Skip to contentSkip to conversation

iommu/vt-d: Correctly check format of page table in debugfs

PATCH Applied Lu Baolu 3 participants 6 messages Jul 20, 2019 – Aug 9, 2019 IOMMU

lore
Share & tools
lore

Lu Baolu

author
#0 lore raw

PATCH 1/1 iommu/vt-d: Correctly check format of page table in debugfs

PASID support and enable bit in the context entry isn't the right indicator for the type of tables (legacy or scalable mode). Check the DMA_RTADDR_SMT bit in the root context pointer instead.

Cc Ashok Raj <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Kevin Tian <[email protected]>
Cc Sai Praneeth <[email protected]>
Fixes dd5142ca5d24b ("iommu/vt-d: Add debugfs support to show scalable mode DMAR table internals")
---
 drivers/iommu/intel-iommu-debugfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/intel-iommu-debugfs.c +1 -1
index 73a552914455..e31c3b416351 100644
@@ -235,7 +235,7 @@ static void ctx_tbl_walk(struct seq_file *m, struct intel_iommu *iommu, u16 bus)
 		tbl_wlk.ctx_entry = context;
 		m->private = &tbl_wlk;
 
-		if (pasid_supported(iommu) && is_pasid_enabled(context)) {
+		if (dmar_readq(iommu->reg + DMAR_RTADDR_REG) & DMA_RTADDR_SMT) {
 			pasid_dir_ptr = context->lo & VTD_PAGE_MASK;
 			pasid_dir_size = get_pasid_dir_size(context);
 			pasid_dir_walk(m, pasid_dir_ptr, pasid_dir_size);
Signature · 6 lines
2.17.1

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
#1 lore raw

Hi Allen,

drivers/iommu/intel-iommu-debugfs.c static void ctx_tbl_walk(struct seq_file *m, struct · line 240

Quoted 13 lines
drivers/iommu/intel- renamed from drivers/iommu/intel-iommu-debugfs.c +0 -0

iommu-debugfs.c
index 73a552914455..e31c3b416351 100644

drivers/iommu/intel-iommu-debugfs.c +1 -1
@@ -235,7 +235,7 @@ static void ctx_tbl_walk(struct seq_file *m, struct
 intel_iommu *iommu, u16 bus)
 		tbl_wlk.ctx_entry = context;
 		m->private = &tbl_wlk;
 
-		if (pasid_supported(iommu) && is_pasid_enabled(context)) {
+		if (dmar_readq(iommu->reg + DMAR_RTADDR_REG) &
 DMA_RTADDR_SMT) {

Thanks for adding this, I do believe this is a good addition but I also think that we might need "is_pasid_enabled()" as well. It checks if PASIDE bit in context entry is enabled or not.

I am thinking that even though DMAR might be using scalable root and context table, the entry itself should have PASIDE bit set. Did I miss something here?

And I also think a macro would be better so that it could reused elsewhere (if need be).

Regards,
Sai

author
#2 lore raw

Hi Sai,

Quoted 19 lines · Sai Praneeth

Hi Allen,

drivers/iommu/intel-iommu-debugfs.c static void ctx_tbl_walk(struct seq_file *m, struct · line 240

Quoted 13 lines
drivers/iommu/intel- renamed from drivers/iommu/intel-iommu-debugfs.c +0 -0

iommu-debugfs.c
index 73a552914455..e31c3b416351 100644

drivers/iommu/intel-iommu-debugfs.c +1 -1
@@ -235,7 +235,7 @@ static void ctx_tbl_walk(struct seq_file *m, struct
 intel_iommu *iommu, u16 bus)
  		tbl_wlk.ctx_entry = context;
  		m->private = &tbl_wlk;
 
-		if (pasid_supported(iommu) && is_pasid_enabled(context)) {
+		if (dmar_readq(iommu->reg + DMAR_RTADDR_REG) &
 DMA_RTADDR_SMT) {

Thanks for adding this, I do believe this is a good addition but I also think that we might need "is_pasid_enabled()" as well. It checks if PASIDE bit in context entry is enabled or not.

Quoted 2 lines

I am thinking that even though DMAR might be using scalable root and context table, the entry itself should have PASIDE bit set. Did I miss something here?

No matter the PASIDE bit set or not, IOMMU always uses the scalable mode page table if scalable mode is enabled. If PASIDE is set, requests with PASID will be handled. Otherwise, requests with PASID will be blocked (but request without PASID will always be handled).

We are dumpling the page table of the IOMMU, so we only care about what page table format it is using. Do I understand it right>

Best regards,
Baolu

Quoted 6 lines

And I also think a macro would be better so that it could reused elsewhere (if need be).

Regards,
Sai

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

#3 lore raw
Quoted 25 lines · Sai Praneeth
Quoted 12 lines

Hi Allen,

Quoted 10 lines
(quoted diff) +0 -0

b/drivers/iommu/intel- iommu-debugfs.c index
73a552914455..e31c3b416351 100644

drivers/iommu/intel-iommu-debugfs.c +0 -0
@@ -235,7 +235,7 @@ static void ctx_tbl_walk(struct seq_file *m,
 struct intel_iommu *iommu, u16 bus)
  		tbl_wlk.ctx_entry = context;
  		m->private = &tbl_wlk;
 
Quoted 6 lines

drivers/iommu/intel-iommu-debugfs.c static void ctx_tbl_walk(struct seq_file *m, · line 240

Quoted 3 lines
drivers/iommu/intel-iommu-debugfs.c +1 -1
-		if (pasid_supported(iommu) && is_pasid_enabled(context)) {
+		if (dmar_readq(iommu->reg + DMAR_RTADDR_REG) &
 DMA_RTADDR_SMT) {

Thanks for adding this, I do believe this is a good addition but I also think that we might need "is_pasid_enabled()" as well. It checks if PASIDE

bit in context entry is enabled or not.

Quoted 3 lines

I am thinking that even though DMAR might be using scalable root and context table, the entry itself should have PASIDE bit set. Did I miss something

here?

Quoted 7 lines

No matter the PASIDE bit set or not, IOMMU always uses the scalable mode page table if scalable mode is enabled. If PASIDE is set, requests with PASID will be handled. Otherwise, requests with PASID will be blocked (but request without PASID will always be handled).

We are dumpling the page table of the IOMMU, so we only care about what page table format it is using. Do I understand it right>

Thanks! Baolu, for the explanation. Yes, it makes sense and I agree that we don't need the extra check for PASIDE bit.

I have tested this change on scalable/legacy DMAR's with/without iommu=pt and it works :)

Regards,
Sai
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

author
#4 lore raw
In reply to Lu Baolu

Hi Joerg,

Just a friendly reminder. What do you think of this fix?

Best regards,
Lu Baolu

Quoted 20 lines · Lu Baolu

PASID support and enable bit in the context entry isn't the right indicator for the type of tables (legacy or scalable mode). Check the DMA_RTADDR_SMT bit in the root context pointer instead.

Cc Ashok Raj <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Kevin Tian <[email protected]>
Cc Sai Praneeth <[email protected]>
Fixes dd5142ca5d24b ("iommu/vt-d: Add debugfs support to show scalable mode DMAR table internals")
---
  drivers/iommu/intel-iommu-debugfs.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/intel-iommu-debugfs.c +0 -0
index 73a552914455..e31c3b416351 100644

drivers/iommu/intel-iommu-debugfs.c static void ctx_tbl_walk(struct seq_file *m, struct intel_iommu *iommu, u16 bus) · line 241

Quoted 8 lines
drivers/iommu/intel-iommu-debugfs.c +1 -1
@@ -235,7 +235,7 @@ static void ctx_tbl_walk(struct seq_file *m, struct intel_iommu *iommu, u16 bus)
  		tbl_wlk.ctx_entry = context;
  		m->private = &tbl_wlk;
 
-		if (pasid_supported(iommu) && is_pasid_enabled(context)) {
+		if (dmar_readq(iommu->reg + DMAR_RTADDR_REG) & DMA_RTADDR_SMT) {
  			pasid_dir_ptr = context->lo & VTD_PAGE_MASK;
  			pasid_dir_size = get_pasid_dir_size(context);
  			pasid_dir_walk(m, pasid_dir_ptr, pasid_dir_size);

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

maintainer Applied
#5 lore raw
In reply to Lu Baolu
Quoted 2 lines · Lu Baolu
 drivers/iommu/intel-iommu-debugfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks.