[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] xen/arm: gicv2: Export GICv2m register frames to domain0 by device tree



Hi Julien,

On 23 April 2016 at 02:29, Julien Grall <julien.grall@xxxxxxx> wrote:
Hi Wei,


On 21/04/16 09:24, Wei Chen wrote:
This patch adds v2m extension support in GIC-v2 driver. The GICv2 driver
detects the MSI frames from device tree and creates corresponding device
tree nodes in Domain0's DTB. It also provides one hw_ops callback to map
v2m MMIO regions to domain0 and route v2m SPIs to domain0.

With this GICv2m extension support, the domain0 kernel can do GICv2m
frame setup and initialization.

This patch is based on the GICv2m patch of Suravee Suthikulpanit:
[PATCH 2/2] xen/arm: gicv2: Adding support for GICv2m in Dom0
http://lists.xen.org/archives/html/xen-devel/2015-04/msg02613.html

Signed-off-by: Wei Chen <Wei.Chen@xxxxxxxxxx>
---
  xen/arch/arm/domain_build.c |   5 +
  xen/arch/arm/gic-v2.c       | 285 ++++++++++++++++++++++++++++++++++++++++++++
  xen/arch/arm/gic.c          |   9 ++
  xen/include/asm-arm/gic.h   |   3 +
  4 files changed, 302 insertions(+)

diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
index aba714c..b5d82fc 100644
--- a/xen/arch/arm/domain_build.c
+++ b/xen/arch/arm/domain_build.c
@@ -2176,6 +2176,11 @@ int construct_dom0(struct domain *d)
      if ( rc < 0 )
          return rc;

+    /* Map extra GIC MMIO, irqs and other hw stuffs to domain0. */

NIT: s/domain0/dom0/

+    rc = gic_map_hwdom_extra_mappings(d);
+    if ( rc < 0 )
+        return rc;
+
      rc = platform_specific_mapping(d);
      if ( rc < 0 )
          return rc;
diff --git a/xen/arch/arm/gic-v2.c b/xen/arch/arm/gic-v2.c
index 450755f..4a7d65a 100644
--- a/xen/arch/arm/gic-v2.c
+++ b/xen/arch/arm/gic-v2.c
@@ -21,6 +21,7 @@
  #include <xen/lib.h>
  #include <xen/init.h>
  #include <xen/mm.h>
+#include <xen/vmap.h>

Why do you need to include this header?

This header is used for "iounmap" API. 

  #include <xen/irq.h>
  #include <xen/iocap.h>
  #include <xen/sched.h>
@@ -66,6 +67,41 @@
  #define GICH_V2_VMCR_PRIORITY_MASK   0x1f
  #define GICH_V2_VMCR_PRIORITY_SHIFT  27

+/* GICv2m extension register definitions. */
+/*
+* MSI_TYPER:
+*     [31:26] Reserved
+*     [25:16] lowest SPI assigned to MSI
+*     [15:10] Reserved
+*     [9:0]   Number of SPIs assigned to MSI
+*/
+#define V2M_MSI_TYPER               0x008
+#define V2M_MSI_TYPER_BASE_SHIFT    16
+#define V2M_MSI_TYPER_BASE_MASK     0x3FF
+#define V2M_MSI_TYPER_NUM_MASK      0x3FF
+#define V2M_MSI_SETSPI_NS           0x040
+#define V2M_MIN_SPI                 32
+#define V2M_MAX_SPI                 1019
+#define V2M_MSI_IIDR                0xFCC
+
+#define V2M_MSI_TYPER_BASE_SPI(x)   \
+                (((x) >> V2M_MSI_TYPER_BASE_SHIFT) & V2M_MSI_TYPER_BASE_MASK)
+
+#define V2M_MSI_TYPER_NUM_SPI(x)    ((x) & V2M_MSI_TYPER_NUM_MASK)
+
+struct v2m_data {
+    struct list_head entry;
+    /* Pointer to the DT node representing the v2m frame */
+    const struct dt_device_node *dt_node;
+    paddr_t addr; /* Register frame base */
+    paddr_t size; /* Register frame size */
+    u32 spi_start; /* The SPI number that MSIs start */
+    u32 nr_spis; /* The number of SPIs for MSIs */
+};
+
+/* v2m extension register frame information list */
+static LIST_HEAD(gicv2m_info);
+
  /* Global state */
  static struct {
      void __iomem * map_dbase; /* IO mapped Address of distributor registers */
@@ -551,6 +587,173 @@ static void gicv2_irq_set_affinity(struct irq_desc *desc, const cpumask_t *cpu_m
      spin_unlock(&gicv2.lock);
  }

+static int gicv2_map_hwdown_extra_mappings(struct domain *d)
+{
+    const struct v2m_data *v2m_data;
+
+    /*
+     * For the moment, we'll assign all v2m frames to a single domain
+     * (i.e Domain0).

NIT: s/Domain0/hardware domain/

Also, I would drop the i.e and say "... frames to the hardware domain."

+     */
+    list_for_each_entry( v2m_data, &gicv2m_info, entry )
+    {
+        int ret, spi;

Please use "u32" for the variable spi.

+
+        printk("GICv2: Mapping v2m frame to domain%d: addr=0x%lx size=0x%lx spi_base=%u num_spis=%u\n",

NIT: You can shorten domain%d to d%d

+               d->domain_id, v2m_data->addr, v2m_data->size,
+               v2m_data->spi_start, v2m_data->nr_spis);
+
+        ret = map_mmio_regions(d, paddr_to_pfn(v2m_data->addr),
+                               DIV_ROUND_UP(v2m_data->size, PAGE_SIZE),
+                               paddr_to_pfn(v2m_data->addr));
+        if ( ret )
+        {
+            printk(XENLOG_ERR
+                  "GICv2: Map v2m frame to domain%d failed.\n",

Ditto.

Also the indentation looks wrong.
It's very strange. This indentation looks well in the patch file, and all indentations are space chars.
I'll try to send a revised patch.


+                   d->domain_id);
+            return ret;
+        }
+
+        /*
+         * Map all SPIs that are allocated to MSIs in a single frame to the

s/in a single frame/for the frame/


+         * domain.
+         */
+        for ( spi = v2m_data->spi_start;
+              spi < (v2m_data->spi_start + v2m_data->nr_spis); spi++ )
+        {
+            /*
+             * MSIs are always edge-triggered. Configure the associated SPIs
+             * to be edge-rising.
+             */
+            ret = irq_set_spi_type(spi, IRQ_TYPE_EDGE_RISING);
+            if ( ret )
+            {
+                printk(XENLOG_ERR
+                       "GICv2: Failed to set v2m MSI SPI[%d] type.\n", spi);
+                return ret;
+            }
+
+            /* Route a SPI that is allocated to MSI to the domain. */
+            ret = route_irq_to_guest(d, spi, spi, "v2m");
+            if ( ret )
+            {
+                printk(XENLOG_ERR
+                       "GICv2: Failed to route v2m MSI SPI[%d] to Dom%d.\n",
+                        spi, d->domain_id);
+                return ret;
+            }
+
+            /* Reserve a SPI that is allocated to MSI for the domain. */
+            if ( !vgic_reserve_virq(d, spi) )
+            {
+                printk(XENLOG_ERR
+                       "GICv2: Failed to reserve v2m MSI SPI[%d] for Dom%d.\n",
+                        spi, d->domain_id);
+                return -EINVAL;
+            }
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * Set up gic v2m DT sub-node.
+ * Please refer to the binding document:
+ * https://www.kernel.org/doc/Documentation/devicetree/bindings/interrupt-controller/arm,gic.txt
+ */
+static int gicv2m_make_dt_node(const struct domain *d,
+                              const struct dt_device_node *gic,
+                              void *fdt)

The indentation looks wrong to me. It should be:

foo(const struct ...,
    const struct ...
    void *...)

Ditto. 
+{
+    u32 len;
+    int res;
+    const void *prop = NULL;
+    const struct dt_device_node *v2m = NULL;
+    const struct v2m_data *v2m_data;
+
+    /* The sub-nodes require the ranges property */
+    prop = dt_get_property(gic, "ranges", &len);
+    if ( !prop )
+    {
+        printk(XENLOG_ERR "Can't find ranges property for the gic node\n");
+        return -FDT_ERR_XEN(ENOENT);
+    }
+
+    res = fdt_property(fdt, "ranges", prop, len);
+    if ( res )
+        return res;
+
+    list_for_each_entry( v2m_data, &gicv2m_info, entry )
+    {
+        v2m = v2m_data->dt_node;
+
+        printk("GICv2: Creating v2m DT node for domain%d: addr=0x%lx size=0x%lx spi_base=%u num_spis=%u\n",

NIT: s/domain%d/d%d/

+               d->domain_id, v2m_data->addr, v2m_data->size,
+               v2m_data->spi_start, v2m_data->nr_spis);
+
+        res = fdt_begin_node(fdt, v2m->name);
+        if ( res )
+            return res;
+
+        res = fdt_property_string(fdt, "compatible", "arm,gic-v2m-frame");
+        if ( res )
+            return res;
+
+        res = fdt_property(fdt, "msi-controller", NULL, 0);
+        if ( res )
+            return res;
+
+        if ( v2m->phandle )
+        {
+            res = fdt_property_cell(fdt, "phandle", v2m->phandle);
+            if ( res )
+                return res;
+        }
+
+        /* Use the same reg regions as v2m node in DTB. */

s/DTB/host DTB/ to make clear we are speaking about the host one and not the guest one.

+        prop = dt_get_property(v2m, "reg", &len);
+        if ( !prop )
+        {
+            printk(XENLOG_ERR "GICv2: Can't find v2m reg property.\n");
+            res = -FDT_ERR_XEN(ENOENT);
+            return res;
+        }
+
+        res = fdt_property(fdt, "reg", prop, len);
+        if ( res )
+            return res;
+
+        /*
+         * Create msi-base-spi and msi-num-spis properties in Guest DT to
+         * override the hardware setting. From the binding document, we can
+         * find that these two properties are optional. But if these two
+         * properties are present in the v2m DT node, the guest v2m driver
+         * should read the configuration from these two properties instead
+         * of reading from hardware register.

This paragraph is not clear. I would reword;

"The properties msi-base-spi and msi-num-spis are used to override the hardware settings. Therefore it is fine to always write them in the guest DT.".


+         */
+        res = fdt_property_u32(fdt, "arm,msi-base-spi", v2m_data->spi_start);
+        if ( res )
+        {
+            printk(XENLOG_ERR
+                   "GICv2: Failed to create v2m msi-base-spi in Guest DT.\n");
+            return res;
+        }
+
+        res = fdt_property_u32(fdt, "arm,msi-num-spis", v2m_data->nr_spis);
+        if ( res )
+        {
+            printk(XENLOG_ERR
+                   "GICv2: Failed to create v2m msi-num-spis in Guest DT.\n");
+            return res;
+        }
+
+        fdt_end_node(fdt);
+    }
+
+    return res;
+}
+
  static int gicv2_make_hwdom_dt_node(const struct domain *d,
                                      const struct dt_device_node *gic,
                                      void *fdt)
@@ -587,6 +790,10 @@ static int gicv2_make_hwdom_dt_node(const struct domain *d,
      len *= 2;

      res = fdt_property(fdt, "reg", regs, len);
+    if ( res )
+        return res;
+
+    res = gicv2m_make_dt_node(d, gic, fdt);

      return res;
  }
@@ -632,6 +839,77 @@ static bool_t gicv2_is_aliased(paddr_t cbase, paddr_t csize)
      return ((val_low & 0xfff0fff) == 0x0202043B && val_low == val_high);
  }

+static void gicv2_extension_dt_init(const struct dt_device_node *node)
+{
+    int res;
+    u32 spi_start, nr_spis;
+    paddr_t addr, size;
+    const struct dt_device_node *v2m = NULL;
+
+    /*
+     * Check whether this GIC implements the v2m extension. If so,
+     * add v2m register frames to gicv2m_info.
+     */
+    dt_for_each_child_node(node, v2m)
+    {
+        struct v2m_data *v2m_data;
+
+        if ( !dt_device_is_compatible(v2m, "arm,gic-v2m-frame") )
+            continue;
+
+        v2m_data = xzalloc_bytes(sizeof(struct v2m_data));
+        if ( !v2m_data )
+            panic("GICv2: Cannot allocate memory for v2m frame");
+
+        /*
+         * Check whether DT uses msi-base-spi and msi-num-spis properties to
+         * override the hardware setting.
+         */
+        if ( !dt_property_read_u32(v2m, "arm,msi-base-spi", &spi_start) ||
+             !dt_property_read_u32(v2m, "arm,msi-num-spis", &nr_spis) )
+        {
+            u32 msi_typer;
+            void __iomem *base;
+
+            /*
+            * DT does not override the hardware setting, so we have to read
+            * base_spi and num_spis from hardware registers to reserve irqs.
+            */
+            res = dt_device_get_address(v2m, 0, &addr, &size);
+            if ( res )
+                panic("GICv2: Cannot find a valid v2m frame address");
+
+            base = ioremap_nocache(addr, size);
+            if ( !base )
+                panic("GICv2: Cannot remap v2m register frame");
+
+            msi_typer = readl_relaxed(base + V2M_MSI_TYPER);
+            spi_start = V2M_MSI_TYPER_BASE_SPI(msi_typer);
+            nr_spis = V2M_MSI_TYPER_NUM_SPI(msi_typer);
+
+            iounmap(base);
+        }
+
+        /* Initialize a new entry to record new frame infomation. */
+        INIT_LIST_HEAD(&v2m_data->entry);
+        v2m_data->addr = addr;
+        v2m_data->size = size;
+        v2m_data->spi_start = spi_start;
+        v2m_data->nr_spis = nr_spis;
+        v2m_data->dt_node = v2m;
+
+        printk("GICv2m extension register frame:\n"
+               "        gic_v2m_addr=%"PRIpaddr"\n"
+               "        gic_v2m_size=%"PRIpaddr"\n"
+               "        gic_v2m_spi_base=%u\n"
+               "        gic_v2m_num_spis=%u\n",
+               v2m_data->addr, v2m_data->size,
+               v2m_data->spi_start, v2m_data->nr_spis);
+
+        list_add_tail(&v2m_data->entry, &gicv2m_info);

I would pull non-DT code in a separate function, so we can re-use the code later for the support of GICv2m with ACPI.


+    }
+}
+
  static paddr_t __initdata hbase, dbase, cbase, csize, vbase;

  static void __init gicv2_dt_init(void)
@@ -683,6 +961,12 @@ static void __init gicv2_dt_init(void)
      if ( csize != vsize )
          panic("GICv2: Sizes of GICC (%#"PRIpaddr") and GICV (%#"PRIpaddr") don't match\n",
                 csize, vsize);
+
+    /*
+     * Check whether this GIC implements the v2m extension. If so,
+     * add v2m register frames to gicv2_extension_info.
+     */
+    gicv2_extension_dt_init(node);
  }

  static int gicv2_iomem_deny_access(const struct domain *d)
@@ -936,6 +1220,7 @@ const static struct gic_hw_operations gicv2_ops = {
      .read_apr            = gicv2_read_apr,
      .make_hwdom_dt_node  = gicv2_make_hwdom_dt_node,
      .make_hwdom_madt     = gicv2_make_hwdom_madt,
+    .map_hwdom_extra_mappings = gicv2_map_hwdown_extra_mappings,
      .iomem_deny_access   = gicv2_iomem_deny_access,
  };

diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
index 2bfe4de..12bb0ab 100644
--- a/xen/arch/arm/gic.c
+++ b/xen/arch/arm/gic.c
@@ -230,6 +230,15 @@ int gic_irq_xlate(const u32 *intspec, unsigned int intsize,
      return 0;
  }

+/* Map extra GIC MMIO, irqs and other hw stuffs to the hardware domain. */
+int gic_map_hwdom_extra_mappings(struct domain *d)
+{
+    if ( gic_hw_ops->map_hwdom_extra_mappings )
+        return gic_hw_ops->map_hwdom_extra_mappings(d);
+
+    return 0;
+}
+
  static void __init gic_dt_preinit(void)
  {
      int rc;
diff --git a/xen/include/asm-arm/gic.h b/xen/include/asm-arm/gic.h
index cd97bb2..836f1ad 100644
--- a/xen/include/asm-arm/gic.h
+++ b/xen/include/asm-arm/gic.h
@@ -360,6 +360,8 @@ struct gic_hw_operations {
                                const struct dt_device_node *gic, void *fdt);
      /* Create MADT table for the hardware domain */
      int (*make_hwdom_madt)(const struct domain *d, u32 offset);
+    /* Map extra GIC MMIO, irqs and other hw stuffs to the hardware domain. */
+    int (*map_hwdom_extra_mappings)(struct domain *d);
      /* Deny access to GIC regions */
      int (*iomem_deny_access)(const struct domain *d);
  };
@@ -369,6 +371,7 @@ int gic_make_hwdom_dt_node(const struct domain *d,
                             const struct dt_device_node *gic,
                             void *fdt);
  int gic_make_hwdom_madt(const struct domain *d, u32 offset);
+int gic_map_hwdom_extra_mappings(struct domain *d);
  int gic_iomem_deny_access(const struct domain *d);

  #endif /* __ASSEMBLY__ */


Regards,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.