Skip to content

Commit

Permalink
iommu/qcom: Fix bogus detach logic
Browse files Browse the repository at this point in the history
Currently, the implementation of qcom_iommu_domain_free() is guaranteed
to do one of two things: WARN() and leak everything, or dereference NULL
and crash. That alone is terrible, but in fact the whole idea of trying
to track the liveness of a domain via the qcom_domain->iommu pointer as
a sanity check is full of fundamentally flawed assumptions. Make things
robust and actually functional by not trying to be quite so clever.

Reported-by: Brian Masney <[email protected]>
Tested-by: Brian Masney <[email protected]>
Reported-by: Naresh Kamboju <[email protected]>
Fixes: 0ae349a ("iommu/qcom: Add qcom_iommu")
Signed-off-by: Robin Murphy <[email protected]>
Tested-by: Stephan Gerhold <[email protected]>
Cc: [email protected] # v4.14+
Signed-off-by: Joerg Roedel <[email protected]>
  • Loading branch information
rmurphy-arm authored and joergroedel committed Feb 19, 2020
1 parent 3dfee47 commit faf305c
Showing 1 changed file with 12 additions and 16 deletions.
28 changes: 12 additions & 16 deletions drivers/iommu/qcom_iommu.c
Original file line number Diff line number Diff line change
Expand Up @@ -344,21 +344,19 @@ static void qcom_iommu_domain_free(struct iommu_domain *domain)
{
struct qcom_iommu_domain *qcom_domain = to_qcom_iommu_domain(domain);

if (WARN_ON(qcom_domain->iommu)) /* forgot to detach? */
return;

iommu_put_dma_cookie(domain);

/* NOTE: unmap can be called after client device is powered off,
* for example, with GPUs or anything involving dma-buf. So we
* cannot rely on the device_link. Make sure the IOMMU is on to
* avoid unclocked accesses in the TLB inv path:
*/
pm_runtime_get_sync(qcom_domain->iommu->dev);

free_io_pgtable_ops(qcom_domain->pgtbl_ops);

pm_runtime_put_sync(qcom_domain->iommu->dev);
if (qcom_domain->iommu) {
/*
* NOTE: unmap can be called after client device is powered
* off, for example, with GPUs or anything involving dma-buf.
* So we cannot rely on the device_link. Make sure the IOMMU
* is on to avoid unclocked accesses in the TLB inv path:
*/
pm_runtime_get_sync(qcom_domain->iommu->dev);
free_io_pgtable_ops(qcom_domain->pgtbl_ops);
pm_runtime_put_sync(qcom_domain->iommu->dev);
}

kfree(qcom_domain);
}
Expand Down Expand Up @@ -404,7 +402,7 @@ static void qcom_iommu_detach_dev(struct iommu_domain *domain, struct device *de
struct qcom_iommu_domain *qcom_domain = to_qcom_iommu_domain(domain);
unsigned i;

if (!qcom_domain->iommu)
if (WARN_ON(!qcom_domain->iommu))
return;

pm_runtime_get_sync(qcom_iommu->dev);
Expand All @@ -417,8 +415,6 @@ static void qcom_iommu_detach_dev(struct iommu_domain *domain, struct device *de
ctx->domain = NULL;
}
pm_runtime_put_sync(qcom_iommu->dev);

qcom_domain->iommu = NULL;
}

static int qcom_iommu_map(struct iommu_domain *domain, unsigned long iova,
Expand Down

0 comments on commit faf305c

Please sign in to comment.