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

Re: [PATCH V7 2/2] libxl: Introduce basic virtio-mmio support on Arm




On 19.04.22 00:41, Stefano Stabellini wrote:


Hello Stefano

On Fri, 8 Apr 2022, Oleksandr Tyshchenko wrote:
From: Julien Grall <julien.grall@xxxxxxx>

This patch introduces helpers to allocate Virtio MMIO params
(IRQ and memory region) and create specific device node in
the Guest device-tree with allocated params. In order to deal
with multiple Virtio devices, reserve corresponding ranges.
For now, we reserve 1MB for memory regions and 10 SPIs.

As these helpers should be used for every Virtio device attached
to the Guest, call them for Virtio disk(s).

Please note, with statically allocated Virtio IRQs there is
a risk of a clash with a physical IRQs of passthrough devices.
For the first version, it's fine, but we should consider allocating
the Virtio IRQs automatically. Thankfully, we know in advance which
IRQs will be used for passthrough to be able to choose non-clashed
ones.

Signed-off-by: Julien Grall <julien.grall@xxxxxxx>
Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@xxxxxxxx>
Tested-by: Jiamei Xie <Jiamei.xie@xxxxxxx>
Reviewed-by: Henry Wang <Henry.Wang@xxxxxxx>
I realize that we are at v7 and I haven't reviewed this before, so I'll
limit my comments. I'll clarify the ones that I think are more important
from the one that are less important.

thank you




---
@Jiamei, @Henry I decided to leave your T-b and R-b tags with the minor
change I made, are you still happy with that?

s/if (disk->virtio)/if (disk->protocol == LIBXL_DISK_PROTOCOL_VIRTIO_MMIO)

Please note, this is a split/cleanup/hardening of Julien's PoC:
"Add support for Guest IO forwarding to a device emulator"

Changes RFC -> V1:
    - was squashed with:
      "[RFC PATCH V1 09/12] libxl: Handle virtio-mmio irq in more correct way"
      "[RFC PATCH V1 11/12] libxl: Insert "dma-coherent" property into virtio-mmio 
device node"
      "[RFC PATCH V1 12/12] libxl: Fix duplicate memory node in DT"
    - move VirtIO MMIO #define-s to xen/include/public/arch-arm.h

Changes V1 -> V2:
    - update the author of a patch

Changes V2 -> V3:
    - no changes

Changes V3 -> V4:
    - no changes

Changes V4 -> V5:
    - split the changes, change the order of the patches
    - drop an extra "virtio" configuration option
    - update patch description
    - use CONTAINER_OF instead of own implementation
    - reserve ranges for Virtio MMIO params and put them
      in correct location
    - create helpers to allocate Virtio MMIO params, add
      corresponding sanity-сhecks
    - add comment why MMIO size 0x200 is chosen
    - update debug print
    - drop Wei's T-b

Changes V5 -> V6:
    - rebase on current staging

Changes V6 -> V7:
    - rebase on current staging
    - add T-b and R-b tags
    - update according to the recent changes to
      "libxl: Add support for Virtio disk configuration"
---
  tools/libs/light/libxl_arm.c  | 131 +++++++++++++++++++++++++++++++++++++++++-
  xen/include/public/arch-arm.h |   7 +++
  2 files changed, 136 insertions(+), 2 deletions(-)

diff --git a/tools/libs/light/libxl_arm.c b/tools/libs/light/libxl_arm.c
index eef1de0..8132a47 100644
--- a/tools/libs/light/libxl_arm.c
+++ b/tools/libs/light/libxl_arm.c
@@ -8,6 +8,56 @@
  #include <assert.h>
  #include <xen/device_tree_defs.h>
+/*
+ * There is no clear requirements for the total size of Virtio MMIO region.
+ * The size of control registers is 0x100 and device-specific configuration
+ * registers starts at the offset 0x100, however it's size depends on the 
device
+ * and the driver. Pick the biggest known size at the moment to cover most
+ * of the devices (also consider allowing the user to configure the size via
+ * config file for the one not conforming with the proposed value).
+ */
+#define VIRTIO_MMIO_DEV_SIZE   xen_mk_ullong(0x200)
Actually, I don't think we need to make this generic. We only support
virtio-disk now and I think it is fine if this was called
VIRTIO_DISK_MMIO_DEV_SIZE and the size was exactly the size needed by
virtio-disk. When/if we support another virtio protocol we can add set
the appropriate size of that one as well.

I would prefer if we keep this generic, although we are going to support virtio-blk as the first virtio-mmio device, there is nothing disk specific in that value. The same value (0x200) is used by SW which also generates virtio-mmio device tree nodes like we do in Xen libxl, I am speaking about kvmtool, Qemu here.



At the moment as you can see below:

      */
     for (i = 0; i < d_config->num_disks; i++) {
         libxl_device_disk *disk = &d_config->disks[i];

         if (disk->protocol == LIBXL_DISK_PROTOCOL_VIRTIO_MMIO) {

it is all very virtio-disk specific, so we might as well exploit it for
the better :-)

Given that we are at v7 of this series, I don't think this is very
important, so I think it is OK if we keep the code as is.

ok




+static uint64_t virtio_mmio_base;
+static uint32_t virtio_mmio_irq;
No need for these two variables to be global in this file, they could be
local variables in libxl__arch_domain_prepare_config.

I think that this is a change we should make if possible before
committing these patches.


I agree with that, will do





+static void init_virtio_mmio_params(void)
+{
+    virtio_mmio_base = GUEST_VIRTIO_MMIO_BASE;
+    virtio_mmio_irq = GUEST_VIRTIO_MMIO_SPI_FIRST;
+}
+
+static uint64_t alloc_virtio_mmio_base(libxl__gc *gc)
+{
+    uint64_t base = virtio_mmio_base;
If virtio_mmio_base is local in libxl__arch_domain_prepare_config, then
it could be passed here as parameter (as pointer).


yes, good point, will do



If we make the dev MMIO size (GUEST_VIRTIO_MMIO_SIZE) specific to
virtio-disk, then we could pass it also as argument to this function:

static uint64_t alloc_virtio_mmio_base(libxl__gc *gc,
                                        uint64 *virtio_mmio_base,
                                        uint64_t dev_mmio_size)


Please see my comment regarding VIRTIO_MMIO_DEV_SIZE above





+    /* Make sure we have enough reserved resources */
+    if ((virtio_mmio_base + VIRTIO_MMIO_DEV_SIZE >
+        GUEST_VIRTIO_MMIO_BASE + GUEST_VIRTIO_MMIO_SIZE)) {
+        LOG(ERROR, "Ran out of reserved range for Virtio MMIO BASE 
0x%"PRIx64"\n",
+            virtio_mmio_base);
+        return 0;
+    }
+    virtio_mmio_base += VIRTIO_MMIO_DEV_SIZE;
+
+    return base;
+}
+
+static uint32_t alloc_virtio_mmio_irq(libxl__gc *gc)
+{
+    uint32_t irq = virtio_mmio_irq;
+
+    /* Make sure we have enough reserved resources */
+    if (virtio_mmio_irq > GUEST_VIRTIO_MMIO_SPI_LAST) {
+        LOG(ERROR, "Ran out of reserved range for Virtio MMIO IRQ %u\n",
+            virtio_mmio_irq);
+        return 0;
+    }
+    virtio_mmio_irq++;
+
+    return irq;
+}
+
  static const char *gicv_to_string(libxl_gic_version gic_version)
  {
      switch (gic_version) {
@@ -26,8 +76,8 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc,
  {
      uint32_t nr_spis = 0;
      unsigned int i;
-    uint32_t vuart_irq;
-    bool vuart_enabled = false;
+    uint32_t vuart_irq, virtio_irq = 0;
+    bool vuart_enabled = false, virtio_enabled = false;
/*
       * If pl011 vuart is enabled then increment the nr_spis to allow 
allocation
@@ -39,6 +89,35 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc,
          vuart_enabled = true;
      }
+ /*
+     * Virtio MMIO params are non-unique across the whole system and must be
+     * initialized for every new guest.
+     */
+    init_virtio_mmio_params();
No need to initialize them in init_virtio_mmio_params, you could just
do:

   uint64_t virtio_mmio_base = GUEST_VIRTIO_MMIO_BASE;
   uint32_t virtio_mmio_irq = GUEST_VIRTIO_MMIO_SPI_FIRST;

at the top of libxl__arch_domain_prepare_config


yes, will do




+    for (i = 0; i < d_config->num_disks; i++) {
+        libxl_device_disk *disk = &d_config->disks[i];
+
+        if (disk->protocol == LIBXL_DISK_PROTOCOL_VIRTIO_MMIO) {
+            disk->base = alloc_virtio_mmio_base(gc);
+            if (!disk->base)
+                return ERROR_FAIL;
+
+            disk->irq = alloc_virtio_mmio_irq(gc);
+            if (!disk->irq)
+                return ERROR_FAIL;
+
+            if (virtio_irq < disk->irq)
+                virtio_irq = disk->irq;
+            virtio_enabled = true;
+
+            LOG(DEBUG, "Allocate Virtio MMIO params for Vdev %s: IRQ %u BASE 
0x%"PRIx64,
+                disk->vdev, disk->irq, disk->base);
+        }
+    }
+
+    if (virtio_enabled)
+        nr_spis += (virtio_irq - 32) + 1;
+
      for (i = 0; i < d_config->b_info.num_irqs; i++) {
          uint32_t irq = d_config->b_info.irqs[i];
          uint32_t spi;
@@ -58,6 +137,13 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc,
              return ERROR_FAIL;
          }
+ /* The same check as for vpl011 */
+        if (virtio_enabled &&
+           (irq >= GUEST_VIRTIO_MMIO_SPI_FIRST && irq <= virtio_irq)) {
NIT: alignment:

         if (virtio_enabled &&
             (irq >= GUEST_VIRTIO_MMIO_SPI_FIRST && irq <= virtio_irq)) {


ok





+            LOG(ERROR, "Physical IRQ %u conflicting with Virtio MMIO IRQ 
range\n", irq);
+            return ERROR_FAIL;
+        }
+
          if (irq < 32)
              continue;
@@ -787,6 +873,39 @@ static int make_vpci_node(libxl__gc *gc, void *fdt,
      return 0;
  }
+
+static int make_virtio_mmio_node(libxl__gc *gc, void *fdt,
+                                 uint64_t base, uint32_t irq)
+{
+    int res;
+    gic_interrupt intr;
+    /* Placeholder for virtio@ + a 64-bit number + \0 */
+    char buf[24];
+
+    snprintf(buf, sizeof(buf), "virtio@%"PRIx64, base);
+    res = fdt_begin_node(fdt, buf);
+    if (res) return res;
+
+    res = fdt_property_compat(gc, fdt, 1, "virtio,mmio");
+    if (res) return res;
+
+    res = fdt_property_regs(gc, fdt, GUEST_ROOT_ADDRESS_CELLS, 
GUEST_ROOT_SIZE_CELLS,
+                            1, base, VIRTIO_MMIO_DEV_SIZE);
+    if (res) return res;
+
+    set_interrupt(intr, irq, 0xf, DT_IRQ_TYPE_EDGE_RISING);
+    res = fdt_property_interrupts(gc, fdt, &intr, 1);
+    if (res) return res;
+
+    res = fdt_property(fdt, "dma-coherent", NULL, 0);
+    if (res) return res;
+
+    res = fdt_end_node(fdt);
+    if (res) return res;
+
+    return 0;
+}
+
  static const struct arch_info *get_arch_info(libxl__gc *gc,
                                               const struct xc_dom_image *dom)
  {
@@ -988,6 +1107,7 @@ static int libxl__prepare_dtb(libxl__gc *gc, 
libxl_domain_config *d_config,
      size_t fdt_size = 0;
      int pfdt_size = 0;
      libxl_domain_build_info *const info = &d_config->b_info;
+    unsigned int i;
const libxl_version_info *vers;
      const struct arch_info *ainfo;
@@ -1094,6 +1214,13 @@ next_resize:
          if (d_config->num_pcidevs)
              FDT( make_vpci_node(gc, fdt, ainfo, dom) );
+ for (i = 0; i < d_config->num_disks; i++) {
+            libxl_device_disk *disk = &d_config->disks[i];
+
+            if (disk->protocol == LIBXL_DISK_PROTOCOL_VIRTIO_MMIO)
+                FDT( make_virtio_mmio_node(gc, fdt, disk->base, disk->irq) );
+        }
+
          if (pfdt)
              FDT( copy_partial_fdt(gc, fdt, pfdt) );
diff --git a/xen/include/public/arch-arm.h b/xen/include/public/arch-arm.h
index ab05fe1..c8b6058 100644
--- a/xen/include/public/arch-arm.h
+++ b/xen/include/public/arch-arm.h
@@ -407,6 +407,10 @@ typedef uint64_t xen_callback_t;
/* Physical Address Space */ +/* Virtio MMIO mappings */
+#define GUEST_VIRTIO_MMIO_BASE   xen_mk_ullong(0x02000000)
+#define GUEST_VIRTIO_MMIO_SIZE   xen_mk_ullong(0x00100000)
+
  /*
   * vGIC mappings: Only one set of mapping is used by the guest.
   * Therefore they can overlap.
@@ -493,6 +497,9 @@ typedef uint64_t xen_callback_t;
#define GUEST_VPL011_SPI 32 +#define GUEST_VIRTIO_MMIO_SPI_FIRST 33
+#define GUEST_VIRTIO_MMIO_SPI_LAST    43
+
  /* PSCI functions */
  #define PSCI_cpu_suspend 0
  #define PSCI_cpu_off     1
--
2.7.4

--
Regards,

Oleksandr Tyshchenko




 


Rackspace

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