On 01/06/2017 15:42, Peter Lieven wrote:
>>>
>>>
>>> Any idea whats the cause of this?
>> It may depend on memory fragmentation.
>
> This sounds like nothing that can be influenced or changed?
>
> I forgot to mention, if I specify bs=512k the requests go through also
> with virtio-scsi
Can you repeat it many times, perhaps with different amount
of memory in the VMs, and trace the number of buffers. I'll
send this patch shortly.
diff --git a/hw/virtio/trace-events b/hw/virtio/trace-events
index 1f7a7c1ae1..e24d8fa997 100644
--- a/hw/virtio/trace-events
+++ b/hw/virtio/trace-events
@@ -1,6 +1,7 @@
# See docs/tracing.txt for syntax documentation.
# hw/virtio/virtio.c
+virtqueue_alloc_element(void *elem, size_t sz, unsigned in_num, unsigned
out_num) "elem %p size %zd in_num %u out_num %u"
virtqueue_fill(void *vq, const void *elem, unsigned int len, unsigned int idx)
"vq %p elem %p len %u idx %u"
virtqueue_flush(void *vq, unsigned int count) "vq %p count %u"
virtqueue_pop(void *vq, void *elem, unsigned int in_num, unsigned int out_num)
"vq %p elem %p in_num %u out_num %u"
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 890b4d7eb7..286446b261 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -816,6 +816,7 @@ static void *virtqueue_alloc_element(size_t sz, unsigned
out_num, unsigned in_nu
assert(sz >= sizeof(VirtQueueElement));
elem = g_malloc(out_sg_end);
+ trace_virtqueue_alloc_element(elem, sz, in_num, out_num);
elem->out_num = out_num;
elem->in_num = in_num;
elem->in_addr = (void *)elem + in_addr_ofs;
Thanks,
Paolo
