Prevent RESV_DIRECT devices from user assignment
Subsystems (3)
Lu Baolu
maintainerv2 cover Prevent RESV_DIRECT devices from user assignment
These are follow-up patches on this discussion:
I just summarized the ideas and code into a real patch series. Please help to review and merge.
Change log:
v2:
- Move "pg_size == 0" check out of the loop.
- Rebase on the top of v6.5-rc1.
v1: https://lore.kernel.org/linux-iommu/[email protected]/
Best regards,
baolu
Lu Baolu (2):
iommu: Prevent RESV_DIRECT devices from blocking domains iommu/vt-d: Remove rmrr check in domain attaching device path
include/linux/iommu.h | 2 ++ drivers/iommu/intel/iommu.c | 58 ------------------------------------- drivers/iommu/iommu.c | 37 ++++++++++++++++------- 3 files changed, 29 insertions(+), 68 deletions(-)
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH v2 1/2 iommu: Prevent RESV_DIRECT devices from blocking domains
The IOMMU_RESV_DIRECT flag indicates that a memory region must be mapped 1:1 at all times. This means that the region must always be accessible to the device, even if the device is attached to a blocking domain. This is equal to saying that IOMMU_RESV_DIRECT flag prevents devices from being attached to blocking domains.
This also implies that devices that implement RESV_DIRECT regions will be prevented from being assigned to user space since taking the DMA ownership immediately switches to a blocking domain.
The rule of preventing devices with the IOMMU_RESV_DIRECT regions from being assigned to user space has existed in the Intel IOMMU driver for a long time. Now, this rule is being lifted up to a general core rule, as other architectures like AMD and ARM also have RMRR-like reserved regions. This has been discussed in the community mailing list and refer to below link for more details.
Other places using unmanaged domains for kernel DMA must follow the iommu_get_resv_regions() and setup IOMMU_RESV_DIRECT - we do not restrict them in the core code.
--- include/linux/iommu.h | 2 ++ drivers/iommu/iommu.c | 37 +++++++++++++++++++++++++++---------- 2 files changed, 29 insertions(+), 10 deletions(-)
include/linux/iommu.h +2 -0
@@ -409,6 +409,7 @@ struct iommu_fault_param {* @priv: IOMMU Driver private data * @max_pasids: number of PASIDs this device can consume * @attach_deferred: the dma domain attachment is deferred +* @requires_direct: The driver requested IOMMU_RESV_DIRECT * * TODO: migrate other per device data pointers under iommu_dev_data, e.g. * struct iommu_group *iommu_group;
@@ -422,6 +423,7 @@ struct dev_iommu {void *priv;u32 max_pasids; u32 attach_deferred:1; + u32 requires_direct:1; }; int iommu_device_register(struct iommu_device *iommu,
drivers/iommu/iommu.c +27 -10
@@ -959,14 +959,12 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain,unsigned long pg_size;int ret =0 ; -if (!iommu_is_dma_domain (domain)) -return 0 ; - -BUG_ON (!domain->pgsize_bitmap); - - pg_size =1UL <<__ffs (domain->pgsize_bitmap); + pg_size = domain->pgsize_bitmap ?1UL <<__ffs (domain->pgsize_bitmap) :0 ;INIT_LIST_HEAD (&mappings); +if (WARN_ON_ONCE (iommu_is_dma_domain (domain) && !pg_size)) +return -EINVAL ; +iommu_get_resv_regions (dev, &mappings);/* We need to consider overlapping regions for different devices */
@@ -974,13 +972,17 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain,dma_addr_t start, end, addr;size_t map_size =0 ; +if (entry->type ==IOMMU_RESV_DIRECT ) + dev->iommu->requires_direct =1 ; + +if ((entry->type !=IOMMU_RESV_DIRECT && + entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) || + !iommu_is_dma_domain (domain)) +continue ; + start =ALIGN (entry->start, pg_size); end =ALIGN (entry->start + entry->length, pg_size); -if (entry->type !=IOMMU_RESV_DIRECT && - entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) -continue ; -for (addr = start; addr <= end; addr += pg_size) {phys_addr_t phys_addr;
@@ -2121,6 +2123,21 @@ static int __iommu_device_set_domain(struct iommu_group *group, {int ret; +/* +* If the driver has requested IOMMU_RESV_DIRECT then we cannot allow +* the blocking domain to be attached as it does not contain the +* required 1:1 mapping. This test effectively exclusive the device from +* being used with iommu_group_claim_dma_owner() which will block vfio +* and iommufd as well. +*/ +if (dev->iommu->requires_direct && + (new_domain->type ==IOMMU_DOMAIN_BLOCKED || + new_domain == group->blocking_domain)) { +dev_warn (dev, +"Firmware has requested this device have a 1:1 IOMMU mapping, rejecting configuring the device without a 1:1 mapping. Contact your platform vendor.\n" ); +return -EINVAL ; + } +if (dev->iommu->attach_deferred) {if (new_domain == group->default_domain)return 0 ;
Signature · 1 line
2.34.1
Lu Baolu
maintainerPATCH v2 2/2 iommu/vt-d: Remove rmrr check in domain attaching device path
In reply to Lu BaoluThe core code now prevents devices with RMRR regions from being assigned to user space. There is no need to check for this condition in individual drivers. Remove it to avoid duplicate code.
--- drivers/iommu/intel/iommu.c | 58 ------------------------------------- 1 file changed, 58 deletions(-)
drivers/iommu/intel/iommu.c +0 -58
@@ -2446,30 +2446,6 @@ static int dmar_domain_attach_device(struct dmar_domain *domain,return 0 ; } -static bool device_has_rmrr (struct device *dev) -{ -struct dmar_rmrr_unit *rmrr; -struct device *tmp; -int i; - -rcu_read_lock (); -for_each_rmrr_units (rmrr) { -/* -* Return TRUE if this RMRR contains the device that -* is passed in. -*/ -for_each_active_dev_scope (rmrr->devices, - rmrr->devices_cnt, i, tmp) -if (tmp == dev || -is_downstream_to_pci_bridge (dev, tmp)) { -rcu_read_unlock (); -return true; - } - } -rcu_read_unlock (); -return false; -} -/** * device_rmrr_is_relaxable - Test whether the RMRR of this device * is relaxable (ie. is allowed to be not enforced under some conditions)
@@ -2499,34 +2475,6 @@ static bool device_rmrr_is_relaxable(struct device *dev)return false; } -/* -* There are a couple cases where we need to restrict the functionality of -* devices associated with RMRRs. The first is when evaluating a device for -* identity mapping because problems exist when devices are moved in and out -* of domains and their respective RMRR information is lost. This means that -* a device with associated RMRRs will never be in a "passthrough" domain. -* The second is use of the device through the IOMMU API. This interface -* expects to have full control of the IOVA space for the device. We cannot -* satisfy both the requirement that RMRR access is maintained and have an -* unencumbered IOVA space. We also have no ability to quiesce the device's -* use of the RMRR space or even inform the IOMMU API user of the restriction. -* We therefore prevent devices associated with an RMRR from participating in -* the IOMMU API, which eliminates them from device assignment. -* -* In both cases, devices which have relaxable RMRRs are not concerned by this -* restriction. See device_rmrr_is_relaxable comment. -*/ -static bool device_is_rmrr_locked (struct device *dev) -{ -if (!device_has_rmrr (dev)) -return false; - -if (device_rmrr_is_relaxable (dev)) -return false; - -return true; -} -/* * Return the required default domain type for a specific device. *
@@ -4139,12 +4087,6 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,struct device_domain_info *info =dev_iommu_priv_get (dev);int ret; -if (domain->type ==IOMMU_DOMAIN_UNMANAGED && -device_is_rmrr_locked (dev)) { -dev_warn (dev,"Device is ineligible for IOMMU domain attach due to platform RMRR requirement. Contact your platform vendor.\n" ); -return -EPERM ; - } -if (info->domain)device_block_translation (dev);
Signature · 1 line
2.34.1
RE: [PATCH v2 1/2] iommu: Prevent RESV_DIRECT devices from blocking domains
In reply to Lu Baoluinclude/linux/iommu.h struct iommu_fault_param { · line 412
Quoted 8 lines
From: Lu Baolu <[email protected]>
Sent: Thursday, July 13, 2023 12:33 PMinclude/linux/iommu.h +1
-0@@ -409,6 +409,7 @@ struct iommu_fault_param {* @priv: IOMMU Driver private data * @max_pasids: number of PASIDs this device can consume * @attach_deferred: the dma domain attachment is deferred +* @requires_direct: The driver requested IOMMU_RESV_DIRECT
it's not accurate to say "driver requested" as it's a device attribute.
s/requires_direct/require_direct/
what about "has_resv_direct"?
Quoted 34 lines
drivers/iommu/iommu.c +4
-6@@ -959,14 +959,12 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, unsigned long pg_size;int ret =0 ; -if (!iommu_is_dma_domain (domain)) -return 0 ; - -BUG_ON (!domain->pgsize_bitmap); - - pg_size =1UL <<__ffs (domain->pgsize_bitmap); + pg_size = domain->pgsize_bitmap ?1UL <<__ffs (domain- >pgsize_bitmap) : 0;INIT_LIST_HEAD (&mappings); +if (WARN_ON_ONCE (iommu_is_dma_domain (domain) && !pg_size)) +return -EINVAL ; +iommu_get_resv_regions (dev, &mappings);drivers/iommu/iommu.c +7
-0/* We need to consider overlapping regions for different devices */ @@ -974,13 +972,17 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, dma_addr_t start, end, addr;size_t map_size =0 ; +if (entry->type ==IOMMU_RESV_DIRECT ) + dev->iommu->requires_direct =1 ; + +if ((entry->type !=IOMMU_RESV_DIRECT && + entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) || + !iommu_is_dma_domain (domain)) +continue ;
drivers/iommu/iommu.c static int · line 988
Quoted 10 lines
drivers/iommu/iommu.c +1
-4+ start =ALIGN (entry->start, pg_size); end =ALIGN (entry->start + entry->length, pg_size); -if (entry->type !=IOMMU_RESV_DIRECT && - entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) -continue ; -for (addr = start; addr <= end; addr += pg_size) {phys_addr_t phys_addr;
piggybacking a device attribute detection in a function which tries to populate domain mappings is a bit confusing.
Does it work better to introduce a new function to detect this attribute and has it directly called in the probe path?
drivers/iommu/iommu.c static int __iommu_device_set_domain(struct · line 2128
Quoted 7 lines
drivers/iommu/iommu.c +2
-0@@ -2121,6 +2123,21 @@ static int __iommu_device_set_domain(struct iommu_group *group, {int ret; +/* +* If the driver has requested IOMMU_RESV_DIRECT then we cannot
ditto. It's not requested by the driver.
Review
Quoted 3 lines
allow
(quoted diff) +2
-0+ * the blocking domain to be attached as it does not contain the + * required 1:1 mapping. This test effectively exclusive the device
s/exclusive/excludes/
RE: [PATCH v2 2/2] iommu/vt-d: Remove rmrr check in domain attaching device path
In reply to Lu BaoluQuoted 9 lines
From: Lu Baolu <[email protected]>
Sent: Thursday, July 13, 2023 12:33 PMThe core code now prevents devices with RMRR regions from being assigned to user space. There is no need to check for this condition in individual drivers. Remove it to avoid duplicate code.
Signed-off-by Lu Baolu <[email protected]>Reviewed-by Jason Gunthorpe <[email protected]>
A side note.
device_rmrr_is_relaxable() marks USB/GPU as relaxable. I wonder whether this policy is specific to Intel platform or generally applied. If the latter probably that could be moved to iommu core too.
Re: [PATCH v2 1/2] iommu: Prevent RESV_DIRECT devices from blocking domains
In reply to Tian, KevinQuoted 5 lines · Kevin
Quoted 5 lines
drivers/iommu/iommu.c +0
-0@@ -974,13 +972,17 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, dma_addr_t start, end, addr;size_t map_size =0 ;
Quoted 10 lines
drivers/iommu/iommu.c static int iommu_create_device_direct_mappings(struct iommu_domain *domain, · line 981
Quoted 7 lines
drivers/iommu/iommu.c +7
-0+if (entry->type ==IOMMU_RESV_DIRECT ) + dev->iommu->requires_direct =1 ; + +if ((entry->type !=IOMMU_RESV_DIRECT && + entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) || + !iommu_is_dma_domain (domain)) +continue ;piggybacking a device attribute detection in a function which tries to populate domain mappings is a bit confusing.
It is, but to do otherwise we'd want to have the caller obtain the reserved regions list and iterate it twice. Not sure it is worth the trouble right now.
Jason
Re: [PATCH v2 1/2] iommu: Prevent RESV_DIRECT devices from blocking domains
In reply to Lu BaoluQuoted 27 lines · Lu Baolu
The
IOMMU_RESV_DIRECTflag indicates that a memory region must be mapped 1:1 at all times. This means that the region must always be accessible to the device, even if the device is attached to a blocking domain. This is equal to saying thatIOMMU_RESV_DIRECTflag prevents devices from being attached to blocking domains.This also implies that devices that implement
RESV_DIRECTregions will be prevented from being assigned to user space since taking the DMA ownership immediately switches to a blocking domain.The rule of preventing devices with the
IOMMU_RESV_DIRECTregions from being assigned to user space has existed in the Intel IOMMU driver for a long time. Now, this rule is being lifted up to a general core rule, as other architectures like AMD and ARM also have RMRR-like reserved regions. This has been discussed in the community mailing list and refer to below link for more details.Other places using unmanaged domains for kernel DMA must follow the
iommu_get_resv_regions()and setupIOMMU_RESV_DIRECT- we do not restrict them in the core code.Cc Robin Murphy <[email protected]>Cc Alex Williamson <[email protected]>Cc Kevin Tian <[email protected]>Signed-off-by Jason Gunthorpe <[email protected]>Link https://lore.kernel.org/linux-iommu/BN9PR11MB5276E84229B5BD952D78E9598C639@BN9PR11MB5276.namprd11.prod.outlook.com open hereSigned-off-by Lu Baolu <[email protected]>
Quoted 4 lines
--- include/linux/iommu.h | 2 ++ drivers/iommu/iommu.c | 37 +++++++++++++++++++++++++++---------- 2 files changed, 29 insertions(+), 10 deletions(-)
Jason
Baolu Lu
maintainerRe: [PATCH v2 1/2] iommu: Prevent RESV_DIRECT devices from blocking domains
In reply to Tian, KevinQuoted 11 lines · Kevin
include/linux/iommu.h struct iommu_fault_param { · line 412
Quoted 8 lines
From: Lu Baolu <[email protected]>
Sent: Thursday, July 13, 2023 12:33 PMinclude/linux/iommu.h +1
-0@@ -409,6 +409,7 @@ struct iommu_fault_param {* @priv: IOMMU Driver private data * @max_pasids: number of PASIDs this device can consume * @attach_deferred: the dma domain attachment is deferred +* @requires_direct: The driver requested IOMMU_RESV_DIRECT it's not accurate to say "driver requested" as it's a device attribute.
Quoted 3 lines
s/requires_direct/require_direct/
what about "
has_resv_direct"?
How about
* @require_direct: device requires IOMMU_RESV_DIRECT reserved regions
?
Quoted 49 lines
Quoted 35 lines
drivers/iommu/iommu.c +4
-6@@ -959,14 +959,12 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, unsigned long pg_size;int ret =0 ; -if (!iommu_is_dma_domain (domain)) -return 0 ; - -BUG_ON (!domain->pgsize_bitmap); - - pg_size =1UL <<__ffs (domain->pgsize_bitmap); + pg_size = domain->pgsize_bitmap ?1UL <<__ffs (domain- > pgsize_bitmap) : 0;INIT_LIST_HEAD (&mappings); +if (WARN_ON_ONCE (iommu_is_dma_domain (domain) && !pg_size)) +return -EINVAL ; +iommu_get_resv_regions (dev, &mappings);/* We need to consider overlapping regions for different devices */
drivers/iommu/iommu.c +7
-0@@ -974,13 +972,17 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, dma_addr_t start, end, addr;size_t map_size =0 ; +if (entry->type ==IOMMU_RESV_DIRECT ) + dev->iommu->requires_direct =1 ; + +if ((entry->type !=IOMMU_RESV_DIRECT && + entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) || + !iommu_is_dma_domain (domain)) +continue ;drivers/iommu/iommu.c static int · line 988
Quoted 10 lines
drivers/iommu/iommu.c +1
-4+ start =ALIGN (entry->start, pg_size); end =ALIGN (entry->start + entry->length, pg_size); -if (entry->type !=IOMMU_RESV_DIRECT && - entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) -continue ; -for (addr = start; addr <= end; addr += pg_size) {phys_addr_t phys_addr;piggybacking a device attribute detection in a function which tries to populate domain mappings is a bit confusing.
Quoted 2 lines
Does it work better to introduce a new function to detect this attribute and has it directly called in the probe path?
Jason answered this.
Quoted 11 lines
drivers/iommu/iommu.c static int __iommu_device_set_domain(struct · line 2128
Quoted 8 lines
drivers/iommu/iommu.c +2
-0@@ -2121,6 +2123,21 @@ static int __iommu_device_set_domain(struct iommu_group *group, {int ret; +/* +* If the driver has requested IOMMU_RESV_DIRECT then we cannot ditto. It's not requested by the driver.
Quoted 6 lines
Review
Quoted 3 lines
allow
(quoted diff) +2
-0+ * the blocking domain to be attached as it does not contain the + * required 1:1 mapping. This test effectively exclusive the devices/exclusive/excludes/
Updated. Thanks!
Best regards,
baolu
RE: [PATCH v2 1/2] iommu: Prevent RESV_DIRECT devices from blocking domains
In reply to Jason GunthorpeQuoted 20 lines
From: Jason Gunthorpe <[email protected]>
Sent: Friday, July 21, 2023 11:10 PMQuoted 5 lines · Kevin
Quoted 5 lines
drivers/iommu/iommu.c +0
-0@@ -974,13 +972,17 @@ static int iommu_create_device_direct_mappings(struct iommu_domain *domain, dma_addr_t start, end, addr;size_t map_size =0 ;Quoted 10 lines
drivers/iommu/iommu.c static int iommu_create_device_direct_mappings(struct iommu_domain *domain, · line 981
Quoted 7 lines
drivers/iommu/iommu.c +7
-0+if (entry->type ==IOMMU_RESV_DIRECT ) + dev->iommu->requires_direct =1 ; + +if ((entry->type !=IOMMU_RESV_DIRECT && + entry->type !=IOMMU_RESV_DIRECT_RELAXABLE ) || + !iommu_is_dma_domain (domain)) +continue ;piggybacking a device attribute detection in a function which tries to populate domain mappings is a bit confusing.
Quoted 4 lines
It is, but to do otherwise we'd want to have the caller obtain the reserved regions list and iterate it twice. Not sure it is worth the trouble right now.
Not a strong opinion but It's a slow path and readability is more preferable to me. 😊