Two enhancements to iommu_at[de]tach_device_pasid()
Subsystems (4)
v2 cover Two enhancements to iommu_at[de]tach_device_pasid()
There are minor mistakes in the iommu set_dev_pasid() and remove_dev_pasid() paths. The set_dev_pasid() path updates the group->pasid_array first, and then call into remove_dev_pasid() in error handling when there are devices within the group that failed to set_dev_pasid. The remove_dev_pasid() callbacks of the underlying iommu drivers get the domain for pasid from the group->pasid_array. So the remove_dev_pasid() callback may get a wrong domain in the set_dev_pasid() path. [1] Even if the group is singleton, the existing code logic would have unnecessary warnings in the error handling of the set_dev_pasid() path. e.g. intel iommu driver.
The above issue can be fixed by improving the error handling in the set_dev_pasid() path. Also, this reminds that it is not reliable for the underlying iommu driver callback to get the domain from group->pasid_array. So, the second patch of this series passes the domain to remove_dev_pasid op.
[1] https://lore.kernel.org/linux-iommu/[email protected]/
Change log:
v2:
- Make clear that the patch 1/2 of v1 does not fix the problem (Kevin)
- Swap the order of patch 1/2 and 2/2 of v1. In this new series, patch 1/2 fixes the real issue, patch 2/2 is to avoid potential issue in the future.
v1: https://lore.kernel.org/linux-iommu/[email protected]/
Regards,
Yi Liu
Yi Liu (2):
iommu: Undo pasid attachment only for the devices that have succeeded
iommu: Pass domain to remove_dev_pasid() op
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++----- drivers/iommu/intel/iommu.c | 11 +++----- drivers/iommu/iommu.c | 28 ++++++++++++++------- include/linux/iommu.h | 3 ++- 4 files changed, 26 insertions(+), 25 deletions(-)
Signature · 1 line
2.34.1
PATCH v2 1/2 iommu: Undo pasid attachment only for the devices that have succeeded
There is no error handling now in __iommu_set_group_pasid(), it relies on its caller to loop all the devices to undo the pasid attachment. This is not self-contained and has drawbacks. It would result in unnecessary remove_dev_pasid() calls on the devices that have not been attached to the new domain. But the remove_dev_pasid() callback would get the new domain from the group->pasid_array. So for such devices, the iommu driver won't find the attachment under the domain, hence unable to do cleanup. This may not be a real problem today. But it depends on the implementation of the underlying iommu driver. e.g. the intel iommu driver would warn for such devices. Such warnings are unnecessary.
To solve the above problem, it is necessary to handle the error within __iommu_set_group_pasid(). It only loops the devices that have attached to the new domain, and undo it.
--- drivers/iommu/iommu.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-)
drivers/iommu/iommu.c +15 -6
@@ -3317,15 +3317,26 @@ EXPORT_SYMBOL_GPL(iommu_group_dma_owner_claimed);static int __iommu_set_group_pasid (struct iommu_domain *domain,struct iommu_group *group,ioasid_t pasid) { -struct group_device *device; -int ret =0 ; +struct group_device *device, *last_gdev; +int ret;for_each_group_device (group, device) { ret = domain->ops->set_dev_pasid (domain, device->dev, pasid);if (ret) -break ; + goto err_revert; } +return 0 ; + +err_revert: + last_gdev = device; +for_each_group_device (group, device) { +const struct iommu_ops *ops =dev_iommu_ops (device->dev); + +if (device == last_gdev) +break ; + ops->remove_dev_pasid (device->dev, pasid); + }return ret; }
@@ -3374,10 +3385,8 @@ int iommu_attach_device_pasid(struct iommu_domain *domain, } ret =__iommu_set_group_pasid (domain, group, pasid); -if (ret) { -__iommu_remove_group_pasid (group, pasid); +if (ret)xa_erase (&group->pasid_array, pasid); - } out_unlock:mutex_unlock (&group->mutex);return ret;
Signature · 1 line
2.34.1
Re: [PATCH v2 1/2] iommu: Undo pasid attachment only for the devices that have succeeded
Quoted 15 lines · Yi Liu
There is no error handling now in
__iommu_set_group_pasid(), it relies on its caller to loop all the devices to undo the pasid attachment. This is not self-contained and has drawbacks. It would result in unnecessaryremove_dev_pasid()calls on the devices that have not been attached to the new domain. But theremove_dev_pasid()callback would get the new domain from thegroup->pasid_array. So for such devices, the iommu driver won't find the attachment under the domain, hence unable to do cleanup. This may not be a real problem today. But it depends on the implementation of the underlying iommu driver. e.g. the intel iommu driver would warn for such devices. Such warnings are unnecessary.To solve the above problem, it is necessary to handle the error within
__iommu_set_group_pasid(). It only loops the devices that have attached to the new domain, and undo it.
Quoted 5 lines
Fixes 16603704559c ("iommu: Add attach/detach_dev_pasid iommu interfaces")Suggested-by Jason Gunthorpe<[email protected]>Reviewed-by Jason Gunthorpe<[email protected]>Reviewed-by Kevin Tian<[email protected]>Signed-off-by Yi Liu<[email protected]>
Best regards,
baolu
PATCH v2 2/2 iommu: Pass domain to remove_dev_pasid() op
In reply to Yi LiuExisting remove_dev_pasid() callbacks of the underlying iommu drivers get the attached domain from the group->pasid_array. However, the domain stored in group->pasid_array is not always correct in all scenarios. A wrong domain may result in failure in remove_dev_pasid() callback. To avoid such problems, it is more reliable to pass the domain to the remove_dev_pasid() op.
--- drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++------- drivers/iommu/intel/iommu.c | 11 +++-------- drivers/iommu/iommu.c | 9 +++++---- include/linux/iommu.h | 3 ++- 4 files changed, 12 insertions(+), 20 deletions(-)
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c +2 -7
@@ -3044,14 +3044,9 @@ static int arm_smmu_def_domain_type(struct device *dev)return 0 ; } -static void arm_smmu_remove_dev_pasid (struct device *dev,ioasid_t pasid) +static void arm_smmu_remove_dev_pasid (struct device *dev,ioasid_t pasid, +struct iommu_domain *domain) { -struct iommu_domain *domain; - - domain =iommu_get_domain_for_dev_pasid (dev, pasid,IOMMU_DOMAIN_SVA ); -if (WARN_ON (IS_ERR (domain)) || !domain) -return ; -arm_smmu_sva_remove_dev_pasid (domain, dev, pasid); }
drivers/iommu/intel/iommu.c +3 -8
@@ -4587,19 +4587,15 @@ static int intel_iommu_iotlb_sync_map(struct iommu_domain *domain,return 0 ; } -static void intel_iommu_remove_dev_pasid (struct device *dev,ioasid_t pasid) +static void intel_iommu_remove_dev_pasid (struct device *dev,ioasid_t pasid, +struct iommu_domain *domain) {struct device_domain_info *info =dev_iommu_priv_get (dev); +struct dmar_domain *dmar_domain =to_dmar_domain (domain);struct dev_pasid_info *curr, *dev_pasid =NULL ;struct intel_iommu *iommu = info->iommu; -struct dmar_domain *dmar_domain; -struct iommu_domain *domain;unsigned long flags; - domain =iommu_get_domain_for_dev_pasid (dev, pasid,0 ); -if (WARN_ON_ONCE (!domain)) - goto out_tear_down; -/* * The SVA implementation needs to handle its own stuffs like the mm * notification. Before consolidating that code into iommu core, let
@@ -4610,7 +4606,6 @@ static void intel_iommu_remove_dev_pasid(struct device *dev, ioasid_t pasid) goto out_tear_down; } - dmar_domain =to_dmar_domain (domain);spin_lock_irqsave (&dmar_domain->lock, flags);list_for_each_entry (curr, &dmar_domain->dev_pasids, link_domain) {if (curr->dev == dev && curr->pasid == pasid) {
drivers/iommu/iommu.c +5 -4
@@ -3335,20 +3335,21 @@ static int __iommu_set_group_pasid(struct iommu_domain *domain,if (device == last_gdev)break ; - ops->remove_dev_pasid (device->dev, pasid); + ops->remove_dev_pasid (device->dev, pasid, domain); }return ret; }static void __iommu_remove_group_pasid (struct iommu_group *group, -ioasid_t pasid) +ioasid_t pasid, +struct iommu_domain *domain) {struct group_device *device;const struct iommu_ops *ops;for_each_group_device (group, device) { ops =dev_iommu_ops (device->dev); - ops->remove_dev_pasid (device->dev, pasid); + ops->remove_dev_pasid (device->dev, pasid, domain); } }
@@ -3409,7 +3410,7 @@ void iommu_detach_device_pasid(struct iommu_domain *domain, struct device *dev,struct iommu_group *group = dev->iommu_group;mutex_lock (&group->mutex); -__iommu_remove_group_pasid (group, pasid); +__iommu_remove_group_pasid (group, pasid, domain);WARN_ON (xa_erase (&group->pasid_array, pasid) != domain);mutex_unlock (&group->mutex); }
include/linux/iommu.h +2 -1
@@ -578,7 +578,8 @@ struct iommu_ops { struct iommu_page_response *msg);int (*def_domain_type)(struct device *dev); -void (*remove_dev_pasid)(struct device *dev,ioasid_t pasid); +void (*remove_dev_pasid)(struct device *dev,ioasid_t pasid, +struct iommu_domain *domain);const struct iommu_domain_ops *default_domain_ops;unsigned long pgsize_bitmap;
Signature · 1 line
2.34.1
RE: [PATCH v2 2/2] iommu: Pass domain to remove_dev_pasid() op
Quoted 10 lines
From: Liu, Yi L <[email protected]>
Sent: Thursday, March 28, 2024 8:30 PMExisting
remove_dev_pasid()callbacks of the underlying iommu drivers get the attached domain from thegroup->pasid_array. However, the domain stored ingroup->pasid_arrayis not always correct in all scenarios. A wrong domain may result in failure inremove_dev_pasid()callback. To avoid such problems, it is more reliable to pass the domain to theremove_dev_pasid()op.
Quoted 2 lines
Suggested-by Jason Gunthorpe <[email protected]>Signed-off-by Yi Liu <[email protected]>
Baolu Lu
maintainerRe: [PATCH v2 2/2] iommu: Pass domain to remove_dev_pasid() op
In reply to Yi Liuinclude/linux/iommu.h struct iommu_ops { · line 582
Quoted 11 lines · Yi Liu
include/linux/iommu.h +2
-1@@ -578,7 +578,8 @@ struct iommu_ops { struct iommu_page_response *msg);int (*def_domain_type)(struct device *dev); -void (*remove_dev_pasid)(struct device *dev,ioasid_t pasid); +void (*remove_dev_pasid)(struct device *dev,ioasid_t pasid, +struct iommu_domain *domain);
Previously, this callback said "Hey, remove any domain associated with this pasid".
Now this callback changes to "Hey, please remove *this* domain from the pasid". What the driver should do if it doesn't match the previously attached domain, or the pasid hasn't been attached to any domain?
Best regards,
baolu
Re: [PATCH v2 2/2] iommu: Pass domain to remove_dev_pasid() op
Quoted 15 lines · Baolu Lu
include/linux/iommu.h struct iommu_ops { · line 581
Quoted 10 lines · Yi Liu
include/linux/iommu.h +2
-1@@ -578,7 +578,8 @@ struct iommu_ops { struct iommu_page_response *msg); int (*def_domain_type)(struct device *dev); - void (*remove_dev_pasid)(struct device *dev, ioasid_t pasid); + void (*remove_dev_pasid)(struct device *dev, ioasid_t pasid, + struct iommu_domain *domain);Previously, this callback said "Hey, remove any domain associated with this pasid".
Quoted 3 lines
Now this callback changes to "Hey, please remove *this* domain from the pasid". What the driver should do if it doesn't match the previously attached domain, or the pasid hasn't been attached to any domain?
I think the caller of this callback should know very well whether a pasid has been attached to this domain or not. So the problem you described should not happen at all. Otherwise, it is a bug in the iommu layer.
Actually, there is similar concern in the iommu_detach_device_pasid(). The input domain may be different with what iommu layer tracks. If so there is a warn. This means the external callers of this API are buggy. While, I have more faith on iommu layer. :)
Signature · 2 lines
Regards, Yi Liu
Re: [PATCH v2 2/2] iommu: Pass domain to remove_dev_pasid() op
Quoted 25 lines · Yi Liu
Quoted 15 lines · Baolu Lu
include/linux/iommu.h struct iommu_ops { · line 581
Quoted 10 lines · Yi Liu
include/linux/iommu.h +2
-1@@ -578,7 +578,8 @@ struct iommu_ops { struct iommu_page_response *msg); int (*def_domain_type)(struct device *dev); - void (*remove_dev_pasid)(struct device *dev, ioasid_t pasid); + void (*remove_dev_pasid)(struct device *dev, ioasid_t pasid, + struct iommu_domain *domain);Previously, this callback said "Hey, remove any domain associated with this pasid".
Quoted 3 lines
Now this callback changes to "Hey, please remove *this* domain from the pasid". What the driver should do if it doesn't match the previously attached domain, or the pasid hasn't been attached to any domain?
I think the caller of this callback should know very well whether a pasid has been attached to this domain or not. So the problem you described should not happen at all. Otherwise, it is a bug in the iommu layer.
Quoted 4 lines
Actually, there is similar concern in the
iommu_detach_device_pasid(). The input domain may be different with what iommu layer tracks. If so there is a warn. This means the external callers of this API are buggy. While, I have more faith on iommu layer. :)
Yeah, the iommu layer should obtain the domain from the pasid xarray and every driver today also obtains the domain from the pasid xarray. It is not something different, it is just moving code around.
The core code should guarentee the invariant.
Jason
Quoted 7 lines
-----Original Message-----
From: Liu, Yi L <[email protected]>
Subject: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()
Quoted 5 lines
There are minor mistakes in the iommu
set_dev_pasid()andremove_dev_pasid()
paths. Theset_dev_pasid()path updates thegroup->pasid_arrayfirst, and then call intoremove_dev_pasid()in error handling when there are devices within the group that failed toset_dev_pasid.
Not related to this patch, just curious in which cases some of the devices In same group failed to set_dev_pasid while others succeed?
Thanks
Zhenzhong
Quoted 43 lines
The
remove_dev_pasid()
callbacks of the underlying iommu drivers get the domain for pasid from thegroup->pasid_array. So theremove_dev_pasid()callback may get a wrong domain
in theset_dev_pasid()path. [1] Even if the group is singleton, the existing code logic would have unnecessary warnings in the error handling of theset_dev_pasid()path. e.g. intel iommu driver.The above issue can be fixed by improving the error handling in the
set_dev_pasid()path. Also, this reminds that it is not reliable for the underlying iommu driver callback to get the domain from group-Quoted 1 line
pasid_array.So, the second patch of this series passes the domain to
remove_dev_pasidop.[1] https://lore.kernel.org/linux-
iommu/[email protected]/Change log:
v2:
- Make clear that the patch 1/2 of v1 does not fix the problem (Kevin)
- Swap the order of patch 1/2 and 2/2 of v1. In this new series, patch 1/2 fixes the real issue, patch 2/2 is to avoid potential issue in the future.v1: https://lore.kernel.org/linux-iommu/20240327125433.248946-1-
[email protected]/Regards,
Yi LiuYi Liu (2):
iommu: Undo pasid attachment only for the devices that have succeeded
iommu: Pass domain toremove_dev_pasid()opdrivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++----- drivers/iommu/intel/iommu.c | 11 +++----- drivers/iommu/iommu.c | 28 ++++++++++++++------- include/linux/iommu.h | 3 ++- 4 files changed, 26 insertions(+), 25 deletions(-)--
2.34.1
Quoted 7 lines · Zhenzhong
Quoted 7 lines
-----Original Message-----
From: Liu, Yi L <[email protected]>
Subject: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()
Quoted 8 lines
Quoted 5 lines
There are minor mistakes in the iommu
set_dev_pasid()andremove_dev_pasid()
paths. Theset_dev_pasid()path updates thegroup->pasid_arrayfirst, and then call intoremove_dev_pasid()in error handling when there are devices within the group that failed toset_dev_pasid.Not related to this patch, just curious in which cases some of the devices In same group failed to
set_dev_pasidwhile others succeed?
There are multiple failure reasons. Given to the fact of some devices have already succeeded, the most typical error may be no memory. Not sure about other reasons.
Regards,
Yi Liu
Quoted 47 lines
Thanks
ZhenzhongQuoted 43 lines
The
remove_dev_pasid()
callbacks of the underlying iommu drivers get the domain for pasid from thegroup->pasid_array. So theremove_dev_pasid()callback may get a wrong domain
in theset_dev_pasid()path. [1] Even if the group is singleton, the existing code logic would have unnecessary warnings in the error handling of theset_dev_pasid()path. e.g. intel iommu driver.The above issue can be fixed by improving the error handling in the
set_dev_pasid()path. Also, this reminds that it is not reliable for the underlying iommu driver callback to get the domain from group-Quoted 1 line
pasid_array.So, the second patch of this series passes the domain to
remove_dev_pasidop.[1] https://lore.kernel.org/linux-
iommu/[email protected]/Change log:
v2:
- Make clear that the patch 1/2 of v1 does not fix the problem (Kevin)
- Swap the order of patch 1/2 and 2/2 of v1. In this new series, patch 1/2 fixes the real issue, patch 2/2 is to avoid potential issue in the future.v1: https://lore.kernel.org/linux-iommu/20240327125433.248946-1-
[email protected]/Regards,
Yi LiuYi Liu (2):
iommu: Undo pasid attachment only for the devices that have succeeded
iommu: Pass domain toremove_dev_pasid()opdrivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++----- drivers/iommu/intel/iommu.c | 11 +++----- drivers/iommu/iommu.c | 28 ++++++++++++++------- include/linux/iommu.h | 3 ++- 4 files changed, 26 insertions(+), 25 deletions(-)--
2.34.1
Signature · 2 lines
Regards, Yi Liu
Quoted 21 lines
-----Original Message-----
From: Liu, Yi L <[email protected]>
Subject: Re: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()Quoted 10 lines · Zhenzhong
Quoted 10 lines
-----Original Message-----
From: Liu, Yi L <[email protected]>
Subject: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()There are minor mistakes in the iommu
set_dev_pasid()andremove_dev_pasid()
paths. Theset_dev_pasid()path updates thegroup->pasid_arrayfirst,and
Quoted 1 line
Quoted 1 line
then call into
remove_dev_pasid()in error handling when there aredevices
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
within the group that failed to
set_dev_pasid.Not related to this patch, just curious in which cases some of the devices In same group failed to
set_dev_pasidwhile others succeed?There are multiple failure reasons. Given to the fact of some devices have already succeeded, the most typical error may be no memory. Not sure about
other reasons.
Oh, the no memory case, clear, thanks Yi.
BRs.
Zhenzhong
Baolu Lu
maintainerQuoted 6 lines · Zhenzhong
Quoted 6 lines
-----Original Message-----
From: Liu, Yi L<[email protected]>
Subject: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()
Quoted 7 lines
Quoted 5 lines
There are minor mistakes in the iommu
set_dev_pasid()andremove_dev_pasid()
paths. Theset_dev_pasid()path updates thegroup->pasid_arrayfirst, and then call intoremove_dev_pasid()in error handling when there are devices within the group that failed toset_dev_pasid.Not related to this patch, just curious in which cases some of the devices In same group failed to
set_dev_pasidwhile others succeed?
The failure cases could be checked in the set_dev_pasid implementation of the individual iommu driver. For x86 platforms, which are PCI fabric- based, there's no such case as PCI/PASID requires a singleton iommu group.
Best regards,
baolu
Quoted 20 lines
-----Original Message-----
From: Baolu Lu <[email protected]>
Subject: Re: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()Quoted 9 lines · Zhenzhong
Quoted 9 lines
-----Original Message-----
From: Liu, Yi L<[email protected]>
Subject: [PATCH v2 0/2] Two enhancements to
iommu_at[de]tach_device_pasid()There are minor mistakes in the iommu
set_dev_pasid()andremove_dev_pasid()
paths. Theset_dev_pasid()path updates thegroup->pasid_arrayfirst,and
Quoted 1 line
Quoted 1 line
then call into
remove_dev_pasid()in error handling when there aredevices
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
within the group that failed to
set_dev_pasid.Not related to this patch, just curious in which cases some of the devices In same group failed to
set_dev_pasidwhile others succeed?The failure cases could be checked in the
set_dev_pasidimplementation of the individual iommu driver. For x86 platforms, which are PCI fabric- based, there's no such case as PCI/PASID requires a singleton iommu group.
Clear, thanks Baolu.
BRs.
Zhenzhong
Quoted 9 lines · Yi Liu
Yi Liu (2):
iommu: Undo pasid attachment only for the devices that have succeeded
iommu: Pass domain toremove_dev_pasid()opdrivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++----- drivers/iommu/intel/iommu.c | 11 +++----- drivers/iommu/iommu.c | 28 ++++++++++++++------- include/linux/iommu.h | 3 ++- 4 files changed, 26 insertions(+), 25 deletions(-)
Applied, thanks.