RE: [PATCH v5 05/20] vfio/iommufd: Add support for iova_ranges and pgsizes

[Duan, Zhenzhong]

Hi Cédric,

>-----Original Message-----
>From: Cédric Le Goater <clg@redhat.com>
>Sent: Friday, November 10, 2023 5:36 PM
>Subject: Re: [PATCH v5 05/20] vfio/iommufd: Add support for iova_ranges and
>pgsizes
>
>On 11/9/23 12:45, Zhenzhong Duan wrote:
>> Some vIOMMU such as virtio-iommu use iova ranges from host side to
>> setup reserved ranges for passthrough device, so that guest will not
>> use an iova range beyond host support.
>>
>> Use an uAPI of IOMMUFD to get iova ranges of host side and pass to
>> vIOMMU just like the legacy backend.
>>
>> Also use out_iova_alignment returned from uAPI as pgsizes instead of
>> qemu_real_host_page_size() as a fallback.
>>
>> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
>> ---
>> v5: Add missed pgsizes initialization in vfio_get_info_iova_range
>>
>>   hw/vfio/iommufd.c | 48
>+++++++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 48 insertions(+)
>>
>> diff --git a/hw/vfio/iommufd.c b/hw/vfio/iommufd.c
>> index ea4e23f4ec..958c3e794f 100644
>> --- a/hw/vfio/iommufd.c
>> +++ b/hw/vfio/iommufd.c
>> @@ -267,6 +267,53 @@ static int iommufd_ram_block_discard_disable(bool
>state)
>>       return ram_block_uncoordinated_discard_disable(state);
>>   }
>>
>> +static int vfio_get_info_iova_range(VFIOIOMMUFDContainer *container,
>> +                                    uint32_t ioas_id)
>> +{
>> +    VFIOContainerBase *bcontainer = &container->bcontainer;
>> +    struct iommu_ioas_iova_ranges *info;
>> +    struct iommu_iova_range *iova_ranges;
>> +    int ret, sz, fd = container->be->fd;
>> +
>> +    info = g_malloc0(sizeof(*info));
>> +    info->size = sizeof(*info);
>> +    info->ioas_id = ioas_id;
>> +
>> +    ret = ioctl(fd, IOMMU_IOAS_IOVA_RANGES, info);
>> +    if (ret && errno != EMSGSIZE) {
>> +        goto error;
>> +    }
>> +
>> +    sz = info->num_iovas * sizeof(struct iommu_iova_range);
>> +    info = g_realloc(info, sizeof(*info) + sz);
>> +    info->allowed_iovas = (uintptr_t)(info + 1);
>> +
>> +    ret = ioctl(fd, IOMMU_IOAS_IOVA_RANGES, info);
>> +    if (ret) {
>> +        goto error;
>> +    }
>> +
>> +    iova_ranges = (struct iommu_iova_range *)(uintptr_t)info->allowed_iovas;
>> +
>> +    for (int i = 0; i < info->num_iovas; i++) {
>> +        Range *range = g_new(Range, 1);
>> +
>> +        range_set_bounds(range, iova_ranges[i].start, iova_ranges[i].last);
>> +        bcontainer->iova_ranges =
>> +            range_list_insert(bcontainer->iova_ranges, range);
>> +    }
>> +    bcontainer->pgsizes = info->out_iova_alignment;
>> +
>> +    g_free(info);
>> +    return 0;
>> +
>> +error:
>> +    ret = -errno;
>> +    g_free(info);
>> +    error_report("vfio/iommufd: Cannot get iova ranges: %m");
>
>Can we propagate the error ?

Do you mean propagating the error to call site and call error_report?
In fact, getting iova range from host is a better to have, not a must.
If fails we fallback to 64bit range.

Thanks
Zhenzhong