> -----Original Message----- > From: Lorenzo Pieralisi [mailto:lorenzo.pieralisi@xxxxxxx] > Sent: Friday, September 22, 2017 3:28 PM > To: Shameerali Kolothum Thodi <shameerali.kolothum.thodi@xxxxxxxxxx> > Cc: marc.zyngier@xxxxxxx; sudeep.holla@xxxxxxx; will.deacon@xxxxxxx; > robin.murphy@xxxxxxx; joro@xxxxxxxxxx; mark.rutland@xxxxxxx; > hanjun.guo@xxxxxxxxxx; Gabriele Paoloni <gabriele.paoloni@xxxxxxxxxx>; > John Garry <john.garry@xxxxxxxxxx>; iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx; > linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-acpi@xxxxxxxxxxxxxxx; > devicetree@xxxxxxxxxxxxxxx; devel@xxxxxxxxxx; Linuxarm > <linuxarm@xxxxxxxxxx>; Wangzhou (B) <wangzhou1@xxxxxxxxxxxxx>; > Guohanjun (Hanjun Guo) <guohanjun@xxxxxxxxxx> > Subject: Re: [PATCH v7 3/5] iommu/of: Add msi address regions reservation > helper > > John, Shameer, > > On Thu, Sep 14, 2017 at 01:57:54PM +0100, Shameer Kolothum wrote: > > From: John Garry <john.garry@xxxxxxxxxx> > > > > On some platforms msi-controller address regions have to be excluded > > from normal IOVA allocation in that they are detected and decoded in > > a HW specific way by system components and so they cannot be > considered > > normal IOVA address space. > > > > Add a helper function that retrieves msi address regions through device > > tree msi mapping, so that these regions will not be translated by IOMMU > > and will be excluded from IOVA allocations. > > > > Signed-off-by: John Garry <john.garry@xxxxxxxxxx> > > Signed-off-by: Shameer Kolothum > <shameerali.kolothum.thodi@xxxxxxxxxx> > > --- > > drivers/iommu/of_iommu.c | 117 > +++++++++++++++++++++++++++++++++++++++++++++++ > > include/linux/of_iommu.h | 10 ++++ > > 2 files changed, 127 insertions(+) > > > > diff --git a/drivers/iommu/of_iommu.c b/drivers/iommu/of_iommu.c > > index 8cb6082..f2d1a76 100644 > > --- a/drivers/iommu/of_iommu.c > > +++ b/drivers/iommu/of_iommu.c > > @@ -21,6 +21,7 @@ > > #include <linux/iommu.h> > > #include <linux/limits.h> > > #include <linux/of.h> > > +#include <linux/of_address.h> > > #include <linux/of_iommu.h> > > #include <linux/of_pci.h> > > #include <linux/slab.h> > > @@ -246,6 +247,122 @@ const struct iommu_ops > *of_iommu_configure(struct device *dev, > > return ops; > > } > > > > +/** > > + * of_iommu_msi_get_resv_regions - Reserved region driver helper > > + * @dev: Device from iommu_get_resv_regions() > > + * @list: Reserved region list from iommu_get_resv_regions() > > + * > > + * Returns: Number of reserved regions on success (0 if no associated > > + * msi parent), appropriate error value otherwise. > > + */ > > +int of_iommu_msi_get_resv_regions(struct device *dev, struct list_head > *head) > > +{ > > + int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO; > > + struct iommu_resv_region *region; > > + struct device_node *np; > > + struct resource res; > > + int i, resv = 0, mappings = 0; > > + > > + if (dev_is_pci(dev)) { > > + struct device *dma_dev, *bridge; > > + struct of_phandle_args iommu_spec; > > + struct pci_dev *pdev = to_pci_dev(dev); > > + int err, count; > > + u32 rid, map_mask; > > + const __be32 *msi_map; > > + > > + bridge = pci_get_host_bridge_device(pdev); > > + dma_dev = bridge->parent; > > + pci_put_host_bridge_device(bridge); > > + > > + if (!dma_dev->of_node) > > + return -ENODEV; > > + > > + iommu_spec.args_count = 1; > > + np = iommu_spec.np = dma_dev->of_node; > > + pci_for_each_dma_alias(pdev, __get_pci_rid, > &iommu_spec); > > + > > + rid = iommu_spec.args[0]; > > + if (!of_property_read_u32(np, "msi-map-mask", > &map_mask)) > > + rid &= map_mask; > > + > > + msi_map = of_get_property(np, "msi-map", NULL); > > + if (!msi_map) > > + return -ENODEV; > > + > > + mappings = of_count_phandle_with_args(np, "msi-map", > NULL) / 4; > > + > > + for (i = 0, count = mappings; i < count; i++, msi_map += 4) { > > + struct device_node *msi_node; > > + u32 rid_base, rid_len, phandle; > > + > > + rid_base = be32_to_cpup(msi_map + 0); > > + phandle = be32_to_cpup(msi_map + 1); > > + rid_len = be32_to_cpup(msi_map + 3); > > + > > + /* check rid is within range */ > > + if (rid < rid_base || rid >= rid_base + rid_len) { > > + mappings--; > > + continue; > > + } > > + > > + msi_node = of_find_node_by_phandle(phandle); > > + if (!msi_node) > > + return -ENODEV; > > This is basically of_pci_map_rid(), I wonder whether there is not > a way to consolidate some code here - duplicating certainly does not > help. To make MSI reservations generic this is probably the only way > to do it but it would be nice to reuse some OF MSI code. > > With the current kernel API there is a way but it is a bit whacky. > > Just loop over "msi-controller" nodes and try to map the device to > them through of_pci_map_rid, if mapping succeeds reserve region for > the target node. > > Not a big fan of what I am proposing but it certainly helps reuse > some existing code that makes no sense to duplicate. Right, lot of this is of_pci_map_rid() code. And just to confirm, I think the proposal is to make use of the @target node param in of_pci_map_rid() for a matching "msi-controller". > > + err = of_address_to_resource(msi_node, 0, &res); > > + of_node_put(msi_node); > > + if (err) > > + return err; > > + > > + region = iommu_alloc_resv_region(res.start, > > + resource_size(&res), > > + prot, > IOMMU_RESV_MSI); > > + if (region) { > > + list_add_tail(®ion->list, head); > > + resv++; > > + } > > + } > > + } else if (dev->of_node) { > > + struct device_node *msi_np; > > + int index = 0; > > + int tuples; > > + > > + np = dev->of_node; > > + > > + tuples = of_count_phandle_with_args(np, "msi-parent", > NULL); > > + > > + while (index < tuples) { > > Would not be easier to have an of_parse_phandle_with_args() loop here ? Ok. Many thanks for going through this. We will rework this based on your suggestions and send out the next revision (rebased on -rc1). Thanks, Shameer -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
