Skip to contentSkip to conversation

iommu:Check that iommu_device_create has completed successfully in alloc_iommu

PATCH In review Nicholas Krause 3 participants 6 messages Dec 29, 2015 – Jan 4, 2016 2 messages not in the archive IOMMU

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iommu:Check that iommu_device_create has completed successfully in alloc_iommu

This adds the proper check to alloc_iommu to make sure that the call to iommu_device_create has completed successfully and if not return to the caller the error code returned after freeing up resources allocated previously by alloc_iommu.

---
 drivers/iommu/dmar.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/iommu/dmar.c +4 -1
index 80e3c17..27333b6 100644
@@ -1069,9 +1069,12 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
 		iommu->iommu_dev = iommu_device_create(NULL, iommu,
 						       intel_iommu_groups,
 						       "%s", iommu->name);
+	if (IS_ERR(iommu->iommu_dev)) {
+		err = PTR_ERR(iommu->iommu_dev);
+		goto err_unmap;
+	}
 
 	return 0;
-
 err_unmap:
 	unmap_iommu(iommu);
 error_free_seq_id:
Signature · 1 line
2.5.0
#1 lore raw

2015-12-29 13:10 GMT+08:00 Nicholas Krause <[email protected]>:

Quoted 15 lines

This adds the proper check to alloc_iommu to make sure that the call to iommu_device_create has completed successfully and if not return to the caller the error code returned after freeing up resources allocated previously by alloc_iommu.

---
 drivers/iommu/dmar.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/iommu/dmar.c +0 -0
index 80e3c17..27333b6 100644

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1075

Quoted 7 lines
drivers/iommu/dmar.c +4 -0
@@ -1069,9 +1069,12 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
                iommu->iommu_dev = iommu_device_create(NULL, iommu,
                                                       intel_iommu_groups,
                                                       "%s", iommu->name);
+       if (IS_ERR(iommu->iommu_dev)) {
+               err = PTR_ERR(iommu->iommu_dev);
+               goto err_unmap;
+       }

If so, will this bad 'iommu->iommu_dev' break your iommu work? It seems not necessary.

Quoted 13 lines
drivers/iommu/dmar.c +0 -2
        return 0;
-
 err_unmap:
        unmap_iommu(iommu);
 error_free_seq_id:
--

2.5.0

_______________________________________________
iommu mailing list iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Signature · 3 lines
---
Vincent Wan(Zongshun)
www.mcuos.com

A message that is not in the archive (sent off-list or not yet received): CAKT61h-n6=enWPAmEkPE36VZvrGFn9_Zj0y4+DdkOB33jyELUw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

author
#3 lore raw
Quoted 11 lines · Wan ZongShun

2015-12-29 13:10 GMT+08:00 Nicholas Krause <[email protected]>:

Quoted 10 lines

This adds the proper check to alloc_iommu to make sure that the call to iommu_device_create has completed successfully and if not return to the caller the error code returned after freeing up resources allocated previously by alloc_iommu.

---
 drivers/iommu/dmar.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
Quoted 16 lines

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1075

Quoted 12 lines
drivers/iommu/dmar.c +4 -0
index 80e3c17..27333b6 100644
@@ -1069,9 +1069,12 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
                iommu->iommu_dev = iommu_device_create(NULL, iommu,
                                                       intel_iommu_groups,
                                                       "%s", iommu->name);
+       if (IS_ERR(iommu->iommu_dev)) {
+               err = PTR_ERR(iommu->iommu_dev);
+               goto err_unmap;
+       }

If so, will this bad 'iommu->iommu_dev' break your iommu work? It seems not necessary.

My concern is that if the iommu device is allocated successfully but not loaded successfully into sysfs does this not make the device useless to some degree and therefore there is no point in having this device being allocated without it's information being exported in sysfs. Again if it's fine without being exported in sysfs that's fine and in that case we should at least print to the console with KERN_CRIT in order to warn of this critical error to the user. Again I took the first approach but if you/the maintainer Joerg Roedel feel that the second idea is better I will be glad to send a patch doing that :). There are also other callers doing this as I found here so depending on how we solve this one I will do the others in a patch series:
http://lxr.free-electrons.com/ident?i=iommu_device_create
Hopes that this explains the patch,
Nick

Quoted 16 lines
Quoted 13 lines
drivers/iommu/dmar.c +0 -2
        return 0;
-
 err_unmap:
        unmap_iommu(iommu);
 error_free_seq_id:
--

2.5.0

_______________________________________________
iommu mailing list iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu

maintainer
#4 lore raw
In reply to Nicholas Krause
Quoted 15 lines · Nicholas Krause

This adds the proper check to alloc_iommu to make sure that the call to iommu_device_create has completed successfully and if not return to the caller the error code returned after freeing up resources allocated previously by alloc_iommu.

---
 drivers/iommu/dmar.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/iommu/dmar.c +0 -0
index 80e3c17..27333b6 100644

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1075

Quoted 7 lines
drivers/iommu/dmar.c +4 -0
@@ -1069,9 +1069,12 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
 		iommu->iommu_dev = iommu_device_create(NULL, iommu,
 						       intel_iommu_groups,
 						       "%s", iommu->name);
+	if (IS_ERR(iommu->iommu_dev)) {
+		err = PTR_ERR(iommu->iommu_dev);
+		goto err_unmap;
+	}

This leaves a dangling pointer to the iommu in drhd->iommu.

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1074

Quoted 3 lines
drivers/iommu/dmar.c +0 -1
 	return 0;
-

That blank line improves readability of the code, please leave it in.

Quoted 3 lines
drivers/iommu/dmar.c +0 -0
 err_unmap:
 	unmap_iommu(iommu);
 error_free_seq_id:

A message that is not in the archive (sent off-list or not yet received): [email protected]

#6 lore raw
Quoted 17 lines · Joerg Roedel

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit

Quoted 15 lines · Nicholas Krause

This adds the proper check to alloc_iommu to make sure that the call to iommu_device_create has completed successfully and if not return to the caller the error code returned after freeing up resources allocated previously by alloc_iommu.

---
 drivers/iommu/dmar.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/iommu/dmar.c +0 -0
index 80e3c17..27333b6 100644
@@ -1069,9 +1069,12 @@ static int alloc_iommu(struct dmar_drhd_unit

*drhd)

Quoted 9 lines

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1075

Quoted 7 lines
drivers/iommu/dmar.c +4 -0
 		iommu->iommu_dev = iommu_device_create(NULL, iommu,
 						       intel_iommu_groups,
 						       "%s", iommu->name);
+	if (IS_ERR(iommu->iommu_dev)) {
+		err = PTR_ERR(iommu->iommu_dev);
+		goto err_unmap;
+	}

This leaves a dangling pointer to the iommu in drhd->iommu.

Yes that's true and was wondering if you want me to fix it up to set the pointer back to null first or if preferred around safe state.
Nick

Quoted 10 lines

drivers/iommu/dmar.c static int alloc_iommu(struct dmar_drhd_unit *drhd) · line 1074

Quoted 4 lines
drivers/iommu/dmar.c +0 -1
 	return 0;
-

That blank line improves readability of the code, please leave it in.

Quoted 3 lines
drivers/iommu/dmar.c +0 -0
 err_unmap:
 	unmap_iommu(iommu);
 error_free_seq_id:
maintainer
#7 lore raw
Quoted 2 lines · Nicholas Krause

Yes that's true and was wondering if you want me to fix it up to set the pointer back to null first or if preferred around safe state.

The best solution is to set the dmar->iommu pointer after the error check for iommu_device_create().

Joerg