The WPP Event Tracing can be disabled which means return to the good old DbgPrintEx routine. To disable WPP, the following steps need to be done:
- commenting the line `#define EVENT_TRACING` in project's trace.h file,
- setting the "Run Wpp Tracing" option in project settings to No.
Signed-off-by: Martin Drab <[email protected]>
The leverages WSK to create a set of client and server threads. The server thread listen for connection and exchange a message with connecting clients. Cryptographic hash function (SHA256) is used to verify that message data are not damaged during their transmission.
Server threads are working in non-blocking mode. Such behavior is necessary in order to be able to cancel pending WskAccept calls since the viosock.sys driver does not permit their cancellation (unlike send/recv requests which ARE cancellable by design).
Signed-off-by: Martin Drab <[email protected]>
This call translates to an IOCTL_SOCKET_BIND request to the VIOSOCK driver, followed with IOCTL_SOCKET_LISTEN one if the socket in question is a listening one. This needs to be done because the WskListen function was not part of the WSK interface until Windows 10 version 1703.
Contrary to the TCP/IP WSK implementation in Windows, it is not necessary to call WskBind for non-listening sockets.
ref: https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/wsk/nc-wsk-pfn_wsk_bind
Signed-off-by: Martin Drab <[email protected]>
If called at IRQL = PASSIVE_LEVEL, this call blocks until a connection is accepted. This is due to the fact that the VIOSOCK driver implements the accept() operation through an IRP_MJ_CREATE (ZwCreateFile) which is usually synchronous. To make WskAccept non-blocking, the client driver can call it at higher IRQL, the accept() call into the VIOSOCK driver is then deferred to a workitem.
ref: https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/wsk/nc-wsk-pfn_wsk_accept
Signed-off-by: Martin Drab <[email protected]>
Routines for building IRP_MJ_WRITE and IRP_MJ_WRITE IRPs are also par of this commit since the VIOSOCK driver implements send/recv operations through these requests. It is also necessary to deal with the fact that the WSK interface transfers send/recv buffers through WSK_BUF structures, very similar to WSABUF in usermode, that support MDL chaining. Since the VIOSOCK driver does not support MDL chaining, the WskSend initiates a sequence of IRP_MJ_WRITE requests, one for each MDL in the chain.
ref: [WskSend](https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/wsk/nc-wsk-pfn_wsk_send_to)
Signed-off-by: Martin Drab <[email protected]>
Most of the work done by this WSK implementation consists of packing client driver's calls into IOCTL and read/write request and passing them to the VIOSOCK driver. This commit introduces routines for creating IOCTL requests and sending them to VIOSOCK. Also, a concept of WSK Completion Contexts is introduces. The purpose of these contexts is to transfer information between code sending IRPs and their completion routines. Also, these contexts can be used to implement one WSK interface call as a sequence of requests to the VIOSOCK driver.
Signed-off-by: Martin Drab <[email protected]>
Unlike the Windows WSK implementation, this library needs to be initialized through VioWskModuleInit before calling its WsRegister equivalent. This approach allows the client driver to provide the library with an optional device object that is then used to prevent IRP completion routines and workitems used by the library to execute after the device deletion. If the client application decides not to provide such a device object, the library can live with it but no unload protection is then enforced.
Also, the client driver must provide pointer to its DRIVER_OBJECT which is used during WskCaptureProviderNPI to connect to the VIOSOCK driver. Instead of having the VIOSOCK device name hard-coded, the library takes advantage of IoRegisterPlugPlayNotification to search for the GUID_DEVINTERFACE_VIOSOCK device interface (which is enabled by the VIOSOCK driver) dynamically.
Signed-off-by: Martin Drab <[email protected]>
These macros will be used mostly when entering into or exiting from a function, sending routine name nad values of its arguments to a potential kernel debugger. These macros are NOPs on Release build, however, their usage can be invaluable when debugging driver crashes.
Signed-off-by: Martin Drab <[email protected]>
- WSK project added to the main solution,
- removed Debug configurations (to be in sync with other projects)
Signed-off-by: Martin Drab <[email protected]>
When an other end (peer) of the socket reports a shutdown, the recv request queue is not processed. This means that some recv calls may get stuck forever. This commit changes the behavior -- when a shutdown from its peer is received, the recv queue gets processed and all requests that would get stuck are completed with STATUS_SUCCESS (which usually ends in receiving zero bytes).
Signed-off-by: Martin Drab <[email protected]>
The Socket driver stores socket-related data in memory associated with file objects representing the sockets. It uses WdfObjectReference/WdfObjectDereference to track ownership and protect sockets being in use from deletion. However, these routines do not affect reference count of the underlying FILE_OBJECT structures (served by Object Manager) -- they work only with the reference count of the WDF object wrapped around the FILE_OBJECT structure. Thus, when the underlying FILE_OBJECT is deleted, WdfObjectReference does not protect the socket context from deallocation.
This commit fixes this issue by introducing VioSockReference and VioSockDereference which also work with reference count of the wrapped FILE_OBJECT structures. Also, VioSockClose is replaced with VioSockCleanup -- the latter one is invoked when the last handle for a file object (FILE_OBJECT, not WDFFILEOBJECT) gets closed (the former one is invoked when the last reference is gone -- which does not happen if VioSock(De)reference are used). For usermode, this makes no difference since user applications work only with handles, not pointer references to file objects.
Signed-off-by:¨Martin Drab <[email protected]>
Split tx,rx and rss statistics
Add property for [initial] configuration data
Make reset of statistics selective and explicit
Add metrics for:
- minimal number of rx buffers
- indications with 'resource' flag
- amount of allocated shared memory
Signed-off-by: Yuri Benditovich <[email protected]>
Implemented a workaround for rx packet loss encountered with TCP_NODELAY. By default,
this feature is turned off to minimize the impact on CPU performance and network bandwidth.
To activate the feature, go to the network card properties and set the MinRxBufferPercent
to a value above 0. A good starting point is to set MinRxBufferPercent to 25% and then monitor
for any packet loss. If packet loss continues, incrementally adjust the value to 50%, then 75%,
and if necessary, up to the maximum of 100%.
In extreme cases, it may be required to not only adjust the MinRxBufferPercent but also to increase
the number of queues and the size of each queue (queue_size).
It is important to note that this feature is only applicable to rx packet loss.
For more details, see [this thread](https://github.com/virtio-win/kvm-guest-drivers-windows/issues/1012).
Signed-off-by: junjiehua <[email protected]>
Removal of redundant/unneded lines from the .vcxproj
They also caused by IDE warning on loading of any project
including netkvmco.
Signed-off-by: Yuri Benditovich <[email protected]>
https://issues.redhat.com/browse/RHEL-19430
The number of buffers is always limited by queue size,
so with the default QEMU setting (256) the driver will do
the same. If QEMU is configured in command-line to have
bigger queue (it allows up to 1024), the previous driver
required to change also driver property to really use
full size of QEMU RX queue. Now this will not be required
(up to 1024 buffers).
Signed-off-by: Yuri Benditovich <[email protected]>
This commit updates the CNB::BindToDescriptor method signature, changing
the return type from a boolean to the newly introduced NBMappingStatus
enum. The updated method now returns one of the following statuses:
NBMappingStatus::SUCCESS
NBMappingStatus::NO_RESOURCE
NBMappingStatus::FAILURE
This change allows for more detailed reporting of the binding operation
outcome, enabling the caller to distinguish between different failure
reasons beyond the previous binary success/failure indication. The
NBMappingStatus enum provides a clearer, more expressive way to handle
the results of the binding process.
Signed-off-by: Zhang JianMing <[email protected]>
This commit introduces the CopyFromMdlChain function to the CNB class. The function is responsible for copying data from a Memory Descriptor List (MDL) chain to a specified destination buffer. A special focus has been given to ensuring the function handles the traversal of the MDL chain correctly, updating the Source MDL pointer and the Offset within the current MDL as it progresses through the chain.
Signed-off-by: Zhang JianMing <[email protected]>
This commit introduces the MAX_FRAGMENTS_IN_ONE_NB macro, which
defines the maximum number of fragments in a single NET_BUFFER. It
is used to constrain the SGTableCapacity value, ensuring it does
not exceed the meaningful limit set by MAX_FRAGMENTS_IN_ONE_NB.
Signed-off-by: Zhang JianMing <[email protected]>
This commit introduces extra pages within the TXManager module,
which are intended to support the copy mechanism that will be
implemented in subsequent updates. The addition of these pages lays
the groundwork for the new functionality and ensures that the
TXManager is prepared for the enhancements.
Signed-off-by: Zhang JianMing <[email protected]>
https://issues.redhat.com/browse/RHEL-20575
Currently the driver get RX descriptor from the rx virtQ,
analyses it (L2/L3), calculates hash and delivers to the
proper RSS queue. Later it extract the packet from the RSS
queue and checks whether it passes RX filters.
Now we move the filtering to the earliest stage - just after
basic analysis.
Note that in case of vhost there are no receive filters
configured by the host and typically the host does not work
in promiscuous mode, so it is redundant to calculate the hash
and trigger the RSS queue for packets that should be dropped.
Signed-off-by: Yuri Benditovich <[email protected]>
Kostiantyn Kostiuk