* [PATCH 0/2] iommu/qcom+msm: Use identity domain for display/GPU
@ 2026-10-29 21:23 Rob Clark
2026-10-29 21:23 ` [PATCH 1/2] iommu/qcom: Use identity domain for GPU and display Rob Clark
2026-10-29 21:23 ` [PATCH 2/2] iommu/msm: " Rob Clark
0 siblings, 2 replies; 3+ messages in thread
From: Rob Clark @ 2026-10-29 21:23 UTC (permalink / raw)
To: iommu
Cc: dri-devel, linux-arm-msm, freedreno, Dmitry Baryshkov, Rob Clark,
Joerg Roedel (AMD),
moderated list:ARM SMMU DRIVERS, open list, Robin Murphy,
Will Deacon
Similar to what is already done in arm-smmu-qcom, match display and GPU
compatible strings to pick identity domain type for them.
So far, we have an "evil twin" domain kicking around after drm/msm
attaches it's own domain. Until now, that has been relatively harmless.
But it becomes a problem with [1] when we start using sg_dma_len()/
sg_dma_address() for setting up mappings. Which is required to avoid
problems with CONFIG_DMABUF_DEBUG (dma_buf_wrap_sg_table()).
The first patch is at least compile tested.. msm_iommu seems to be
armv7-only.
If there are devices with bootloader splash, we may also need some
work on qcom_iommu_identity_attach() and/or msm_iommu_identity_attach()
to put the context bank into bypass instead of disabling it? Not sure,
but that is probably going to require someone with such hardware at
hand.
[1] https://patchwork.freedesktop.org/series/175619/
Rob Clark (2):
iommu/qcom: Use identity domain for GPU and display
iommu/msm: Use identity domain for GPU and display
drivers/iommu/arm/arm-smmu/qcom_iommu.c | 18 ++++++++++++++++++
drivers/iommu/msm_iommu.c | 16 ++++++++++++++++
2 files changed, 34 insertions(+)
--
2.56.0
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 1/2] iommu/qcom: Use identity domain for GPU and display
2026-10-29 21:23 [PATCH 0/2] iommu/qcom+msm: Use identity domain for display/GPU Rob Clark
@ 2026-10-29 21:23 ` Rob Clark
2026-10-29 21:23 ` [PATCH 2/2] iommu/msm: " Rob Clark
1 sibling, 0 replies; 3+ messages in thread
From: Rob Clark @ 2026-10-29 21:23 UTC (permalink / raw)
To: iommu
Cc: dri-devel, linux-arm-msm, freedreno, Dmitry Baryshkov, Rob Clark,
Will Deacon, Robin Murphy, Joerg Roedel (AMD),
moderated list:ARM SMMU DRIVERS, open list
Similar to commit 0e764a01015d ("iommu/arm-smmu: Allow client devices to
select direct mapping") for arm-smmu-qcom, match GPU and display compat
strings to pick identity domain time.
The drm/msm driver attaches it's own domain, which it manages directly,
so we don't want an iommu managed domain interfering.
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
---
drivers/iommu/arm/arm-smmu/qcom_iommu.c | 18 ++++++++++++++++++
1 file changed, 18 insertions(+)
diff --git a/drivers/iommu/arm/arm-smmu/qcom_iommu.c b/drivers/iommu/arm/arm-smmu/qcom_iommu.c
index 21d18ce67b98..5b10feb419d8 100644
--- a/drivers/iommu/arm/arm-smmu/qcom_iommu.c
+++ b/drivers/iommu/arm/arm-smmu/qcom_iommu.c
@@ -22,6 +22,7 @@
#include <linux/init.h>
#include <linux/mutex.h>
#include <linux/of.h>
+#include <linux/of_device.h>
#include <linux/of_platform.h>
#include <linux/platform_device.h>
#include <linux/pm.h>
@@ -589,6 +590,22 @@ static int qcom_iommu_of_xlate(struct device *dev,
return iommu_fwspec_add_ids(dev, &asid, 1);
}
+static const struct of_device_id qcom_iommu_client_of_match[] = {
+ { .compatible = "qcom,adreno" },
+ { .compatible = "qcom,mdp4" },
+ { .compatible = "qcom,mdp5" },
+ { .compatible = "qcom,mdss" },
+ { }
+};
+
+static int qcom_iommu_def_domain_type(struct device *dev)
+{
+ const struct of_device_id *match =
+ of_match_device(qcom_iommu_client_of_match, dev);
+
+ return match ? IOMMU_DOMAIN_IDENTITY : 0;
+}
+
static const struct iommu_ops qcom_iommu_ops = {
.identity_domain = &qcom_iommu_identity_domain,
.capable = qcom_iommu_capable,
@@ -596,6 +613,7 @@ static const struct iommu_ops qcom_iommu_ops = {
.probe_device = qcom_iommu_probe_device,
.device_group = generic_device_group,
.of_xlate = qcom_iommu_of_xlate,
+ .def_domain_type = qcom_iommu_def_domain_type,
.default_domain_ops = &(const struct iommu_domain_ops) {
.attach_dev = qcom_iommu_attach_dev,
.map_pages = qcom_iommu_map,
--
2.56.0
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 2/2] iommu/msm: Use identity domain for GPU and display
2026-10-29 21:23 [PATCH 0/2] iommu/qcom+msm: Use identity domain for display/GPU Rob Clark
2026-10-29 21:23 ` [PATCH 1/2] iommu/qcom: Use identity domain for GPU and display Rob Clark
@ 2026-10-29 21:23 ` Rob Clark
1 sibling, 0 replies; 3+ messages in thread
From: Rob Clark @ 2026-10-29 21:23 UTC (permalink / raw)
To: iommu
Cc: dri-devel, linux-arm-msm, freedreno, Dmitry Baryshkov, Rob Clark,
Joerg Roedel (AMD),
Will Deacon, Robin Murphy, open list
Similar to commit 0e764a01015d ("iommu/arm-smmu: Allow client devices to
select direct mapping") for arm-smmu-qcom, match GPU and display compat
strings to pick identity domain time.
The drm/msm driver attaches it's own domain, which it manages directly,
so we don't want an iommu managed domain interfering.
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
---
drivers/iommu/msm_iommu.c | 16 ++++++++++++++++
1 file changed, 16 insertions(+)
diff --git a/drivers/iommu/msm_iommu.c b/drivers/iommu/msm_iommu.c
index 4ef1e8c8244b..1c5011432628 100644
--- a/drivers/iommu/msm_iommu.c
+++ b/drivers/iommu/msm_iommu.c
@@ -12,6 +12,7 @@
#include <linux/io.h>
#include <linux/io-pgtable.h>
#include <linux/interrupt.h>
+#include <linux/of_device.h>
#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/slab.h>
@@ -691,12 +692,27 @@ irqreturn_t msm_iommu_fault_handler(int irq, void *dev_id)
return 0;
}
+static const struct of_device_id msm_iommu_client_of_match[] = {
+ { .compatible = "qcom,adreno" },
+ { .compatible = "qcom,mdp4" },
+ { }
+};
+
+static int msm_iommu_def_domain_type(struct device *dev)
+{
+ const struct of_device_id *match =
+ of_match_device(msm_iommu_client_of_match, dev);
+
+ return match ? IOMMU_DOMAIN_IDENTITY : 0;
+}
+
static struct iommu_ops msm_iommu_ops = {
.identity_domain = &msm_iommu_identity_domain,
.domain_alloc_paging = msm_iommu_domain_alloc_paging,
.probe_device = msm_iommu_probe_device,
.device_group = generic_device_group,
.of_xlate = qcom_iommu_of_xlate,
+ .def_domain_type = msm_iommu_def_domain_type,
.default_domain_ops = &(const struct iommu_domain_ops) {
.attach_dev = msm_iommu_attach_dev,
.map_pages = msm_iommu_map,
--
2.56.0
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-10-29 21:23 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-10-29 21:23 [PATCH 0/2] iommu/qcom+msm: Use identity domain for display/GPU Rob Clark
2026-10-29 21:23 ` [PATCH 1/2] iommu/qcom: Use identity domain for GPU and display Rob Clark
2026-10-29 21:23 ` [PATCH 2/2] iommu/msm: " Rob Clark
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®