net: revert virtio-net queue size to 256

Changing the queue size from 256 to 512 increases RX throughput but
seems to regress TX throughput by 3% on average. We are not sure what is
causing this. Revert this change until we have enough time to
investigate and understand better the mechanisms that are in play.

Signed-off-by: Babis Chalios <[email protected]>

CHANGELOG.md

@@ -33,10 +33,6 @@ and this project adheres to
 
 ### Changed
 
-- [#4875](https://github.com/firecracker-microvm/firecracker/pull/4875):
-  Increase default queue size for the `virtio-net` device from 256 to 512. This
-  decreases wait time between guest and vmm threads for network packets
-  processing and allows for more throughput.
 - [#4844](https://github.com/firecracker-microvm/firecracker/pull/4844): Upgrade
   `virtio-net` device to use `readv` syscall to avoid unnecessary memory copies
   on RX path, increasing the RX performance.

src/vmm/src/devices/virtio/net/mod.rs

@@ -6,7 +6,7 @@
 use std::io;
 
 /// Maximum size of the queue for network device.
-pub const NET_QUEUE_MAX_SIZE: u16 = 512;
+pub const NET_QUEUE_MAX_SIZE: u16 = 256;
 /// Maximum size of the frame buffers handled by this device.
 pub const MAX_BUFFER_SIZE: usize = 65562;
 /// The number of queues of the network device.