On Mon, Jul 29, 2019 at 06:41:27PM +0200, Stefano Garzarella wrote:

On Mon, Jul 29, 2019 at 12:01:37PM -0400, Michael S. Tsirkin wrote:

quoted

On Mon, Jul 29, 2019 at 05:36:56PM +0200, Stefano Garzarella wrote:

quoted

On Mon, Jul 29, 2019 at 10:04:29AM -0400, Michael S. Tsirkin wrote:

quoted

On Wed, Jul 17, 2019 at 01:30:26PM +0200, Stefano Garzarella wrote:

quoted

Since virtio-vsock was introduced, the buffers filled by the host
and pushed to the guest using the vring, are directly queued in
a per-socket list. These buffers are preallocated by the guest
with a fixed size (4 KB).

The maximum amount of memory used by each socket should be
controlled by the credit mechanism.
The default credit available per-socket is 256 KB, but if we use
only 1 byte per packet, the guest can queue up to 262144 of 4 KB
buffers, using up to 1 GB of memory per-socket. In addition, the
guest will continue to fill the vring with new 4 KB free buffers
to avoid starvation of other sockets.

This patch mitigates this issue copying the payload of small
packets (< 128 bytes) into the buffer of last packet queued, in
order to avoid wasting memory.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>

This is good enough for net-next, but for net I think we
should figure out how to address the issue completely.
Can we make the accounting precise? What happens to
performance if we do?

In order to do more precise accounting maybe we can use the buffer size,
instead of payload size when we update the credit available.
In this way, the credit available for each socket will reflect the memory
actually used.

I should check better, because I'm not sure what happen if the peer sees
1KB of space available, then it sends 1KB of payload (using a 4KB
buffer).
The other option is to copy each packet in a new buffer like I did in
the v2 [2], but this forces us to make a copy for each packet that does
not fill the entire buffer, perhaps too expensive.

[2] https://patchwork.kernel.org/patch/10938741/

So one thing we can easily do is to under-report the
available credit. E.g. if we copy up to 256bytes,
then report just 256bytes for every buffer in the queue.

Ehm sorry, I got lost :(
Can you explain better?


Thanks,
Stefano

I think I suggested a better idea more recently.
But to clarify this option: we are adding a 4K buffer.
Let's say we know we will always copy 128 bytes.

So we just tell remote we have 128.
If we add another 4K buffer we add another 128 credits.

So we are charging local socket 16x more (4k for a 128 byte packet) but
we are paying remote 16x less (128 credits for 4k byte buffer). It evens
out.

Way less credits to go around so I'm not sure it's a good idea,
at least as the only solution. Can be combined with other
optimizations and probably in a less drastic fashion
(e.g. 2x rather than 16x).


-- 
MST