pw_transfer#

pw_transfer is a reliable data transfer protocol which runs on top of Pigweed RPC.

Attention

pw_transfer is under construction and so is its documentation.

Usage#

C++#

Transfer thread#

To run transfers as either a client or server (or both), a dedicated thread is required. The transfer thread is used to process all transfer-related events safely. The same transfer thread can be shared by a transfer client and service running on the same system.

Note

All user-defined transfer callbacks (i.e. the virtual interface of a Handler or completion function in a transfer client) will be invoked from the transfer thread’s context.

In order to operate, a transfer thread requires two buffers:

  • The first is a chunk buffer. This is used to stage transfer packets received by the RPC system to be processed by the transfer thread. It must be large enough to store the largest possible chunk the system supports.

  • The second is an encode buffer. This is used by the transfer thread to encode outgoing RPC packets. It is necessarily larger than the chunk buffer. Typically, this is sized to the system’s maximum transmission unit at the transport layer.

A transfer thread is created by instantiating a pw::transfer::Thread. This class derives from pw::thread::ThreadCore, allowing it to directly be used when creating a system thread. Refer to Thread Creation for additional information.

Example thread configuration

#include "pw_transfer/transfer_thread.h"

namespace {

// The maximum number of concurrent transfers the thread should support as
// either a client or a server. These can be set to 0 (if only using one or
// the other).
constexpr size_t kMaxConcurrentClientTransfers = 5;
constexpr size_t kMaxConcurrentServerTransfers = 3;

// The maximum payload size that can be transmitted by the system's
// transport stack. This would typically be defined within some transport
// header.
constexpr size_t kMaxTransmissionUnit = 512;

// The maximum amount of data that should be sent within a single transfer
// packet. By necessity, this should be less than the max transmission unit.
//
// pw_transfer requires some additional per-packet overhead, so the actual
// amount of data it sends may be lower than this.
constexpr size_t kMaxTransferChunkSizeBytes = 480;

// Buffers for storing and encoding chunks (see documentation above).
std::array<std::byte, kMaxTransferChunkSizeBytes> chunk_buffer;
std::array<std::byte, kMaxTransmissionUnit> encode_buffer;

pw::transfer::Thread<kMaxConcurrentClientTransfers,
                     kMaxConcurrentServerTransfers>
    transfer_thread(chunk_buffer, encode_buffer);

}  // namespace

// pw::transfer::TransferThread is the generic, non-templated version of the
// Thread class. A Thread can implicitly convert to a TransferThread.
pw::transfer::TransferThread& GetSystemTransferThread() {
  return transfer_thread;
}

Transfer server#

pw_transfer provides an RPC service for running transfers through an RPC server.

To know how to read data from or write data to device, a Handler interface is defined (pw_transfer/public/pw_transfer/handler.h). Transfer handlers represent a transferable resource, wrapping a stream reader and/or writer with initialization and completion code. Custom transfer handler implementations should derive from ReadOnlyHandler, WriteOnlyHandler, or ReadWriteHandler as appropriate and override Prepare and Finalize methods if necessary.

A transfer handler should be implemented and instantiated for each unique resource that can be transferred to or from a device. Each instantiated handler must have a globally-unique integer ID used to identify the resource.

Handlers are registered with the transfer service. This may be done during system initialization (for static resources), or dynamically at runtime to support ephemeral transfer resources.

Example transfer handler implementation

#include "pw_stream/memory_stream.h"
#include "pw_transfer/transfer.h"

// A simple transfer handler which reads data from an in-memory buffer.
class SimpleBufferReadHandler : public pw::transfer::ReadOnlyHandler {
 public:
  SimpleReadTransfer(uint32_t resource_id, pw::ConstByteSpan data)
      : ReadOnlyHandler(resource_id), reader_(data) {
    set_reader(reader_);
  }

 private:
  pw::stream::MemoryReader reader_;
};

Handlers may optionally implement a GetStatus method, which allows clients to query the status of a resource with a handler registered. The application layer above transfer can choose how to fill and interpret this information. The status information is readable_offset, writeable_offset, read_checksum, and write_checksum.

Example GetStatus implementation

Status GetStatus(uint64_t& readable_offset,
                 uint64_t& writeable_offset,
                 uint64_t& read_checksum,
                 uint64_t& write_checksum) {
  readable_offset = resource.get_size();
  writeable_offset = resource.get_writeable_offset();
  read_checksum = resource.get_crc();
  write_checksum = resource.calculate_crc(0, writeable_offset);

  return pw::OkStatus();
}

The transfer service is instantiated with a reference to the system’s transfer thread and registered with the system’s RPC server.

Example transfer service initialization

#include "pw_transfer/transfer.h"

namespace {

// In a write transfer, the maximum number of bytes to receive at one time
// (potentially across multiple chunks), unless specified otherwise by the
// transfer handler's stream::Writer. Should be set reasonably high; the
// transfer will attempt to determine an optimal window size based on the
// link.
constexpr size_t kDefaultMaxBytesToReceive = 16384;

pw::transfer::TransferService transfer_service(
    GetSystemTransferThread(), kDefaultMaxBytesToReceive);

// Instantiate a handler for the data to be transferred. The resource ID will
// be used by the transfer client and server to identify the handler.
constexpr uint32_t kMagicBufferResourceId = 1;
char magic_buffer_to_transfer[256] = { /* ... */ };
SimpleBufferReadHandler magic_buffer_handler(
    kMagicBufferResourceId, magic_buffer_to_transfer);

}  // namespace

void InitTransferService() {
  // Register the handler with the transfer service, then the transfer service
  // with an RPC server.
  transfer_service.RegisterHandler(magic_buffer_handler);
  GetSystemRpcServer().RegisterService(transfer_service);
}

Transfer client#

pw_transfer provides a transfer client capable of running transfers through an RPC client.

Note

Currently, a transfer client is only capable of running transfers on a single RPC channel. This may be expanded in the future.

The transfer client provides the following APIs for managing data transfers:

Result<pw::Transfer::Client::Handle> pw::transfer::Client::Read(
uint32_t resource_id,
pw::stream::Writer &output,
CompletionFunc &&on_completion,
pw::transfer::ProtocolVersion version = kDefaultProtocolVersion,
pw::chrono::SystemClock::duration timeout = cfg::kDefaultClientTimeout,
pw::chrono::SystemClock::duration initial_chunk_timeout = cfg::kDefaultInitialChunkTimeout,
uint32_t initial_offset = 0u,
)#

Reads data from a transfer server to the specified pw::stream::Writer. Invokes the provided callback function with the overall status of the transfer.

Due to the asynchronous nature of transfer operations, this function will only return a non-OK status if it is called with bad arguments. Otherwise, it will return OK and errors will be reported through the completion callback.

For using the offset parameter, please see Non-zero Starting Offset Transfers.

Result<pw::Transfer::Client::Handle> pw::transfer::Client::Write(
uint32_t resource_id,
pw::stream::Reader &input,
CompletionFunc &&on_completion,
pw::transfer::ProtocolVersion version = kDefaultProtocolVersion,
pw::chrono::SystemClock::duration timeout = cfg::kDefaultClientTimeout,
pw::chrono::SystemClock::duration initial_chunk_timeout = cfg::kDefaultInitialChunkTimeout,
uint32_t initial_offset = 0u,
)#

Writes data from a source pw::stream::Reader to a transfer server. Invokes the provided callback function with the overall status of the transfer.

Due to the asynchronous nature of transfer operations, this function will only return a non-OK status if it is called with bad arguments. Otherwise, it will return OK and errors will be reported through the completion callback.

For using the offset parameter, please see Non-zero Starting Offset Transfers.

Transfer handles#

Each transfer session initiated by a client returns a Handle object which is used to manage the transfer. These handles support the following operations:

pw::Transfer::Client::Handle::Cancel()#

Terminates the ongoing transfer.

pw::Transfer::Client::Handle::SetTransferSize(size_t size_bytes)#

In a write transfer, indicates the total size of the transfer resource.

Example client setup

#include "pw_transfer/client.h"

namespace {

// RPC channel on which transfers should be run.
constexpr uint32_t kChannelId = 42;

// In a read transfer, the maximum number of bytes to receive at one time
// (potentially across multiple chunks), unless specified otherwise by the
// transfer's stream. Should be set reasonably high; the transfer will
// attempt to determine an optimal window size based on the link.
constexpr size_t kDefaultMaxBytesToReceive = 16384;

pw::transfer::Client transfer_client(GetSystemRpcClient(),
                                     kChannelId,
                                     GetSystemTransferThread(),
                                     kDefaultMaxBytesToReceive);

}  // namespace

Status ReadMagicBufferSync(pw::ByteSpan sink) {
  pw::stream::Writer writer(sink);

  struct {
    pw::sync::ThreadNotification notification;
    pw::Status status;
  } transfer_state;

  Result<pw::transfer::Client::Handle> handle = transfer_client.Read(
      kMagicBufferResourceId,
      writer,
      [&transfer_state](pw::Status status) {
        transfer_state.status = status;
        transfer_state.notification.release();
      });
  if (!handle.ok()) {
    return handle.status();
  }

  // Block until the transfer completes.
  transfer_state.notification.acquire();
  return transfer_state.status;
}

Specifying Resource Sizes#

Transfer data is sent and received through the pw::Stream interface, which does not have a concept of overall stream size. Users of transfers that are fixed-size may optionally indicate this to the transfer client and server, which will be shared with the transfer peer to enable features such as progress reporting.

The transfer size can only be set on the transmitting side of the transfer; that is, the client in a Write transfer or the server in a Read transfer.

If the specified resource size is smaller than the available transferrable data, only a slice of the data up to the resource size will be transferred. If the specified size is equal to or larger than the data size, all of the data will be sent.

Setting a transfer size from a transmitting client

Result<pw::transfer::Client::Handle> handle = client.Write(...);
if (handle.ok()) {
  handle->SetTransferSize(kMyResourceSize);
}

Setting a transfer size on a server resource

The TransferHandler interface allows overriding its ResourceSize function to return the size of its transfer resource.

 class MyResourceHandler : public pw::transfer::ReadOnlyHandler {
  public:
   Status PrepareRead() final;

   virtual size_t ResourceSize() const final {
     return kMyResourceSize;
   }

};

Atomic File Transfer Handler#

Transfers are handled using the generic Handler interface. A specialized Handler, AtomicFileTransferHandler is available to handle file transfers with atomic semantics. It guarantees that the target file of the transfer is always in a correct state. A temporary file is written to prior to updating the target file. If any transfer failure occurs, the transfer is aborted and the target file is either not created or not updated.

Module Configuration Options#

The following configurations can be adjusted via compile-time configuration of this module, see the module documentation for more details.

PW_TRANSFER_DEFAULT_MAX_CLIENT_RETRIES#

The default maximum number of times a transfer client should retry sending a chunk when no response is received. Can later be configured per-transfer when starting one.

PW_TRANSFER_DEFAULT_MAX_SERVER_RETRIES#

The default maximum number of times a transfer server should retry sending a chunk when no response is received.

In typical setups, retries are driven by the client, and timeouts on the server are used only to clean up resources, so this defaults to 0.

PW_TRANSFER_DEFAULT_MAX_LIFETIME_RETRIES#

The default maximum number of times a transfer should retry sending any chunk over the course of its entire lifetime.

This number should be high, particularly if long-running transfers are expected. Its purpose is to prevent transfers from getting stuck in an infinite loop.

PW_TRANSFER_DEFAULT_CLIENT_TIMEOUT_MS#

The default amount of time, in milliseconds, to wait for a chunk to arrive in a transfer client before retrying. This can later be configured per-transfer.

PW_TRANSFER_DEFAULT_SERVER_TIMEOUT_MS#

The default amount of time, in milliseconds, to wait for a chunk to arrive on the server before retrying. This can later be configured per-transfer.

PW_TRANSFER_DEFAULT_INITIAL_TIMEOUT_MS#

The default amount of time, in milliseconds, to wait for an initial server response to a transfer before retrying. This can later be configured per-transfer.

This is set separately to PW_TRANSFER_DEFAULT_TIMEOUT_MS as transfers may require additional time for resource initialization (e.g. erasing a flash region before writing to it).

PW_TRANSFER_DEFAULT_EXTEND_WINDOW_DIVISOR#

The fractional position within a window at which a receive transfer should extend its window size to minimize the amount of time the transmitter spends blocked.

For example, a divisor of 2 will extend the window when half of the requested data has been received, a divisor of three will extend at a third of the window, and so on.

PW_TRANSFER_LOG_DEFAULT_CHUNKS_BEFORE_RATE_LIMIT#

Number of chunks to send repetitive logs at full rate before reducing to rate_limit. Retransmit parameter chunks will restart at this chunk count limit. Default is first 10 parameter logs will be sent, then reduced to one log every PW_TRANSFER_RATE_PERIOD_MS

PW_TRANSFER_LOG_DEFAULT_RATE_PERIOD_MS#

The minimum time between repetative logs after the rate limit has been applied (after CHUNKS_BEFORE_RATE_LIMIT parameter chunks). Default is to reduce repetative logs to once every 10 seconds after CHUNKS_BEFORE_RATE_LIMIT parameter chunks have been sent.

PW_TRANSFER_CONFIG_LOG_LEVEL#

Configurable log level for the entire transfer module.

PW_TRANSFER_CONFIG_DEBUG_CHUNKS#

Turns on logging of individual non-data or non-parameter chunks. Default is false, to disable logging.

PW_TRANSFER_CONFIG_DEBUG_DATA_CHUNKS#

Turns on logging of individual data and parameter chunks. Default is false to disable logging. These chunks are moderated (rate-limited) by the same PW_TRANSFER_RATE_PERIOD_MS as other repetitive logs.

Non-zero Starting Offset Transfers#

pw_transfer provides for transfers which read from or write to a server resource starting from a point after the beginning. Handling of read/write/erase boundaries of the resource storage backend must be handled by the user through the transfer handler interfaces of GetStatus and PrepareRead/Write(uint32_t offset).

A resource can be read or written from a non-zero starting offset simply by having the transfer client calling read() or write() with an offset parameter. The offset gets included in the starting handshake.

Note

The data or stream passed to read() or write() will be used as-is. I.e. no seeking will be applied; the user is expected to seek to the desired location.

On the server side, the offset is accepted, and passed to the transfer handler’s Prepare(uint32_t) method. This method must be implemented specifically by the handler in order to support the offset transfer. The transfer handler confirms that the start offset is valid for the read/write operation, and the server responds with the offset to confirm the non-zero transfer operation. Older server sw will ignore the offset, so the clients check that the server has accepted the non-zero offset during the handshake, so users may elect to catch such errors. Clients return Status.UNIMPLEMENTED in such cases.

Due to the need to seek streams by the handler to support the non-zero offset, it is recommended to return Status.RESOURCE_EXHAUSTED if a seek is requested past the end of the stream.

See the transfer handler documentation for further information about configuring resources for non-zero transfers and the interface documentation in pw/transfer/public/pw_transfer/handler.h

Python#

Provides a simple interface for transferring bulk data over pw_rpc.

exception pw_transfer.Error(resource_id: int, status: Status)#

Exception raised when a transfer fails.

Stores the ID of the failed transfer resource and the error that occurred.

__init__(resource_id: int, status: Status)#
class pw_transfer.Manager(
rpc_transfer_service,
*,
default_response_timeout_s: float = 2.0,
initial_response_timeout_s: float = 4.0,
max_retries: int = 3,
max_lifetime_retries: int = 1500,
max_chunk_size_bytes: int = 1024,
default_protocol_version=ProtocolVersion.VERSION_TWO,
)#

A manager for transmitting data through an RPC TransferService.

This should be initialized with an active Manager over an RPC channel. Only one instance of this class should exist for a configured RPC TransferService – the Manager supports multiple simultaneous transfers.

When created, a Manager starts a separate thread in which transfer communications and events are handled.

__init__(
rpc_transfer_service,
*,
default_response_timeout_s: float = 2.0,
initial_response_timeout_s: float = 4.0,
max_retries: int = 3,
max_lifetime_retries: int = 1500,
max_chunk_size_bytes: int = 1024,
default_protocol_version=ProtocolVersion.VERSION_TWO,
)#

Initializes a Manager on top of a TransferService.

Parameters:

  • rpc_transfer_service – the pw_rpc transfer service client

  • default_response_timeout_s – max time to wait between receiving packets

  • initial_response_timeout_s – timeout for the first packet; may be longer to account for transfer handler initialization

  • max_retries – number of times to retry a single package after a timeout

  • max_lifetime_retires – Cumulative maximum number of times to retry over the course of the transfer before giving up.

  • max_chunk_size_bytes – In a read transfer, the maximum size of data the server should send within a single packet.

  • default_protocol_version – Version of the pw_transfer protocol to use. Defaults to the latest, but can be set to legacy for projects which use legacy devices.

read(
resource_id: int,
progress_callback: Callable[[ProgressStats], Any] | None = None,
protocol_version: ProtocolVersion | None = None,
chunk_timeout_s: float | None = None,
initial_timeout_s: float | None = None,
initial_offset: int = 0,
) → bytes#

Receives (“downloads”) data from the server.

Parameters:

  • resource_id – ID of the resource from which to read.

  • progress_callback – Optional callback periodically invoked throughout the transfer with the transfer state. Can be used to provide user- facing status updates such as progress bars.

  • protocol_version – The desired protocol version to use for this transfer. Defaults to the version the manager was initialized (typically VERSION_TWO).

  • chunk_timeout_s – Timeout for any individual chunk.

  • initial_timeout_s – Timeout for the first chunk, overrides chunk_timeout_s.

  • initial_offset – Initial offset to start reading from. Must be supported by the transfer handler. All transfers support starting from 0, the default. Returned bytes will not have any padding related to this initial offset. No seeking is done in the transfer operation on the client side.

Raises:

Error – the transfer failed to complete

write(
resource_id: int,
data: bytes | str,
progress_callback: Callable[[ProgressStats], Any] | None = None,
protocol_version: ProtocolVersion | None = None,
chunk_timeout_s: Any | None = None,
initial_timeout_s: Any | None = None,
initial_offset: int = 0,
) → None#

Transmits (“uploads”) data to the server.

Parameters:

  • resource_id – ID of the resource to which to write.

  • data – Data to send to the server.

  • progress_callback – Optional callback periodically invoked throughout the transfer with the transfer state. Can be used to provide user- facing status updates such as progress bars.

  • protocol_version – The desired protocol version to use for this transfer. Defaults to the version the manager was initialized (defaults to LATEST).

  • chunk_timeout_s – Timeout for any individual chunk.

  • initial_timeout_s – Timeout for the first chunk, overrides chunk_timeout_s.

  • initial_offset – Initial offset to start writing to. Must be supported by the transfer handler. All transfers support starting from 0, the default. data arg should start with the data you want to see starting at this initial offset on the server. No seeking is done in the transfer operation on the client side.

Raises:

Error – the transfer failed to complete

class pw_transfer.ProgressStats(bytes_sent: int, bytes_confirmed_received: int, total_size_bytes: int | None)#
__init__(bytes_sent: int, bytes_confirmed_received: int, total_size_bytes: int | None) → None#
class pw_transfer.ProtocolVersion(value)#

Supported versions of pw_transfer’s RPC data transfer protocol.

Example

import pw_transfer

# Initialize a Pigweed RPC client; see pw_rpc docs for more info.
rpc_client = CustomRpcClient()
rpcs = rpc_client.channel(1).rpcs

transfer_service = rpcs.pw.transfer.Transfer
transfer_manager = pw_transfer.Manager(transfer_service)

try:
  # Read the transfer resource with ID 3 from the server.
  data = transfer_manager.read(3)
except pw_transfer.Error as err:
  print('Failed to read:', err.status)

try:
  # Send some data to the server. The transfer manager does not have to be
  # reinitialized.
  transfer_manager.write(2, b'hello, world')
except pw_transfer.Error as err:
  print('Failed to write:', err.status)

Typescript#

Provides a simple interface for transferring bulk data over pw_rpc.

Example

import { pw_transfer } from 'pigweedjs';
const { Manager } from pw_transfer;

const client = new CustomRpcClient();
service = client.channel()!.service('pw.transfer.Transfer')!;

const manager = new Manager(service, DEFAULT_TIMEOUT_S);

manager.read(3, (stats: ProgressStats) => {
  console.log(`Progress Update: ${stats}`);
}).then((data: Uint8Array) => {
  console.log(`Completed read: ${data}`);
}).catch(error => {
  console.log(`Failed to read: ${error.status}`);
});

manager.write(2, textEncoder.encode('hello world'))
  .catch(error => {
    console.log(`Failed to read: ${error.status}`);
  });

Java#

pw_transfer provides a Java client. The transfer client returns a ListenableFuture to represent the results of a read or write transfer.

import dev.pigweed.pw_transfer.TransferClient;

public class TheClass  {
  public void DoTransfer(MethodClient transferReadMethodClient,
                         MethodClient transferWriteMethodClient) {
    // Create a new transfer client.
    TransferClient client = new TransferClient(
        transferReadMethodClient,
        transferWriteMethodClient,
        TransferTimeoutSettings.builder()
            .setTimeoutMillis(TRANSFER_TIMEOUT_MS)
            .setMaxRetries(MAX_RETRIES)
            .build());

    // Start a read transfer.
    ListenableFuture<byte[]> readTransfer = client.read(123);

    // Start a write transfer.
    ListenableFuture<Void> writeTransfer = client.write(123, dataToWrite);

    // Get the data from the read transfer.
    byte[] readData = readTransfer.get();

    // Wait for the write transfer to complete.
    writeTransfer.get();
  }
}

Protocol#

Chunks#

Transfers run as a series of chunks exchanged over an RPC stream. Chunks can contain transferable data, metadata, and control parameters. Each chunk has an associated type, which determines what information it holds and the semantics of its fields.

The chunk is a protobuf message, whose definition can be found here.

Resources and sessions#

Transfers are run for a specific resource — a stream of data which can be read from or written to. Resources have a system-specific integral identifier defined by the implementers of the server-side transfer node.

The series of chunks exchanged in an individual transfer operation for a resource constitute a transfer session. The session runs from its opening chunk until either a terminating chunk is received or the transfer times out. Sessions are assigned IDs by the client that starts them, which are unique over the RPC channel between the client and server, allowing the server to identify transfers across multiple clients.

Reliability#

pw_transfer attempts to be a reliable data transfer protocol.

As Pigweed RPC is considered an unreliable communications system, pw_transfer implements its own mechanisms for reliability. These include timeouts, data retransmissions, and handshakes.

Note

A transfer can only be reliable if its underlying data stream is seekable. A non-seekable stream could prematurely terminate a transfer following a packet drop.

Opening handshake#

Transfers begin with a three-way handshake, whose purpose is to identify the resource being transferred, assign a session ID, and synchronize the protocol version to use.

A read or write transfer for a resource is initiated by a transfer client. The client sends the ID of the resource to the server alongside a unique session ID in a START chunk, indicating that it wishes to begin a new transfer. This chunk additionally encodes the protocol version which the client is configured to use.

Upon receiving a START chunk, the transfer server checks whether the requested resource is available. If so, it prepares the resource for the operation, which typically involves opening a data stream, alongside any additional user-specified setup. The server accepts the client’s session ID, then responds to the client with a START_ACK chunk containing the resource, session, and configured protocol version for the transfer.

Windowing#

Throughout a transfer, the receiver maintains a window of how much data it can receive at a given time. This window is a multiple of the maximum size of a single data chunk, and is adjusted dynamically in response to the ongoing status of the transfer.

pw_transfer uses a congestion control algorithm similar to that of TCP (RFC 5681 §3.1), adapted to pw_transfer’s mode of operation that tunes parameters per window.

Once a portion of a window has successfully been received, it is acknowledged by the reciever and the window size is extended. Transfers begin in a “slow start” phase, during which the window is doubled on each ACK. This continues until the transfer detects a packet loss. Once this occurs, the window size is halved and the transfer enters a “congestion avoidance” phase for the remainder of its run. During this phase, successful ACKs increase the window size by a single chunk, whereas packet loss continues to half it.

Transfer completion#

Either side of a transfer can terminate the operation at any time by sending a COMPLETION chunk containing the final status of the transfer. When a COMPLETION chunk is sent, the terminator of the transfer performs local cleanup, then waits for its peer to acknowledge the completion.

Upon receving a COMPLETION chunk, the transfer peer cancels any pending operations, runs its set of cleanups, and responds with a COMPLETION_ACK, fully ending the session from the peer’s side.

The terminator’s session remains active waiting for a COMPLETION_ACK. If not received after a timeout, it re-sends its COMPLETION chunk. The session ends either following receipt of the acknowledgement or if a maximum number of retries is hit.

Server to client transfer (read)#

../_images/read.svg

Client to server transfer (write)#

../_images/write.svg

Protocol buffer definition#

syntax = "proto3";

package pw.transfer;

option java_multiple_files = true;
option java_package = "dev.pigweed.pw_transfer";

// The transfer RPC service is used to send data between the client and server.
service Transfer {
  // Transfer data from the server to the client; a "download" from the client's
  // perspective.
  rpc Read(stream Chunk) returns (stream Chunk);

  // Transfer data from the client to the server; an "upload" from the client's
  // perspective.
  rpc Write(stream Chunk) returns (stream Chunk);

  // Query the status of a resource. Can be used for partially completed
  // transfers
  rpc GetResourceStatus(ResourceStatusRequest) returns (ResourceStatus);
}

// Represents a chunk of data sent by the transfer service. Includes fields for
// configuring the transfer parameters.
//
// Notation: (Read|Write) (→|←)
//   X → Means client sending data to the server.
//   X ← Means server sending data to the client.
message Chunk {
  // Represents the source or destination of the data. May be ephemeral or
  // stable depending on the implementation. Sent in every request to identify
  // the transfer target.
  //
  // LEGACY FIELD ONLY. Split into resource_id and session_id in transfer v2.
  //
  //  Read → ID of transfer
  //  Read ← ID of transfer
  // Write → ID of transfer
  // Write ← ID of transfer
  uint32 transfer_id = 1;

  // Used by the receiver to indicate how many bytes it can accept. The
  // transmitter sends this much data, divided into chunks no larger than
  // max_chunk_size_bytes. The receiver then starts another window by sending
  // request_bytes again with a new offset.
  //
  //  Read → The client requests this many bytes to be sent.
  //  Read ← N/A
  // Write → N/A
  // Write ← The server requests this many bytes to be sent.
  optional uint32 pending_bytes = 2;

  // Maximum size of an individual chunk. The transmitter may send smaller
  // chunks if required.
  //
  //  Read → Set maximum size for subsequent chunks.
  //  Read ← N/A
  // Write → N/A
  // Write ← Set maximum size for subsequent chunks.
  optional uint32 max_chunk_size_bytes = 3;

  // Minimum required delay between chunks. The transmitter may delay longer if
  // desired.
  //
  //  Read → Set minimum delay for subsequent chunks.
  //  Read ← N/A
  // Write → N/A
  // Write ← Set minimum delay for subsequent chunks.
  optional uint32 min_delay_microseconds = 4;

  // On writes, the offset of the data. On reads, the offset at which to read.
  //
  //  Read → Read data starting at this offset.
  //  Read ← Offset of the data.
  // Write → Offset of the data.
  // Write ← Write data starting at this offset.
  uint64 offset = 5;

  // The data that was read or the data to write.
  //
  //  Read → N/A
  //  Read ← Data read
  // Write → Data to write
  // Write ← N/A
  bytes data = 6;

  // Estimated bytes remaining to read/write. Optional except for the last data
  // chunk, for which remaining_bytes must be set to 0.
  //
  // The sender can set remaining_bytes at the beginning of a read/write so that
  // the receiver can track progress or cancel the transaction if the value is
  // too large.
  //
  //  Read → N/A
  //  Read ← Remaining bytes to read, excluding any data in this chunk. Set to
  //         0 for the last chunk.
  // Write → Remaining bytes to write, excluding any data in is chunk. Set to
  //         0 for the last chunk.
  // Write ← N/A
  optional uint64 remaining_bytes = 7;

  // Pigweed status code indicating the completion of a transfer. This is only
  // present in the final packet sent by either the transmitter or receiver.
  //
  // The possible status codes and their meanings are listed below:
  //
  //   OK: Transfer completed successfully.
  //   DATA_LOSS: Transfer data could not be read/written (e.g. corruption).
  //   INVALID_ARGUMENT: Received malformed chunk.
  //   NOT_FOUND: The requested resource ID is not registered (read/write).
  //   OUT_OF_RANGE: The requested offset is larger than the data (read/write).
  //   RESOURCE_EXHAUSTED: Concurrent transfer limit reached.
  //   UNIMPLEMENTED: Resource ID does not support requested operation (e.g.
  //       trying to write to a read-only transfer).
  //
  //  Read → Transfer complete.
  //  Read ← Transfer complete.
  // Write → Transfer complete.
  // Write ← Transfer complete.
  optional uint32 status = 8;

  // The offset up to which the transmitter can send data before waiting for the
  // receiver to acknowledge.
  //
  //  Read → Offset up to which the server can send without blocking.
  //  Read ← N/A
  // Write → N/A
  // Write ← Offset up to which the client can send without blocking.
  //
  // TODO(frolv): This will replace the pending_bytes field. Once all uses of
  // transfer are migrated, that field should be removed.
  uint32 window_end_offset = 9;

  enum Type {
    // Chunk containing transfer data.
    DATA = 0;

    // First chunk of a transfer (only sent by the client).
    START = 1;

    // Transfer parameters indicating that the transmitter should retransmit
    // from the specified offset.
    PARAMETERS_RETRANSMIT = 2;

    // Transfer parameters telling the transmitter to continue sending up to
    // index `offset + pending_bytes` of data. If the transmitter is already
    // beyond `offset`, it does not have to rewind.
    PARAMETERS_CONTINUE = 3;

    // Sender of the chunk is terminating the transfer.
    COMPLETION = 4;

    // Acknowledge the completion of a transfer. Currently unused.
    // TODO(konkers): Implement this behavior.
    COMPLETION_ACK = 5;

    // Acknowledges a transfer start request, accepting the session ID for the
    // transfer and optionally negotiating the protocol version. Sent from
    // server to client.
    START_ACK = 6;

    // Confirmation of a START_ACK's negotiated parameters, sent by the client
    // to the server. Initiates the data transfer proper.
    START_ACK_CONFIRMATION = 7;
  };

  // The type of this chunk. This field should only be processed when present.
  // TODO(frolv): Update all users of pw_transfer and remove the optional
  // semantics from this field.
  //
  //  Read → Chunk type (start/parameters).
  //  Read ← Chunk type (data).
  // Write → Chunk type (data).
  // Write ← Chunk type (start/parameters).
  optional Type type = 10;

  // Unique identifier for the source or destination of transfer data. May be
  // stable or ephemeral depending on the implementation. Only sent during the
  // initial handshake phase of a version 2 or higher transfer.
  //
  //  Read → ID of transferable resource
  //  Read ← ID of transferable resource
  // Write → ID of transferable resource
  // Write ← ID of transferable resource
  optional uint32 resource_id = 11;

  // Unique identifier for a specific transfer session. Chosen by the transfer
  // client during the initial handshake phase, and persists for the remainder
  // of that transfer operation.
  //
  //  Read → ID of transfer session
  //  Read ← ID of transfer session
  // Write → ID of transfer session
  // Write ← ID of transfer session
  optional uint32 session_id = 12;

  // The protocol version to use for this transfer. Only sent during the initial
  // handshake phase of a version 2 or higher transfer to negotiate a common
  // protocol version between the client and server.
  //
  //  Read → Desired (START) or configured (START_ACK_CONFIRMATION) version.
  //  Read ← Configured protocol version (START_ACK).
  // Write → Desired (START) or configured (START_ACK_CONFIRMATION) version.
  // Write ← Configured protocol version (START_ACK).
  optional uint32 protocol_version = 13;

  // Unique identifier for a specific transfer session. Chosen by the transfer
  // client during the initial handshake phase. This field is used to request a
  // session during the handshake, after which the regular session_id field is
  // used.
  //
  //  Read → Requested ID of transfer session
  //  Read ← N/A
  // Write → Requested ID of transfer session
  // Write ← N/A
  optional uint32 desired_session_id = 14;

  // The initial offset to start the transfer from. Can be used for read or
  // write transfers. Set by the client during start handshake.
  // Needs to be accepted by the resource transfer handler in order for the
  // non-zero offset transfer to start from the initial_offset.
  //
  //  Read → Requested initial offset for the session
  //  Read ← Confirmed (matches) or denied (zero) initial offset
  // Write → Requested initial offset for the session
  // Write ← Confirmed (matches) or denied (zero) initial offset
  uint64 initial_offset = 15;
}

// Request for GetResourceStatus, indicating the resource to get status from.
message ResourceStatusRequest {
  uint32 resource_id = 1;
}

// Response for GetResourceStatus
message ResourceStatus {
  // Resource id, matching request
  uint32 resource_id = 1;

  // Status of the resource, returns Unimplemented by default.
  uint32 status = 2;
  // The offset that can be written to (other than 0).
  uint64 writeable_offset = 3;
  // The offset that can be read from (other than 0).
  uint64 readable_offset = 4;
  // The checksum at the given write offset.
  optional uint64 write_checksum = 5;
  // The checksum at the given read offset.
  optional uint64 read_checksum = 6;
}

Errors#

Protocol errors#

The following table describes the meaning of each status code when sent by the sender or the receiver (see Transfer roles).

Status

Sent by sender

Sent by receiver

OK

(not sent)

All data was received and handled successfully.

ABORTED

The service aborted the transfer because the client restarted it. This status is passed to the transfer handler, but not sent to the client because it restarted the transfer.

CANCELLED

The client cancelled the transfer.

DATA_LOSS

Failed to read the data to send. The Reader returned an error.

Failed to write the received data. The Writer returned an error.

FAILED_PRECONDITION

Received chunk for transfer that is not active.

INVALID_ARGUMENT

Received a malformed packet.

INTERNAL

An assumption of the protocol was violated. Encountering INTERNAL indicates that there is a bug in the service or client implementation.

PERMISSION_DENIED

The transfer does not support the requested operation (either reading or writing).

RESOURCE_EXHAUSTED

The receiver requested zero bytes, indicating their storage is full, but there is still data to send.

Storage is full.

UNAVAILABLE

The service is busy with other transfers and cannot begin a new transfer at this time.

UNIMPLEMENTED

Out-of-order chunk was requested, but seeking is not supported.

(not sent)

Transfer roles#

Every transfer has two participants: the sender and the receiver. The sender transmits data to the receiver. The receiver controls how the data is transferred and sends the final status when the transfer is complete.

In read transfers, the client is the receiver and the service is the sender. In write transfers, the client is the sender and the service is the receiver.

Sender flow#

graph TD start([Client initiates<br>transfer]) -->data_request data_request[Receive transfer<br>parameters]-->send_chunk send_chunk[Send chunk]-->sent_all sent_all{Sent final<br>chunk?} -->|yes|wait sent_all-->|no|sent_requested sent_requested{Sent all<br>pending?}-->|yes|data_request sent_requested-->|no|send_chunk wait[Wait for receiver]-->is_done is_done{Received<br>final chunk?}-->|yes|done is_done-->|no|data_request done([Transfer complete])

Receiver flow#

graph TD start([Client initiates<br>transfer]) -->request_bytes request_bytes[Set transfer<br>parameters]-->wait wait[Wait for chunk]-->received_chunk received_chunk{Received<br>chunk by<br>deadline?}-->|no|request_bytes received_chunk-->|yes|check_chunk check_chunk{Correct<br>offset?} -->|yes|process_chunk check_chunk --> |no|request_bytes process_chunk[Process chunk]-->final_chunk final_chunk{Final<br>chunk?}-->|yes|signal_completion final_chunk{Final<br>chunk?}-->|no|received_requested received_requested{Received all<br>pending?}-->|yes|request_bytes received_requested-->|no|wait signal_completion[Signal completion]-->done done([Transfer complete])

Legacy protocol#

pw_transfer was initially released into production prior to several of the reliability improvements of its modern protocol. As a result of this, transfer implementations support a “legacy” protocol mode, in which transfers run without utilizing these features.

The primary differences between the legacy and modern protocols are listed below.

  • There is no distinction between a transfer resource and session — a single transfer_id field represents both. Only one transfer for a given resource can run at a time, and it is not possible to determine where one transfer for a resource ends and the next begins.

  • The legacy protocol has no opening handshake phase. The client initiates with a transfer ID and starting transfer parameters (during a read), and the data transfer phase begins immediately.

  • The legacy protocol has no terminating handshake phase. When either end completes a transfer by sending a status chunk, it does not wait for the peer to acknowledge. Resources used by the transfer are immediately freed, and there is no guarantee that the peer is notified of completion.

Transfer clients request the latest transfer protocol version by default, but may be configured to request the legacy protocol. Transfer server and client implementations detect if their transfer peer is running the legacy protocol and automatically switch to it if required, even if they requested a newer protocol version. It is strongly unadvised to use the legacy protocol in new code.

Integration tests#

The pw_transfer module has a set of integration tests that verify the correctness of implementations in different languages. Test source code.

To run the tests on your machine, run

$ bazel test \
      pw_transfer/integration_test:cross_language_small_test \
      pw_transfer/integration_test:cross_language_medium_test

Note

There is a large test that tests transfers that are megabytes in size. These are not run automatically, but can be run manually via the pw_transfer/integration_test:cross_language_large_test test. These are VERY slow, but exist for manual validation of real-world use cases.

The integration tests permit injection of client/server/proxy binaries to use when running the tests. This allows manual testing of older versions of pw_transfer against newer versions.

# Test a newer version of pw_transfer against an old C++ client that was
# backed up to another directory.
$ bazel run pw_transfer/integration_test:cross_language_medium_test -- \
    --cpp-client-binary ../old_pw_transfer_version/cpp_client

Backwards compatibility tests#

pw_transfer includes a suite of backwards-compatibility tests that are intended to continuously validate a degree of backwards-compatibility with older pw_transfer servers and clients. This is done by retrieving older binaries hosted in CIPD and running tests between the older client/server binaries and the latest binaries.

The CIPD package contents can be created with this command:

To update the CIPD package itself, follow the internal documentation for updating a CIPD package.

CI/CQ integration#

Current status of the test in CI.

By default, these tests are not run in CQ (on presubmit) because they are too slow. However, you can request that the tests be run in presubmit on your change by adding to following line to the commit message footer:

Cq-Include-Trybots: luci.pigweed.try:pigweed-linux-bzl-integration

Running the tests many times#

Because the tests bind to network ports, you cannot run more than one instance of each test in parallel. However, you might want to do so, e.g. to debug flakes. This section describes a manual process that makes this possible.

Linux#

On Linux, you can add the "block-network" tag to the tests (example). This enables network isolation for the tests, allowing you to run them in parallel via,

bazel test --runs_per_test=10 //pw_transfer/integration_tests/...

MacOS#

Network isolation is not supported on MacOS because the OS doesn’t support network virtualization (gh#2669). The best you can do is to tag the tests "exclusive". This allows you to use --runs_per_test, but will force each test to run by itself, with no parallelism.

Why is this manual?#

Ideally, we would apply either the "block-network" or "exclusive" tag to the tests depending on the OS. But this is not supported, gh#2971.

We don’t want to tag the tests "exclusive" by default because that will prevent different tests from running in parallel, significantly slowing them down.