* [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan
@ 2015-12-14 16:31 Prem Mallappa
[not found] ` <1450110687-32207-1-git-send-email-pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
0 siblings, 1 reply; 6+ messages in thread
From: Prem Mallappa @ 2015-12-14 16:31 UTC (permalink / raw)
To: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA,
will.deacon-5wv7dgnIgG8,
linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r
Cc: jchandra-dY08KVG/lbpWk0Htik3J/w, Prem Mallappa
Vulcan SMMUv3 looks for AARCH64 and S2PS inroder to validate the STE entry,
which is a overkill, but when proper encoding not found; the SMMU stops
processing PCIe read/write requests. Giving the h/w what it wants.
Signed-off-by: Prem Mallappa <pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/iommu/arm-smmu-v3.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c
index d4af50d..dfda564 100644
--- a/drivers/iommu/arm-smmu-v3.c
+++ b/drivers/iommu/arm-smmu-v3.c
@@ -260,6 +260,7 @@
#define STRTAB_STE_2_S2VMID_MASK 0xffffUL
#define STRTAB_STE_2_VTCR_SHIFT 32
#define STRTAB_STE_2_VTCR_MASK 0x7ffffUL
+#define STRTAB_STE_2_S2PS_SHIFT 48
#define STRTAB_STE_2_S2AA64 (1UL << 51)
#define STRTAB_STE_2_S2ENDI (1UL << 52)
#define STRTAB_STE_2_S2PTW (1UL << 54)
@@ -577,6 +578,7 @@ struct arm_smmu_device {
u32 features;
#define ARM_SMMU_OPT_SKIP_PREFETCH (1 << 0)
+#define ARM_SMMU_OPT_BROKEN_STE_VALID (1 << 1)
u32 options;
struct arm_smmu_cmdq cmdq;
@@ -641,6 +643,7 @@ struct arm_smmu_option_prop {
static struct arm_smmu_option_prop arm_smmu_options[] = {
{ ARM_SMMU_OPT_SKIP_PREFETCH, "hisilicon,broken-prefetch-cmd" },
+ { ARM_SMMU_OPT_BROKEN_STE_VALID, "broadcom,broken-ste-valid-check" },
{ 0, NULL},
};
@@ -1046,6 +1049,15 @@ static void arm_smmu_write_strtab_ent(struct arm_smmu_device *smmu, u32 sid,
: STRTAB_STE_0_CFG_BYPASS;
dst[0] = cpu_to_le64(val);
dst[2] = 0; /* Nuke the VMID */
+
+ if (smmu && (smmu->options & ARM_SMMU_OPT_BROKEN_STE_VALID)) {
+#define SMMU_STE_OAS_44_BITS 0x4UL
+ WARN_ON(smmu->oas != 44);
+ dst[1] = STRTAB_STE_1_EATS_TRANS << STRTAB_STE_1_EATS_SHIFT;
+ dst[2] |= cpu_to_le64((SMMU_STE_OAS_44_BITS << STRTAB_STE_2_S2PS_SHIFT) |
+ STRTAB_STE_2_S2AA64);
+ }
+
if (ste_live)
arm_smmu_sync_ste_for_sid(smmu, sid);
return;
--
2.6.0
^ permalink raw reply related [flat|nested] 6+ messages in thread[parent not found: <1450110687-32207-1-git-send-email-pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>]
* Re: [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan [not found] ` <1450110687-32207-1-git-send-email-pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org> @ 2015-12-15 13:40 ` Will Deacon [not found] ` <20151215134010.GH9452-5wv7dgnIgG8@public.gmane.org> 0 siblings, 1 reply; 6+ messages in thread From: Will Deacon @ 2015-12-15 13:40 UTC (permalink / raw) To: Prem Mallappa Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA, jchandra-dY08KVG/lbpWk0Htik3J/w, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r Hi Prem, On Mon, Dec 14, 2015 at 10:01:27PM +0530, Prem Mallappa wrote: > Vulcan SMMUv3 looks for AARCH64 and S2PS inroder to validate the STE entry, 'inroder' ? > which is a overkill, but when proper encoding not found; the SMMU stops > processing PCIe read/write requests. Giving the h/w what it wants. > > Signed-off-by: Prem Mallappa <pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org> > --- > drivers/iommu/arm-smmu-v3.c | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index d4af50d..dfda564 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -260,6 +260,7 @@ > #define STRTAB_STE_2_S2VMID_MASK 0xffffUL > #define STRTAB_STE_2_VTCR_SHIFT 32 > #define STRTAB_STE_2_VTCR_MASK 0x7ffffUL > +#define STRTAB_STE_2_S2PS_SHIFT 48 > #define STRTAB_STE_2_S2AA64 (1UL << 51) > #define STRTAB_STE_2_S2ENDI (1UL << 52) > #define STRTAB_STE_2_S2PTW (1UL << 54) > @@ -577,6 +578,7 @@ struct arm_smmu_device { > u32 features; > > #define ARM_SMMU_OPT_SKIP_PREFETCH (1 << 0) > +#define ARM_SMMU_OPT_BROKEN_STE_VALID (1 << 1) > u32 options; > > struct arm_smmu_cmdq cmdq; > @@ -641,6 +643,7 @@ struct arm_smmu_option_prop { > > static struct arm_smmu_option_prop arm_smmu_options[] = { > { ARM_SMMU_OPT_SKIP_PREFETCH, "hisilicon,broken-prefetch-cmd" }, > + { ARM_SMMU_OPT_BROKEN_STE_VALID, "broadcom,broken-ste-valid-check" }, This looks like a more specific problem than "broken ste valid check". Maybe we could call the prop ARM_SMMU_OPT_BCOM_BROKEN_STE_VALID? You also need to update the devicetree binding in Documentation/devicetree/bindings/iommu/arm,smmu-v3.txt. > { 0, NULL}, > }; > > @@ -1046,6 +1049,15 @@ static void arm_smmu_write_strtab_ent(struct arm_smmu_device *smmu, u32 sid, > : STRTAB_STE_0_CFG_BYPASS; > dst[0] = cpu_to_le64(val); > dst[2] = 0; /* Nuke the VMID */ > + > + if (smmu && (smmu->options & ARM_SMMU_OPT_BROKEN_STE_VALID)) { > +#define SMMU_STE_OAS_44_BITS 0x4UL Please don't add a #define here. Can we instead use the oas field that we've extracted from IDR5? I think we need to be doing that anyway when we're using stage-2 translation, looking at the spec... > + WARN_ON(smmu->oas != 44); > + dst[1] = STRTAB_STE_1_EATS_TRANS << STRTAB_STE_1_EATS_SHIFT; Why are you enabling ATS? > + dst[2] |= cpu_to_le64((SMMU_STE_OAS_44_BITS << STRTAB_STE_2_S2PS_SHIFT) | > + STRTAB_STE_2_S2AA64); > + } Will ^ permalink raw reply [flat|nested] 6+ messages in thread
[parent not found: <20151215134010.GH9452-5wv7dgnIgG8@public.gmane.org>]
* Re: [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan [not found] ` <20151215134010.GH9452-5wv7dgnIgG8@public.gmane.org> @ 2015-12-15 15:48 ` Will Deacon 2015-12-16 4:37 ` Prem (Premachandra) Mallappa 1 sibling, 0 replies; 6+ messages in thread From: Will Deacon @ 2015-12-15 15:48 UTC (permalink / raw) To: Prem Mallappa Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA, jchandra-dY08KVG/lbpWk0Htik3J/w, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r On Tue, Dec 15, 2015 at 01:40:10PM +0000, Will Deacon wrote: > On Mon, Dec 14, 2015 at 10:01:27PM +0530, Prem Mallappa wrote: > > @@ -1046,6 +1049,15 @@ static void arm_smmu_write_strtab_ent(struct arm_smmu_device *smmu, u32 sid, > > : STRTAB_STE_0_CFG_BYPASS; > > dst[0] = cpu_to_le64(val); > > dst[2] = 0; /* Nuke the VMID */ > > + > > + if (smmu && (smmu->options & ARM_SMMU_OPT_BROKEN_STE_VALID)) { > > +#define SMMU_STE_OAS_44_BITS 0x4UL > > Please don't add a #define here. Can we instead use the oas field that > we've extracted from IDR5? I think we need to be doing that anyway when > we're using stage-2 translation, looking at the spec... ... which we already are doing, thanks to the vtcr initialisation in the io-pgtable code, which populates the PS field based on the OAS. Will ^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan [not found] ` <20151215134010.GH9452-5wv7dgnIgG8@public.gmane.org> 2015-12-15 15:48 ` Will Deacon @ 2015-12-16 4:37 ` Prem (Premachandra) Mallappa [not found] ` <CC0BC9CB4553B146B290EC1D4DC5C35901094BCE-HXj2mutaA2rOzKo2G4nwXJr/X4hKkxxPpWgKQ6/u3Fg@public.gmane.org> 1 sibling, 1 reply; 6+ messages in thread From: Prem (Premachandra) Mallappa @ 2015-12-16 4:37 UTC (permalink / raw) To: Will Deacon Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org, Jayachandran Chandrashekaran Nair, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org Hi Will, Thanks for the review. > On Mon, Dec 14, 2015 at 10:01:27PM +0530, Prem Mallappa wrote: > > Vulcan SMMUv3 looks for AARCH64 and S2PS inroder to validate the STE > > entry, > > 'inroder' ? > In order > > which is a overkill, but when proper encoding not found; the SMMU > > stops processing PCIe read/write requests. Giving the h/w what it wants. > > > > Signed-off-by: Prem Mallappa <pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org> > > --- > > drivers/iommu/arm-smmu-v3.c | 12 ++++++++++++ > > 1 file changed, 12 insertions(+) > > > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu- > v3.c > > index d4af50d..dfda564 100644 > > --- a/drivers/iommu/arm-smmu-v3.c > > +++ b/drivers/iommu/arm-smmu-v3.c > > @@ -260,6 +260,7 @@ > > #define STRTAB_STE_2_S2VMID_MASK 0xffffUL > > #define STRTAB_STE_2_VTCR_SHIFT 32 > > #define STRTAB_STE_2_VTCR_MASK 0x7ffffUL > > +#define STRTAB_STE_2_S2PS_SHIFT 48 > > #define STRTAB_STE_2_S2AA64 (1UL << 51) > > #define STRTAB_STE_2_S2ENDI (1UL << 52) > > #define STRTAB_STE_2_S2PTW (1UL << 54) > > @@ -577,6 +578,7 @@ struct arm_smmu_device { > > u32 features; > > > > #define ARM_SMMU_OPT_SKIP_PREFETCH (1 << 0) > > +#define ARM_SMMU_OPT_BROKEN_STE_VALID (1 << 1) > > u32 options; > > > > struct arm_smmu_cmdq cmdq; > > @@ -641,6 +643,7 @@ struct arm_smmu_option_prop { > > > > static struct arm_smmu_option_prop arm_smmu_options[] = { > > { ARM_SMMU_OPT_SKIP_PREFETCH, "hisilicon,broken-prefetch- > cmd" }, > > + { ARM_SMMU_OPT_BROKEN_STE_VALID, "broadcom,broken-ste- > valid-check" > > +}, > > This looks like a more specific problem than "broken ste valid check". > Maybe we could call the prop > ARM_SMMU_OPT_BCOM_BROKEN_STE_VALID? > > You also need to update the devicetree binding in > Documentation/devicetree/bindings/iommu/arm,smmu-v3.txt. > Sure I'll send out v2. > > { 0, NULL}, > > }; > > > > @@ -1046,6 +1049,15 @@ static void arm_smmu_write_strtab_ent(struct > arm_smmu_device *smmu, u32 sid, > > : STRTAB_STE_0_CFG_BYPASS; > > dst[0] = cpu_to_le64(val); > > dst[2] = 0; /* Nuke the VMID */ > > + > > + if (smmu && (smmu->options & > ARM_SMMU_OPT_BROKEN_STE_VALID)) { > > +#define SMMU_STE_OAS_44_BITS 0x4UL > > Please don't add a #define here. Can we instead use the oas field that we've > extracted from IDR5? I think we need to be doing that anyway when we're > using stage-2 translation, looking at the spec... > As you mentioned in other email, "OAS" is already taken care by io-pgtable vtcr. > > + WARN_ON(smmu->oas != 44); > > + dst[1] = STRTAB_STE_1_EATS_TRANS << > STRTAB_STE_1_EATS_SHIFT; > > Why are you enabling ATS? > Again, based on feedback from h/w team which expects these fields populated (hence broken :) ), I am not aware of the h/w doing anything beyond checking this for STE.VALID() (i.o.w, if STE is illegal). /Prem ^ permalink raw reply [flat|nested] 6+ messages in thread
[parent not found: <CC0BC9CB4553B146B290EC1D4DC5C35901094BCE-HXj2mutaA2rOzKo2G4nwXJr/X4hKkxxPpWgKQ6/u3Fg@public.gmane.org>]
* Re: [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan [not found] ` <CC0BC9CB4553B146B290EC1D4DC5C35901094BCE-HXj2mutaA2rOzKo2G4nwXJr/X4hKkxxPpWgKQ6/u3Fg@public.gmane.org> @ 2015-12-16 10:17 ` Will Deacon [not found] ` <20151216101735.GB4308-5wv7dgnIgG8@public.gmane.org> 0 siblings, 1 reply; 6+ messages in thread From: Will Deacon @ 2015-12-16 10:17 UTC (permalink / raw) To: Prem (Premachandra) Mallappa Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org, Jayachandran Chandrashekaran Nair, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org On Wed, Dec 16, 2015 at 04:37:07AM +0000, Prem (Premachandra) Mallappa wrote: > > On Mon, Dec 14, 2015 at 10:01:27PM +0530, Prem Mallappa wrote: > > > Vulcan SMMUv3 looks for AARCH64 and S2PS inroder to validate the STE > > > entry, > > > > 'inroder' ? > > > In order Got it! > > > which is a overkill, but when proper encoding not found; the SMMU > > > stops processing PCIe read/write requests. Giving the h/w what it wants. > > > > > > Signed-off-by: Prem Mallappa <pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org> > > > --- > > > drivers/iommu/arm-smmu-v3.c | 12 ++++++++++++ > > > 1 file changed, 12 insertions(+) > > > > > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu- > > v3.c > > > index d4af50d..dfda564 100644 > > > --- a/drivers/iommu/arm-smmu-v3.c > > > +++ b/drivers/iommu/arm-smmu-v3.c > > > @@ -260,6 +260,7 @@ > > > #define STRTAB_STE_2_S2VMID_MASK 0xffffUL > > > #define STRTAB_STE_2_VTCR_SHIFT 32 > > > #define STRTAB_STE_2_VTCR_MASK 0x7ffffUL > > > +#define STRTAB_STE_2_S2PS_SHIFT 48 > > > #define STRTAB_STE_2_S2AA64 (1UL << 51) > > > #define STRTAB_STE_2_S2ENDI (1UL << 52) > > > #define STRTAB_STE_2_S2PTW (1UL << 54) > > > @@ -577,6 +578,7 @@ struct arm_smmu_device { > > > u32 features; > > > > > > #define ARM_SMMU_OPT_SKIP_PREFETCH (1 << 0) > > > +#define ARM_SMMU_OPT_BROKEN_STE_VALID (1 << 1) > > > u32 options; > > > > > > struct arm_smmu_cmdq cmdq; > > > @@ -641,6 +643,7 @@ struct arm_smmu_option_prop { > > > > > > static struct arm_smmu_option_prop arm_smmu_options[] = { > > > { ARM_SMMU_OPT_SKIP_PREFETCH, "hisilicon,broken-prefetch- > > cmd" }, > > > + { ARM_SMMU_OPT_BROKEN_STE_VALID, "broadcom,broken-ste- > > valid-check" > > > +}, > > > > This looks like a more specific problem than "broken ste valid check". > > Maybe we could call the prop > > ARM_SMMU_OPT_BCOM_BROKEN_STE_VALID? > > > > You also need to update the devicetree binding in > > Documentation/devicetree/bindings/iommu/arm,smmu-v3.txt. > > > Sure I'll send out v2. > > > > { 0, NULL}, > > > }; > > > > > > @@ -1046,6 +1049,15 @@ static void arm_smmu_write_strtab_ent(struct > > arm_smmu_device *smmu, u32 sid, > > > : STRTAB_STE_0_CFG_BYPASS; > > > dst[0] = cpu_to_le64(val); > > > dst[2] = 0; /* Nuke the VMID */ > > > + > > > + if (smmu && (smmu->options & > > ARM_SMMU_OPT_BROKEN_STE_VALID)) { > > > +#define SMMU_STE_OAS_44_BITS 0x4UL > > > > Please don't add a #define here. Can we instead use the oas field that we've > > extracted from IDR5? I think we need to be doing that anyway when we're > > using stage-2 translation, looking at the spec... > > > As you mentioned in other email, "OAS" is already taken care by io-pgtable > vtcr. Ok -- so what is the value reported by your IDR5? > > > + WARN_ON(smmu->oas != 44); > > > + dst[1] = STRTAB_STE_1_EATS_TRANS << > > STRTAB_STE_1_EATS_SHIFT; > > > > Why are you enabling ATS? > > > Again, based on feedback from h/w team which expects these fields > populated (hence broken :) ), Ok, I just didn't see any mention of that in the commit message and wanted to make sure it wasn't an accidental unrelated change. Will ^ permalink raw reply [flat|nested] 6+ messages in thread
[parent not found: <20151216101735.GB4308-5wv7dgnIgG8@public.gmane.org>]
* RE: [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan [not found] ` <20151216101735.GB4308-5wv7dgnIgG8@public.gmane.org> @ 2015-12-16 10:49 ` Prem (Premachandra) Mallappa 0 siblings, 0 replies; 6+ messages in thread From: Prem (Premachandra) Mallappa @ 2015-12-16 10:49 UTC (permalink / raw) To: Will Deacon Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org, Jayachandran Chandrashekaran Nair, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org > Ok -- so what is the value reported by your IDR5? Our IDR5, OAS=0x4 (44bits) > > Ok, I just didn't see any mention of that in the commit message and wanted > to make sure it wasn't an accidental unrelated change. > Got it. /Prem ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2015-12-16 10:49 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-12-14 16:31 [PATCH] IOMMU: arm-smmu-v3: fix broken S2PS and AARCH64 in Broadcom Vulcan Prem Mallappa
[not found] ` <1450110687-32207-1-git-send-email-pmallapp-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
2015-12-15 13:40 ` Will Deacon
[not found] ` <20151215134010.GH9452-5wv7dgnIgG8@public.gmane.org>
2015-12-15 15:48 ` Will Deacon
2015-12-16 4:37 ` Prem (Premachandra) Mallappa
[not found] ` <CC0BC9CB4553B146B290EC1D4DC5C35901094BCE-HXj2mutaA2rOzKo2G4nwXJr/X4hKkxxPpWgKQ6/u3Fg@public.gmane.org>
2015-12-16 10:17 ` Will Deacon
[not found] ` <20151216101735.GB4308-5wv7dgnIgG8@public.gmane.org>
2015-12-16 10:49 ` Prem (Premachandra) Mallappa
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox