2024-06-10 12:24:49

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 00/21] iommu: Refactoring domain allocation interface

The IOMMU subsystem has undergone some changes, including the removal
of iommu_ops from the bus structure. Consequently, the existing domain
allocation interface, which relies on a bus type argument, is no longer
relevant:

struct iommu_domain *iommu_domain_alloc(struct bus_type *bus)

This series is designed to refactor the use of this interface. It
proposes a new interface iommu_paging_domain_alloc() to replace
iommu_domain_alloc(). This interface is for allocating iommu paging
domains for kernel drivers. It takes a device pointer as a parameter,
which better reflects the current design of the IOMMU subsystem.

The majority of device drivers currently using iommu_domain_alloc() do
so to allocate a domain for a specific device and then attach that
domain to the device. These cases can be straightforwardly migrated to
the new interface.

The drm/tegra driver is a bit different in that the device pointer
passed to the helper, which allocates the iommu domain, is not the one
that will be used for the kernel DMA API. Move the existing logic in
iommu_domain_alloc() into the driver to ensure it works as intended.

Now that all consumers of iommu_domain_alloc() have switched to the new
interface, we can finally remove iommu_domain_alloc(). This removal
paves the way for the IOMMU subsystem to support multiple iommu drivers.
Additionally, the individual iommu driver implementation for domain
allocation could also be simplified, as there will always be a valid
device pointer passed along the path.

The whole series is also available on GitHub:
https://github.com/LuBaolu/intel-iommu/commits/iommu-domain-allocation-refactor-v3

Change log:

v3:
- Remove the proposed iommu_user_domain_alloc() interface.

v2: https://lore.kernel.org/linux-iommu/[email protected]/
- Drop the vt-d patches which implement paging domain support from this
series. I will post them in a separate series later.
- Convert all drivers that call iommu_domain_alloc() to use the new
interface and remove iommu_domain_alloc() from the tree.
- For the drm/msm driver, make the code compatible with the no-IOMMU
case.
- Various cleanups and refinements.

v1: https://lore.kernel.org/linux-iommu/[email protected]/

Lu Baolu (20):
iommu: Add iommu_paging_domain_alloc() interface
iommufd: Use iommu_paging_domain_alloc()
vfio/type1: Use iommu_paging_domain_alloc()
vhost-vdpa: Use iommu_paging_domain_alloc()
drm/msm: Use iommu_paging_domain_alloc()
drm/nouveau/tegra: Use iommu_paging_domain_alloc()
gpu: host1x: Use iommu_paging_domain_alloc()
media: nvidia: tegra: Use iommu_paging_domain_alloc()
media: venus: firmware: Use iommu_paging_domain_alloc()
wifi: ath10k: Use iommu_paging_domain_alloc()
wifi: ath11k: Use iommu_paging_domain_alloc()
remoteproc: Use iommu_paging_domain_alloc()
soc/fsl/qbman: Use iommu_paging_domain_alloc()
RDMA/usnic: Use iommu_paging_domain_alloc()
iommu/vt-d: Add helper to allocate paging domain
ARM: dma-mapping: Use iommu_paging_domain_alloc()
drm/rockchip: Use iommu_paging_domain_alloc()
drm/tegra: Remove call to iommu_domain_alloc()
iommu: Remove iommu_present()
iommu: Remove iommu_domain_alloc()

Robin Murphy (1):
ARM: dma-mapping: Pass device to arm_iommu_create_mapping()

include/linux/iommu.h | 12 +--
arch/arm/include/asm/dma-iommu.h | 2 +-
arch/arm/mm/dma-mapping.c | 12 +--
drivers/gpu/drm/exynos/exynos_drm_dma.c | 2 +-
drivers/gpu/drm/msm/msm_iommu.c | 7 +-
.../drm/nouveau/nvkm/engine/device/tegra.c | 4 +-
drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 10 ++-
drivers/gpu/drm/tegra/drm.c | 34 ++++++--
drivers/gpu/host1x/dev.c | 7 +-
drivers/infiniband/hw/usnic/usnic_uiom.c | 6 +-
drivers/iommu/intel/iommu.c | 87 +++++++++++++++++--
drivers/iommu/iommu.c | 63 +++-----------
drivers/iommu/iommufd/hw_pagetable.c | 7 +-
drivers/iommu/ipmmu-vmsa.c | 3 +-
drivers/iommu/mtk_iommu_v1.c | 3 +-
.../media/platform/nvidia/tegra-vde/iommu.c | 7 +-
drivers/media/platform/qcom/venus/firmware.c | 6 +-
drivers/media/platform/ti/omap3isp/isp.c | 2 +-
drivers/net/wireless/ath/ath10k/snoc.c | 6 +-
drivers/net/wireless/ath/ath11k/ahb.c | 6 +-
drivers/remoteproc/remoteproc_core.c | 6 +-
drivers/soc/fsl/qbman/qman_portal.c | 5 +-
drivers/vfio/vfio_iommu_type1.c | 7 +-
drivers/vhost/vdpa.c | 14 ++-
24 files changed, 183 insertions(+), 135 deletions(-)

--
2.34.1



2024-06-10 12:37:59

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 08/21] media: nvidia: tegra: Use iommu_paging_domain_alloc()

An iommu domain is allocated in tegra_vde_iommu_init() and is attached to
vde->dev. Use iommu_paging_domain_alloc() to make it explicit.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/media/platform/nvidia/tegra-vde/iommu.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/media/platform/nvidia/tegra-vde/iommu.c b/drivers/media/platform/nvidia/tegra-vde/iommu.c
index 5521ed3e465f..b1d9d841d944 100644
--- a/drivers/media/platform/nvidia/tegra-vde/iommu.c
+++ b/drivers/media/platform/nvidia/tegra-vde/iommu.c
@@ -78,9 +78,10 @@ int tegra_vde_iommu_init(struct tegra_vde *vde)
arm_iommu_release_mapping(mapping);
}
#endif
- vde->domain = iommu_domain_alloc(&platform_bus_type);
- if (!vde->domain) {
- err = -ENOMEM;
+ vde->domain = iommu_paging_domain_alloc(dev);
+ if (IS_ERR(vde->domain)) {
+ err = PTR_ERR(vde->domain);
+ vde->domain = NULL;
goto put_group;
}

--
2.34.1


2024-06-10 12:38:31

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 07/21] gpu: host1x: Use iommu_paging_domain_alloc()

An iommu domain is allocated in host1x_iommu_attach() and is attached to
host->dev. Use iommu_paging_domain_alloc() to make it explicit.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/gpu/host1x/dev.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/host1x/dev.c b/drivers/gpu/host1x/dev.c
index 3a0aaa68ac8d..f86a6b12f24a 100644
--- a/drivers/gpu/host1x/dev.c
+++ b/drivers/gpu/host1x/dev.c
@@ -404,9 +404,10 @@ static struct iommu_domain *host1x_iommu_attach(struct host1x *host)
if (err < 0)
goto put_group;

- host->domain = iommu_domain_alloc(&platform_bus_type);
- if (!host->domain) {
- err = -ENOMEM;
+ host->domain = iommu_paging_domain_alloc(host->dev);
+ if (IS_ERR(host->domain)) {
+ err = PTR_ERR(host->domain);
+ host->domain = NULL;
goto put_cache;
}

--
2.34.1


2024-06-10 12:54:31

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 12/21] remoteproc: Use iommu_paging_domain_alloc()

An iommu domain is allocated in rproc_enable_iommu() and is attached to
rproc->dev.parent in the same function.

Use iommu_paging_domain_alloc() to make it explicit.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/remoteproc/remoteproc_core.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
index f276956f2c5c..eb66f78ec8b7 100644
--- a/drivers/remoteproc/remoteproc_core.c
+++ b/drivers/remoteproc/remoteproc_core.c
@@ -109,10 +109,10 @@ static int rproc_enable_iommu(struct rproc *rproc)
return 0;
}

- domain = iommu_domain_alloc(dev->bus);
- if (!domain) {
+ domain = iommu_paging_domain_alloc(dev);
+ if (IS_ERR(domain)) {
dev_err(dev, "can't alloc iommu domain\n");
- return -ENOMEM;
+ return PTR_ERR(domain);
}

iommu_set_fault_handler(domain, rproc_iommu_fault, rproc);
--
2.34.1


2024-06-10 13:00:03

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 15/21] iommu/vt-d: Add helper to allocate paging domain

The domain_alloc_user operation is currently implemented by allocating a
paging domain using iommu_domain_alloc(). This is because it needs to fully
initialize the domain before return. Add a helper to do this to avoid using
iommu_domain_alloc().

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/iommu/intel/iommu.c | 87 +++++++++++++++++++++++++++++++++----
1 file changed, 78 insertions(+), 9 deletions(-)

diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
index 2e9811bf2a4e..ccde5f5972e4 100644
--- a/drivers/iommu/intel/iommu.c
+++ b/drivers/iommu/intel/iommu.c
@@ -3633,6 +3633,79 @@ static struct iommu_domain blocking_domain = {
}
};

+static int iommu_superpage_capability(struct intel_iommu *iommu, bool first_stage)
+{
+ if (!intel_iommu_superpage)
+ return 0;
+
+ if (first_stage)
+ return cap_fl1gp_support(iommu->cap) ? 2 : 1;
+
+ return fls(cap_super_page_val(iommu->cap));
+}
+
+static struct dmar_domain *paging_domain_alloc(struct device *dev, bool first_stage)
+{
+ struct device_domain_info *info = dev_iommu_priv_get(dev);
+ struct intel_iommu *iommu = info->iommu;
+ struct dmar_domain *domain;
+ int addr_width;
+
+ domain = kzalloc(sizeof(*domain), GFP_KERNEL);
+ if (!domain)
+ return ERR_PTR(-ENOMEM);
+
+ INIT_LIST_HEAD(&domain->devices);
+ INIT_LIST_HEAD(&domain->dev_pasids);
+ INIT_LIST_HEAD(&domain->cache_tags);
+ spin_lock_init(&domain->lock);
+ spin_lock_init(&domain->cache_lock);
+ xa_init(&domain->iommu_array);
+
+ domain->nid = dev_to_node(dev);
+ domain->has_iotlb_device = info->ats_enabled;
+ domain->use_first_level = first_stage;
+
+ /* calculate the address width */
+ addr_width = agaw_to_width(iommu->agaw);
+ if (addr_width > cap_mgaw(iommu->cap))
+ addr_width = cap_mgaw(iommu->cap);
+ domain->gaw = addr_width;
+ domain->agaw = iommu->agaw;
+ domain->max_addr = __DOMAIN_MAX_ADDR(addr_width);
+
+ /* iommu memory access coherency */
+ domain->iommu_coherency = iommu_paging_structure_coherency(iommu);
+
+ /* pagesize bitmap */
+ domain->domain.pgsize_bitmap = SZ_4K;
+ domain->iommu_superpage = iommu_superpage_capability(iommu, first_stage);
+ domain->domain.pgsize_bitmap |= domain_super_pgsize_bitmap(domain);
+
+ /*
+ * IOVA aperture: First-level translation restricts the input-address
+ * to a canonical address (i.e., address bits 63:N have the same value
+ * as address bit [N-1], where N is 48-bits with 4-level paging and
+ * 57-bits with 5-level paging). Hence, skip bit [N-1].
+ */
+ domain->domain.geometry.force_aperture = true;
+ domain->domain.geometry.aperture_start = 0;
+ if (first_stage)
+ domain->domain.geometry.aperture_end = __DOMAIN_MAX_ADDR(domain->gaw - 1);
+ else
+ domain->domain.geometry.aperture_end = __DOMAIN_MAX_ADDR(domain->gaw);
+
+ /* always allocate the top pgd */
+ domain->pgd = iommu_alloc_page_node(domain->nid, GFP_KERNEL);
+ if (!domain->pgd) {
+ kfree(domain);
+ return ERR_PTR(-ENOMEM);
+ }
+ domain_flush_cache(domain, domain->pgd, PAGE_SIZE);
+
+ return domain;
+}
+
static struct iommu_domain *intel_iommu_domain_alloc(unsigned type)
{
struct dmar_domain *dmar_domain;
@@ -3695,15 +3768,11 @@ intel_iommu_domain_alloc_user(struct device *dev, u32 flags,
if (user_data || (dirty_tracking && !ssads_supported(iommu)))
return ERR_PTR(-EOPNOTSUPP);

- /*
- * domain_alloc_user op needs to fully initialize a domain before
- * return, so uses iommu_domain_alloc() here for simple.
- */
- domain = iommu_domain_alloc(dev->bus);
- if (!domain)
- return ERR_PTR(-ENOMEM);
-
- dmar_domain = to_dmar_domain(domain);
+ /* Do not use first stage for user domain translation. */
+ dmar_domain = paging_domain_alloc(dev, false);
+ if (IS_ERR(dmar_domain))
+ return ERR_CAST(dmar_domain);
+ domain = &dmar_domain->domain;

if (nested_parent) {
dmar_domain->nested_parent = true;
--
2.34.1


2024-06-10 13:09:16

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 19/21] drm/tegra: Remove call to iommu_domain_alloc()

Commit <17de3f5fdd35> ("iommu: Retire bus ops") removes iommu ops from
the bus structure. The iommu subsystem no longer relies on bus for
operations. So iommu_domain_alloc() interface is no longer relevant.

Normally, iommu_paging_domain_alloc() could be a replacement for
iommu_domain_alloc() if the caller has the right device for IOMMU API
use. Unfortunately, this is not the case for this driver.

Iterate the devices on the platform bus and find a suitable device
whose device DMA is translated by an IOMMU. Then use this device to
allocate an iommu domain. The iommu subsystem prevents domains
allocated by one iommu driver from being attached to devices managed
by any different iommu driver.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/gpu/drm/tegra/drm.c | 34 +++++++++++++++++++++++++---------
1 file changed, 25 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
index 03d1c76aec2d..ee391f859992 100644
--- a/drivers/gpu/drm/tegra/drm.c
+++ b/drivers/gpu/drm/tegra/drm.c
@@ -1133,6 +1133,17 @@ static bool host1x_drm_wants_iommu(struct host1x_device *dev)
return domain != NULL;
}

+static int iommu_mapped_device(struct device *dev, void *data)
+{
+ struct device **iommu_dev = data;
+
+ if (!device_iommu_mapped(dev))
+ return 0;
+
+ *iommu_dev = dev;
+ return 1;
+}
+
static int host1x_drm_probe(struct host1x_device *dev)
{
struct tegra_drm *tegra;
@@ -1149,16 +1160,21 @@ static int host1x_drm_probe(struct host1x_device *dev)
goto put;
}

- if (host1x_drm_wants_iommu(dev) && iommu_present(&platform_bus_type)) {
- tegra->domain = iommu_domain_alloc(&platform_bus_type);
- if (!tegra->domain) {
- err = -ENOMEM;
- goto free;
+ if (host1x_drm_wants_iommu(dev)) {
+ struct device *iommu_dev = NULL;
+
+ bus_for_each_dev(&platform_bus_type, NULL, &iommu_dev, iommu_mapped_device);
+ if (iommu_dev) {
+ tegra->domain = iommu_paging_domain_alloc(iommu_dev);
+ if (IS_ERR(tegra->domain)) {
+ err = PTR_ERR(tegra->domain);
+ goto free;
+ }
+
+ err = iova_cache_get();
+ if (err < 0)
+ goto domain;
}
-
- err = iova_cache_get();
- if (err < 0)
- goto domain;
}

mutex_init(&tegra->clients_lock);
--
2.34.1


2024-06-10 13:17:15

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 06/21] drm/nouveau/tegra: Use iommu_paging_domain_alloc()

In nvkm_device_tegra_probe_iommu(), a paging domain is allocated for @dev
and attached to it on success. Use iommu_paging_domain_alloc() to make it
explicit.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c
index 87caa4a72921..763c4c2925f9 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c
@@ -120,8 +120,8 @@ nvkm_device_tegra_probe_iommu(struct nvkm_device_tegra *tdev)
mutex_init(&tdev->iommu.mutex);

if (device_iommu_mapped(dev)) {
- tdev->iommu.domain = iommu_domain_alloc(&platform_bus_type);
- if (!tdev->iommu.domain)
+ tdev->iommu.domain = iommu_paging_domain_alloc(dev);
+ if (IS_ERR(tdev->iommu.domain))
goto error;

/*
--
2.34.1


2024-06-10 13:17:32

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 09/21] media: venus: firmware: Use iommu_paging_domain_alloc()

An iommu domain is allocated in venus_firmware_init() and is attached to
core->fw.dev in the same function. Use iommu_paging_domain_alloc() to
make it explicit.

Signed-off-by: Lu Baolu <[email protected]>
---
drivers/media/platform/qcom/venus/firmware.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/media/platform/qcom/venus/firmware.c b/drivers/media/platform/qcom/venus/firmware.c
index fe7da2b30482..66a18830e66d 100644
--- a/drivers/media/platform/qcom/venus/firmware.c
+++ b/drivers/media/platform/qcom/venus/firmware.c
@@ -316,10 +316,10 @@ int venus_firmware_init(struct venus_core *core)

core->fw.dev = &pdev->dev;

- iommu_dom = iommu_domain_alloc(&platform_bus_type);
- if (!iommu_dom) {
+ iommu_dom = iommu_paging_domain_alloc(core->fw.dev);
+ if (IS_ERR(iommu_dom)) {
dev_err(core->fw.dev, "Failed to allocate iommu domain\n");
- ret = -ENOMEM;
+ ret = PTR_ERR(iommu_dom);
goto err_unregister;
}

--
2.34.1


2024-06-10 14:19:35

by Lu Baolu

[permalink] [raw]
Subject: [PATCH v3 14/21] RDMA/usnic: Use iommu_paging_domain_alloc()

usnic_uiom_alloc_pd() allocates a paging domain for a given device.
In this case, iommu_domain_alloc(dev->bus) is equivalent to 
iommu_paging_domain_alloc(dev). Replace it as iommu_domain_alloc()
has been deprecated.

Signed-off-by: Lu Baolu <[email protected]>
Acked-by: Jason Gunthorpe <[email protected]>
---
drivers/infiniband/hw/usnic/usnic_uiom.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/infiniband/hw/usnic/usnic_uiom.c b/drivers/infiniband/hw/usnic/usnic_uiom.c
index 84e0f41e7dfa..f948b76f984d 100644
--- a/drivers/infiniband/hw/usnic/usnic_uiom.c
+++ b/drivers/infiniband/hw/usnic/usnic_uiom.c
@@ -443,11 +443,11 @@ struct usnic_uiom_pd *usnic_uiom_alloc_pd(struct device *dev)
if (!pd)
return ERR_PTR(-ENOMEM);

- pd->domain = domain = iommu_domain_alloc(dev->bus);
- if (!domain) {
+ pd->domain = domain = iommu_paging_domain_alloc(dev);
+ if (IS_ERR(domain)) {
usnic_err("Failed to allocate IOMMU domain");
kfree(pd);
- return ERR_PTR(-ENOMEM);
+ return ERR_CAST(domain);
}

iommu_set_fault_handler(pd->domain, usnic_uiom_dma_fault, NULL);
--
2.34.1