Re: [dpdk-stable] [PATCH v3 2/3] vhost: convert buffer addresses to GPA for logging

[Adrian Moreno <amorenoz@redhat.com>]

On 10/8/19 10:01 AM, Tiwei Bie wrote:
> On Tue, Oct 08, 2019 at 09:43:48AM +0200, Maxime Coquelin wrote:
>> On 10/8/19 9:38 AM, Adrian Moreno wrote:
>>> On 10/8/19 8:30 AM, Tiwei Bie wrote:
>>>> On Fri, Sep 27, 2019 at 08:37:58PM +0200, Adrian Moreno wrote:
>>>>> Note that the API facing rte_vhost_log_write is not modified.
>>>>> So, make explicit that it expects the address in GPA space.
>>>>>
>>>>> Fixes: 69c90e98f483 ("vhost: enable IOMMU support")
>>>>> Cc: maxime.coquelin@redhat.com
>>>>> Cc: stable@dpdk.org
>>>>>
>>>>> Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
>>>>> ---
>>>>>  lib/librte_vhost/rte_vhost.h  |  2 +-
>>>>>  lib/librte_vhost/vdpa.c       |  4 +++-
>>>>>  lib/librte_vhost/vhost.c      | 21 +++++++++++++++++++++
>>>>>  lib/librte_vhost/vhost.h      | 16 ++++++++++++++++
>>>>>  lib/librte_vhost/virtio_net.c | 12 +++++++-----
>>>>>  5 files changed, 48 insertions(+), 7 deletions(-)
>>>>>
>>>>> diff --git a/lib/librte_vhost/rte_vhost.h b/lib/librte_vhost/rte_vhost.h
>>>>> index 7fb172912..16c7e4243 100644
>>>>> --- a/lib/librte_vhost/rte_vhost.h
>>>>> +++ b/lib/librte_vhost/rte_vhost.h
>>>>> @@ -265,7 +265,7 @@ rte_vhost_va_from_guest_pa(struct rte_vhost_memory *mem,
>>>>>   * @param vid
>>>>>   *  vhost device ID
>>>>>   * @param addr
>>>>> - *  the starting address for write
>>>>> + *  the starting address for write (in guest physical address space)
>>>>>   * @param len
>>>>>   *  the length to write
>>>>>   */
>>>>> diff --git a/lib/librte_vhost/vdpa.c b/lib/librte_vhost/vdpa.c
>>>>> index 8e45ce9f8..11fd5aede 100644
>>>>> --- a/lib/librte_vhost/vdpa.c
>>>>> +++ b/lib/librte_vhost/vdpa.c
>>>>> @@ -201,7 +201,9 @@ rte_vdpa_relay_vring_used(int vid, uint16_t qid, void *vring_m)
>>>>>  				goto fail;
>>>>>  			desc = desc_ring[desc_id];
>>>>>  			if (desc.flags & VRING_DESC_F_WRITE)
>>>>> -				vhost_log_write(dev, desc.addr, desc.len);
>>>>> +				vhost_log_cache_write_iova(dev, vq,
>>>>> +							   desc.addr,
>>>>> +							   desc.len);
>>>>
>>>> This will only log the dirty pages in cache.
>>> You're right. Thanks for spotting it. I will change it in the next version.
>>>>
>>>>
>>>>>  			desc_id = desc.next;
>>>>>  		} while (desc.flags & VRING_DESC_F_NEXT);
>>>>>  
>>>>> diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
>>>>> index e57dda22f..78457a77c 100644
>>>>> --- a/lib/librte_vhost/vhost.c
>>>>> +++ b/lib/librte_vhost/vhost.c
>>>>> @@ -200,6 +200,27 @@ __vhost_log_cache_write(struct virtio_net *dev, struct vhost_virtqueue *vq,
>>>>>  	}
>>>>>  }
>>>>>  
>>>>> +void
>>>>> +__vhost_log_cache_write_iova(struct virtio_net *dev, struct vhost_virtqueue *vq,
>>>>> +			     uint64_t iova, uint64_t len)
>>>>> +{
>>>>> +	uint64_t hva, gpa, map_len;
>>>>> +	map_len = len;
>>>>> +
>>>>> +	hva = __vhost_iova_to_vva(dev, vq, iova, &map_len, VHOST_ACCESS_RW);
>>>>> +	if (map_len != len) {
>>>>> +		RTE_LOG(ERR, VHOST_CONFIG,
>>>>> +			"Failed to write log for IOVA 0x%0lx."
>>>>
>>>> PRIx64 should be used here, otherwise the build will be
>>>> broken on i686.
>>>>
>>>> It's better to add a space after '.'
>>> Thank you. Will do.
>>>>
>>>>
>>>>> +			"No IOTLB entry found\n",
>>
>> I personally prefer not to split error messages even it is is slightly
>> above 80 chars. It makes grepping for error messages easier.
> 
> +1
> 
Makes perfect sense. Addressed in the new version.
Thanks!

> 
>>
>>>>> +			iova);
>>>>> +		return;
>>>>> +	}
>>>>> +
>>>>> +	gpa = hva_to_gpa(dev, hva, len);
>>>>> +	if (gpa)
>>>>> +		__vhost_log_cache_write(dev, vq, gpa, len);
>>>>> +}
>>>>> +
>>>

-Adrián