Re: [Qemu-devel] [patch v5 06/12] vfio: add check host bus reset is support or not

[Alex Williamson <alex.williamson@redhat.com>]

[resending, first copy had the list address incorrect, apologies to
those on the cc list]

On Wed, 23 Mar 2016 18:08:16 +0800
Cao jin <caoj.fnst@cn.fujitsu.com> wrote:

> From: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
> 
> When assigning a vfio device with AER enabled, we must check whether
> the device supports a host bus reset (ie. hot reset) as this may be
> used by the guest OS in order to recover the device from an AER
> error.  QEMU must therefore have the ability to perform a physical
> host bus reset using the existing vfio APIs in response to a virtual
> bus reset in the VM.  A physical bus reset affects all of the devices
> on the host bus, therefore we place a few simplifying configuration
> restriction on the VM:
> 
>  - All physical devices affected by a bus reset must be assigned to
>    the VM with AER enabled on each and be configured on the same
>    virtual bus in the VM.
> 
>  - No devices unaffected by the bus reset, be they physical, emulated,
>    or paravirtual may be configured on the same virtual bus as a
>    device supporting AER signaling through vfio.
> 
> In other words users wishing to enable AER on a multifunction device
> need to assign all functions of the device to the same virtual bus
> and enable AER support for each device.  The easiest way to
> accomplish this is to identity map the physical functions to virtual
> functions with multifunction enabled on the virtual device.
> 
> Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
> ---
>  hw/vfio/pci.c | 276 +++++++++++++++++++++++++++++++++++++++++++++++++++++-----
>  hw/vfio/pci.h |   1 +
>  2 files changed, 254 insertions(+), 23 deletions(-)
> 
> diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> index 5b23a86..939b764 100644
> --- a/hw/vfio/pci.c
> +++ b/hw/vfio/pci.c
> @@ -1716,6 +1716,41 @@ static void vfio_check_af_flr(VFIOPCIDevice *vdev, uint8_t pos)
>      }
>  }
>  
> +static int vfio_pci_name_to_addr(const char *name, PCIHostDeviceAddress *addr)
> +{
> +    if (strlen(name) != 12 ||
> +        sscanf(name, "%04x:%02x:%02x.%1x", &addr->domain,
> +               &addr->bus, &addr->slot, &addr->function) != 4) {
> +        return -EINVAL;
> +    }
> +
> +    return 0;
> +}
> +
> +static bool vfio_pci_host_match(PCIHostDeviceAddress *addr, const char *name)
> +{
> +    PCIHostDeviceAddress tmp;
> +
> +    if (vfio_pci_name_to_addr(name, &tmp)) {
> +        return false;
> +    }
> +
> +    return (tmp.domain == addr->domain && tmp.bus == addr->bus &&
> +            tmp.slot == addr->slot && tmp.function == addr->function);
> +}
> +
> +static bool vfio_pci_host_match_slot(PCIHostDeviceAddress *addr, const char *name)
> +{
> +    PCIHostDeviceAddress tmp;
> +
> +    if (vfio_pci_name_to_addr(name, &tmp)) {
> +        return false;
> +    }
> +
> +    return (tmp.domain == addr->domain && tmp.bus == addr->bus &&
> +            tmp.slot == addr->slot);
> +}
> +
>  /*
>   * return negative with errno, return 0 on success.
>   * if success, the point of ret_info fill with the affected device reset info.
> @@ -1877,6 +1912,199 @@ static int vfio_add_std_cap(VFIOPCIDevice *vdev, uint8_t pos)
>      return 0;
>  }
>  
> +static int vfio_device_range_limit(PCIBus *bus)
> +{
> +    PCIDevice *br;
> +
> +    br = pci_bridge_get_device(bus);
> +    if (!br ||
> +        !pci_is_express(br) ||
> +        !(br->exp.exp_cap) ||
> +        pcie_cap_is_arifwd_enabled(br)) {
> +        return 255;
> +    }

Perhaps something like vfio_pci_bus_devfn_limit() might be a better
name.  This probably also deserves a comment.  If I understand
correctly, you're trying to only return 8 for an express bus without
ARI enabled.  It seems like it would be easier to do that with the
function pci_bus_is_express().  I think that would reduce to:

if (!br || !pci_bus_is_express(bus) || pcie_cap_is_arifwd_enabled(br)) {
    return 255;
}
...

> +
> +    return 8;
> +}
> +
> +static void vfio_check_hot_bus_reset(VFIOPCIDevice *vdev, Error **errp)
> +{
> +    PCIBus *bus = vdev->pdev.bus;
> +    struct vfio_pci_hot_reset_info *info = NULL;
> +    struct vfio_pci_dependent_device *devices;
> +    VFIOGroup *group;
> +    int ret, i, devfn, range_limit;
> +
> +    ret = vfio_get_hot_reset_info(vdev, &info);
> +    if (ret) {
> +        error_setg(errp, "vfio: Cannot enable AER for device %s,"
> +                   " device does not support hot reset.",
> +                   vdev->vbasedev.name);
> +        return;
> +    }
> +
> +    /* List all affected devices by bus reset */
> +    devices = &info->devices[0];
> +
> +    /* Verify that we have all the groups required */
> +    for (i = 0; i < info->count; i++) {
> +        PCIHostDeviceAddress host;
> +        VFIOPCIDevice *tmp;
> +        VFIODevice *vbasedev_iter;
> +        bool found = false;
> +
> +        host.domain = devices[i].segment;
> +        host.bus = devices[i].bus;
> +        host.slot = PCI_SLOT(devices[i].devfn);
> +        host.function = PCI_FUNC(devices[i].devfn);
> +
> +        /* Skip the current device */
> +        if (vfio_pci_host_match(&host, vdev->vbasedev.name)) {
> +            continue;
> +        }
> +
> +        /* Ensure we own the group of the affected device */
> +        QLIST_FOREACH(group, &vfio_group_list, next) {
> +            if (group->groupid == devices[i].group_id) {
> +                break;
> +            }
> +        }
> +
> +        if (!group) {
> +            error_setg(errp, "vfio: Cannot enable AER for device %s, "
> +                       "depends on group %d which is not owned.",
> +                       vdev->vbasedev.name, devices[i].group_id);
> +            goto out;
> +        }
> +
> +        /* Ensure affected devices for reset on the same bus */
> +        QLIST_FOREACH(vbasedev_iter, &group->device_list, next) {
> +            if (vbasedev_iter->type != VFIO_DEVICE_TYPE_PCI) {
> +                continue;
> +            }
> +            tmp = container_of(vbasedev_iter, VFIOPCIDevice, vbasedev);
> +            if (vfio_pci_host_match(&host, tmp->vbasedev.name)) {
> +                /*
> +                 * AER errors may be broadcast to all functions of a multi-
> +                 * function endpoint.  If any of those sibling functions are
> +                 * also assigned, they need to have AER enabled or else an
> +                 * error may continue to cause a vm_stop condition.  IOW,
> +                 * AER setup of this function would be pointless.
> +                 */
> +                if (vfio_pci_host_match_slot(&host, vdev->vbasedev.name) &&
> +                    !(tmp->features & VFIO_FEATURE_ENABLE_AER)) {
> +                    error_setg(errp, "vfio: Cannot enable AER for device %s, on same slot"
> +                               " the dependent device %s which does not enable AER.",
> +                               vdev->vbasedev.name, tmp->vbasedev.name);
> +                    goto out;
> +                }
> +
> +                if (tmp->pdev.bus != bus) {
> +                    error_setg(errp, "vfio: Cannot enable AER for device %s, "
> +                               "the dependent device %s is not on the same bus",
> +                               vdev->vbasedev.name, tmp->vbasedev.name);
> +                    goto out;
> +                }
> +                found = true;
> +                break;
> +            }
> +        }
> +
> +        /* Ensure all affected devices assigned to VM */
> +        if (!found) {
> +            error_setg(errp, "vfio: Cannot enable AER for device %s, "
> +                       "the dependent device %04x:%02x:%02x.%x "
> +                       "is not assigned to VM.",
> +                       vdev->vbasedev.name, host.domain, host.bus,
> +                       host.slot, host.function);
> +            goto out;
> +        }
> +    }
> +
> +    /*
> +     * The above code verified that all devices affected by a bus reset
> +     * exist on the same bus in the VM.  To further simplify, we also
> +     * require that there are no additional devices beyond those existing on
> +     * the VM bus.
> +     */
> +    range_limit = vfio_device_range_limit(bus);
> +    for (devfn = 0; devfn < range_limit; devfn++) {
> +        VFIOPCIDevice *tmp;
> +        PCIDevice *dev;
> +        bool found = false;
> +
> +        dev = pci_find_device(bus, pci_bus_num(bus),
> +                  PCI_DEVFN(PCI_SLOT(vdev->pdev.devfn), devfn));

This is incorrect, we're enumerating through all devfns on the bus now,
we don't need PCI_DEVFN().

> +
> +        if (!dev) {
> +            continue;
> +        }
> +
> +        if (!object_dynamic_cast(OBJECT(dev), "vfio-pci")) {
> +            error_setg(errp, "vfio: Cannot enable AER for device %s, device"
> +                             " %s: slot %d function%d cannot be configured"

Can we say "VM address %02x.%d" rather than this slot/function
thing?

> +                             " on the same virtual bus",
> +                             vdev->vbasedev.name, dev->name,
> +                             PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn));
> +            goto out;
> +        }
> +
> +        tmp = DO_UPCAST(VFIOPCIDevice, pdev, dev);
> +        for (i = 0; i < info->count; i++) {
> +            PCIHostDeviceAddress host;
> +
> +            host.domain = devices[i].segment;
> +            host.bus = devices[i].bus;
> +            host.slot = PCI_SLOT(devices[i].devfn);
> +            host.function = PCI_FUNC(devices[i].devfn);
> +
> +            if (vfio_pci_host_match(&host, tmp->vbasedev.name)) {
> +                found = true;
> +                break;
> +            }
> +        }
> +
> +        if (!found) {
> +            error_setg(errp, "vfio: Cannot enable AER for device %s, affected"
> +                             " device %s does not be configured on the same"
> +                             " virtual bus",

And use the same sort of comment here as above:

"vfio: Cannot enable AER for device %s, vfio-pci device at VM address
%02x.%d cannot be configured on the same virtual bus"

> +                       vdev->vbasedev.name, tmp->vbasedev.name);
> +            goto out;
> +        }
> +    }
> +
> +out:
> +    g_free(info);
> +    return;
> +}
> +
> +static void vfio_aer_check_host_bus_reset(Error **errp)
> +{
> +    VFIOGroup *group;
> +    VFIODevice *vbasedev;
> +    VFIOPCIDevice *vdev;
> +    Error *local_err = NULL;
> +
> +    /* Check All vfio-pci devices if have bus reset capability */
> +    QLIST_FOREACH(group, &vfio_group_list, next) {
> +        QLIST_FOREACH(vbasedev, &group->device_list, next) {
> +            if (vbasedev->type != VFIO_DEVICE_TYPE_PCI) {
> +                continue;
> +            }
> +            vdev = container_of(vbasedev, VFIOPCIDevice, vbasedev);
> +            if (vdev->features & VFIO_FEATURE_ENABLE_AER) {
> +                vfio_check_hot_bus_reset(vdev, &local_err);
> +                if (local_err) {
> +                    error_propagate(errp, local_err);
> +                    return;
> +                }
> +            }
> +        }
> +    }
> +
> +    return;
> +}
> +
>  static int vfio_setup_aer(VFIOPCIDevice *vdev, uint8_t cap_ver,
>                            int pos, uint16_t size)
>  {
> @@ -2060,29 +2288,6 @@ static void vfio_pci_post_reset(VFIOPCIDevice *vdev)
>      vfio_intx_enable(vdev);
>  }
>  
> -static int vfio_pci_name_to_addr(const char *name, PCIHostDeviceAddress *addr)
> -{
> -    if (strlen(name) != 12 ||
> -        sscanf(name, "%04x:%02x:%02x.%1x", &addr->domain,
> -               &addr->bus, &addr->slot, &addr->function) != 4) {
> -        return -EINVAL;
> -    }
> -
> -    return 0;
> -}
> -
> -static bool vfio_pci_host_match(PCIHostDeviceAddress *addr, const char *name)
> -{
> -    PCIHostDeviceAddress tmp;
> -
> -    if (vfio_pci_name_to_addr(name, &tmp)) {
> -        return false;
> -    }
> -
> -    return (tmp.domain == addr->domain && tmp.bus == addr->bus &&
> -            tmp.slot == addr->slot && tmp.function == addr->function);
> -}
> -
>  static int vfio_pci_hot_reset(VFIOPCIDevice *vdev, bool single)
>  {
>      VFIOGroup *group;
> @@ -2589,6 +2794,22 @@ static void vfio_unregister_req_notifier(VFIOPCIDevice *vdev)
>      vdev->req_enabled = false;
>  }
>  
> +static void vfio_pci_machine_done_notify(Notifier *notifier, void *unused)
> +{
> +    Error *local_err = NULL;
> +
> +    vfio_aer_check_host_bus_reset(&local_err);
> +    if (local_err) {
> +        fprintf(stderr, "%s\n", error_get_pretty(local_err));
> +        error_free(local_err);
> +        exit(1);
> +    }
> +}
> +
> +static Notifier machine_notifier = {
> +    .notify = vfio_pci_machine_done_notify,
> +};
> +
>  static int vfio_initfn(PCIDevice *pdev)
>  {
>      VFIOPCIDevice *vdev = DO_UPCAST(VFIOPCIDevice, pdev, pdev);
> @@ -2934,6 +3155,15 @@ static const TypeInfo vfio_pci_dev_info = {
>  static void register_vfio_pci_dev_type(void)
>  {
>      type_register_static(&vfio_pci_dev_info);
> +
> +    /*
> +     * The AER configuration may depend on multiple devices, so we cannot
> +     * validate consistency after each device is initialized.  We can only
> +     * depend on function initialization order (function 0 last) for hotplug
> +     * devices, therefore a machine-init-done notifier is used to validate
> +     * the configuration after all cold-plug devices are processed.
> +     */
> +     qemu_add_machine_init_done_notifier(&machine_notifier);
>  }
>  
>  type_init(register_vfio_pci_dev_type)
> diff --git a/hw/vfio/pci.h b/hw/vfio/pci.h
> index 7b3924e..db7c6d5 100644
> --- a/hw/vfio/pci.h
> +++ b/hw/vfio/pci.h
> @@ -15,6 +15,7 @@
>  #include "qemu-common.h"
>  #include "exec/memory.h"
>  #include "hw/pci/pci.h"
> +#include "hw/pci/pci_bus.h"
>  #include "hw/pci/pci_bridge.h"
>  #include "hw/vfio/vfio-common.h"
>  #include "qemu/event_notifier.h"