Re: [PATCH V5 3/5] iommu/dma-iommu: Handle deferred devices

On Sat, 17 Aug 2019 at 04:39, Hillf Danton [off-list ref] wrote:

On Thu, 15 Aug 2019 12:09:41 +0100 Tom Murphy wrote:

quoted

Handle devices which defer their attach to the iommu in the dma-iommu api

Signed-off-by: Tom Murphy <redacted>
---
 drivers/iommu/dma-iommu.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 2712fbc68b28..906b7fa14d3c 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c

@@ -22,6 +22,7 @@
 #include <linux/pci.h>
 #include <linux/scatterlist.h>
 #include <linux/vmalloc.h>
+#include <linux/crash_dump.h>

 struct iommu_dma_msi_page {
      struct list_head        list;

@@ -351,6 +352,21 @@ static int iommu_dma_init_domain(struct iommu_domain *domain, dma_addr_t base,
      return iova_reserve_iommu_regions(dev, domain);
 }

+static int handle_deferred_device(struct device *dev,
+     struct iommu_domain *domain)
+{
+     const struct iommu_ops *ops = domain->ops;
+
+     if (!is_kdump_kernel())
+             return 0;
+
+     if (unlikely(ops->is_attach_deferred &&
+             ops->is_attach_deferred(domain, dev)))
+             return iommu_attach_device(domain, dev);
+
+     return 0;
+}
+
 /**
  * dma_info_to_prot - Translate DMA API directions and attributes to IOMMU API
  *                    page flags.

@@ -463,6 +479,9 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
      size_t iova_off = iova_offset(iovad, phys);
      dma_addr_t iova;

+     if (unlikely(handle_deferred_device(dev, domain)))
+             return DMA_MAPPING_ERROR;
+
      size = iova_align(iovad, size + iova_off);

      iova = iommu_dma_alloc_iova(domain, size, dma_get_mask(dev), dev);

iommu_map_atomic() is applied to __iommu_dma_map() in 2/5.
Is it an atomic context currently given the mutex_lock() in
iommu_attach_device()?

I don't see your point here. __iommu_dma_map isn't called from
iommu_attach_device, why would we care about a mutex in
iommu_attach_device?

__iommu_dma_map can be called from an atomic context (it isn't always
but it does happen). __iommu_dma_map is called by iommu_dma_alloc
which implements the iommu_dma_ops::alloc function which by design
needs to be callable from an atomic context. Does that answer your
question?