iommu/arm-smmu: Avoid pathological RPM behaviour for unmaps

When games, browser, or anything using a lot of GPU buffers exits, there
can be many hundreds or thousands of buffers to unmap and free.  If the
GPU is otherwise suspended, this can cause arm-smmu to resume/suspend
for each buffer, resulting 5-10 seconds worth of reprogramming the
context bank (arm_smmu_write_context_bank()/arm_smmu_write_s2cr()/etc).
To the user it would appear that the system just locked up.

A simple solution is to use pm_runtime_put_autosuspend() instead, so we
don't immediately suspend the SMMU device.

Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>
Signed-off-by: Will Deacon <will@kernel.org>

diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
index 080af0326816dafebb52c6b8f862c5484f37d5eb..a2b1ca55b73e0fb9f691a81312bd480429252cbc 100644 (file)
--- a/drivers/iommu/arm-smmu.c
+++ b/drivers/iommu/arm-smmu.c
@@ -123,7 +123,7 @@ static inline int arm_smmu_rpm_get(struct arm_smmu_device *smmu)
 static inline void arm_smmu_rpm_put(struct arm_smmu_device *smmu)
 {
        if (pm_runtime_enabled(smmu->dev))
-               pm_runtime_put(smmu->dev);
+               pm_runtime_put_autosuspend(smmu->dev);
 }
 
 static struct arm_smmu_domain *to_smmu_domain(struct iommu_domain *dom)
@@ -1167,6 +1167,20 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
        /* Looks ok, so add the device to the domain */
        ret = arm_smmu_domain_add_master(smmu_domain, fwspec);
 
+       /*
+        * Setup an autosuspend delay to avoid bouncing runpm state.
+        * Otherwise, if a driver for a suspended consumer device
+        * unmaps buffers, it will runpm resume/suspend for each one.
+        *
+        * For example, when used by a GPU device, when an application
+        * or game exits, it can trigger unmapping 100s or 1000s of
+        * buffers.  With a runpm cycle for each buffer, that adds up
+        * to 5-10sec worth of reprogramming the context bank, while
+        * the system appears to be locked up to the user.
+        */
+       pm_runtime_set_autosuspend_delay(smmu->dev, 20);
+       pm_runtime_use_autosuspend(smmu->dev);
+
 rpm_put:
        arm_smmu_rpm_put(smmu);
        return ret;