Re: [Qemu-devel] [RFC PATCH 2/5] vfio/quirks: Add generic support for ioveventfds

[Auger Eric <eric.auger@redhat.com>]

Hi Alex,

On 08/02/18 19:33, Alex Williamson wrote:
> On Thu, 8 Feb 2018 12:11:57 +0100
> Auger Eric <eric.auger@redhat.com> wrote:
> 
>> Hi Alex,
>>
>> On 07/02/18 01:26, Alex Williamson wrote:
>>> We might wish to handle some quirks via ioeventfds, add a list of
>>> ioeventfds to the quirk.  
>> The commit title is a bit misleading as we only add the data type and
>> deletion function.
> 
> Unfortunately that's as much as I can add here or else QEMU will fail
> to build because of unused functions :-\  So I pulled as much
> infrastructure as I could here but had to leave some meat for when it
> gets called.  Suggestions welcome for title/commit updates.

Understood. Something like "vfio/quirks: Add a VFIOIOEventFD list in
quirks"?
> 
>>>
>>> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
>>> ---
>>>  hw/vfio/pci-quirks.c |   17 +++++++++++++++++
>>>  hw/vfio/pci.h        |   11 +++++++++++
>>>  2 files changed, 28 insertions(+)
>>>
>>> diff --git a/hw/vfio/pci-quirks.c b/hw/vfio/pci-quirks.c
>>> index 10af23217292..e4cf4ea2dd9c 100644
>>> --- a/hw/vfio/pci-quirks.c
>>> +++ b/hw/vfio/pci-quirks.c
>>> @@ -12,6 +12,7 @@
>>>  
>>>  #include "qemu/osdep.h"
>>>  #include "qemu/error-report.h"
>>> +#include "qemu/main-loop.h"
>>>  #include "qemu/range.h"
>>>  #include "qapi/error.h"
>>>  #include "qapi/visitor.h"
>>> @@ -278,12 +279,24 @@ static const MemoryRegionOps vfio_ati_3c3_quirk = {
>>>  static VFIOQuirk *vfio_quirk_alloc(int nr_mem)
>>>  {
>>>      VFIOQuirk *quirk = g_malloc0(sizeof(*quirk));
>>> +    QLIST_INIT(&quirk->ioeventfds);
>>>      quirk->mem = g_new0(MemoryRegion, nr_mem);
>>>      quirk->nr_mem = nr_mem;
>>>  
>>>      return quirk;
>>>  }
>>>  
>>> +static void vfio_ioeventfd_exit(VFIOIOEventFD *ioeventfd)
>>> +{
>>> +    QLIST_REMOVE(ioeventfd, next);
>>> +    memory_region_del_eventfd(ioeventfd->mr, ioeventfd->addr, ioeventfd->size,
>>> +                              ioeventfd->match_data, ioeventfd->data,
>>> +                              &ioeventfd->e);
>>> +    qemu_set_fd_handler(event_notifier_get_fd(&ioeventfd->e), NULL, NULL, NULL);
>>> +    event_notifier_cleanup(&ioeventfd->e);
>>> +    g_free(ioeventfd);
>>> +}
>>> +
>>>  static void vfio_vga_probe_ati_3c3_quirk(VFIOPCIDevice *vdev)
>>>  {
>>>      VFIOQuirk *quirk;
>>> @@ -1668,6 +1681,10 @@ void vfio_bar_quirk_exit(VFIOPCIDevice *vdev, int nr)
>>>      int i;
>>>  
>>>      QLIST_FOREACH(quirk, &bar->quirks, next) {
>>> +        while (!QLIST_EMPTY(&quirk->ioeventfds)) {
>>> +            vfio_ioeventfd_exit(QLIST_FIRST(&quirk->ioeventfds));
>>> +        }
>>> +
>>>          for (i = 0; i < quirk->nr_mem; i++) {
>>>              memory_region_del_subregion(bar->region.mem, &quirk->mem[i]);
>>>          }
>>> diff --git a/hw/vfio/pci.h b/hw/vfio/pci.h
>>> index f4aa13e021fa..146065c2f715 100644
>>> --- a/hw/vfio/pci.h
>>> +++ b/hw/vfio/pci.h
>>> @@ -24,9 +24,20 @@
>>>  
>>>  struct VFIOPCIDevice;
>>>  
>>> +typedef struct VFIOIOEventFD {
>>> +    QLIST_ENTRY(VFIOIOEventFD) next;
>>> +    MemoryRegion *mr;
>>> +    hwaddr addr;
>>> +    unsigned size;
>>> +    bool match_data;  
>> Shouldn't you add the match_data field also in the kernel uapi?
> 
> In the vfio uapi?  vfio isn't matching anything and has no ability to
> match anything, it's just triggering a pre-programmed write via
> eventfd.  Maybe I'm not understanding the comment.  Thanks,

Yes that not sensible. We have a KVM ioeventfd that triggers an fd only
if the address/data matches and this VFIO "ioeventfd" listens to this fd
and writes the fixed address/data. At the beginning I was thinking about
having the VFIO handler taking in charge any value written at this
address but there is no way to retrieve the value besides doing a fixed
assignment from userspace. Sorry for the noise.

Thanks

Eric
> 
> Alex
>