Description of block&net_queue-handler.md

block&net_queue-handler.md

@@ -0,0 +1,84 @@
+# Description of queue-handler
+
+### Brief about Virtio:  
+It is virtualized driver for network and disk device. It improves the performance of guests by lowering guest I/O latency and
+raising throughput to levels that are almost comparable to those of bare metal in KVM.
+
+### About virtio block queue-handler:
+- The virtio-block device presents a block device to the virtual machine.
+- This object is used to filter the events with given condition of a block device, in case of exceptional events we need to remove the event from the QueueHandler.
+- It has two components:
+- inner - an Inorder Queue handler object which is used to handle requests.
+- ioeventfd - a file descriptor with an underlying count within the kernel.
+```rs
+pub(crate) struct QueueHandler<M: GuestAddressSpace> {
+    pub inner: InOrderQueueHandler<M, SingleFdSignalQueue>,
+    pub ioeventfd: EventFd,
+}
+```
+
+- In this process of its implementation, we match the condition of events and remove the unmatched ones from the EventOps of EventManager as follows:
+  - filtering the events:
+    1. events.event\_set() == EventSet::IN
+    2. events.data() == IOEVENT_DATA
+    3. self.ioeventfd.read().is\_err()
+    4. inorderhander process\_queue() is giving error
+  - Then, we remove the other event.  
+- In this init of its implementation, we initializing op as EventOps by adding event of a file descriptor, associated data and EVENTSET::IN 
+```rs
+impl<M: GuestAddressSpace> MutEventSubscriber for QueueHandler<M> {      
+    fn process(&mut self, events: Events, ops: &mut EventOps) {
+    ...
+    }
+    fn init(&mut self, ops: &mut EventOps) {
+    ...
+    }
+```
+### About virtio net queue-handler:
+- The virtio-net device use to access network adapters to the virtual machine.
+- ![image](https://user-images.githubusercontent.com/70102514/189861275-e2c79fbb-8bd9-4131-9961-6fc9a81c3dfd.png)
+
+- This object is used to filter the events with given condition of a network adapter, in case of exceptional events we need to remove the event from the QueueHandler.
+- It has three components:
+- inner - an Inorder Queue handler object which is used to handle requests.
+- rx_ioevent - a file descriptor with an underlying count within the kernel.
+- tx_ioevent - a file descriptor with an underlying count within the kernel.
+```rs
+pub struct QueueHandler<M: GuestAddressSpace> {
+    pub inner: SimpleHandler<M, SingleFdSignalQueue>,
+    pub rx_ioevent: EventFd,
+    pub tx_ioevent: EventFd,
+}
+```
+
+- In this handle_error of its implementation, we remove all events in ops of EventOps.  
+
+```rs
+impl<M: GuestAddressSpace> QueueHandler<M> {
+    fn handle_error<S: AsRef<str>>(&self, s: S, ops: &mut EventOps) {
+    ...
+    }
+}
+```
+- In this process of each of MutEventSubscriber of its implementation, we match the condition of events and remove the unmatched ones from the EventOps of EventManager as follows:
+  - filtering the events:
+    1. events.event\_set() == EventSet::IN
+    2. no error in inorder handler processing request respectively:
+      - process tap: inorder queue.process_tap()
+      - process rx: inorder queue.process_rxq()
+      - process tx: inorder queue.process_txq()
+    3. no error in respective event read:
+      - rx: rx_ioevent.read()
+      - tx: tx_ioevent.read()
+  - forward for exceptional events to call handle_error to remove them.
+- In this init of its implementation, we initializing op as EventOps by adding event of a file descriptor(tap of inorder handler, rx_ioevent, tx_ioevent), associated data (TAPFD_DATA, RX_IOEVENT_DATA, TX_IOEVENT_DATA) and EVENTSET::IN (in case of tap process EDGE_TRIGGERED is also included)
+```rs
+impl<M: GuestAddressSpace> MutEventSubscriber for QueueHandler<M> {  
+    fn process(&mut self, events: Events, ops: &mut EventOps) {
+    ...
+    }
+    fn init(&mut self, ops: &mut EventOps) {
+    ...
+    }
+```
+