As discussed here is a small series to address the confusing lifetime scheme for the hwpt. This was some leftover from prior rework that was never fully cleaned up.
Make it clear that the ioas and ioas->hwpt_list are associated with the hwpt during creation and never changed until it is destroyed. A hwpt with a positive reference count is always valid for device attachment.
This also improves the selftest handling of the mock domains so that we can implement more testing for the hwpt model.
This is a step toward the nesting and replace series.
I have this on github:
https://github.com/jgunthorpe/linux/commits/iommufd_hwpt
v3: - Split patch to just make the hwpt_item handling consistent - Remove two confusing comments - Four new patches to clean up the confusing 'domain_id' in the selftest v2: https://lore.kernel.org/r/0-v2-406f7ac07936+6a-iommufd_hwpt_jgg@nvidia.com - Basically completely different - Add a test for HWPT cases by enhancing the mock iommu driver to be more complete - Move the device attachment as well into iommufd_hw_pagetable_alloc() so destroy is more symmetric v1: https://lore.kernel.org/r/0-v1-4336b5cb2fe4+1d7-iommufd_hwpt_jgg@nvidia.com
Jason Gunthorpe (12): iommufd: Assert devices_lock for iommufd_hw_pagetable_has_group() iommufd: Add iommufd_lock_obj() around the auto-domains hwpts iommufd: Consistently manage hwpt_item iommufd: Move ioas related HWPT destruction into iommufd_hw_pagetable_destroy() iommufd: Move iommufd_device to iommufd_private.h iommufd: Make iommufd_hw_pagetable_alloc() do iopt_table_add_domain() iommufd/selftest: Rename the sefltest 'device_id' to 'stdev_id' iommufd/selftest: Rename domain_id to stdev_id for FIXTURE iommufd_ioas iommufd/selftest: Rename domain_id to hwpt_id for FIXTURE iommufd_mock_domain iommufd/selftest: Rename the remaining mock device_id's to stdev_id iommufd/selftest: Make selftest create a more complete mock device iommufd/selftest: Add a selftest for iommufd_device_attach() with a hwpt argument
drivers/iommu/iommufd/device.c | 149 +++++-------- drivers/iommu/iommufd/hw_pagetable.c | 70 ++++++- drivers/iommu/iommufd/iommufd_private.h | 35 +++- drivers/iommu/iommufd/iommufd_test.h | 2 +- drivers/iommu/iommufd/selftest.c | 198 +++++++++++++++--- tools/testing/selftests/iommu/iommufd.c | 68 +++--- .../selftests/iommu/iommufd_fail_nth.c | 38 ++-- tools/testing/selftests/iommu/iommufd_utils.h | 16 +- 8 files changed, 368 insertions(+), 208 deletions(-)
base-commit: 8473685c99683f9e2c7e2a42ef3d3f14a79be274
The hwpt->devices list is locked by this, make it clearer.
Reviewed-by: Kevin Tian kevin.tian@intel.com Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 2 ++ 1 file changed, 2 insertions(+)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index a0c66f47a65ada..dcfaf6567420e0 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -186,6 +186,8 @@ static bool iommufd_hw_pagetable_has_group(struct iommufd_hw_pagetable *hwpt, { struct iommufd_device *cur_dev;
+ lockdep_assert_held(&hwpt->devices_lock); + list_for_each_entry(cur_dev, &hwpt->devices, devices_item) if (cur_dev->group == group) return true;
A later patch will require this locking - currently under the ioas mutex the hwpt can not have a 0 reference and be on the list.
Reviewed-by: Kevin Tian kevin.tian@intel.com Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 3 +++ 1 file changed, 3 insertions(+)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index dcfaf6567420e0..0a80ff7b2e0d79 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -282,7 +282,10 @@ static int iommufd_device_auto_get_domain(struct iommufd_device *idev, if (!hwpt->auto_domain) continue;
+ if (!iommufd_lock_obj(&hwpt->obj)) + continue; rc = iommufd_device_do_attach(idev, hwpt); + iommufd_put_object(&hwpt->obj);
/* * -EINVAL means the domain is incompatible with the device.
This should be added immediately after every iopt_table_add_domain(), and deleted after every iopt_table_remove_domain() under the ioas->mutex.
Tidy things to be consistent.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index 0a80ff7b2e0d79..adb73539b39c0c 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -243,6 +243,7 @@ static int iommufd_device_do_attach(struct iommufd_device *idev, hwpt->domain); if (rc) goto out_detach; + list_add_tail(&hwpt->hwpt_item, &hwpt->ioas->hwpt_list); } }
@@ -307,7 +308,6 @@ static int iommufd_device_auto_get_domain(struct iommufd_device *idev, rc = iommufd_device_do_attach(idev, hwpt); if (rc) goto out_abort; - list_add_tail(&hwpt->hwpt_item, &ioas->hwpt_list);
mutex_unlock(&ioas->mutex); iommufd_object_finalize(idev->ictx, &hwpt->obj); @@ -753,6 +753,10 @@ iommufd_device_selftest_attach(struct iommufd_ctx *ictx, if (rc) goto out_hwpt;
+ mutex_lock(&ioas->mutex); + list_add_tail(&hwpt->hwpt_item, &hwpt->ioas->hwpt_list); + mutex_unlock(&ioas->mutex); + refcount_inc(&hwpt->obj.users); iommufd_object_finalize(ictx, &hwpt->obj); return hwpt; @@ -765,7 +769,10 @@ iommufd_device_selftest_attach(struct iommufd_ctx *ictx, void iommufd_device_selftest_detach(struct iommufd_ctx *ictx, struct iommufd_hw_pagetable *hwpt) { + mutex_lock(&hwpt->ioas->mutex); iopt_table_remove_domain(&hwpt->ioas->iopt, hwpt->domain); + list_del(&hwpt->hwpt_item); + mutex_unlock(&hwpt->ioas->mutex); refcount_dec(&hwpt->obj.users); } #endif
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
This should be added immediately after every iopt_table_add_domain(), and deleted after every iopt_table_remove_domain() under the ioas->mutex.
Tidy things to be consistent.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
A HWPT is permanently associated with an IOAS when it is created, remove the strange situation where a refcount != 0 HWPT can have been disconnected from the IOAS by putting all the IOAS related destruction in the object destroy function.
Initializing a HWPT is two stages, we have to allocate it, attach it to a device and then populate the domain. Once the domain is populated it is fully linked to the IOAS.
Arrange things so that all the error unwinds flow through the iommufd_hw_pagetable_destroy() and allow it to handle all cases.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 17 ++--------------- drivers/iommu/iommufd/hw_pagetable.c | 27 +++++++++++++++++++-------- 2 files changed, 21 insertions(+), 23 deletions(-)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index adb73539b39c0c..6787a0d8d6e9f0 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -386,28 +386,19 @@ void iommufd_device_detach(struct iommufd_device *idev) { struct iommufd_hw_pagetable *hwpt = idev->hwpt;
- mutex_lock(&hwpt->ioas->mutex); mutex_lock(&hwpt->devices_lock); list_del(&idev->devices_item); - if (!iommufd_hw_pagetable_has_group(hwpt, idev->group)) { - if (list_empty(&hwpt->devices)) { - iopt_table_remove_domain(&hwpt->ioas->iopt, - hwpt->domain); - list_del(&hwpt->hwpt_item); - } + if (!iommufd_hw_pagetable_has_group(hwpt, idev->group)) iommu_detach_group(hwpt->domain, idev->group); - } iopt_remove_reserved_iova(&hwpt->ioas->iopt, idev->dev); + idev->hwpt = NULL; mutex_unlock(&hwpt->devices_lock); - mutex_unlock(&hwpt->ioas->mutex);
if (hwpt->auto_domain) iommufd_object_destroy_user(idev->ictx, &hwpt->obj); else refcount_dec(&hwpt->obj.users);
- idev->hwpt = NULL; - refcount_dec(&idev->obj.users); } EXPORT_SYMBOL_NS_GPL(iommufd_device_detach, IOMMUFD); @@ -769,10 +760,6 @@ iommufd_device_selftest_attach(struct iommufd_ctx *ictx, void iommufd_device_selftest_detach(struct iommufd_ctx *ictx, struct iommufd_hw_pagetable *hwpt) { - mutex_lock(&hwpt->ioas->mutex); - iopt_table_remove_domain(&hwpt->ioas->iopt, hwpt->domain); - list_del(&hwpt->hwpt_item); - mutex_unlock(&hwpt->ioas->mutex); refcount_dec(&hwpt->obj.users); } #endif diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c index 43d473989a0667..10db1359c067c1 100644 --- a/drivers/iommu/iommufd/hw_pagetable.c +++ b/drivers/iommu/iommufd/hw_pagetable.c @@ -13,7 +13,17 @@ void iommufd_hw_pagetable_destroy(struct iommufd_object *obj)
WARN_ON(!list_empty(&hwpt->devices));
- iommu_domain_free(hwpt->domain); + if (!list_empty(&hwpt->hwpt_item)) { + mutex_lock(&hwpt->ioas->mutex); + list_del(&hwpt->hwpt_item); + mutex_unlock(&hwpt->ioas->mutex); + + iopt_table_remove_domain(&hwpt->ioas->iopt, hwpt->domain); + } + + if (hwpt->domain) + iommu_domain_free(hwpt->domain); + refcount_dec(&hwpt->ioas->obj.users); mutex_destroy(&hwpt->devices_lock); } @@ -37,21 +47,22 @@ iommufd_hw_pagetable_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, if (IS_ERR(hwpt)) return hwpt;
- hwpt->domain = iommu_domain_alloc(dev->bus); - if (!hwpt->domain) { - rc = -ENOMEM; - goto out_abort; - } - INIT_LIST_HEAD(&hwpt->devices); INIT_LIST_HEAD(&hwpt->hwpt_item); mutex_init(&hwpt->devices_lock); /* Pairs with iommufd_hw_pagetable_destroy() */ refcount_inc(&ioas->obj.users); hwpt->ioas = ioas; + + hwpt->domain = iommu_domain_alloc(dev->bus); + if (!hwpt->domain) { + rc = -ENOMEM; + goto out_abort; + } + return hwpt;
out_abort: - iommufd_object_abort(ictx, &hwpt->obj); + iommufd_object_abort_and_destroy(ictx, &hwpt->obj); return ERR_PTR(rc); }
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
A HWPT is permanently associated with an IOAS when it is created, remove the strange situation where a refcount != 0 HWPT can have been disconnected from the IOAS by putting all the IOAS related destruction in the object destroy function.
Initializing a HWPT is two stages, we have to allocate it, attach it to a device and then populate the domain. Once the domain is populated it is fully linked to the IOAS.
Arrange things so that all the error unwinds flow through the iommufd_hw_pagetable_destroy() and allow it to handle all cases.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
hw_pagetable.c will need this in the next patches.
Reviewed-by: Kevin Tian kevin.tian@intel.com Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 17 ----------------- drivers/iommu/iommufd/iommufd_private.h | 18 ++++++++++++++++++ 2 files changed, 18 insertions(+), 17 deletions(-)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index 6787a0d8d6e9f0..4483c06e0ec38b 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -15,23 +15,6 @@ MODULE_PARM_DESC( "Allow IOMMUFD to bind to devices even if the platform cannot isolate " "the MSI interrupt window. Enabling this is a security weakness.");
-/* - * A iommufd_device object represents the binding relationship between a - * consuming driver and the iommufd. These objects are created/destroyed by - * external drivers, not by userspace. - */ -struct iommufd_device { - struct iommufd_object obj; - struct iommufd_ctx *ictx; - struct iommufd_hw_pagetable *hwpt; - /* Head at iommufd_hw_pagetable::devices */ - struct list_head devices_item; - /* always the physical device */ - struct device *dev; - struct iommu_group *group; - bool enforce_cache_coherency; -}; - void iommufd_device_destroy(struct iommufd_object *obj) { struct iommufd_device *idev = diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h index 9d7f71510ca1bc..84fe19a195466b 100644 --- a/drivers/iommu/iommufd/iommufd_private.h +++ b/drivers/iommu/iommufd/iommufd_private.h @@ -12,6 +12,7 @@ struct iommu_domain; struct iommu_group; struct iommu_option; +struct iommufd_device;
struct iommufd_ctx { struct file *file; @@ -257,6 +258,23 @@ iommufd_hw_pagetable_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, struct device *dev); void iommufd_hw_pagetable_destroy(struct iommufd_object *obj);
+/* + * A iommufd_device object represents the binding relationship between a + * consuming driver and the iommufd. These objects are created/destroyed by + * external drivers, not by userspace. + */ +struct iommufd_device { + struct iommufd_object obj; + struct iommufd_ctx *ictx; + struct iommufd_hw_pagetable *hwpt; + /* Head at iommufd_hw_pagetable::devices */ + struct list_head devices_item; + /* always the physical device */ + struct device *dev; + struct iommu_group *group; + bool enforce_cache_coherency; +}; + void iommufd_device_destroy(struct iommufd_object *obj);
struct iommufd_access {
The HWPT is always linked to an IOAS and once a HWPT exists its domain should be fully mapped. This ended up being split up into device.c during a two phase creation that was a bit confusing.
Move the iopt_table_add_domain() into iommufd_hw_pagetable_alloc() by having it call back to device.c to complete the domain attach in the required order.
Calling iommufd_hw_pagetable_alloc() with immediate_attach = false will work on most drivers, but notably the SMMU drivers will fail because they can't decide what kind of domain to create until they are attached. This will be fixed when the domain_alloc function can take in a struct device.
Reviewed-by: Kevin Tian kevin.tian@intel.com Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 87 +++++++++++-------------- drivers/iommu/iommufd/hw_pagetable.c | 45 +++++++++++-- drivers/iommu/iommufd/iommufd_private.h | 6 +- 3 files changed, 85 insertions(+), 53 deletions(-)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index 4483c06e0ec38b..8dc7ed678e3fbb 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -177,13 +177,16 @@ static bool iommufd_hw_pagetable_has_group(struct iommufd_hw_pagetable *hwpt, return false; }
-static int iommufd_device_do_attach(struct iommufd_device *idev, - struct iommufd_hw_pagetable *hwpt) +int iommufd_hw_pagetable_attach(struct iommufd_hw_pagetable *hwpt, + struct iommufd_device *idev) { phys_addr_t sw_msi_start = PHYS_ADDR_MAX; int rc;
- mutex_lock(&hwpt->devices_lock); + lockdep_assert_held(&hwpt->devices_lock); + + if (WARN_ON(idev->hwpt)) + return -EINVAL;
/* * Try to upgrade the domain we have, it is an iommu driver bug to @@ -198,19 +201,18 @@ static int iommufd_device_do_attach(struct iommufd_device *idev, hwpt->domain); if (!hwpt->enforce_cache_coherency) { WARN_ON(list_empty(&hwpt->devices)); - rc = -EINVAL; - goto out_unlock; + return -EINVAL; } }
rc = iopt_table_enforce_group_resv_regions(&hwpt->ioas->iopt, idev->dev, idev->group, &sw_msi_start); if (rc) - goto out_unlock; + return rc;
rc = iommufd_device_setup_msi(idev, hwpt, sw_msi_start); if (rc) - goto out_iova; + goto err_unresv;
/* * FIXME: Hack around missing a device-centric iommu api, only attach to @@ -219,27 +221,35 @@ static int iommufd_device_do_attach(struct iommufd_device *idev, if (!iommufd_hw_pagetable_has_group(hwpt, idev->group)) { rc = iommu_attach_group(hwpt->domain, idev->group); if (rc) - goto out_iova; - - if (list_empty(&hwpt->devices)) { - rc = iopt_table_add_domain(&hwpt->ioas->iopt, - hwpt->domain); - if (rc) - goto out_detach; - list_add_tail(&hwpt->hwpt_item, &hwpt->ioas->hwpt_list); - } + goto err_unresv; } + return 0; +err_unresv: + iopt_remove_reserved_iova(&hwpt->ioas->iopt, idev->dev); + return rc; +} + +void iommufd_hw_pagetable_detach(struct iommufd_hw_pagetable *hwpt, + struct iommufd_device *idev) +{ + if (!iommufd_hw_pagetable_has_group(hwpt, idev->group)) + iommu_detach_group(hwpt->domain, idev->group); + iopt_remove_reserved_iova(&hwpt->ioas->iopt, idev->dev); +} + +static int iommufd_device_do_attach(struct iommufd_device *idev, + struct iommufd_hw_pagetable *hwpt) +{ + int rc; + + mutex_lock(&hwpt->devices_lock); + rc = iommufd_hw_pagetable_attach(hwpt, idev); + if (rc) + goto out_unlock;
idev->hwpt = hwpt; refcount_inc(&hwpt->obj.users); list_add(&idev->devices_item, &hwpt->devices); - mutex_unlock(&hwpt->devices_lock); - return 0; - -out_detach: - iommu_detach_group(hwpt->domain, idev->group); -out_iova: - iopt_remove_reserved_iova(&hwpt->ioas->iopt, idev->dev); out_unlock: mutex_unlock(&hwpt->devices_lock); return rc; @@ -281,23 +291,16 @@ static int iommufd_device_auto_get_domain(struct iommufd_device *idev, goto out_unlock; }
- hwpt = iommufd_hw_pagetable_alloc(idev->ictx, ioas, idev->dev); + hwpt = iommufd_hw_pagetable_alloc(idev->ictx, ioas, idev, true); if (IS_ERR(hwpt)) { rc = PTR_ERR(hwpt); goto out_unlock; } hwpt->auto_domain = true;
- rc = iommufd_device_do_attach(idev, hwpt); - if (rc) - goto out_abort; - mutex_unlock(&ioas->mutex); iommufd_object_finalize(idev->ictx, &hwpt->obj); return 0; - -out_abort: - iommufd_object_abort_and_destroy(idev->ictx, &hwpt->obj); out_unlock: mutex_unlock(&ioas->mutex); return rc; @@ -371,10 +374,8 @@ void iommufd_device_detach(struct iommufd_device *idev)
mutex_lock(&hwpt->devices_lock); list_del(&idev->devices_item); - if (!iommufd_hw_pagetable_has_group(hwpt, idev->group)) - iommu_detach_group(hwpt->domain, idev->group); - iopt_remove_reserved_iova(&hwpt->ioas->iopt, idev->dev); idev->hwpt = NULL; + iommufd_hw_pagetable_detach(hwpt, idev); mutex_unlock(&hwpt->devices_lock);
if (hwpt->auto_domain) @@ -716,28 +717,18 @@ iommufd_device_selftest_attach(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, struct device *mock_dev) { + struct iommufd_device tmp_idev = { .dev = mock_dev }; struct iommufd_hw_pagetable *hwpt; - int rc; - - hwpt = iommufd_hw_pagetable_alloc(ictx, ioas, mock_dev); - if (IS_ERR(hwpt)) - return hwpt; - - rc = iopt_table_add_domain(&hwpt->ioas->iopt, hwpt->domain); - if (rc) - goto out_hwpt;
mutex_lock(&ioas->mutex); - list_add_tail(&hwpt->hwpt_item, &hwpt->ioas->hwpt_list); + hwpt = iommufd_hw_pagetable_alloc(ictx, ioas, &tmp_idev, false); mutex_unlock(&ioas->mutex); + if (IS_ERR(hwpt)) + return hwpt;
refcount_inc(&hwpt->obj.users); iommufd_object_finalize(ictx, &hwpt->obj); return hwpt; - -out_hwpt: - iommufd_object_abort_and_destroy(ictx, &hwpt->obj); - return ERR_PTR(rc); }
void iommufd_device_selftest_detach(struct iommufd_ctx *ictx, diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c index 10db1359c067c1..6cdb6749d359f3 100644 --- a/drivers/iommu/iommufd/hw_pagetable.c +++ b/drivers/iommu/iommufd/hw_pagetable.c @@ -32,17 +32,22 @@ void iommufd_hw_pagetable_destroy(struct iommufd_object *obj) * iommufd_hw_pagetable_alloc() - Get an iommu_domain for a device * @ictx: iommufd context * @ioas: IOAS to associate the domain with - * @dev: Device to get an iommu_domain for + * @idev: Device to get an iommu_domain for + * @immediate_attach: True if idev should be attached to the hwpt * - * Allocate a new iommu_domain and return it as a hw_pagetable. + * Allocate a new iommu_domain and return it as a hw_pagetable. The HWPT + * will be linked to the given ioas and upon return the underlying iommu_domain + * is fully popoulated. */ struct iommufd_hw_pagetable * iommufd_hw_pagetable_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, - struct device *dev) + struct iommufd_device *idev, bool immediate_attach) { struct iommufd_hw_pagetable *hwpt; int rc;
+ lockdep_assert_held(&ioas->mutex); + hwpt = iommufd_object_alloc(ictx, hwpt, IOMMUFD_OBJ_HW_PAGETABLE); if (IS_ERR(hwpt)) return hwpt; @@ -54,14 +59,46 @@ iommufd_hw_pagetable_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, refcount_inc(&ioas->obj.users); hwpt->ioas = ioas;
- hwpt->domain = iommu_domain_alloc(dev->bus); + hwpt->domain = iommu_domain_alloc(idev->dev->bus); if (!hwpt->domain) { rc = -ENOMEM; goto out_abort; }
+ mutex_lock(&hwpt->devices_lock); + + /* + * immediate_attach exists only to accommodate iommu drivers that cannot + * directly allocate a domain. These drivers do not finish creating the + * domain until attach is completed. Thus we must have this call + * sequence. Once those drivers are fixed this should be removed. + */ + if (immediate_attach) { + rc = iommufd_hw_pagetable_attach(hwpt, idev); + if (rc) + goto out_unlock; + } + + rc = iopt_table_add_domain(&hwpt->ioas->iopt, hwpt->domain); + if (rc) + goto out_detach; + list_add_tail(&hwpt->hwpt_item, &hwpt->ioas->hwpt_list); + + if (immediate_attach) { + /* See iommufd_device_do_attach() */ + refcount_inc(&hwpt->obj.users); + idev->hwpt = hwpt; + list_add(&idev->devices_item, &hwpt->devices); + } + + mutex_unlock(&hwpt->devices_lock); return hwpt;
+out_detach: + if (immediate_attach) + iommufd_hw_pagetable_detach(hwpt, idev); +out_unlock: + mutex_unlock(&hwpt->devices_lock); out_abort: iommufd_object_abort_and_destroy(ictx, &hwpt->obj); return ERR_PTR(rc); diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h index 84fe19a195466b..331664e917b771 100644 --- a/drivers/iommu/iommufd/iommufd_private.h +++ b/drivers/iommu/iommufd/iommufd_private.h @@ -255,7 +255,11 @@ struct iommufd_hw_pagetable {
struct iommufd_hw_pagetable * iommufd_hw_pagetable_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas, - struct device *dev); + struct iommufd_device *idev, bool immediate_attach); +int iommufd_hw_pagetable_attach(struct iommufd_hw_pagetable *hwpt, + struct iommufd_device *idev); +void iommufd_hw_pagetable_detach(struct iommufd_hw_pagetable *hwpt, + struct iommufd_device *idev); void iommufd_hw_pagetable_destroy(struct iommufd_object *obj);
/*
It is too confusing now that we have the 'dev_id' as part of the main interface. Make it clear this is the special selftest device object. This object is analogous to the VFIO device FD.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/iommufd_test.h | 2 +- drivers/iommu/iommufd/selftest.c | 2 +- tools/testing/selftests/iommu/iommufd_utils.h | 16 ++++++++-------- 3 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/drivers/iommu/iommufd/iommufd_test.h b/drivers/iommu/iommufd/iommufd_test.h index 1d96a8f466fd29..b3d69cca77295c 100644 --- a/drivers/iommu/iommufd/iommufd_test.h +++ b/drivers/iommu/iommufd/iommufd_test.h @@ -49,7 +49,7 @@ struct iommu_test_cmd { __aligned_u64 length; } add_reserved; struct { - __u32 out_device_id; + __u32 out_stdev_id; __u32 out_hwpt_id; } mock_domain; struct { diff --git a/drivers/iommu/iommufd/selftest.c b/drivers/iommu/iommufd/selftest.c index cfb5fe9a5e0ee8..e05b41059630a7 100644 --- a/drivers/iommu/iommufd/selftest.c +++ b/drivers/iommu/iommufd/selftest.c @@ -302,7 +302,7 @@ static int iommufd_test_mock_domain(struct iommufd_ucmd *ucmd,
/* Userspace must destroy both of these IDs to destroy the object */ cmd->mock_domain.out_hwpt_id = hwpt->obj.id; - cmd->mock_domain.out_device_id = sobj->obj.id; + cmd->mock_domain.out_stdev_id = sobj->obj.id; iommufd_object_finalize(ucmd->ictx, &sobj->obj); iommufd_put_object(&ioas->obj); return iommufd_ucmd_respond(ucmd, sizeof(*cmd)); diff --git a/tools/testing/selftests/iommu/iommufd_utils.h b/tools/testing/selftests/iommu/iommufd_utils.h index 0d1f46369c2a30..85d6662ef8e867 100644 --- a/tools/testing/selftests/iommu/iommufd_utils.h +++ b/tools/testing/selftests/iommu/iommufd_utils.h @@ -38,7 +38,7 @@ static unsigned long BUFFER_SIZE; &test_cmd)); \ })
-static int _test_cmd_mock_domain(int fd, unsigned int ioas_id, __u32 *device_id, +static int _test_cmd_mock_domain(int fd, unsigned int ioas_id, __u32 *stdev_id, __u32 *hwpt_id) { struct iommu_test_cmd cmd = { @@ -52,19 +52,19 @@ static int _test_cmd_mock_domain(int fd, unsigned int ioas_id, __u32 *device_id, ret = ioctl(fd, IOMMU_TEST_CMD, &cmd); if (ret) return ret; - if (device_id) - *device_id = cmd.mock_domain.out_device_id; + if (stdev_id) + *stdev_id = cmd.mock_domain.out_stdev_id; assert(cmd.id != 0); if (hwpt_id) *hwpt_id = cmd.mock_domain.out_hwpt_id; return 0; } -#define test_cmd_mock_domain(ioas_id, device_id, hwpt_id) \ - ASSERT_EQ(0, _test_cmd_mock_domain(self->fd, ioas_id, device_id, \ - hwpt_id)) -#define test_err_mock_domain(_errno, ioas_id, device_id, hwpt_id) \ +#define test_cmd_mock_domain(ioas_id, stdev_id, hwpt_id) \ + ASSERT_EQ(0, \ + _test_cmd_mock_domain(self->fd, ioas_id, stdev_id, hwpt_id)) +#define test_err_mock_domain(_errno, ioas_id, stdev_id, hwpt_id) \ EXPECT_ERRNO(_errno, _test_cmd_mock_domain(self->fd, ioas_id, \ - device_id, hwpt_id)) + stdev_id, hwpt_id))
static int _test_cmd_create_access(int fd, unsigned int ioas_id, __u32 *access_id, unsigned int flags)
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
It is too confusing now that we have the 'dev_id' as part of the main interface. Make it clear this is the special selftest device object. This object is analogous to the VFIO device FD.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
Nothing uses this for anything more than checking if a mock domain had been created. Rename it to stdev_id to match our naming system. Currently domain_id is the hwpt_id.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- tools/testing/selftests/iommu/iommufd.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index fa08209268c42f..000fadf8110833 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -186,7 +186,7 @@ FIXTURE(iommufd_ioas) { int fd; uint32_t ioas_id; - uint32_t domain_id; + uint32_t stdev_id; uint64_t base_iova; };
@@ -212,7 +212,7 @@ FIXTURE_SETUP(iommufd_ioas) }
for (i = 0; i != variant->mock_domains; i++) { - test_cmd_mock_domain(self->ioas_id, NULL, &self->domain_id); + test_cmd_mock_domain(self->ioas_id, &self->stdev_id, NULL); self->base_iova = MOCK_APERTURE_START; } } @@ -249,7 +249,7 @@ TEST_F(iommufd_ioas, ioas_auto_destroy)
TEST_F(iommufd_ioas, ioas_destroy) { - if (self->domain_id) { + if (self->stdev_id) { /* IOAS cannot be freed while a domain is on it */ EXPECT_ERRNO(EBUSY, _test_ioctl_destroy(self->fd, self->ioas_id)); @@ -263,7 +263,7 @@ TEST_F(iommufd_ioas, ioas_area_destroy) { /* Adding an area does not change ability to destroy */ test_ioctl_ioas_map_fixed(buffer, PAGE_SIZE, self->base_iova); - if (self->domain_id) + if (self->stdev_id) EXPECT_ERRNO(EBUSY, _test_ioctl_destroy(self->fd, self->ioas_id)); else @@ -382,7 +382,7 @@ TEST_F(iommufd_ioas, area_auto_iova) for (i = 0; i != 10; i++) { size_t length = PAGE_SIZE * (i + 1);
- if (self->domain_id) { + if (self->stdev_id) { test_ioctl_ioas_map(buffer, length, &iovas[i]); } else { test_ioctl_ioas_map((void *)(1UL << 31), length, @@ -418,7 +418,7 @@ TEST_F(iommufd_ioas, area_auto_iova) ioctl(self->fd, IOMMU_IOAS_ALLOW_IOVAS, &allow_cmd));
/* Allocate from an allowed region */ - if (self->domain_id) { + if (self->stdev_id) { ranges[0].start = MOCK_APERTURE_START + PAGE_SIZE; ranges[0].last = MOCK_APERTURE_START + PAGE_SIZE * 600 - 1; } else { @@ -525,7 +525,7 @@ TEST_F(iommufd_ioas, iova_ranges) /* Range can be read */ ASSERT_EQ(0, ioctl(self->fd, IOMMU_IOAS_IOVA_RANGES, &ranges_cmd)); EXPECT_EQ(1, ranges_cmd.num_iovas); - if (!self->domain_id) { + if (!self->stdev_id) { EXPECT_EQ(0, ranges[0].start); EXPECT_EQ(SIZE_MAX, ranges[0].last); EXPECT_EQ(1, ranges_cmd.out_iova_alignment); @@ -550,7 +550,7 @@ TEST_F(iommufd_ioas, iova_ranges) &test_cmd)); ranges_cmd.num_iovas = BUFFER_SIZE / sizeof(*ranges); ASSERT_EQ(0, ioctl(self->fd, IOMMU_IOAS_IOVA_RANGES, &ranges_cmd)); - if (!self->domain_id) { + if (!self->stdev_id) { EXPECT_EQ(2, ranges_cmd.num_iovas); EXPECT_EQ(0, ranges[0].start); EXPECT_EQ(PAGE_SIZE - 1, ranges[0].last); @@ -565,7 +565,7 @@ TEST_F(iommufd_ioas, iova_ranges) /* Buffer too small */ memset(ranges, 0, BUFFER_SIZE); ranges_cmd.num_iovas = 1; - if (!self->domain_id) { + if (!self->stdev_id) { EXPECT_ERRNO(EMSGSIZE, ioctl(self->fd, IOMMU_IOAS_IOVA_RANGES, &ranges_cmd)); EXPECT_EQ(2, ranges_cmd.num_iovas); @@ -789,7 +789,7 @@ TEST_F(iommufd_ioas, fork_gone) ASSERT_NE(-1, child); ASSERT_EQ(child, waitpid(child, NULL, 0));
- if (self->domain_id) { + if (self->stdev_id) { /* * If a domain already existed then everything was pinned within * the fork, so this copies from one domain to another.
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index fa08209268c42f..000fadf8110833 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -186,7 +186,7 @@ FIXTURE(iommufd_ioas) { int fd; uint32_t ioas_id;
- uint32_t domain_id;
- uint32_t stdev_id;
Probably add a comment that a valid stdev_id implies a domain/hwpt under the ioas. with that...
TEST_F(iommufd_ioas, ioas_destroy) {
- if (self->domain_id) {
- if (self->stdev_id) { /* IOAS cannot be freed while a domain is on it */ EXPECT_ERRNO(EBUSY, _test_ioctl_destroy(self->fd, self->ioas_id));
it's easier to understand this comment.
Reviewed-by: Kevin Tian kevin.tian@intel.com
On Thu, Mar 02, 2023 at 01:23:48AM +0000, Tian, Kevin wrote:
{
- if (self->domain_id) {
- if (self->stdev_id) { /* IOAS cannot be freed while a domain is on it */ EXPECT_ERRNO(EBUSY, _test_ioctl_destroy(self->fd, self->ioas_id));
it's easier to understand this comment.
I changed this comment to use the right language:
if (self->stdev_id) { /* IOAS cannot be freed while a device has a HWPT using it */ EXPECT_ERRNO(EBUSY, _test_ioctl_destroy(self->fd, self->ioas_id));
Thanks, Jason
In this case the domain_id was acting as the hwpt_id, so be clearer.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- tools/testing/selftests/iommu/iommufd.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-)
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index 000fadf8110833..6a61ec84243572 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -988,8 +988,8 @@ FIXTURE(iommufd_mock_domain) { int fd; uint32_t ioas_id; - uint32_t domain_id; - uint32_t domain_ids[2]; + uint32_t hwpt_id; + uint32_t hwpt_ids[2]; int mmap_flags; size_t mmap_buf_size; }; @@ -1008,11 +1008,11 @@ FIXTURE_SETUP(iommufd_mock_domain) ASSERT_NE(-1, self->fd); test_ioctl_ioas_alloc(&self->ioas_id);
- ASSERT_GE(ARRAY_SIZE(self->domain_ids), variant->mock_domains); + ASSERT_GE(ARRAY_SIZE(self->hwpt_ids), variant->mock_domains);
for (i = 0; i != variant->mock_domains; i++) - test_cmd_mock_domain(self->ioas_id, NULL, &self->domain_ids[i]); - self->domain_id = self->domain_ids[0]; + test_cmd_mock_domain(self->ioas_id, NULL, &self->hwpt_ids[i]); + self->hwpt_id = self->hwpt_ids[0];
self->mmap_flags = MAP_SHARED | MAP_ANONYMOUS; self->mmap_buf_size = PAGE_SIZE * 8; @@ -1061,7 +1061,7 @@ FIXTURE_VARIANT_ADD(iommufd_mock_domain, two_domains_hugepage) struct iommu_test_cmd check_map_cmd = { \ .size = sizeof(check_map_cmd), \ .op = IOMMU_TEST_OP_MD_CHECK_MAP, \ - .id = self->domain_id, \ + .id = self->hwpt_id, \ .check_map = { .iova = _iova, \ .length = _length, \ .uptr = (uintptr_t)(_ptr) }, \ @@ -1070,8 +1070,8 @@ FIXTURE_VARIANT_ADD(iommufd_mock_domain, two_domains_hugepage) ioctl(self->fd, \ _IOMMU_TEST_CMD(IOMMU_TEST_OP_MD_CHECK_MAP), \ &check_map_cmd)); \ - if (self->domain_ids[1]) { \ - check_map_cmd.id = self->domain_ids[1]; \ + if (self->hwpt_ids[1]) { \ + check_map_cmd.id = self->hwpt_ids[1]; \ ASSERT_EQ(0, \ ioctl(self->fd, \ _IOMMU_TEST_CMD( \ @@ -1203,9 +1203,9 @@ TEST_F(iommufd_mock_domain, all_aligns_copy) test_ioctl_ioas_map(buf + start, length, &iova);
/* Add and destroy a domain while the area exists */ - old_id = self->domain_ids[1]; + old_id = self->hwpt_ids[1]; test_cmd_mock_domain(self->ioas_id, &mock_device_id, - &self->domain_ids[1]); + &self->hwpt_ids[1]);
check_mock_iova(buf + start, iova, length); check_refs(buf + start / PAGE_SIZE * PAGE_SIZE, @@ -1214,8 +1214,8 @@ TEST_F(iommufd_mock_domain, all_aligns_copy) 1);
test_ioctl_destroy(mock_device_id); - test_ioctl_destroy(self->domain_ids[1]); - self->domain_ids[1] = old_id; + test_ioctl_destroy(self->hwpt_ids[1]); + self->hwpt_ids[1] = old_id;
test_ioctl_ioas_unmap(iova, length); }
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
In this case the domain_id was acting as the hwpt_id, so be clearer.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
For consistency.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- tools/testing/selftests/iommu/iommufd.c | 12 ++++---- .../selftests/iommu/iommufd_fail_nth.c | 30 +++++++++---------- 2 files changed, 21 insertions(+), 21 deletions(-)
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index 6a61ec84243572..a9b232c281c02f 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -605,7 +605,7 @@ TEST_F(iommufd_ioas, access_pin) MOCK_FLAGS_ACCESS_CREATE_NEEDS_PIN_PAGES);
for (npages = 1; npages < BUFFER_SIZE / PAGE_SIZE; npages++) { - uint32_t mock_device_id; + uint32_t mock_stdev_id; uint32_t mock_hwpt_id;
access_cmd.access_pages.length = npages * PAGE_SIZE; @@ -637,14 +637,14 @@ TEST_F(iommufd_ioas, access_pin) ASSERT_EQ(0, ioctl(self->fd, _IOMMU_TEST_CMD(IOMMU_TEST_OP_ACCESS_PAGES), &access_cmd)); - test_cmd_mock_domain(self->ioas_id, &mock_device_id, + test_cmd_mock_domain(self->ioas_id, &mock_stdev_id, &mock_hwpt_id); check_map_cmd.id = mock_hwpt_id; ASSERT_EQ(0, ioctl(self->fd, _IOMMU_TEST_CMD(IOMMU_TEST_OP_MD_CHECK_MAP), &check_map_cmd));
- test_ioctl_destroy(mock_device_id); + test_ioctl_destroy(mock_stdev_id); test_ioctl_destroy(mock_hwpt_id); test_cmd_destroy_access_pages( access_cmd.id, @@ -1197,14 +1197,14 @@ TEST_F(iommufd_mock_domain, all_aligns_copy) for (; end < buf_size; end += MOCK_PAGE_SIZE) { size_t length = end - start; unsigned int old_id; - uint32_t mock_device_id; + uint32_t mock_stdev_id; __u64 iova;
test_ioctl_ioas_map(buf + start, length, &iova);
/* Add and destroy a domain while the area exists */ old_id = self->hwpt_ids[1]; - test_cmd_mock_domain(self->ioas_id, &mock_device_id, + test_cmd_mock_domain(self->ioas_id, &mock_stdev_id, &self->hwpt_ids[1]);
check_mock_iova(buf + start, iova, length); @@ -1213,7 +1213,7 @@ TEST_F(iommufd_mock_domain, all_aligns_copy) start / PAGE_SIZE * PAGE_SIZE, 1);
- test_ioctl_destroy(mock_device_id); + test_ioctl_destroy(mock_stdev_id); test_ioctl_destroy(self->hwpt_ids[1]); self->hwpt_ids[1] = old_id;
diff --git a/tools/testing/selftests/iommu/iommufd_fail_nth.c b/tools/testing/selftests/iommu/iommufd_fail_nth.c index 9713111b820dd7..e7d535680721b2 100644 --- a/tools/testing/selftests/iommu/iommufd_fail_nth.c +++ b/tools/testing/selftests/iommu/iommufd_fail_nth.c @@ -297,7 +297,7 @@ TEST_FAIL_NTH(basic_fail_nth, basic) TEST_FAIL_NTH(basic_fail_nth, map_domain) { uint32_t ioas_id; - __u32 device_id; + __u32 stdev_id; __u32 hwpt_id; __u64 iova;
@@ -313,7 +313,7 @@ TEST_FAIL_NTH(basic_fail_nth, map_domain)
fail_nth_enable();
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id, &hwpt_id)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1;
if (_test_ioctl_ioas_map(self->fd, ioas_id, buffer, 262144, &iova, @@ -321,12 +321,12 @@ TEST_FAIL_NTH(basic_fail_nth, map_domain) IOMMU_IOAS_MAP_READABLE)) return -1;
- if (_test_ioctl_destroy(self->fd, device_id)) + if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; if (_test_ioctl_destroy(self->fd, hwpt_id)) return -1;
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id, &hwpt_id)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1; return 0; } @@ -334,8 +334,8 @@ TEST_FAIL_NTH(basic_fail_nth, map_domain) TEST_FAIL_NTH(basic_fail_nth, map_two_domains) { uint32_t ioas_id; - __u32 device_id2; - __u32 device_id; + __u32 stdev_id2; + __u32 stdev_id; __u32 hwpt_id2; __u32 hwpt_id; __u64 iova; @@ -350,12 +350,12 @@ TEST_FAIL_NTH(basic_fail_nth, map_two_domains) if (_test_ioctl_set_temp_memory_limit(self->fd, 32)) return -1;
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id, &hwpt_id)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1;
fail_nth_enable();
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id2, &hwpt_id2)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id2, &hwpt_id2)) return -1;
if (_test_ioctl_ioas_map(self->fd, ioas_id, buffer, 262144, &iova, @@ -363,19 +363,19 @@ TEST_FAIL_NTH(basic_fail_nth, map_two_domains) IOMMU_IOAS_MAP_READABLE)) return -1;
- if (_test_ioctl_destroy(self->fd, device_id)) + if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; if (_test_ioctl_destroy(self->fd, hwpt_id)) return -1;
- if (_test_ioctl_destroy(self->fd, device_id2)) + if (_test_ioctl_destroy(self->fd, stdev_id2)) return -1; if (_test_ioctl_destroy(self->fd, hwpt_id2)) return -1;
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id, &hwpt_id)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1; - if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id2, &hwpt_id2)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id2, &hwpt_id2)) return -1; return 0; } @@ -518,7 +518,7 @@ TEST_FAIL_NTH(basic_fail_nth, access_pin_domain) { uint32_t access_pages_id; uint32_t ioas_id; - __u32 device_id; + __u32 stdev_id; __u32 hwpt_id; __u64 iova;
@@ -532,7 +532,7 @@ TEST_FAIL_NTH(basic_fail_nth, access_pin_domain) if (_test_ioctl_set_temp_memory_limit(self->fd, 32)) return -1;
- if (_test_cmd_mock_domain(self->fd, ioas_id, &device_id, &hwpt_id)) + if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1;
if (_test_ioctl_ioas_map(self->fd, ioas_id, buffer, BUFFER_SIZE, &iova, @@ -570,7 +570,7 @@ TEST_FAIL_NTH(basic_fail_nth, access_pin_domain) return -1; self->access_id = 0;
- if (_test_ioctl_destroy(self->fd, device_id)) + if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; if (_test_ioctl_destroy(self->fd, hwpt_id)) return -1;
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
For consistency.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
iommufd wants to use more infrastructure, like the iommu_group, that the mock device does not support. Create a more complete mock device that can go through the whole cycle of ownership, blocking domain, and has an iommu_group.
This requires creating a real struct device on a real bus to be able to connect it to a iommu_group. Unfortunately we cannot formally attach the mock iommu driver as an actual driver as the iommu core does not allow more than one driver or provide a general way for busses to link to iommus. This can be solved with a little hack to open code the dev_iommus struct.
With this infrastructure things work exactly the same as the normal domain path, including the auto domains mechanism and direct attach of hwpts. As the created hwpt is now an autodomain it is no longer required to destroy it and trying to do so will trigger a failure.
Reviewed-by: Kevin Tian kevin.tian@intel.com Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- drivers/iommu/iommufd/device.c | 40 +--- drivers/iommu/iommufd/iommufd_private.h | 11 +- drivers/iommu/iommufd/selftest.c | 196 +++++++++++++++--- tools/testing/selftests/iommu/iommufd.c | 2 - .../selftests/iommu/iommufd_fail_nth.c | 8 - 5 files changed, 181 insertions(+), 76 deletions(-)
diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c index 8dc7ed678e3fbb..c6f4852a8a0c08 100644 --- a/drivers/iommu/iommufd/device.c +++ b/drivers/iommu/iommufd/device.c @@ -22,7 +22,8 @@ void iommufd_device_destroy(struct iommufd_object *obj)
iommu_device_release_dma_owner(idev->dev); iommu_group_put(idev->group); - iommufd_ctx_put(idev->ictx); + if (!iommufd_selftest_is_mock_dev(idev->dev)) + iommufd_ctx_put(idev->ictx); }
/** @@ -69,7 +70,8 @@ struct iommufd_device *iommufd_device_bind(struct iommufd_ctx *ictx, goto out_release_owner; } idev->ictx = ictx; - iommufd_ctx_get(ictx); + if (!iommufd_selftest_is_mock_dev(dev)) + iommufd_ctx_get(ictx); idev->dev = dev; idev->enforce_cache_coherency = device_iommu_capable(dev, IOMMU_CAP_ENFORCE_CACHE_COHERENCY); @@ -151,7 +153,8 @@ static int iommufd_device_setup_msi(struct iommufd_device *idev, * operation from the device (eg a simple DMA) cannot trigger an * interrupt outside this iommufd context. */ - if (!iommu_group_has_isolated_msi(idev->group)) { + if (!iommufd_selftest_is_mock_dev(idev->dev) && + !iommu_group_has_isolated_msi(idev->group)) { if (!allow_unsafe_interrupts) return -EPERM;
@@ -706,34 +709,3 @@ int iommufd_access_rw(struct iommufd_access *access, unsigned long iova, return rc; } EXPORT_SYMBOL_NS_GPL(iommufd_access_rw, IOMMUFD); - -#ifdef CONFIG_IOMMUFD_TEST -/* - * Creating a real iommufd_device is too hard, bypass creating a iommufd_device - * and go directly to attaching a domain. - */ -struct iommufd_hw_pagetable * -iommufd_device_selftest_attach(struct iommufd_ctx *ictx, - struct iommufd_ioas *ioas, - struct device *mock_dev) -{ - struct iommufd_device tmp_idev = { .dev = mock_dev }; - struct iommufd_hw_pagetable *hwpt; - - mutex_lock(&ioas->mutex); - hwpt = iommufd_hw_pagetable_alloc(ictx, ioas, &tmp_idev, false); - mutex_unlock(&ioas->mutex); - if (IS_ERR(hwpt)) - return hwpt; - - refcount_inc(&hwpt->obj.users); - iommufd_object_finalize(ictx, &hwpt->obj); - return hwpt; -} - -void iommufd_device_selftest_detach(struct iommufd_ctx *ictx, - struct iommufd_hw_pagetable *hwpt) -{ - refcount_dec(&hwpt->obj.users); -} -#endif diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h index 331664e917b771..d523ef12890e1e 100644 --- a/drivers/iommu/iommufd/iommufd_private.h +++ b/drivers/iommu/iommufd/iommufd_private.h @@ -297,12 +297,6 @@ void iopt_remove_access(struct io_pagetable *iopt, void iommufd_access_destroy_object(struct iommufd_object *obj);
#ifdef CONFIG_IOMMUFD_TEST -struct iommufd_hw_pagetable * -iommufd_device_selftest_attach(struct iommufd_ctx *ictx, - struct iommufd_ioas *ioas, - struct device *mock_dev); -void iommufd_device_selftest_detach(struct iommufd_ctx *ictx, - struct iommufd_hw_pagetable *hwpt); int iommufd_test(struct iommufd_ucmd *ucmd); void iommufd_selftest_destroy(struct iommufd_object *obj); extern size_t iommufd_test_memory_limit; @@ -311,6 +305,7 @@ void iommufd_test_syz_conv_iova_id(struct iommufd_ucmd *ucmd, bool iommufd_should_fail(void); void __init iommufd_test_init(void); void iommufd_test_exit(void); +bool iommufd_selftest_is_mock_dev(struct device *dev); #else static inline void iommufd_test_syz_conv_iova_id(struct iommufd_ucmd *ucmd, unsigned int ioas_id, @@ -327,5 +322,9 @@ static inline void __init iommufd_test_init(void) static inline void iommufd_test_exit(void) { } +static inline bool iommufd_selftest_is_mock_dev(struct device *dev) +{ + return false; +} #endif #endif diff --git a/drivers/iommu/iommufd/selftest.c b/drivers/iommu/iommufd/selftest.c index e05b41059630a7..17cb7b95eb2757 100644 --- a/drivers/iommu/iommufd/selftest.c +++ b/drivers/iommu/iommufd/selftest.c @@ -91,23 +91,50 @@ enum selftest_obj_type { TYPE_IDEV, };
+struct mock_dev { + struct device dev; +}; + struct selftest_obj { struct iommufd_object obj; enum selftest_obj_type type;
union { struct { - struct iommufd_hw_pagetable *hwpt; + struct iommufd_device *idev; struct iommufd_ctx *ictx; - struct device mock_dev; + struct mock_dev *mock_dev; } idev; }; };
+static void mock_domain_blocking_free(struct iommu_domain *domain) +{ +} + +static int mock_domain_nop_attach(struct iommu_domain *domain, + struct device *dev) +{ + return 0; +} + +static const struct iommu_domain_ops mock_blocking_ops = { + .free = mock_domain_blocking_free, + .attach_dev = mock_domain_nop_attach, +}; + +static struct iommu_domain mock_blocking_domain = { + .type = IOMMU_DOMAIN_BLOCKED, + .ops = &mock_blocking_ops, +}; + static struct iommu_domain *mock_domain_alloc(unsigned int iommu_domain_type) { struct mock_iommu_domain *mock;
+ if (iommu_domain_type == IOMMU_DOMAIN_BLOCKED) + return &mock_blocking_domain; + if (WARN_ON(iommu_domain_type != IOMMU_DOMAIN_UNMANAGED)) return NULL;
@@ -236,19 +263,39 @@ static phys_addr_t mock_domain_iova_to_phys(struct iommu_domain *domain, return (xa_to_value(ent) & MOCK_PFN_MASK) * MOCK_IO_PAGE_SIZE; }
+static bool mock_domain_capable(struct device *dev, enum iommu_cap cap) +{ + return cap == IOMMU_CAP_CACHE_COHERENCY; +} + +static void mock_domain_set_plaform_dma_ops(struct device *dev) +{ + /* + * mock doesn't setup default domains because we can't hook into the + * normal probe path + */ +} + static const struct iommu_ops mock_ops = { .owner = THIS_MODULE, .pgsize_bitmap = MOCK_IO_PAGE_SIZE, .domain_alloc = mock_domain_alloc, + .capable = mock_domain_capable, + .set_platform_dma_ops = mock_domain_set_plaform_dma_ops, .default_domain_ops = &(struct iommu_domain_ops){ .free = mock_domain_free, + .attach_dev = mock_domain_nop_attach, .map_pages = mock_domain_map_pages, .unmap_pages = mock_domain_unmap_pages, .iova_to_phys = mock_domain_iova_to_phys, }, };
+struct iommu_device mock_iommu_device = { + .ops = &mock_ops, +}; + static inline struct iommufd_hw_pagetable * get_md_pagetable(struct iommufd_ucmd *ucmd, u32 mockpt_id, struct mock_iommu_domain **mock) @@ -269,48 +316,142 @@ get_md_pagetable(struct iommufd_ucmd *ucmd, u32 mockpt_id, return hwpt; }
+static struct bus_type iommufd_mock_bus_type = { + .name = "iommufd_mock", + .iommu_ops = &mock_ops, +}; + +static void mock_dev_release(struct device *dev) +{ + struct mock_dev *mdev = container_of(dev, struct mock_dev, dev); + + kfree(mdev); +} + +static struct mock_dev *mock_dev_create(void) +{ + struct iommu_group *iommu_group; + struct dev_iommu *dev_iommu; + struct mock_dev *mdev; + int rc; + + mdev = kzalloc(sizeof(*mdev), GFP_KERNEL); + if (!mdev) + return ERR_PTR(-ENOMEM); + + device_initialize(&mdev->dev); + mdev->dev.release = mock_dev_release; + mdev->dev.bus = &iommufd_mock_bus_type; + + iommu_group = iommu_group_alloc(); + if (IS_ERR(iommu_group)) { + rc = PTR_ERR(iommu_group); + goto err_put; + } + + rc = dev_set_name(&mdev->dev, "iommufd_mock%u", + iommu_group_id(iommu_group)); + if (rc) + goto err_group; + + /* + * The iommu core has no way to associate a single device with an iommu + * driver (heck currently it can't even support two iommu_drivers + * registering). Hack it together with an open coded dev_iommu_get(). + * Notice that the normal notifier triggered iommu release process also + * does not work here because this bus is not in iommu_buses. + */ + mdev->dev.iommu = kzalloc(sizeof(*dev_iommu), GFP_KERNEL); + if (!mdev->dev.iommu) { + rc = -ENOMEM; + goto err_group; + } + mutex_init(&mdev->dev.iommu->lock); + mdev->dev.iommu->iommu_dev = &mock_iommu_device; + + rc = device_add(&mdev->dev); + if (rc) + goto err_dev_iommu; + + rc = iommu_group_add_device(iommu_group, &mdev->dev); + if (rc) + goto err_del; + iommu_group_put(iommu_group); + return mdev; + +err_del: + device_del(&mdev->dev); +err_dev_iommu: + kfree(mdev->dev.iommu); + mdev->dev.iommu = NULL; +err_group: + iommu_group_put(iommu_group); +err_put: + put_device(&mdev->dev); + return ERR_PTR(rc); +} + +static void mock_dev_destroy(struct mock_dev *mdev) +{ + iommu_group_remove_device(&mdev->dev); + device_del(&mdev->dev); + kfree(mdev->dev.iommu); + mdev->dev.iommu = NULL; + put_device(&mdev->dev); +} + +bool iommufd_selftest_is_mock_dev(struct device *dev) +{ + return dev->release == mock_dev_release; +} + /* Create an hw_pagetable with the mock domain so we can test the domain ops */ static int iommufd_test_mock_domain(struct iommufd_ucmd *ucmd, struct iommu_test_cmd *cmd) { - static struct bus_type mock_bus = { .iommu_ops = &mock_ops }; - struct iommufd_hw_pagetable *hwpt; + struct iommufd_device *idev; struct selftest_obj *sobj; - struct iommufd_ioas *ioas; + u32 pt_id = cmd->id; + u32 idev_id; int rc;
- ioas = iommufd_get_ioas(ucmd, cmd->id); - if (IS_ERR(ioas)) - return PTR_ERR(ioas); - sobj = iommufd_object_alloc(ucmd->ictx, sobj, IOMMUFD_OBJ_SELFTEST); - if (IS_ERR(sobj)) { - rc = PTR_ERR(sobj); - goto out_ioas; - } + if (IS_ERR(sobj)) + return PTR_ERR(sobj); + sobj->idev.ictx = ucmd->ictx; sobj->type = TYPE_IDEV; - sobj->idev.mock_dev.bus = &mock_bus;
- hwpt = iommufd_device_selftest_attach(ucmd->ictx, ioas, - &sobj->idev.mock_dev); - if (IS_ERR(hwpt)) { - rc = PTR_ERR(hwpt); + sobj->idev.mock_dev = mock_dev_create(); + if (IS_ERR(sobj->idev.mock_dev)) { + rc = PTR_ERR(sobj->idev.mock_dev); goto out_sobj; } - sobj->idev.hwpt = hwpt;
- /* Userspace must destroy both of these IDs to destroy the object */ - cmd->mock_domain.out_hwpt_id = hwpt->obj.id; + idev = iommufd_device_bind(ucmd->ictx, &sobj->idev.mock_dev->dev, + &idev_id); + if (IS_ERR(idev)) { + rc = PTR_ERR(idev); + goto out_mdev; + } + sobj->idev.idev = idev; + + rc = iommufd_device_attach(idev, &pt_id); + if (rc) + goto out_unbind; + + /* Userspace must destroy the device_id to destroy the object */ + cmd->mock_domain.out_hwpt_id = pt_id; cmd->mock_domain.out_stdev_id = sobj->obj.id; iommufd_object_finalize(ucmd->ictx, &sobj->obj); - iommufd_put_object(&ioas->obj); return iommufd_ucmd_respond(ucmd, sizeof(*cmd));
+out_unbind: + iommufd_device_unbind(idev); +out_mdev: + mock_dev_destroy(sobj->idev.mock_dev); out_sobj: iommufd_object_abort(ucmd->ictx, &sobj->obj); -out_ioas: - iommufd_put_object(&ioas->obj); return rc; }
@@ -780,8 +921,9 @@ void iommufd_selftest_destroy(struct iommufd_object *obj)
switch (sobj->type) { case TYPE_IDEV: - iommufd_device_selftest_detach(sobj->idev.ictx, - sobj->idev.hwpt); + iommufd_device_detach(sobj->idev.idev); + iommufd_device_unbind(sobj->idev.idev); + mock_dev_destroy(sobj->idev.mock_dev); break; } } @@ -845,9 +987,11 @@ void __init iommufd_test_init(void) { dbgfs_root = fault_create_debugfs_attr("fail_iommufd", NULL, &fail_iommufd); + WARN_ON(bus_register(&iommufd_mock_bus_type)); }
void iommufd_test_exit(void) { debugfs_remove_recursive(dbgfs_root); + bus_unregister(&iommufd_mock_bus_type); } diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index a9b232c281c02f..113dab3a0cd631 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -645,7 +645,6 @@ TEST_F(iommufd_ioas, access_pin) &check_map_cmd));
test_ioctl_destroy(mock_stdev_id); - test_ioctl_destroy(mock_hwpt_id); test_cmd_destroy_access_pages( access_cmd.id, access_cmd.access_pages.out_access_pages_id); @@ -1214,7 +1213,6 @@ TEST_F(iommufd_mock_domain, all_aligns_copy) 1);
test_ioctl_destroy(mock_stdev_id); - test_ioctl_destroy(self->hwpt_ids[1]); self->hwpt_ids[1] = old_id;
test_ioctl_ioas_unmap(iova, length); diff --git a/tools/testing/selftests/iommu/iommufd_fail_nth.c b/tools/testing/selftests/iommu/iommufd_fail_nth.c index e7d535680721b2..d9afcb23810e1a 100644 --- a/tools/testing/selftests/iommu/iommufd_fail_nth.c +++ b/tools/testing/selftests/iommu/iommufd_fail_nth.c @@ -323,8 +323,6 @@ TEST_FAIL_NTH(basic_fail_nth, map_domain)
if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; - if (_test_ioctl_destroy(self->fd, hwpt_id)) - return -1;
if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1; @@ -365,13 +363,9 @@ TEST_FAIL_NTH(basic_fail_nth, map_two_domains)
if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; - if (_test_ioctl_destroy(self->fd, hwpt_id)) - return -1;
if (_test_ioctl_destroy(self->fd, stdev_id2)) return -1; - if (_test_ioctl_destroy(self->fd, hwpt_id2)) - return -1;
if (_test_cmd_mock_domain(self->fd, ioas_id, &stdev_id, &hwpt_id)) return -1; @@ -572,8 +566,6 @@ TEST_FAIL_NTH(basic_fail_nth, access_pin_domain)
if (_test_ioctl_destroy(self->fd, stdev_id)) return -1; - if (_test_ioctl_destroy(self->fd, hwpt_id)) - return -1; return 0; }
This can now be covered since we have a full struct device.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com --- tools/testing/selftests/iommu/iommufd.c | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/iommu/iommufd.c b/tools/testing/selftests/iommu/iommufd.c index 113dab3a0cd631..7aec1d5386d9bf 100644 --- a/tools/testing/selftests/iommu/iommufd.c +++ b/tools/testing/selftests/iommu/iommufd.c @@ -187,6 +187,7 @@ FIXTURE(iommufd_ioas) int fd; uint32_t ioas_id; uint32_t stdev_id; + uint32_t hwpt_id; uint64_t base_iova; };
@@ -212,7 +213,8 @@ FIXTURE_SETUP(iommufd_ioas) }
for (i = 0; i != variant->mock_domains; i++) { - test_cmd_mock_domain(self->ioas_id, &self->stdev_id, NULL); + test_cmd_mock_domain(self->ioas_id, &self->stdev_id, + &self->hwpt_id); self->base_iova = MOCK_APERTURE_START; } } @@ -259,6 +261,16 @@ TEST_F(iommufd_ioas, ioas_destroy) } }
+TEST_F(iommufd_ioas, hwpt_attach) +{ + /* Create a device attached directly to a hwpt */ + if (self->stdev_id) { + test_cmd_mock_domain(self->hwpt_id, NULL, NULL); + } else { + test_err_mock_domain(ENOENT, self->hwpt_id, NULL, NULL); + } +} + TEST_F(iommufd_ioas, ioas_area_destroy) { /* Adding an area does not change ability to destroy */
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, March 2, 2023 3:30 AM
This can now be covered since we have a full struct device.
Signed-off-by: Jason Gunthorpe jgg@nvidia.com
Reviewed-by: Kevin Tian kevin.tian@intel.com
On Wed, Mar 01, 2023 at 03:30:17PM -0400, Jason Gunthorpe wrote:
As discussed here is a small series to address the confusing lifetime scheme for the hwpt. This was some leftover from prior rework that was never fully cleaned up.
Make it clear that the ioas and ioas->hwpt_list are associated with the hwpt during creation and never changed until it is destroyed. A hwpt with a positive reference count is always valid for device attachment.
This also improves the selftest handling of the mock domains so that we can implement more testing for the hwpt model.
This is a step toward the nesting and replace series.
I have this on github:
https://github.com/jgunthorpe/linux/commits/iommufd_hwpt
v3:
- Split patch to just make the hwpt_item handling consistent
- Remove two confusing comments
- Four new patches to clean up the confusing 'domain_id' in the selftest
v2: https://lore.kernel.org/r/0-v2-406f7ac07936+6a-iommufd_hwpt_jgg@nvidia.com
- Basically completely different
- Add a test for HWPT cases by enhancing the mock iommu driver to be more complete
- Move the device attachment as well into iommufd_hw_pagetable_alloc() so destroy is more symmetric
v1: https://lore.kernel.org/r/0-v1-4336b5cb2fe4+1d7-iommufd_hwpt_jgg@nvidia.com
Applied to iommufd for-next with the noted comment update
Thanks, Jason
linux-kselftest-mirror@lists.linaro.org