Revert "iommu/vt-d: Clean up pasid quirk for pre-production devices"
PATCH 1/1 Revert "iommu/vt-d: Clean up pasid quirk for pre-production devices"
This reverts commit ab96746aaa344fb720a198245a837e266fad3b62.
The commit ab96746aaa34 ("iommu/vt-d: Clean up pasid quirk for pre-production devices") triggers ECS mode on some platforms which have broken ECS support. As the result, graphic device will be inoperable on boot.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107017
--- drivers/iommu/intel-iommu.c | 32 ++++++++++++++++++++++++++++++-- include/linux/intel-iommu.h | 1 + 2 files changed, 31 insertions(+), 2 deletions(-)
drivers/iommu/intel-iommu.c +30 -2
@@ -484,14 +484,37 @@ static int dmar_forcedac;static int intel_iommu_strict;static int intel_iommu_superpage =1 ;static int intel_iommu_ecs =1 ; +static int intel_iommu_pasid28;static int iommu_identity_mapping;#define IDENTMAP_ALL 1#define IDENTMAP_GFX 2#define IDENTMAP_AZALIA 4 -#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap)) -#define pasid_enabled (iommu) (ecs_enabled(iommu) && ecap_pasid(iommu->ecap)) +/* Broadwell and Skylake have broken ECS support — normal so-called "second +* level" translation of DMA requests-without-PASID doesn't actually happen +* unless you also set the NESTE bit in an extended context-entry. Which of +* course means that SVM doesn't work because it's trying to do nested +* translation of the physical addresses it finds in the process page tables, +* through the IOVA->phys mapping found in the "second level" page tables. +* +* The VT-d specification was retroactively changed to change the definition +* of the capability bits and pretend that Broadwell/Skylake never happened... +* but unfortunately the wrong bit was changed. It's ECS which is broken, but +* for some reason it was the PASID capability bit which was redefined (from +* bit 28 on BDW/SKL to bit 40 in future). +* +* So our test for ECS needs to eschew those implementations which set the old +* PASID capabiity bit 28, since those are the ones on which ECS is broken. +* Unless we are working around the 'pasid28' limitations, that is, by putting +* the device into passthrough mode for normal DMA and thus masking the bug. +*/ +#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap) && \ + (intel_iommu_pasid28 || !ecap_broken_pasid(iommu->ecap))) +/* PASID support is thus enabled if ECS is enabled and *either* of the old +* or new capability bits are set. */ +#define pasid_enabled (iommu) (ecs_enabled(iommu) && \ + (ecap_pasid(iommu->ecap) || ecap_broken_pasid(iommu->ecap)))int intel_iommu_gfx_mapped;EXPORT_SYMBOL_GPL (intel_iommu_gfx_mapped);
@@ -554,6 +577,11 @@ static int __init intel_iommu_setup(char *str)printk (KERN_INFO "Intel-IOMMU: disable extended context table support\n" ); intel_iommu_ecs =0 ; + }else if (!strncmp (str,"pasid28" ,7 )) { +printk (KERN_INFO +"Intel-IOMMU: enable pre-production PASID support\n" ); + intel_iommu_pasid28 =1 ; + iommu_identity_mapping |=IDENTMAP_GFX ; }else if (!strncmp (str,"tboot_noforce" ,13 )) {printk (KERN_INFO "Intel-IOMMU: not forcing on after tboot. This could expose security risk for tboot\n" );
include/linux/intel-iommu.h +1 -0
@@ -121,6 +121,7 @@#define ecap_srs (e) ((e >> 31) & 0x1)#define ecap_ers (e) ((e >> 30) & 0x1)#define ecap_prs (e) ((e >> 29) & 0x1) +#define ecap_broken_pasid (e) ((e >> 28) & 0x1)#define ecap_dis (e) ((e >> 27) & 0x1)#define ecap_nest (e) ((e >> 26) & 0x1)#define ecap_mts (e) ((e >> 25) & 0x1)
Signature · 6 lines
2.7.4 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
Hi Joerg,
The graphic guys are looking forward to having this in 4.18. Is it possible to take it in the following rcs?
Best regards,
Lu Baolu
Quoted 77 lines · Lu Baolu
This reverts commit
ab96746aaa344fb720a198245a837e266fad3b62.The commit
ab96746aaa34("iommu/vt-d: Clean up pasid quirk for pre-production devices") triggers ECS mode on some platforms which have broken ECS support. As the result, graphic device will be inoperable on boot.Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107017
Cc Ashok Raj <[email protected]>Signed-off-by Lu Baolu <[email protected]>--- drivers/iommu/intel-iommu.c | 32 ++++++++++++++++++++++++++++++-- include/linux/intel-iommu.h | 1 + 2 files changed, 31 insertions(+), 2 deletions(-)drivers/iommu/intel-iommu.c +30
-2@@ -484,14 +484,37 @@ static int dmar_forcedac;static int intel_iommu_strict;static int intel_iommu_superpage =1 ;static int intel_iommu_ecs =1 ; +static int intel_iommu_pasid28;static int iommu_identity_mapping;#define IDENTMAP_ALL 1#define IDENTMAP_GFX 2#define IDENTMAP_AZALIA 4 -#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap)) -#define pasid_enabled (iommu) (ecs_enabled(iommu) && ecap_pasid(iommu->ecap)) +/* Broadwell and Skylake have broken ECS support — normal so-called "second +* level" translation of DMA requests-without-PASID doesn't actually happen +* unless you also set the NESTE bit in an extended context-entry. Which of +* course means that SVM doesn't work because it's trying to do nested +* translation of the physical addresses it finds in the process page tables, +* through the IOVA->phys mapping found in the "second level" page tables. +* +* The VT-d specification was retroactively changed to change the definition +* of the capability bits and pretend that Broadwell/Skylake never happened... +* but unfortunately the wrong bit was changed. It's ECS which is broken, but +* for some reason it was the PASID capability bit which was redefined (from +* bit 28 on BDW/SKL to bit 40 in future). +* +* So our test for ECS needs to eschew those implementations which set the old +* PASID capabiity bit 28, since those are the ones on which ECS is broken. +* Unless we are working around the 'pasid28' limitations, that is, by putting +* the device into passthrough mode for normal DMA and thus masking the bug. +*/ +#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap) && \ + (intel_iommu_pasid28 || !ecap_broken_pasid(iommu->ecap))) +/* PASID support is thus enabled if ECS is enabled and *either* of the old +* or new capability bits are set. */ +#define pasid_enabled (iommu) (ecs_enabled(iommu) && \ + (ecap_pasid(iommu->ecap) || ecap_broken_pasid(iommu->ecap)))int intel_iommu_gfx_mapped;EXPORT_SYMBOL_GPL (intel_iommu_gfx_mapped);@@ -554,6 +577,11 @@ static int __init intel_iommu_setup(char *str)printk (KERN_INFO "Intel-IOMMU: disable extended context table support\n" ); intel_iommu_ecs =0 ; + }else if (!strncmp (str,"pasid28" ,7 )) { +printk (KERN_INFO +"Intel-IOMMU: enable pre-production PASID support\n" ); + intel_iommu_pasid28 =1 ; + iommu_identity_mapping |=IDENTMAP_GFX ; }else if (!strncmp (str,"tboot_noforce" ,13 )) {printk (KERN_INFO "Intel-IOMMU: not forcing on after tboot. This could expose security risk for tboot\n" );include/linux/intel-iommu.h +0
-0
include/linux/intel-iommu.h line 127
Quoted 7 lines
include/linux/intel-iommu.h +1
-0@@ -121,6 +121,7 @@#define ecap_srs (e) ((e >> 31) & 0x1)#define ecap_ers (e) ((e >> 30) & 0x1)#define ecap_prs (e) ((e >> 29) & 0x1) +#define ecap_broken_pasid (e) ((e >> 28) & 0x1)#define ecap_dis (e) ((e >> 27) & 0x1)#define ecap_nest (e) ((e >> 26) & 0x1)#define ecap_mts (e) ((e >> 25) & 0x1)
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
Zhenyu Wang
Tested-byQuoted 5 lines · Lu Baolu
Hi Joerg,
The graphic guys are looking forward to having this in 4.18. Is it possible to take it in the following rcs?
This breakes intel gfx driver in 4.18 when gfx dmar is on. Please include this fix ASAP.
thanks!
Quoted 87 lines · Lu Baolu
Quoted 84 lines
This reverts commit
ab96746aaa344fb720a198245a837e266fad3b62.The commit
ab96746aaa34("iommu/vt-d: Clean up pasid quirk for pre-production devices") triggers ECS mode on some platforms which have broken ECS support. As the result, graphic device will be inoperable on boot.Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107017
Cc Ashok Raj <[email protected]>--- drivers/iommu/intel-iommu.c | 32 ++++++++++++++++++++++++++++++-- include/linux/intel-iommu.h | 1 + 2 files changed, 31 insertions(+), 2 deletions(-)drivers/iommu/intel-iommu.c +30
-2@@ -484,14 +484,37 @@ static int dmar_forcedac;static int intel_iommu_strict;static int intel_iommu_superpage =1 ;static int intel_iommu_ecs =1 ; +static int intel_iommu_pasid28;static int iommu_identity_mapping;#define IDENTMAP_ALL 1#define IDENTMAP_GFX 2#define IDENTMAP_AZALIA 4 -#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap)) -#define pasid_enabled (iommu) (ecs_enabled(iommu) && ecap_pasid(iommu->ecap)) +/* Broadwell and Skylake have broken ECS support — normal so-called "second +* level" translation of DMA requests-without-PASID doesn't actually happen +* unless you also set the NESTE bit in an extended context-entry. Which of +* course means that SVM doesn't work because it's trying to do nested +* translation of the physical addresses it finds in the process page tables, +* through the IOVA->phys mapping found in the "second level" page tables. +* +* The VT-d specification was retroactively changed to change the definition +* of the capability bits and pretend that Broadwell/Skylake never happened... +* but unfortunately the wrong bit was changed. It's ECS which is broken, but +* for some reason it was the PASID capability bit which was redefined (from +* bit 28 on BDW/SKL to bit 40 in future). +* +* So our test for ECS needs to eschew those implementations which set the old +* PASID capabiity bit 28, since those are the ones on which ECS is broken. +* Unless we are working around the 'pasid28' limitations, that is, by putting +* the device into passthrough mode for normal DMA and thus masking the bug. +*/ +#define ecs_enabled (iommu) (intel_iommu_ecs && ecap_ecs(iommu->ecap) && \ + (intel_iommu_pasid28 || !ecap_broken_pasid(iommu->ecap))) +/* PASID support is thus enabled if ECS is enabled and *either* of the old +* or new capability bits are set. */ +#define pasid_enabled (iommu) (ecs_enabled(iommu) && \ + (ecap_pasid(iommu->ecap) || ecap_broken_pasid(iommu->ecap)))int intel_iommu_gfx_mapped;EXPORT_SYMBOL_GPL (intel_iommu_gfx_mapped);@@ -554,6 +577,11 @@ static int __init intel_iommu_setup(char *str)printk (KERN_INFO "Intel-IOMMU: disable extended context table support\n" ); intel_iommu_ecs =0 ; + }else if (!strncmp (str,"pasid28" ,7 )) { +printk (KERN_INFO +"Intel-IOMMU: enable pre-production PASID support\n" ); + intel_iommu_pasid28 =1 ; + iommu_identity_mapping |=IDENTMAP_GFX ; }else if (!strncmp (str,"tboot_noforce" ,13 )) {printk (KERN_INFO "Intel-IOMMU: not forcing on after tboot. This could expose security risk for tboot\n" );include/linux/intel-iommu.h +1
-0@@ -121,6 +121,7 @@#define ecap_srs (e) ((e >> 31) & 0x1)#define ecap_ers (e) ((e >> 30) & 0x1)#define ecap_prs (e) ((e >> 29) & 0x1) +#define ecap_broken_pasid (e) ((e >> 28) & 0x1)#define ecap_dis (e) ((e >> 27) & 0x1)#define ecap_nest (e) ((e >> 26) & 0x1)#define ecap_mts (e) ((e >> 25) & 0x1)
Signature · 3 lines
Open Source Technology Center, Intel ltd. $gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827
Quoted 7 lines · Lu Baolu
This reverts commit
ab96746aaa344fb720a198245a837e266fad3b62.The commit
ab96746aaa34("iommu/vt-d: Clean up pasid quirk for pre-production devices") triggers ECS mode on some platforms which have broken ECS support. As the result, graphic device will be inoperable on boot.
Quoted 8 lines
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107017
Cc Ashok Raj <[email protected]>Signed-off-by Lu Baolu <[email protected]>--- drivers/iommu/intel-iommu.c | 32 ++++++++++++++++++++++++++++++-- include/linux/intel-iommu.h | 1 + 2 files changed, 31 insertions(+), 2 deletions(-)
Applied to iommu/fixes.