iommu: Add support to change default domain of an iommu
v3 cover iommu: Add support to change default domain of an iommu
Presently, the default domain of a iommu group is allocated during boot time and it cannot be changed later. So, the device would typically be either in identity (pass_through) mode or the device would be in DMA mode as long as the system is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identity mode and DMA mode.
Assume a use case wherein the privileged user would want to use the device in pass-through mode when the device is used for host so that it would be high performing. Presently, this is not supported. Hence add support to change the default domain of an iommu group dynamically.
Support this by writing to a sysfs file, namely
"/sys/kernel/iommu_groups/<grp_id>/type".
Testing:
--------
Tested by dynamically changing storage device (nvme) from
1. identity mode to DMA and making sure file transfer works
2. DMA mode to identity mode and making sure file transfer works Tested only for intel_iommu/vt-d. Would appreciate if someone could test on AMD and ARM based machines.
Based on iommu maintainer's 'next' branch.
Changes from V2:
----------------
1. Change the logic of updating default domain from V2 because ops->probe_finalize() could be used to update dma_ops.
2. Drop 1st and 2nd patch of V2 series because they are no longer needed on iommu maintainer's 'next' branch.
3. Limit this feature to iommu groups with only one device.
4. Hold device_lock and group mutex until the default domain is changed.
Changes from V1:
----------------
1. V1 patch set wasn't updating dma_ops for some vendors (Eg: AMD), hence, change the logic of updating default domain as below (because adding a device to iommu_group automatically updates dma_ops)
a. Allocate a new domain
b. For every device in the group
i. Remove the device from the group
ii. Add the device back to the group
c. Free previous domain
2. Drop 1st patch of V1 (iommu/vt-d: Modify device_def_domain_type() to use at runtime) because "iommu=pt" has no effect on this function anymore.
3. Added a patch to take/release lock while reading iommu_group->default_domain->type because it can be changed any time by user.
4. Before changing default domain type of a group, check if the group is directly assigned for user level access. If so, abort.
5. Sanitize return path (using ternary operator) in iommu_group_store_type()
6. Split 2nd patch of V1 (iommu: Add device_def_domain_type() call back function to iommu_ops) into two patches such that iommu generic changes are now in 1st patch of V2 and vt-d specific changes are in 2nd patch of V2.
7. Rename device_def_domain_type() to dev_def_domain_type()
8. Remove example from documentation
9. Change the value written to file "/sys/kernel/iommu_groups/<grp_id>/type" from "dma" to "DMA".
Changes from RFC:
-----------------
1. Added support for "auto" type, so that kernel selects one among identity or dma mode.
2. Use "system_state" in device_def_domain_type() instead of an argument.
Sai Praneeth Prakhya (3):
iommu: Add support to change default domain of an iommu_group
iommu: Take lock before reading iommu_group default domain type
iommu: Document usage of "/sys/kernel/iommu_groups/<grp_id>/type" file
.../ABI/testing/sysfs-kernel-iommu_groups | 30 +++ drivers/iommu/iommu.c | 213 +++++++++++++++++- 2 files changed, 242 insertions(+), 1 deletion(-)
Signature · 6 lines
2.19.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
PATCH v3 1/3 iommu: Add support to change default domain of an iommu_group
Presently, the default domain of an iommu_group is allocated during boot time (i.e. when a device is being added to a group) and it cannot be changed later. So, the device would typically be either in identity (also known as pass_through) mode (controlled by "iommu=pt" kernel command line argument) or the device would be in DMA mode as long as the machine is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identity mode and DMA mode.
But, assume a use case wherein the user trusts the device and believes that the OS is secure enough and hence wants *only* this device to bypass IOMMU (so that it could be high performing) whereas all the other devices to go through IOMMU (so that the system is protected). Presently, this use case is not supported. It will be helpful if there is some way to change the default domain of a B:D.F dynamically. Hence, add such support.
A privileged user could request the kernel to change the default domain type of a iommu_group by writing to "/sys/kernel/iommu_groups/<grp_id>/type" file. Presently, only three values are supported
1. identity: all the DMA transactions from the device in this group are *not* translated by the iommu
2. DMA: all the DMA transactions from the device in this group are translated by the iommu
3. auto: change to the type the device was booted with
Note:
1. Default domain of an iommu group with two or more devices cannot be changed.
2. The device in the iommu group shouldn't be bound to any driver.
3. The device shouldn't be assigned to user for direct access.
Please see "Documentation/ABI/testing/sysfs-kernel-iommu_groups" for more information.
--- drivers/iommu/iommu.c | 211 +++++++++++++++++++++++++++++++++++++++++- 1 file changed, 210 insertions(+), 1 deletion(-)
drivers/iommu/iommu.c +210 -1
@@ -92,6 +92,8 @@ static void __iommu_detach_group(struct iommu_domain *domain,static int iommu_create_device_direct_mappings (struct iommu_group *group,struct device *dev);static struct iommu_group *iommu_group_get_for_dev (struct device *dev); +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count);#define IOMMU_GROUP_ATTR (_name, _mode, _show, _store) \ struct iommu_group_attribute iommu_group_attr_##_name = \
@@ -524,7 +526,8 @@ static IOMMU_GROUP_ATTR(name, S_IRUGO, iommu_group_show_name, NULL); static IOMMU_GROUP_ATTR(reserved_regions, 0444, iommu_group_show_resv_regions, NULL); -static IOMMU_GROUP_ATTR(type, 0444, iommu_group_show_type, NULL); +static IOMMU_GROUP_ATTR(type, 0644, iommu_group_show_type, + iommu_group_store_type);static void iommu_group_release (struct kobject *kobj) {
@@ -2847,3 +2850,209 @@ int iommu_sva_get_pasid(struct iommu_sva *handle)return ops->sva_get_pasid (handle); }EXPORT_SYMBOL_GPL (iommu_sva_get_pasid); + +/* +* Changes the default domain of a device +* +* @dev: The *only* device in the group +* @group: The group for which the default domain should be changed +* @prev_domain: The previous domain that is being switched from +* @type: The type of the new default domain that gets associated with the group +* +* Returns 0 on success and error code on failure +* +* Note: +* 1. Presently, this function is called only when user requests to change the +* group's default domain type through /sys/kernel/iommu_groups/<grp_id>/type +* Please take a closer look if intended to use for other purposes. +* 2. Assumes that group->mutex is already taken and releases before returning +*/ +static int iommu_change_dev_def_domain (struct device *dev, +struct iommu_group *group, +struct iommu_domain *prev_dom,int type) +{ +int ret =0 ; + +/* Sets group->default_domain to the newly allocated domain */ + ret =iommu_group_alloc_default_domain (dev->bus, group, type); +if (ret) + goto out; + + ret =__iommu_attach_device (group->default_domain, dev); +if (ret) + goto free_new_domain; + + group->domain = group->default_domain; + + ret =iommu_create_device_direct_mappings (group, dev); +if (ret) + goto free_new_domain; + +/* +* Release the mutex here because ops->probe_finalize() call-back of +* some vendor IOMMU drivers calls arm_iommu_attach_device() which +* in-turn might call back into IOMMU core code, where it tries to take +* group->mutex, resulting in a deadlock. +*/ +mutex_unlock (&group->mutex); + +/* Make sure dma_ops is appropriatley set */ +iommu_group_do_probe_finalize (dev, group->default_domain); +iommu_domain_free (prev_dom); +return 0 ; + +free_new_domain: +iommu_domain_free (group->default_domain); + group->default_domain = prev_dom; + group->domain = prev_dom; + +out: +mutex_unlock (&group->mutex); + +return ret; +} + +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count) +{ +int ret, req_type =0 , req_auto =0 , dev_def_dom; +struct iommu_domain *prev_dom; +struct group_device *grp_dev; +const struct iommu_ops *ops; +struct device *dev; + +if (!capable (CAP_SYS_ADMIN ) || !capable (CAP_SYS_RAWIO )) +return -EACCES ; + +if (WARN_ON (!group)) +return -EINVAL ; + +if (sysfs_streq (buf,"identity" )) + req_type =IOMMU_DOMAIN_IDENTITY ; +else if (sysfs_streq (buf,"DMA" )) + req_type =IOMMU_DOMAIN_DMA ; +else if (sysfs_streq (buf,"auto" )) + req_auto =1 ; +else +return -EINVAL ; + +/* Lock the group to make sure that the device count doesn't change */ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { +mutex_unlock (&group->mutex); +pr_err ("Cannot change default domain of a group with two or more devices\n" ); +return -EINVAL ; + } + +/* Since group has only one device */ +list_for_each_entry (grp_dev, &group->devices, list) + dev = grp_dev->dev; + +/* +* Don't hold onto the group mutex because taking group mutex first and +* then the device lock could potentially cause a deadlock as below. +* Assume two threads T1 and T2. T1 is trying to change default domain +* of an iommu group and T2 is trying to hot unplug a device or release +* [1] VF of a PCIe device which is in the same iommu group. T1 takes +* group mutex and before it could take device lock assume T2 has taken +* device lock and is yet to take group mutex. Now, both the threads +* will be waiting for the other thread to release lock. +* +* [1] Typical device release path +* device_lock() from device/driver core code +* -> bus_notifier() +* -> iommu_bus_notifier() +* -> iommu_release_device() +* -> ops->release_device() vendor driver calls back iommu core code +* -> mutex_lock() from iommu core code +*/ +mutex_unlock (&group->mutex); + +/* Check if the device in the group still has a driver bound to it */ +device_lock (dev); +if (device_is_bound (dev)) { +pr_err ("Device is still bound to driver\n" ); + ret = -EBUSY ; + goto dev_unlock; + } + +/* +* iommu group wasn't locked while acquiring device lock. So, make sure +* that the device count hasn't changed while acquiring device lock. +* +* Changing default domain of an iommu group with two or more devices +* isn't supported because there could be a potential deadlock. Consider +* the following scenario. T1 is trying to acquire device locks of all +* the devices in the group and before it could acquire all of them, +* there could be another thread T2 (from different sub-system and use +* case) that has already acquired some of the device locks and might be +* waiting for T1 to release other device locks. +*/ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { +pr_err ("Cannot change default domain of a group with two or more devices\n" ); + ret = -EINVAL ; + goto group_unlock; + } + +/* +* Check if any user level driver (that doesn't use kernel driver like +* VFIO) is directly using the group. VFIO use case is detected by +* device_is_bound() check above +*/ +if (group->default_domain != group->domain) { +pr_err ("Group assigned to user level for direct access\n" ); + ret = -EINVAL ; + goto group_unlock; + } + + prev_dom = group->default_domain; +if (!prev_dom || !prev_dom->ops || !prev_dom->ops->def_domain_type) { +pr_err ("'def_domain_type' call back isn't registered\n" ); + ret = -EINVAL ; + goto group_unlock; + } + + ops = prev_dom->ops; + dev_def_dom = ops->def_domain_type (dev); + +/* Check if user requested domain is supported by the device or not */ +if (!req_auto && dev_def_dom && req_type != dev_def_dom) { +pr_err ("Device cannot be in %s domain\n" , buf); + ret = -EINVAL ; + goto group_unlock; + } + +if (req_auto) { +/* +* If user requested "auto" and the device supports both the +* domains, then default to the domain the device was booted +* with +*/ +if (!dev_def_dom) + req_type = iommu_def_domain_type; +else + req_type = dev_def_dom; + } + +/* +* Switch to a new domain only if the requested domain type is different +* from the existing default domain type +*/ +if (prev_dom->type == req_type) { + ret = count; + goto group_unlock; + } + + ret =iommu_change_dev_def_domain (dev, group, prev_dom, req_type); +device_unlock (dev); +return ret ?: count; + +group_unlock: +mutex_unlock (&group->mutex); + +dev_unlock: +device_unlock (dev); + +return ret; +}
Signature · 6 lines
2.19.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
Re: [PATCH V3 1/3] iommu: Add support to change default domain of an iommu_group
Hi Sai,
Quoted 2 lines · Sai Praneeth Prakhya
Presently, the default domain of an
iommu_groupis allocated during boot time (i.e. when a device is being added to a group) and it cannot be
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This is inaccurate as Joerg's code has deferred default domain allocation and attaching after group allocation. I'd suggest to remove this.
Quoted 2 lines
changed later. So, the device would typically be either in identity (also known as
pass_through) mode (controlled by "iommu=pt" kernel command line
^^^^^^^^
There are other kernel parameters to put device in pass_through mode. I'd suggest to remove this.
Quoted 11 lines
argument) or the device would be in DMA mode as long as the machine is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identity mode and DMA mode.
But, assume a use case wherein the user trusts the device and believes that the OS is secure enough and hence wants *only* this device to bypass IOMMU (so that it could be high performing) whereas all the other devices to go through IOMMU (so that the system is protected). Presently, this use case is not supported. It will be helpful if there is some way to change the default domain of a B:D.F dynamically. Hence, add such support.
^^^^^^^^^^^^^^^^^^^^^^^^^
Currently default domain is per iommu_group, we have no per device default domain yet. Probably, "default domain of an iommu group"?
Quoted 57 lines
A privileged user could request the kernel to change the default domain type of a
iommu_groupby writing to "/sys/kernel/iommu_groups/<grp_id>/type" file. Presently, only three values are supported
1. identity: all the DMA transactions from the device in this group are *not* translated by the iommu
2. DMA: all the DMA transactions from the device in this group are translated by the iommu
3. auto: change to the type the device was booted withNote:
1. Default domain of an iommu group with two or more devices cannot be changed.
2. The device in the iommu group shouldn't be bound to any driver.
3. The device shouldn't be assigned to user for direct access.Please see "
Documentation/ABI/testing/sysfs-kernel-iommu_groups" for more information.Cc Christoph Hellwig <[email protected]>Cc Joerg Roedel <[email protected]>Cc Ashok Raj <[email protected]>Cc Will Deacon <[email protected]>Cc Lu Baolu <[email protected]>Cc Sohil Mehta <[email protected]>Cc Robin Murphy <[email protected]>Cc Jacob Pan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>--- drivers/iommu/iommu.c | 211 +++++++++++++++++++++++++++++++++++++++++- 1 file changed, 210 insertions(+), 1 deletion(-)drivers/iommu/iommu.c +4
-1@@ -92,6 +92,8 @@ static void __iommu_detach_group(struct iommu_domain *domain,static int iommu_create_device_direct_mappings (struct iommu_group *group,struct device *dev);static struct iommu_group *iommu_group_get_for_dev (struct device *dev); +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count);#define IOMMU_GROUP_ATTR (_name, _mode, _show, _store) \ struct iommu_group_attribute iommu_group_attr_##_name = \@@ -524,7 +526,8 @@ static IOMMU_GROUP_ATTR(name, S_IRUGO, iommu_group_show_name, NULL); static IOMMU_GROUP_ATTR(reserved_regions, 0444, iommu_group_show_resv_regions, NULL); -static IOMMU_GROUP_ATTR(type, 0444, iommu_group_show_type, NULL); +static IOMMU_GROUP_ATTR(type, 0644, iommu_group_show_type, + iommu_group_store_type);static void iommu_group_release (struct kobject *kobj) {
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2857
Quoted 8 lines
drivers/iommu/iommu.c +5
-0@@ -2847,3 +2850,209 @@ int iommu_sva_get_pasid(struct iommu_sva *handle)return ops->sva_get_pasid (handle); }EXPORT_SYMBOL_GPL (iommu_sva_get_pasid); + +/* +* Changes the default domain of a device +* +* @dev: The *only* device in the group
All devices in a group have been linked in group->devices, so you don't need to pass it in as a parameter. Otherwise, you also need some sanity check against @dev, right?
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2859
Quoted 2 lines
drivers/iommu/iommu.c +2
-0+* @group: The group for which the default domain should be changed +* @prev_domain: The previous domain that is being switched from
The same as @dev, the default domain in use is saved in group->default_domain. No need a separated parameter to save space and sanity check.
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2868
Quoted 9 lines
drivers/iommu/iommu.c +9
-0+* @type: The type of the new default domain that gets associated with the group +* +* Returns 0 on success and error code on failure +* +* Note: +* 1. Presently, this function is called only when user requests to change the +* group's default domain type through /sys/kernel/iommu_groups/<grp_id>/type +* Please take a closer look if intended to use for other purposes. +* 2. Assumes that group->mutex is already taken and releases before returning
You could assume that the caller should hold the group->mutex and shouldn't be released in this helper. A reference design could be found here.
Quoted 57 lines
drivers/iommu/iommu.c +57
-0+*/ +static int iommu_change_dev_def_domain (struct device *dev, +struct iommu_group *group, +struct iommu_domain *prev_dom,int type) +{ +int ret =0 ; + +/* Sets group->default_domain to the newly allocated domain */ + ret =iommu_group_alloc_default_domain (dev->bus, group, type); +if (ret) + goto out; + + ret =__iommu_attach_device (group->default_domain, dev); +if (ret) + goto free_new_domain; + + group->domain = group->default_domain; + + ret =iommu_create_device_direct_mappings (group, dev); +if (ret) + goto free_new_domain; + +/* +* Release the mutex here because ops->probe_finalize() call-back of +* some vendor IOMMU drivers calls arm_iommu_attach_device() which +* in-turn might call back into IOMMU core code, where it tries to take +* group->mutex, resulting in a deadlock. +*/ +mutex_unlock (&group->mutex); + +/* Make sure dma_ops is appropriatley set */ +iommu_group_do_probe_finalize (dev, group->default_domain); +iommu_domain_free (prev_dom); +return 0 ; + +free_new_domain: +iommu_domain_free (group->default_domain); + group->default_domain = prev_dom; + group->domain = prev_dom; + +out: +mutex_unlock (&group->mutex); + +return ret; +} + +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count) +{ +int ret, req_type =0 , req_auto =0 , dev_def_dom; +struct iommu_domain *prev_dom; +struct group_device *grp_dev; +const struct iommu_ops *ops; +struct device *dev; + +if (!capable (CAP_SYS_ADMIN ) || !capable (CAP_SYS_RAWIO )) +return -EACCES ;
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2935
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+ +if (WARN_ON (!group)) +return -EINVAL ; + +if (sysfs_streq (buf,"identity" )) + req_type =IOMMU_DOMAIN_IDENTITY ; +else if (sysfs_streq (buf,"DMA" )) + req_type =IOMMU_DOMAIN_DMA ; +else if (sysfs_streq (buf,"auto" )) + req_auto =1 ;
How about let req_type = 0 for auto case. This keeps it consistent with the return value of iommu_ops->def_domain_type(). With
if (!req_type)
req_type = ops->def_domain_type(dev);
you don't need req_auto and dev_def_dom any more.
Quoted 24 lines
drivers/iommu/iommu.c +24
-0+else +return -EINVAL ; + +/* Lock the group to make sure that the device count doesn't change */ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { +mutex_unlock (&group->mutex); +pr_err ("Cannot change default domain of a group with two or more devices\n" ); +return -EINVAL ; + } + +/* Since group has only one device */ +list_for_each_entry (grp_dev, &group->devices, list) + dev = grp_dev->dev; + +/* +* Don't hold onto the group mutex because taking group mutex first and +* then the device lock could potentially cause a deadlock as below. +* Assume two threads T1 and T2. T1 is trying to change default domain +* of an iommu group and T2 is trying to hot unplug a device or release +* [1] VF of a PCIe device which is in the same iommu group. T1 takes +* group mutex and before it could take device lock assume T2 has taken +* device lock and is yet to take group mutex. Now, both the threads +* will be waiting for the other thread to release lock.
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2969
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+* +* [1] Typical device release path +* device_lock() from device/driver core code +* -> bus_notifier() +* -> iommu_bus_notifier() +* -> iommu_release_device() +* -> ops->release_device() vendor driver calls back iommu core code +* -> mutex_lock() from iommu core code +*/ +mutex_unlock (&group->mutex);
Since you only support single device group, how about
-> lock_device
-> lock group
-> change default domain
-> unlock group-> unlock device
?
Quoted 26 lines
drivers/iommu/iommu.c +26
-0+ +/* Check if the device in the group still has a driver bound to it */ +device_lock (dev); +if (device_is_bound (dev)) { +pr_err ("Device is still bound to driver\n" ); + ret = -EBUSY ; + goto dev_unlock; + } + +/* +* iommu group wasn't locked while acquiring device lock. So, make sure +* that the device count hasn't changed while acquiring device lock. +* +* Changing default domain of an iommu group with two or more devices +* isn't supported because there could be a potential deadlock. Consider +* the following scenario. T1 is trying to acquire device locks of all +* the devices in the group and before it could acquire all of them, +* there could be another thread T2 (from different sub-system and use +* case) that has already acquired some of the device locks and might be +* waiting for T1 to release other device locks. +*/ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { +pr_err ("Cannot change default domain of a group with two or more devices\n" ); + ret = -EINVAL ; + goto group_unlock;
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3005
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+ } + +/* +* Check if any user level driver (that doesn't use kernel driver like +* VFIO) is directly using the group. VFIO use case is detected by +* device_is_bound() check above +*/ +if (group->default_domain != group->domain) { +pr_err ("Group assigned to user level for direct access\n" ); + ret = -EINVAL ;
-EBUSY?
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3015
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+ goto group_unlock; + } + + prev_dom = group->default_domain; +if (!prev_dom || !prev_dom->ops || !prev_dom->ops->def_domain_type) { +pr_err ("'def_domain_type' call back isn't registered\n" ); + ret = -EINVAL ; + goto group_unlock; + } +
--- start ---
Quoted 11 lines
drivers/iommu/iommu.c +11
-0+ ops = prev_dom->ops; + dev_def_dom = ops->def_domain_type (dev); + +/* Check if user requested domain is supported by the device or not */ +if (!req_auto && dev_def_dom && req_type != dev_def_dom) { +pr_err ("Device cannot be in %s domain\n" , buf); + ret = -EINVAL ; + goto group_unlock; + } + +if (req_auto) {
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3036
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+/* +* If user requested "auto" and the device supports both the +* domains, then default to the domain the device was booted +* with +*/ +if (!dev_def_dom) + req_type = iommu_def_domain_type; +else + req_type = dev_def_dom; + }
--- end ---
Isn't this could could be simply replaced with:
if (!req_type)
req_type = ops->default_domain_type(dev) ? : iommu_def_domain_type;
Quoted 12 lines
drivers/iommu/iommu.c +11
-0+ +/* +* Switch to a new domain only if the requested domain type is different +* from the existing default domain type +*/ +if (prev_dom->type == req_type) { + ret = count; + goto group_unlock; + } > + + ret =iommu_change_dev_def_domain (dev, group, prev_dom, req_type); +device_unlock (dev);
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3058
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+return ret ?: count; + +group_unlock: +mutex_unlock (&group->mutex); + +dev_unlock: +device_unlock (dev); + +return ret; +}
Best regards,
baolu
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
RE: [PATCH V3 1/3] iommu: Add support to change default domain of an iommu_group
Hi Baolu,
Quoted 14 lines
-----Original Message-----
From: Lu Baolu <[email protected]>
Sent: Thursday, May 28, 2020 7:43 PM
To: Prakhya, Sai Praneeth <[email protected]>; [email protected]
Cc: [email protected]; Christoph Hellwig <[email protected]>; Joerg Roedel <[email protected]>; Raj, Ashok <[email protected]>; Will Deacon <[email protected]>; Mehta, Sohil <[email protected]>; Robin Murphy <[email protected]>; Jacob Pan <[email protected]>
Subject: Re: [PATCH V3 1/3] iommu: Add support to change default domain of aniommu_groupHi Sai,
Quoted 8 lines
Quoted 3 lines · Sai Praneeth Prakhya
Presently, the default domain of an
iommu_groupis allocated during boot time (i.e. when a device is being added to a group) and it cannot be^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^This is inaccurate as Joerg's code has deferred default domain allocation and attaching after group allocation. I'd suggest to remove this.
Ok.. makes sense. I will remove it. I think it should have been like below to accurately describe Joerg's changes, is that correct?
"Presently, the default domain of an iommu_group is allocated during
boot time (i.e. during device probe and after the device is added to a group) and it cannot be"
Quoted 7 lines
Quoted 3 lines
changed later. So, the device would typically be either in identity (also known as
pass_through) mode (controlled by "iommu=pt" kernel command line^^^^^^^^There are other kernel parameters to put device in
pass_throughmode. I'd suggest to remove this.
Ok.. makes sense. I will remove it.
I think, you were talking about ARM parameters (iommu.passthrough).. am I right?
Quoted 11 lines
Quoted 11 lines
argument) or the device would be in DMA mode as long as the machine is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identity mode and DMA mode.
But, assume a use case wherein the user trusts the device and believes that the OS is secure enough and hence wants *only* this device to bypass IOMMU (so that it could be high performing) whereas all the other devices to go through IOMMU (so that the system is protected). Presently, this use case is not supported. It will be helpful if there is some way to change the default domain of a B:D.F dynamically. Hence, add
Quoted 5 lines
such support.
^^^^^^^^^^^^^^^^^^^^^^^^^Currently default domain is per
iommu_group, we have no per device default domain yet. Probably, "default domain of an iommu group"?
Makes sense. I will change it.
Quoted 63 lines
Quoted 28 lines
A privileged user could request the kernel to change the default domain type of a
iommu_groupby writing to "/sys/kernel/iommu_groups/<grp_id>/type" file. Presently, only three values are supported 1. identity: all the DMA transactions from the device in this group are
*not* translated by the iommu 2. DMA: all the DMA
transactions from the device in this group are
translated by the iommu
3. auto: change to the type the device was booted withNote:
1. Default domain of an iommu group with two or more devices cannot be changed.
2. The device in the iommu group shouldn't be bound to any driver.
3. The device shouldn't be assigned to user for direct access.Please see "
Documentation/ABI/testing/sysfs-kernel-iommu_groups" for more information.Cc Christoph Hellwig <[email protected]>Cc Joerg Roedel <[email protected]>Cc Ashok Raj <[email protected]>Cc Will Deacon <[email protected]>Cc Lu Baolu <[email protected]>Cc Sohil Mehta <[email protected]>Cc Robin Murphy <[email protected]>Cc Jacob Pan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>Quoted 2 lines
--- drivers/iommu/iommu.c | 211+++++++++++++++++++++++++++++++++++++++++-
drivers/iommu/iommu.c static void __iommu_detach_group(struct
Quoted 7 lines
1 file changed, 210 insertions(+), 1 deletion(-)drivers/iommu/iommu.c index renamed from drivers/iommu/iommu.c +0
-0
a4c2f122eb8b..2b6cca799055100644drivers/iommu/iommu.c +0
-0@@ -92,6 +92,8 @@ static void __iommu_detach_group(struct
iommu_domain*domain,drivers/iommu/iommu.c static void __iommu_detach_group(struct · line 92
Quoted 1 line
drivers/iommu/iommu.c +0
-0static int iommu_create_device_direct_mappings(struct iommu_group*group,
drivers/iommu/iommu.c static void __iommu_detach_group(struct iommu_domain *domain,
Quoted 7 lines
drivers/iommu/iommu.c +2
-0struct device *dev);static struct iommu_group *iommu_group_get_for_dev (struct device *dev); +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count);#define IOMMU_GROUP_ATTR (_name, _mode, _show, _store)\drivers/iommu/iommu.c static IOMMU_GROUP_ATTR(name, S_IRUGO,
Quoted 2 lines
struct iommu_group_attribute iommu_group_attr_##_name = \drivers/iommu/iommu.c +0
-0@@ -524,7 +526,8 @@ static IOMMU_GROUP_ATTR(name, S_IRUGO,
iommu_group_show_name, NULL);drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva
Quoted 10 lines
drivers/iommu/iommu.c +2
-1static IOMMU_GROUP_ATTR(reserved_regions, 0444, iommu_group_show_resv_regions, NULL); -static IOMMU_GROUP_ATTR(type, 0444, iommu_group_show_type, NULL); +static IOMMU_GROUP_ATTR(type, 0644, iommu_group_show_type, + iommu_group_store_type);static void iommu_group_release (struct kobject *kobj) {@@ -2847,3 +2850,209 @@ int iommu_sva_get_pasid(struct iommu_sva*handle)
Quoted 12 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2857
Quoted 8 lines
drivers/iommu/iommu.c +5
-0return ops->sva_get_pasid (handle); }EXPORT_SYMBOL_GPL (iommu_sva_get_pasid); + +/* +* Changes the default domain of a device +* +* @dev: The *only* device in the group All devices in a group have been linked in
group->devices, so you don't need to pass it in as a parameter. Otherwise, you also need some sanity check against @dev, right?
Yes, I agree that we could get "dev" from group->devices. But, I passed it as a parameter because it's already done by iommu_group_store_type() (as below) and I thought that I could save from duplicating code by passing it as a parameter. Also, iommu_change_dev_def_domain() is presently just a helper function (and not an API) for iommu_group_store_type() and hence thought that error checking for "dev" (i.e. if (!dev)) is not needed in iommu_change_dev_def_domain() because iommu_group_store_type() may never pass a null value.
/* Since group has only one device */ list_for_each_entry(grp_dev, &group->devices, list)
dev = grp_dev->dev;
Just wanted to put out the rationale behind why I decided to go this way rather than getting "dev" from group->devices. Please let me know if you think otherwise, I am happy to change it.
Quoted 6 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2859
Quoted 2 lines
drivers/iommu/iommu.c +2
-0+* @group: The group for which the default domain should be changed +* @prev_domain: The previous domain that is being switched from The same as @dev, the default domain in use is saved in
group->default_domain. No need a separated parameter to save space and sanity check.
Same as above. Just wanted to avoid the below check as it's already done by iommu_group_store_type() if (group->default_domain)
prev_dom = group->default_domain;
I am happy to change it if you prefer to see less parameters.
Quoted 9 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 8 lines
drivers/iommu/iommu.c +8
-0+ * @type: The type of the new default domain that gets associated + with the group + * + * Returns 0 on success and error code on failure + * + * Note: + * 1. Presently, this function is called only when user requests to change the + * group's default domain type through/sys/kernel/iommu_groups/<grp_id>/type
Quoted 8 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva · line 2870
Quoted 3 lines
drivers/iommu/iommu.c +3
-0+* Please take a closer look if intended to use for other purposes. +* 2. Assumes that group->mutex is already taken and releases before +returning You could assume that the caller should hold the
group->mutexand shouldn't be released in this helper. A reference design could be found here.
Ok. I have looked at it.. and it does hold group mutex. Just wondering if you would like me to change anything here?
Quoted 70 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2872
Quoted 4 lines
drivers/iommu/iommu.c +4
-0+*/ +static int iommu_change_dev_def_domain (struct device *dev, +struct iommu_group *group, +struct iommu_domain *prev_dom,int type){
Quoted 51 lines
drivers/iommu/iommu.c +51
-0+int ret =0 ; + +/* Sets group->default_domain to the newly allocated domain */ + ret =iommu_group_alloc_default_domain (dev->bus, group, type); +if (ret) + goto out; + + ret =__iommu_attach_device (group->default_domain, dev); +if (ret) + goto free_new_domain; + + group->domain = group->default_domain; + + ret =iommu_create_device_direct_mappings (group, dev); +if (ret) + goto free_new_domain; + +/* +* Release the mutex here because ops->probe_finalize() call-back of +* some vendor IOMMU drivers calls arm_iommu_attach_device() which +* in-turn might call back into IOMMU core code, where it tries to take +* group->mutex, resulting in a deadlock. +*/ +mutex_unlock (&group->mutex); + +/* Make sure dma_ops is appropriatley set */ +iommu_group_do_probe_finalize (dev, group->default_domain); +iommu_domain_free (prev_dom); +return 0 ; + +free_new_domain: +iommu_domain_free (group->default_domain); + group->default_domain = prev_dom; + group->domain = prev_dom; + +out: +mutex_unlock (&group->mutex); + +return ret; +} + +static ssize_t iommu_group_store_type (struct iommu_group *group, +const char *buf,size_t count) { +int ret, req_type =0 , req_auto =0 , dev_def_dom; +struct iommu_domain *prev_dom; +struct group_device *grp_dev; +const struct iommu_ops *ops; +struct device *dev; + +if (!capable (CAP_SYS_ADMIN ) || !capable (CAP_SYS_RAWIO )) +return -EACCES ;drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2935
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+ +if (WARN_ON (!group)) +return -EINVAL ; + +if (sysfs_streq (buf,"identity" )) + req_type =IOMMU_DOMAIN_IDENTITY ; +else if (sysfs_streq (buf,"DMA" )) + req_type =IOMMU_DOMAIN_DMA ; +else if (sysfs_streq (buf,"auto" )) + req_auto =1 ;How about let
req_type= 0 for auto case. This keeps it consistent with the return value ofiommu_ops->def_domain_type(). With
Quoted 4 lines
if (!req_type)
req_type = ops->def_domain_type(dev);you don't need
req_autoanddev_def_domany more.
Please see below where you had comments for changing req_type.
Quoted 39 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 8 lines
drivers/iommu/iommu.c +8
-0+else +return -EINVAL ; + +/* Lock the group to make sure that the device count doesn't change */ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { +mutex_unlock (&group->mutex); + pr_err("Cannot change default domain of a group with two ormore devices\n");
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 9 lines
drivers/iommu/iommu.c +9
-0+return -EINVAL ; + } + +/* Since group has only one device */ +list_for_each_entry (grp_dev, &group->devices, list) + dev = grp_dev->dev; + +/* +* Don't hold onto the group mutex because taking group mutex first and
Quoted 7 lines
drivers/iommu/iommu.c +7
-0+* then the device lock could potentially cause a deadlock as below. +* Assume two threads T1 and T2. T1 is trying to change default domain +* of an iommu group and T2 is trying to hot unplug a device or release +* [1] VF of a PCIe device which is in the same iommu group. T1 takes +* group mutex and before it could take device lock assume T2 has taken +* device lock and is yet to take group mutex. Now, both the threads +* will be waiting for the other thread to release lock. drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 2969
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+* +* [1] Typical device release path +* device_lock() from device/driver core code +* -> bus_notifier() +* -> iommu_bus_notifier() +* -> iommu_release_device() +* -> ops->release_device() vendor driver calls back iommu core code +* -> mutex_lock() from iommu core code +*/ +mutex_unlock (&group->mutex);Since you only support single device group, how about
Quoted 5 lines
->
lock_device-> lock group -> change default domain -> unlock group-> unlock device
Sure! I did the same thing below i.e. first take device_lock() and then take group mutex. I acquired/released group mutex above to
1. Verify that iommu group has only one device
2. Get "struct dev *" of that device.
Just wondering if you would like me to change anything here?
Quoted 27 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 11 lines
drivers/iommu/iommu.c +11
-0+ +/* Check if the device in the group still has a driver bound to it */ +device_lock (dev); +if (device_is_bound (dev)) { +pr_err ("Device is still bound to driver\n" ); + ret = -EBUSY ; + goto dev_unlock; + } + +/* +* iommu group wasn't locked while acquiring device lock. So, make sure
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 3 lines
drivers/iommu/iommu.c +3
-0+* that the device count hasn't changed while acquiring device lock. +* +* Changing default domain of an iommu group with two or more devices
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle)
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+* isn't supported because there could be a potential deadlock. Consider +* the following scenario. T1 is trying to acquire device locks of all +* the devices in the group and before it could acquire all of them, +* there could be another thread T2 (from different sub-system and use +* case) that has already acquired some of the device locks and might be +* waiting for T1 to release other device locks. +*/ +mutex_lock (&group->mutex); +if (iommu_group_device_count (group) !=1 ) { + pr_err("Cannot change default domain of a group with two ormore devices\n");
Quoted 14 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3005
Quoted 12 lines
drivers/iommu/iommu.c +12
-0+ ret = -EINVAL ; + goto group_unlock; + } + +/* +* Check if any user level driver (that doesn't use kernel driver like +* VFIO) is directly using the group. VFIO use case is detected by +* device_is_bound() check above +*/ +if (group->default_domain != group->domain) { +pr_err ("Group assigned to user level for direct access\n" ); + ret = -EINVAL ;
-EBUSY?
Ok.. I will change it.
Quoted 36 lines
drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3015
Quoted 11 lines
drivers/iommu/iommu.c +5
-0+ goto group_unlock; + } + + prev_dom = group->default_domain; + if (!prev_dom || !prev_dom->ops || !prev_dom->ops-
def_domain_type) {drivers/iommu/iommu.c +5
-0+pr_err ("'def_domain_type' call back isn't registered\n" ); + ret = -EINVAL ; + goto group_unlock; + } +--- start ---
Quoted 11 lines
drivers/iommu/iommu.c +11
-0+ ops = prev_dom->ops; + dev_def_dom = ops->def_domain_type (dev); + +/* Check if user requested domain is supported by the device or not */ +if (!req_auto && dev_def_dom && req_type != dev_def_dom) { +pr_err ("Device cannot be in %s domain\n" , buf); + ret = -EINVAL ; + goto group_unlock; + } + +if (req_auto) {drivers/iommu/iommu.c int iommu_sva_get_pasid(struct iommu_sva *handle) · line 3036
Quoted 10 lines
drivers/iommu/iommu.c +10
-0+/* +* If user requested "auto" and the device supports both the +* domains, then default to the domain the device was booted +* with +*/ +if (!dev_def_dom) + req_type = iommu_def_domain_type; +else + req_type = dev_def_dom; + }--- end ---
Quoted 5 lines
Isn't this could could be simply replaced with:
if (!req_type)
req_type = ops->default_domain_type(dev) ? :
iommu_def_domain_type;
I think, replacing this has two issues
1. We might miss validating user request i.e. we need to make sure that when user requests for a specific type of domain (E.g: DMA), we need to validate that the device could actually be in that domain. This is done above by
/* Check if user requested domain is supported by the device or not */
if (!req_auto && dev_def_dom && req_type != dev_def_dom) {
pr_err("Device cannot be in %s domain\n", buf);
ret = -EINVAL;
goto group_unlock;
}2. req_type could end up having 0 if "ops->default_domain_type(dev)" returns 0.
But, after this point, req_type should _only_ be either DMA type or identity type and cannot be 0 because req_type is passed to iommu_ops->domain_alloc() to allocate domain of requested type and it cannot be 0.
Regards,
Sai
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
PATCH v3 2/3 iommu: Take lock before reading iommu_group default domain type
In reply to Sai Praneeth Prakhya"/sys/kernel/iommu_groups/<grp_id>/type" file could be read to find out the default domain type of an iommu group. The default domain of an iommu group doesn't change after booting and hence could be read directly. But, after addding support to dynamically change iommu group default domain, the above assumption no longer stays valid.
iommu group default domain type could be changed at any time by writing to "/sys/kernel/iommu_groups/<grp_id>/type". So, take group mutex before reading iommu group default domain type so that the user wouldn't see stale values or iommu_group_show_type() doesn't try to derefernce stale pointers.
--- drivers/iommu/iommu.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/iommu/iommu.c +2 -0
@@ -500,6 +500,7 @@ static ssize_t iommu_group_show_type(struct iommu_group *group, {char *type ="unknown\n" ; +mutex_lock (&group->mutex);if (group->default_domain) {switch (group->default_domain->type) {case IOMMU_DOMAIN_BLOCKED :
@@ -516,6 +517,7 @@ static ssize_t iommu_group_show_type(struct iommu_group *group,break ; } } +mutex_unlock (&group->mutex);strcpy (buf, type);return strlen (type);
Signature · 6 lines
2.19.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
PATCH v3 3/3 iommu: Document usage of "/sys/kernel/iommu_groups/<grp_id>/type" file
In reply to Sai Praneeth PrakhyaThe default domain type of an iommu group can be changed by writing to "/sys/kernel/iommu_groups/<grp_id>/type" file. Hence, document it's usage and more importantly spell out its limitations.
--- .../ABI/testing/sysfs-kernel-iommu_groups | 30 +++++++++++++++++++ 1 file changed, 30 insertions(+)
Documentation/ABI/testing/sysfs-kernel-iommu_groups +30 -0
@@ -33,3 +33,33 @@ Description: In case an RMRR is used only by graphics or USB devices it is now exposed as "direct-relaxable" instead of "direct". In device assignment use case, for instance, those RMRR are considered to be relaxable and safe. + +What: /sys/kernel/iommu_groups/<grp_id>/type +Date: June 2020 +KernelVersion: v5.8 +Contact: Sai Praneeth Prakhya <[email protected]> +Description: Lets the user know the type of default domain in use by iommu + for this group. A privileged user could request kernel to change + the group type by writing to this file. Presently, only three + types are supported + 1. DMA: All the DMA transactions from the device in this group + are translated by the iommu. + 2. identity: All the DMA transactions from the device in this + group are *not* translated by the iommu. + 3. auto: Change to the type the device was booted with. When the + user reads the file he would never see "auto". This is + just a write only value. + Note: + ----- + A group type could be modified only when + 1. The group has *only* one device + 2. The device in the group is not bound to any device driver. + So, the user must first unbind the appropriate driver and + then change the default domain type. + Caution: + -------- + Unbinding a device driver will take away the driver's control + over the device and if done on devices that host root file + system could lead to catastrophic effects (the user might + need to reboot the machine to get it to normal state). So, it's + expected that the user understands what he is doing.
Signature · 6 lines
2.19.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
Hi Sai,
Quoted 6 lines · Sai Praneeth Prakhya
Presently, the default domain of a iommu group is allocated during boot time and it cannot be changed later. So, the device would typically be either in identity (
pass_through) mode or the device would be in DMA mode as long as the system is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identity mode and DMA mode.
Quoted 7 lines
Assume a use case wherein the privileged user would want to use the device in pass-through mode when the device is used for host so that it would be high performing. Presently, this is not supported. Hence add support to change the default domain of an iommu group dynamically.
Support this by writing to a sysfs file, namely
"/sys/kernel/iommu_groups/<grp_id>/type".
The email subject
[PATCH V3 0/3] iommu: Add support to change default domain of an iommu
probably should be changed to
[PATCH V3 0/3] iommu: Add support to change default domain of an iommu group
Best regards,
baolu
Quoted 60 lines
Testing:
--------
Tested by dynamically changing storage device (nvme) from
1. identity mode to DMA and making sure file transfer works
2. DMA mode to identity mode and making sure file transfer works Tested only for intel_iommu/vt-d. Would appreciate if someone could test on AMD and ARM based machines.Based on iommu maintainer's 'next' branch.
Changes from V2:
----------------
1. Change the logic of updating default domain from V2 becauseops->probe_finalize() could be used to update dma_ops.2. Drop 1st and 2nd patch of V2 series because they are no longer needed on iommu maintainer's 'next' branch.
3. Limit this feature to iommu groups with only one device.
4. Holddevice_lockand group mutex until the default domain is changed.Changes from V1:
----------------
1. V1 patch set wasn't updatingdma_opsfor some vendors (Eg: AMD), hence, change the logic of updating default domain as below (because adding a device toiommu_groupautomatically updatesdma_ops)
a. Allocate a new domain
b. For every device in the group
i. Remove the device from the group
ii. Add the device back to the group
c. Free previous domain
2. Drop 1st patch of V1 (iommu/vt-d: Modifydevice_def_domain_type()to use at runtime) because "iommu=pt" has no effect on this function anymore.
3. Added a patch to take/release lock while readingiommu_group->default_domain->typebecause it can be changed any time by user.
4. Before changing default domain type of a group, check if the group is directly assigned for user level access. If so, abort.
5. Sanitize return path (using ternary operator) iniommu_group_store_type()
6. Split 2nd patch of V1 (iommu: Adddevice_def_domain_type()call back function toiommu_ops) into two patches such that iommu generic changes are now in 1st patch of V2 and vt-d specific changes are in 2nd patch of V2.
7. Renamedevice_def_domain_type()todev_def_domain_type()
8. Remove example from documentation
9. Change the value written to file "/sys/kernel/iommu_groups/<grp_id>/type" from "dma" to "DMA".Changes from RFC:
-----------------
1. Added support for "auto" type, so that kernel selects one among identity or dma mode.
2. Use "system_state" indevice_def_domain_type()instead of an argument.Sai Praneeth Prakhya (3):
iommu: Add support to change default domain of aniommu_group
iommu: Take lock before readingiommu_groupdefault domain type
iommu: Document usage of "/sys/kernel/iommu_groups/<grp_id>/type" file.../ABI/testing/sysfs-kernel-iommu_groups | 30 +++ drivers/iommu/iommu.c | 213 +++++++++++++++++- 2 files changed, 242 insertions(+), 1 deletion(-)
Quoted 10 lines
Cc Christoph Hellwig <[email protected]>Cc Joerg Roedel <[email protected]>Cc Ashok Raj <[email protected]>Cc Will Deacon <[email protected]>Cc Lu Baolu <[email protected]>Cc Sohil Mehta <[email protected]>Cc Robin Murphy <[email protected]>Cc Jacob Pan <[email protected]>Signed-off-by Sai Praneeth Prakhya <[email protected]>
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
Hi Baolu,
Quoted 31 lines
-----Original Message-----
From: Lu Baolu <[email protected]>
Sent: Thursday, May 28, 2020 6:52 PM
To: Prakhya, Sai Praneeth <[email protected]>; [email protected]
Cc: [email protected]; Christoph Hellwig <[email protected]>; Joerg Roedel <[email protected]>; Raj, Ashok <[email protected]>; Will Deacon <[email protected]>; Mehta, Sohil <[email protected]>; Robin Murphy <[email protected]>; Jacob Pan <[email protected]>
Subject: Re: [PATCH V3 0/3] iommu: Add support to change default domain of an iommuHi Sai,
Quoted 6 lines · Sai Praneeth Prakhya
Presently, the default domain of a iommu group is allocated during boot time and it cannot be changed later. So, the device would typically be either in identity
(pass_through) mode or the device would be in DMA mode as long as the system is up and running. There is no way to change the default domain type dynamically i.e. after booting, a device cannot switch between identitymode and DMA mode.
Quoted 8 lines
Assume a use case wherein the privileged user would want to use the device in pass-through mode when the device is used for host so that it would be high performing. Presently, this is not supported. Hence add support to change the default domain of an iommu group dynamically.
Support this by writing to a sysfs file, namely
"/sys/kernel/iommu_groups/<grp_id>/type".
Quoted 8 lines
The email subject
[PATCH V3 0/3] iommu: Add support to change default domain of an iommu
probably should be changed to
[PATCH V3 0/3] iommu: Add support to change default domain of an iommu group
Oops.. my bad. I have it in the patches that I have on my dev machine. I added "group" to a new line because the subject line crossed 80 character limit. Probably, I will just put it in the same line for the next version. Thanks for bringing this up.
Regards,
Sai
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu