OvmfPkg/VirtioNetDxe/TechNotes.txt

   1 ## @file
   2 #
   3 # Technical notes for the virtio-net driver.
   4 #
   5 # Copyright (C) 2013, Red Hat, Inc.
   6 #
   7 # This program and the accompanying materials are licensed and made available
   8 # under the terms and conditions of the BSD License which accompanies this
   9 # distribution. The full text of the license may be found at
  10 # http://opensource.org/licenses/bsd-license.php
  11 #
  12 # THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, WITHOUT
  13 # WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
  14 #
  15 ##
  16
  17 Disclaimer
  18 ----------
  19
  20 All statements concerning standards and specifications are informative and not
  21 normative. They are made in good faith. Corrections are most welcome on the
  22 edk2-devel mailing list.
  23
  24 The following documents have been perused while writing the driver and this
  25 document:
  26 - Unified Extensible Firmware Interface Specification, Version 2.3.1, Errata C;
  27   June 27, 2012
  28 - Driver Writer's Guide for UEFI 2.3.1, 03/08/2012, Version 1.01;
  29 - Virtio PCI Card Specification, v0.9.5 DRAFT, 2012 May 7.
  30
  31
  32 Summary
  33 -------
  34
  35 The VirtioNetDxe UEFI_DRIVER implements the Simple Network Protocol for
  36 virtio-net devices. Higher level protocols are automatically installed on top
  37 of it by the DXE Core / the ConnectController() boot service, enabling for
  38 virtio-net devices eg. DHCP configuration, TCP transfers with edk2 StdLib
  39 applications, and PXE booting in OVMF.
  40
  41
  42 UEFI driver structure
  43 ---------------------
  44
  45 A driver instance, belonging to a given virtio-net device, can be in one of
  46 four states at any time. The states stack up as follows below. The state
  47 transitions are labeled with the primary function (and its important callees
  48 faithfully indented) that implement the transition.
  49
  50                                |  ^
  51                                |  |
  52    [DriverBinding.c]           |  | [DriverBinding.c]
  53    VirtioNetDriverBindingStart |  | VirtioNetDriverBindingStop
  54      VirtioNetSnpPopulate      |  |   VirtioNetSnpEvacuate
  55        VirtioNetGetFeatures    |  |
  56                                v  |
  57                    +-------------------------+
  58                    | EfiSimpleNetworkStopped |
  59                    +-------------------------+
  60                                |  ^
  61                 [SnpStart.c]   |  | [SnpStop.c]
  62                 VirtioNetStart |  | VirtioNetStop
  63                                |  |
  64                                v  |
  65                    +-------------------------+
  66                    | EfiSimpleNetworkStarted |
  67                    +-------------------------+
  68                                |  ^
  69   [SnpInitialize.c]            |  | [SnpShutdown.c]
  70   VirtioNetInitialize          |  | VirtioNetShutdown
  71     VirtioNetInitRing {Rx, Tx} |  |   VirtioNetShutdownRx [SnpSharedHelpers.c]
  72       VirtioRingInit           |  |   VirtioNetShutdownTx [SnpSharedHelpers.c]
  73       VirtioRingMap            |  |   VirtioNetUninitRing [SnpSharedHelpers.c]
  74     VirtioNetInitTx            |  |                       {Tx, Rx}
  75     VirtioNetInitRx            |  |     VirtIo->UnmapSharedBuffer
  76                                |  |     VirtioRingUninit
  77                                v  |
  78                   +-----------------------------+
  79                   | EfiSimpleNetworkInitialized |
  80                   +-----------------------------+
  81
  82 The state at the top means "nonexistent" and is hence unnamed on the diagram --
  83 a driver instance actually doesn't exist at that point. The transition
  84 functions out of and into that state implement the Driver Binding Protocol.
  85
  86 The lower three states characterize an existent driver instance and are all
  87 states defined by the Simple Network Protocol. The transition functions between
  88 them are member functions of the Simple Network Protocol.
  89
  90 Each transition function validates its expected source state and its
  91 parameters. For example, VirtioNetDriverBindingStop will refuse to disconnect
  92 from the controller unless it's in EfiSimpleNetworkStopped.
  93
  94
  95 Driver instance states (Simple Network Protocol)
  96 ------------------------------------------------
  97
  98 In the EfiSimpleNetworkStopped state, the virtio-net device is (has been)
  99 re-set. No resources are allocated for networking / traffic purposes. The MAC
 100 address and other device attributes have been retrieved from the device (this
 101 is necessary for completing the VirtioNetDriverBindingStart transition).
 102
 103 The EfiSimpleNetworkStarted is completely identical to the
 104 EfiSimpleNetworkStopped state for virtio-net, in the functional and
 105 resource-usage sense. This state is mandated / provided by the Simple Network
 106 Protocol for flexibility that the virtio-net driver doesn't exploit.
 107
 108 In particular, the EfiSimpleNetworkStarted state is the target of the Shutdown
 109 SNP member function, and must therefore correspond to a hardware configuration
 110 where "[it] is safe for another driver to initialize". (Clearly another UEFI
 111 driver could not do that due to the exclusivity of the driver binding that
 112 VirtioNetDriverBindingStart() installs, but a later OS driver might qualify.)
 113
 114 The EfiSimpleNetworkInitialized state is the live state of the virtio NIC / the
 115 driver instance. Virtio and other resources required for network traffic have
 116 been allocated, and the following SNP member functions are available (in
 117 addition to VirtioNetShutdown which leaves the state):
 118
 119 - VirtioNetReceive [SnpReceive.c]: poll the virtio NIC for an Rx packet that
 120   may have arrived asynchronously;
 121
 122 - VirtioNetTransmit [SnpTransmit.c]: queue a Tx packet for asynchronous
 123   transmission (meant to be used together with VirtioNetGetStatus);
 124
 125 - VirtioNetGetStatus [SnpGetStatus.c]: query link status and status of pending
 126   Tx packets;
 127
 128 - VirtioNetMcastIpToMac [SnpMcastIpToMac.c]: transform a multicast IPv4/IPv6
 129   address into a multicast MAC address;
 130
 131 - VirtioNetReceiveFilters [SnpReceiveFilters.c]: emulate unicast / multicast /
 132   broadcast filter configuration (not their actual effect -- a more liberal
 133   filter setting than requested is allowed by the UEFI specification).
 134
 135 The following SNP member functions are not supported [SnpUnsupported.c]:
 136
 137 - VirtioNetReset: reinitialize the virtio NIC without shutting it down (a loop
 138   from/to EfiSimpleNetworkInitialized);
 139
 140 - VirtioNetStationAddress: assign a new MAC address to the virtio NIC,
 141
 142 - VirtioNetStatistics: collect statistics,
 143
 144 - VirtioNetNvData: access non-volatile data on the virtio NIC.
 145
 146 Missing support for these functions is allowed by the UEFI specification and
 147 doesn't seem to trip up higher level protocols.
 148
 149
 150 Events and task priority levels
 151 -------------------------------
 152
 153 The UEFI specification defines a sophisticated mechanism for asynchronous
 154 events / callbacks (see "6.1 Event, Timer, and Task Priority Services" for
 155 details). Such callbacks work like software interrupts, and some notion of
 156 locking / masking is important to implement critical sections (atomic or
 157 exclusive access to data or a device). This notion is defined as Task Priority
 158 Levels.
 159
 160 The virtio-net driver for OVMF must concern itself with events for two reasons:
 161
 162 - The Simple Network Protocol provides its clients with a (non-optional) WAIT
 163   type event called WaitForPacket: it allows them to check or wait for Rx
 164   packets by polling or blocking on this event. (This functionality overlaps
 165   with the Receive member function.) The event is available to clients starting
 166   with EfiSimpleNetworkStopped (inclusive).
 167
 168   The virtio-net driver is informed about such client polling or blockage by
 169   receiving an asynchronous callback (a software interrupt). In the callback
 170   function the driver must interrogate the driver instance state, and if it is
 171   EfiSimpleNetworkInitialized, access the Rx queue and see if any packets are
 172   available for consumption. If so, it must signal the WaitForPacket WAIT type
 173   event, waking the client.
 174
 175   For simplicity and safety, all parts of the virtio-net driver that access any
 176   bit of the driver instance (data or device) run at the TPL_CALLBACK level.
 177   This is the highest level allowed for an SNP implementation, and all code
 178   protected in this manner satisfies even stricter non-blocking requirements
 179   than what's documented for TPL_CALLBACK.
 180
 181   The task priority level for the WaitForPacket callback too is set by the
 182   driver, the choice is TPL_CALLBACK again. This in effect serializes  the
 183   WaitForPacket callback (VirtioNetIsPacketAvailable [Events.c]) with "normal"
 184   parts of the driver.
 185
 186 - According to the Driver Writer's Guide, a network driver should install a
 187   callback function for the global EXIT_BOOT_SERVICES event (a special NOTIFY
 188   type event). When the ExitBootServices() boot service has cleaned up internal
 189   firmware state and is about to pass control to the OS, any network driver has
 190   to stop any in-flight DMA transfers, lest it corrupts OS memory. For this
 191   reason EXIT_BOOT_SERVICES is emitted and the network driver must abort
 192   in-flight DMA transfers.
 193
 194   This callback (VirtioNetExitBoot) is synchronized with the rest of the driver
 195   code just the same as explained for WaitForPacket. In
 196   EfiSimpleNetworkInitialized state it resets the virtio NIC, halting all data
 197   transfer. After the callback returns, no further driver code is expected to
 198   be scheduled.
 199
 200
 201 Virtio internals -- Rx
 202 ----------------------
 203
 204 Requests (Rx and Tx alike) are always submitted by the guest and processed by
 205 the host. For Tx, processing means transmission. For Rx, processing means
 206 filling in the request with an incoming packet. Submitted requests exist on the
 207 "Available Ring", and answered (processed) requests show up on the "Used Ring".
 208
 209 Packet data includes the media (Ethernet) header: destination MAC, source MAC,
 210 and Ethertype (14 bytes total).
 211
 212 The following structures implement packet reception. Most of them are defined
 213 in the Virtio specification, the only driver-specific trait here is the static
 214 pre-configuration of the two-part descriptor chains, in VirtioNetInitRx. The
 215 diagram is simplified.
 216
 217                      Available Index       Available Index
 218                      last processed          incremented
 219                        by the host           by the guest
 220                            v       ------->        v
 221 Available  +-------+-------+-------+-------+-------+
 222 Ring       |DescIdx|DescIdx|DescIdx|DescIdx|DescIdx|
 223            +-------+-------+-------+-------+-------+
 224                               =D6     =D2
 225
 226        D2         D3          D4         D5          D6         D7
 227 Descr. +----------+----------++----------+----------++----------+----------+
 228 Table  |Adr:Len:Nx|Adr:Len:Nx||Adr:Len:Nx|Adr:Len:Nx||Adr:Len:Nx|Adr:Len:Nx|
 229        +----------+----------++----------+----------++----------+----------+
 230         =A2    =D3 =A3         =A4    =D5 =A5         =A6    =D7 =A7
 231
 232
 233             A2        A3     A4       A5     A6       A7
 234 Receive     +---------------+---------------+---------------+
 235 Destination |vnet hdr:packet|vnet hdr:packet|vnet hdr:packet|
 236 Area        +---------------+---------------+---------------+
 237
 238                 Used Index                               Used Index incremented
 239         last processed by the guest                            by the host
 240                     v                    ------->                   v
 241 Used    +-----------+-----------+-----------+-----------+-----------+
 242 Ring    |DescIdx:Len|DescIdx:Len|DescIdx:Len|DescIdx:Len|DescIdx:Len|
 243         +-----------+-----------+-----------+-----------+-----------+
 244                                      =D4
 245
 246 In VirtioNetInitRx, the guest allocates the fixed size Receive Destination
 247 Area, which accommodates all packets delivered asynchronously by the host. To
 248 each packet, a slice of this area is dedicated; each slice is further
 249 subdivided into virtio-net request header and network packet data. The
 250 (guest-physical) addresses of these sub-slices are denoted with A2, A3, A4 and
 251 so on. Importantly, an even-subscript "A" always belongs to a virtio-net
 252 request header, while an odd-subscript "A" always belongs to a packet
 253 sub-slice.
 254
 255 Furthermore, the guest lays out a static pattern in the Descriptor Table. For
 256 each packet that can be in-flight or already arrived from the host,
 257 VirtioNetInitRx sets up a separate, two-part descriptor chain. For packet N,
 258 the Nth descriptor chain is set up as follows:
 259
 260 - the first (=head) descriptor, with even index, points to the fixed-size
 261   sub-slice receiving the virtio-net request header,
 262
 263 - the second descriptor (with odd index) points to the fixed (1514 byte) size
 264   sub-slice receiving the packet data,
 265
 266 - a link from the first (head) descriptor in the chain is established to the
 267   second (tail) descriptor in the chain.
 268
 269 Finally, the guest populates the Available Ring with the indices of the head
 270 descriptors. All descriptor indices on both the Available Ring and the Used
 271 Ring are even.
 272
 273 Packet reception occurs as follows:
 274
 275 - The host consumes a descriptor index off the Available Ring. This index is
 276   even (=2*N), and fingers the head descriptor of the chain belonging to packet
 277   N.
 278
 279 - The host reads the descriptors D(2*N) and -- following the Next link there
 280   --- D(2*N+1), and stores the virtio-net request header at A(2*N), and the
 281   packet data at A(2*N+1).
 282
 283 - The host places the index of the head descriptor, 2*N, onto the Used Ring,
 284   and sets the Len field in the same Used Ring Element to the total number of
 285   bytes transferred for the entire descriptor chain. This enables the guest to
 286   identify the length of Rx packets.
 287
 288 - VirtioNetReceive polls the Used Ring. If a new Used Ring Element shows up, it
 289   copies the data out to the caller, and recycles the index of the head
 290   descriptor (ie. 2*N) to the Available Ring.
 291
 292 - Because the host can process (answer) Rx requests in any order theoretically,
 293   the order of head descriptor indices on each of the Available Ring and the
 294   Used Ring is virtually random. (Except right after the initial population in
 295   VirtioNetInitRx, when the Available Ring is full and increasing, and the Used
 296   Ring is empty.)
 297
 298 - If the Available Ring is empty, the host is forced to drop packets. If the
 299   Used Ring is empty, VirtioNetReceive returns EFI_NOT_READY (no packet
 300   available).
 301
 302
 303 Virtio internals -- Tx
 304 ----------------------
 305
 306 The transmission structure erected by VirtioNetInitTx is similar, it differs
 307 in the following:
 308
 309 - There is no Receive Destination Area.
 310
 311 - Each head descriptor, D(2*N), points to a read-only virtio-net request header
 312   that is shared by all of the head descriptors. This virtio-net request header
 313   is never modified by the host.
 314
 315 - Each tail descriptor is re-pointed to the caller-supplied packet buffer
 316   whenever VirtioNetTransmit places the corresponding head descriptor on the
 317   Available Ring. The caller is responsible to hang on to the unmodified buffer
 318   until it is reported transmitted by VirtioNetGetStatus.
 319
 320 Steps of packet transmission:
 321
 322 - Client code calls VirtioNetTransmit. VirtioNetTransmit tracks free descriptor
 323   chains by keeping the indices of their head descriptors in a stack that is
 324   private to the driver instance. All elements of the stack are even.
 325
 326 - If the stack is empty (that is, each descriptor chain, in isolation, is
 327   either pending transmission, or has been processed by the host but not
 328   yet recycled by a VirtioNetGetStatus call), then VirtioNetTransmit returns
 329   EFI_NOT_READY.
 330
 331 - Otherwise the index of a free chain's head descriptor is popped from the
 332   stack. The linked tail descriptor is re-pointed as discussed above. The head
 333   descriptor's index is pushed on the Available Ring.
 334
 335 - The host moves the head descriptor index from the Available Ring to the Used
 336   Ring when it transmits the packet.
 337
 338 - Client code calls VirtioNetGetStatus. In case the Used Ring is empty, the
 339   function reports no Tx completion. Otherwise, a head descriptor's index is
 340   consumed from the Used Ring and recycled to the private stack. The client
 341   code's original packet buffer address is fetched from the tail descriptor
 342   (where it has been stored at VirtioNetTransmit time) and returned to the
 343   caller.
 344
 345 - The Len field of the Used Ring Element is not checked. The host is assumed to
 346   have transmitted the entire packet -- VirtioNetTransmit had forced it below
 347   1514 bytes (inclusive). The Virtio specification suggests this packet size is
 348   always accepted (and a lower MTU could be encountered on any later hop as
 349   well). Additionally, there's no good way to report a short transmit via
 350   VirtioNetGetStatus; EFI_DEVICE_ERROR seems too serious from the specification
 351   and higher level protocols could interpret it as a fatal condition.
 352
 353 - The host can theoretically reorder head descriptor indices when moving them
 354   from the Available Ring to the Used Ring (out of order transmission). Because
 355   of this (and the choice of a stack over a list for free descriptor chain
 356   tracking) the order of head descriptor indices on either Ring is
 357   unpredictable.