however, as it will refuse to use MCFG regions that overlap with memory map regions different than reserved.

Quoted 6 lines

When dom0 boots it can prevent access to misconfigured MCFG regions by using the PHYSDEVOP_pci_mmcfg_reserved hypercall. This brings Xen's early usage of MCFG (prior to ACPI AML parsing) more in line with the implementation in Linux.