Re: [Qemu-devel] [PATCH v3 2/2] remove function during multi-function hot-add

[Alex Williamson]

On Tue, 2015-10-13 at 16:41 +0800, Cao jin wrote:
> In case user regret when hot-adding multi-function, should roll back,
> device_del the function added but not exposed to the guest.

As Michael suggests, this patch should come first, before we actually
enable multi-function hot-add.

> Signed-off-by: Cao jin <address@hidden>
> ---
>  hw/pci/pci_host.c |  6 +++++-
>  hw/pci/pcie.c     | 22 +++++++++++++++++-----
>  2 files changed, 22 insertions(+), 6 deletions(-)
> 
> diff --git a/hw/pci/pci_host.c b/hw/pci/pci_host.c
> index 3e26f92..35e5cf3 100644
> --- a/hw/pci/pci_host.c
> +++ b/hw/pci/pci_host.c
> @@ -20,6 +20,7 @@
>  
>  #include "hw/pci/pci.h"
>  #include "hw/pci/pci_host.h"
> +#include "hw/pci/pci_bus.h"
>  #include "trace.h"
>  
>  /* debug PCI */
> @@ -88,10 +89,13 @@ void pci_data_write(PCIBus *s, uint32_t addr, uint32_t 
> val, int len)
>  uint32_t pci_data_read(PCIBus *s, uint32_t addr, int len)
>  {
>      PCIDevice *pci_dev = pci_dev_find_by_addr(s, addr);
> +    PCIDevice *f0 = NULL;
>      uint32_t config_addr = addr & (PCI_CONFIG_SPACE_SIZE - 1);
>      uint32_t val;
> +    uint8_t slot = (addr >> 11) & 0x1F;
>  
> -    if (!pci_dev) {
> +    f0 = s->devices[PCI_DEVFN(slot, 0)];
> +    if (!pci_dev || (!f0 && pci_dev)) {


This uses a lot more variables and operations than it needs to:

if (!pci_dev || !s->devices[PCI_DEVFN(PCI_SLOT(pci_dev->devfn), 0)]) {

Shouldn't we do the same on pci_data_write()?  A well behaved guest
won't blindly write to config space, but not all guests are well
behaved.

Comments in the code would be nice here to explain that non-zero
functions are only exposed when function zero is present, allowing
direct removal of unexposed devices.

I imagine that due to qemu locking that we don't have a race here, but
note that devices[] is populated early in the core pci realize function,
prior to the device initialize function, and there are any number of
reasons that failure could still occur, which would create a window
where the function is accessible.  I doubt this is an issue, but simply
note it for completeness.

>          return ~0x0;
>      }
>  
> diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> index 89bf61b..58d2153 100644
> --- a/hw/pci/pcie.c
> +++ b/hw/pci/pcie.c
> @@ -261,13 +261,30 @@ void pcie_cap_slot_hotplug_cb(HotplugHandler 
> *hotplug_dev, DeviceState *dev,
>      }
>  }
>  
> +static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque)
> +{
> +    object_unparent(OBJECT(dev));
> +}
> +
>  void pcie_cap_slot_hot_unplug_request_cb(HotplugHandler *hotplug_dev,
>                                           DeviceState *dev, Error **errp)
>  {
>      uint8_t *exp_cap;
> +    PCIDevice *pci_dev = PCI_DEVICE(dev);
> +    PCIBus *bus = pci_dev->bus;
>  
>      pcie_cap_slot_hotplug_common(PCI_DEVICE(hotplug_dev), dev, &exp_cap, 
> errp);
>  
> +    /* In case user regret when hot-adding multi function, remove the 
> function
> +     * that is unexposed to guest individually, without interaction with 
> guest.
> +     */
> +    if (PCI_FUNC(pci_dev->devfn) > 0 &&
> +            bus->devices[PCI_DEVFN(PCI_SLOT(pci_dev->devfn), 0)] == NULL) {

Similarly,

if (PCI_FUNC(pci_dev->devfn) && 
!bus->devices[PCI_DEVFN(PCI_SLOT(pci_dev->devfn), 0)]) {

> +        pcie_unplug_device(bus, pci_dev, NULL);
> +
> +        return;
> +    }
> +
>      pcie_cap_slot_push_attention_button(PCI_DEVICE(hotplug_dev));
>  }
>  
> @@ -378,11 +395,6 @@ void pcie_cap_slot_reset(PCIDevice *dev)
>      hotplug_event_update_event_status(dev);
>  }
>  
> -static void pcie_unplug_device(PCIBus *bus, PCIDevice *dev, void *opaque)
> -{
> -    object_unparent(OBJECT(dev));
> -}
> -
>  void pcie_cap_slot_write_config(PCIDevice *dev,
>                                  uint32_t addr, uint32_t val, int len)
>  {