Re: [PATCH v3 4/5] s390x/pci: Add routine to get the vfio dma available count

[Cornelia Huck]

On Wed, 16 Sep 2020 08:55:00 -0400
Matthew Rosato <mjrosato@linux.ibm.com> wrote:

> On 9/16/20 6:27 AM, Cornelia Huck wrote:
> > On Wed, 16 Sep 2020 09:21:39 +0200
> > Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
> >   
> >> On 9/15/20 9:14 PM, Matthew Rosato wrote:  
> >>> Create new files for separating out vfio-specific work for s390
> >>> pci. Add the first such routine, which issues VFIO_IOMMU_GET_INFO
> >>> ioctl to collect the current dma available count.
> >>>
> >>> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> >>> ---
> >>>   hw/s390x/meson.build     |  1 +
> >>>   hw/s390x/s390-pci-vfio.c | 54 
> >>> ++++++++++++++++++++++++++++++++++++++++++++++++
> >>>   hw/s390x/s390-pci-vfio.h | 17 +++++++++++++++
> >>>   3 files changed, 72 insertions(+)
> >>>   create mode 100644 hw/s390x/s390-pci-vfio.c
> >>>   create mode 100644 hw/s390x/s390-pci-vfio.h
> >>>  
> > 
> > (...)
> >   
> >>> diff --git a/hw/s390x/s390-pci-vfio.c b/hw/s390x/s390-pci-vfio.c
> >>> new file mode 100644
> >>> index 0000000..75e3ac1
> >>> --- /dev/null
> >>> +++ b/hw/s390x/s390-pci-vfio.c
> >>> @@ -0,0 +1,54 @@
> >>> +/*
> >>> + * s390 vfio-pci interfaces
> >>> + *
> >>> + * Copyright 2020 IBM Corp.
> >>> + * Author(s): Matthew Rosato <mjrosato@linux.ibm.com>
> >>> + *
> >>> + * This work is licensed under the terms of the GNU GPL, version 2 or (at
> >>> + * your option) any later version. See the COPYING file in the top-level
> >>> + * directory.
> >>> + */
> >>> +
> >>> +#include <sys/ioctl.h>
> >>> +
> >>> +#include "qemu/osdep.h"
> >>> +#include "s390-pci-vfio.h"
> >>> +#include "hw/vfio/vfio-common.h"
> >>> +
> >>> +/*
> >>> + * Get the current DMA available count from vfio.  Returns true if vfio 
> >>> is
> >>> + * limiting DMA requests, false otherwise.  The current available count 
> >>> read
> >>> + * from vfio is returned in avail.
> >>> + */
> >>> +bool s390_pci_update_dma_avail(int fd, unsigned int *avail)
> >>> +{
> >>> +    g_autofree struct vfio_iommu_type1_info *info;
> >>> +    uint32_t argsz;
> >>> +    int ret;
> >>> +
> >>> +    assert(avail);
> >>> +
> >>> +    argsz = sizeof(struct vfio_iommu_type1_info);
> >>> +    info = g_malloc0(argsz);
> >>> +    info->argsz = argsz;
> >>> +    /*
> >>> +     * If the specified argsz is not large enough to contain all
> >>> +     * capabilities it will be updated upon return.  In this case
> >>> +     * use the updated value to get the entire capability chain.
> >>> +     */
> >>> +    ret = ioctl(fd, VFIO_IOMMU_GET_INFO, info);
> >>> +    if (argsz != info->argsz) {
> >>> +        argsz = info->argsz;
> >>> +        info = g_realloc(info, argsz);  
> >>
> >> Do we need to bzero [sizeof(struct vfio_iommu_type1_info)..argsz[?  
> > 
> > If we do, I think we need to do the equivalent in
> > vfio_get_region_info() as well?
> >   
> 
> I agree that it would need to be in both places or neither -- I would 
> expect the re-driven ioctl to overwrite the prior contents of info 
> (unless we get a bad ret, but in this case we don't care what is in info)?
> 
> Perhaps the fundamental difference between this code and 
> vfio_get_region_info is that the latter checks for only a growing argsz 
> and retries, whereas this code checks for != so it's technically 
> possible for a smaller argsz to trigger the retry here, and we wouldn't 
> know for sure that all bytes from the first ioctl call were overwritten.

Nod. Relying on overwriting should be fine.

> 
> What if I adjust this code to look like vfio_get_region_info:
> 
> retry:
>       info->argsz = argsz;
> 
>       if (ioctl(fd, VFIO_IOMMU_GET_INFO, info)) {
>               // no need to g_free() bc of g_autofree
>               return false;   
>       }
> 
>       if (info->argsz > argsz) {
>               argsz = info->argsz;
>               info = g_realloc(info, argsz);
>               goto retry;
>       }
> 
>       /* If the capability exists, update with the current value */
>       return vfio_get_info_dma_avail(info, avail);
> 
> Now we would only trigger when we are told by the host that the buffer 
> must be larger.

I think that makes sense.

> 
> > (Also, shouldn't we check ret before looking at info->argsz?)
> >   
> 
> Yes, you are correct.  The above proposal would fix that issue too.
> 
> >>  
> >>> +        info->argsz = argsz;
> >>> +        ret = ioctl(fd, VFIO_IOMMU_GET_INFO, info);
> >>> +    }
> >>> +
> >>> +    if (ret) {
> >>> +        return false;
> >>> +    }
> >>> +
> >>> +    /* If the capability exists, update with the current value */
> >>> +    return vfio_get_info_dma_avail(info, avail);
> >>> +}
> >>> +  
> > 
> > (...)
> > 
> >   
>