openethereum/hw/src/trezor.rs

481 lines
16 KiB
Rust
Raw Normal View History

Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
// Copyright 2015-2017 Parity Technologies (UK) Ltd.
// This file is part of Parity.
// Parity is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
// Parity is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
// You should have received a copy of the GNU General Public License
// along with Parity. If not, see <http://www.gnu.org/licenses/>.
//! Trezor hardware wallet module. Supports Trezor v1.
//! See http://doc.satoshilabs.com/trezor-tech/api-protobuf.html
//! and https://github.com/trezor/trezor-common/blob/master/protob/protocol.md
//! for protocol details.
use super::{WalletInfo, TransactionInfo, KeyPath};
use std::cmp::{min, max};
use std::fmt;
use std::sync::{Arc, Weak};
use std::time::{Duration, Instant};
use ethereum_types::{U256, H256, Address};
use ethkey::Signature;
use hidapi;
use libusb;
use parking_lot::{Mutex, RwLock};
use protobuf;
use protobuf::{Message, ProtobufEnum};
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
use trezor_sys::messages::{EthereumAddress, PinMatrixAck, MessageType, EthereumTxRequest, EthereumSignTx, EthereumGetAddress, EthereumTxAck, ButtonAck};
/// Trezor v1 vendor ID
pub const TREZOR_VID: u16 = 0x534c;
/// Trezor product IDs
pub const TREZOR_PIDS: [u16; 1] = [0x0001];
const ETH_DERIVATION_PATH: [u32; 5] = [0x8000002C, 0x8000003C, 0x80000000, 0, 0]; // m/44'/60'/0'/0/0
const ETC_DERIVATION_PATH: [u32; 5] = [0x8000002C, 0x8000003D, 0x80000000, 0, 0]; // m/44'/61'/0'/0/0
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
/// Hardware wallet error.
#[derive(Debug)]
pub enum Error {
/// Ethereum wallet protocol error.
Protocol(&'static str),
/// Hidapi error.
Usb(hidapi::HidError),
/// Device with request key is not available.
KeyNotFound,
/// Signing has been cancelled by user.
UserCancel,
/// The Message Type given in the trezor RPC call is not something we recognize
BadMessageType,
/// Trying to read from a closed device at the given path
LockedDevice(String),
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
}
impl fmt::Display for Error {
fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> {
match *self {
Error::Protocol(ref s) => write!(f, "Trezor protocol error: {}", s),
Error::Usb(ref e) => write!(f, "USB communication error: {}", e),
Error::KeyNotFound => write!(f, "Key not found"),
Error::UserCancel => write!(f, "Operation has been cancelled"),
Error::BadMessageType => write!(f, "Bad Message Type in RPC call"),
Error::LockedDevice(ref s) => write!(f, "Device is locked, needs PIN to perform operations: {}", s),
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
}
}
}
impl From<hidapi::HidError> for Error {
fn from(err: hidapi::HidError) -> Error {
Error::Usb(err)
}
}
impl From<protobuf::ProtobufError> for Error {
fn from(_: protobuf::ProtobufError) -> Error {
Error::Protocol(&"Could not read response from Trezor Device")
}
}
/// Ledger device manager
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
pub struct Manager {
usb: Arc<Mutex<hidapi::HidApi>>,
devices: RwLock<Vec<Device>>,
locked_devices: RwLock<Vec<String>>,
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
key_path: RwLock<KeyPath>,
}
#[derive(Debug)]
struct Device {
path: String,
info: WalletInfo,
}
/// HID Version used for the Trezor device
enum HidVersion {
V1,
V2,
}
impl Manager {
/// Create a new instance.
pub fn new(hidapi: Arc<Mutex<hidapi::HidApi>>) -> Manager {
Manager {
usb: hidapi,
devices: RwLock::new(Vec::new()),
locked_devices: RwLock::new(Vec::new()),
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
key_path: RwLock::new(KeyPath::Ethereum),
}
}
/// Re-populate device list
pub fn update_devices(&self) -> Result<usize, Error> {
let mut usb = self.usb.lock();
usb.refresh_devices();
let devices = usb.devices();
let mut new_devices = Vec::new();
let mut locked_devices = Vec::new();
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
let mut error = None;
for usb_device in devices {
let is_trezor = usb_device.vendor_id == TREZOR_VID;
let is_supported_product = TREZOR_PIDS.contains(&usb_device.product_id);
let is_valid = usb_device.usage_page == 0xFF00 || usb_device.interface_number == 0;
trace!(
"Checking device: {:?}, trezor: {:?}, prod: {:?}, valid: {:?}",
usb_device,
is_trezor,
is_supported_product,
is_valid,
);
if !is_trezor || !is_supported_product || !is_valid {
continue;
}
match self.read_device_info(&usb, &usb_device) {
Ok(device) => new_devices.push(device),
Err(Error::LockedDevice(path)) => locked_devices.push(path.to_string()),
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
Err(e) => {
warn!("Error reading device: {:?}", e);
error = Some(e);
}
}
}
let count = new_devices.len();
trace!("Got devices: {:?}, closed: {:?}", new_devices, locked_devices);
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
*self.devices.write() = new_devices;
*self.locked_devices.write() = locked_devices;
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
match error {
Some(e) => Err(e),
None => Ok(count),
}
}
fn read_device_info(&self, usb: &hidapi::HidApi, dev_info: &hidapi::HidDeviceInfo) -> Result<Device, Error> {
let handle = self.open_path(|| usb.open_path(&dev_info.path))?;
let manufacturer = dev_info.manufacturer_string.clone().unwrap_or("Unknown".to_owned());
let name = dev_info.product_string.clone().unwrap_or("Unknown".to_owned());
let serial = dev_info.serial_number.clone().unwrap_or("Unknown".to_owned());
match self.get_address(&handle) {
Ok(Some(addr)) => {
Ok(Device {
path: dev_info.path.clone(),
info: WalletInfo {
name: name,
manufacturer: manufacturer,
serial: serial,
address: addr,
},
})
}
Ok(None) => Err(Error::LockedDevice(dev_info.path.clone())),
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
Err(e) => Err(e),
}
}
/// Select key derivation path for a known chain.
pub fn set_key_path(&self, key_path: KeyPath) {
*self.key_path.write() = key_path;
}
/// List connected wallets. This only returns wallets that are ready to be used.
pub fn list_devices(&self) -> Vec<WalletInfo> {
self.devices.read().iter().map(|d| d.info.clone()).collect()
}
pub fn list_locked_devices(&self) -> Vec<String> {
(*self.locked_devices.read()).clone()
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
}
/// Get wallet info.
pub fn device_info(&self, address: &Address) -> Option<WalletInfo> {
self.devices.read().iter().find(|d| &d.info.address == address).map(|d| d.info.clone())
}
fn open_path<R, F>(&self, f: F) -> Result<R, Error>
where F: Fn() -> Result<R, &'static str>
{
f().map_err(Into::into)
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
}
pub fn pin_matrix_ack(&self, device_path: &str, pin: &str) -> Result<bool, Error> {
let unlocked = {
let usb = self.usb.lock();
let device = self.open_path(|| usb.open_path(&device_path))?;
let t = MessageType::MessageType_PinMatrixAck;
let mut m = PinMatrixAck::new();
m.set_pin(pin.to_string());
self.send_device_message(&device, &t, &m)?;
let (resp_type, _) = self.read_device_response(&device)?;
match resp_type {
// Getting an Address back means it's unlocked, this is undocumented behavior
MessageType::MessageType_EthereumAddress => Ok(true),
// Getting anything else means we didn't unlock it
_ => Ok(false),
}
};
self.update_devices()?;
unlocked
}
fn get_address(&self, device: &hidapi::HidDevice) -> Result<Option<Address>, Error> {
let typ = MessageType::MessageType_EthereumGetAddress;
let mut message = EthereumGetAddress::new();
match *self.key_path.read() {
KeyPath::Ethereum => message.set_address_n(ETH_DERIVATION_PATH.to_vec()),
KeyPath::EthereumClassic => message.set_address_n(ETC_DERIVATION_PATH.to_vec()),
}
message.set_show_display(false);
self.send_device_message(&device, &typ, &message)?;
let (resp_type, bytes) = self.read_device_response(&device)?;
match resp_type {
MessageType::MessageType_EthereumAddress => {
let response: EthereumAddress = protobuf::core::parse_from_bytes(&bytes)?;
Ok(Some(From::from(response.get_address())))
}
_ => Ok(None),
}
}
/// Sign transaction data with wallet managing `address`.
pub fn sign_transaction(&self, address: &Address, t_info: &TransactionInfo) -> Result<Signature, Error> {
let usb = self.usb.lock();
let devices = self.devices.read();
let device = devices.iter().find(|d| &d.info.address == address).ok_or(Error::KeyNotFound)?;
let handle = self.open_path(|| usb.open_path(&device.path))?;
let msg_type = MessageType::MessageType_EthereumSignTx;
let mut message = EthereumSignTx::new();
match *self.key_path.read() {
KeyPath::Ethereum => message.set_address_n(ETH_DERIVATION_PATH.to_vec()),
KeyPath::EthereumClassic => message.set_address_n(ETC_DERIVATION_PATH.to_vec()),
}
message.set_nonce(self.u256_to_be_vec(&t_info.nonce));
message.set_gas_limit(self.u256_to_be_vec(&t_info.gas_limit));
message.set_gas_price(self.u256_to_be_vec(&t_info.gas_price));
message.set_value(self.u256_to_be_vec(&t_info.value));
match t_info.to {
Some(addr) => {
message.set_to(addr.to_vec())
}
None => (),
}
let first_chunk_length = min(t_info.data.len(), 1024);
let chunk = &t_info.data[0..first_chunk_length];
message.set_data_initial_chunk(chunk.to_vec());
message.set_data_length(t_info.data.len() as u32);
if let Some(c_id) = t_info.chain_id {
message.set_chain_id(c_id as u32);
}
self.send_device_message(&handle, &msg_type, &message)?;
self.signing_loop(&handle, &t_info.chain_id, &t_info.data[first_chunk_length..])
}
fn u256_to_be_vec(&self, val: &U256) -> Vec<u8> {
let mut buf = [0u8; 32];
val.to_big_endian(&mut buf);
buf.iter().skip_while(|x| **x == 0).cloned().collect()
}
fn signing_loop(&self, handle: &hidapi::HidDevice, chain_id: &Option<u64>, data: &[u8]) -> Result<Signature, Error> {
let (resp_type, bytes) = self.read_device_response(&handle)?;
match resp_type {
MessageType::MessageType_Cancel => Err(Error::UserCancel),
MessageType::MessageType_ButtonRequest => {
self.send_device_message(handle, &MessageType::MessageType_ButtonAck, &ButtonAck::new())?;
// Signing loop goes back to the top and reading blocks
// for up to 5 minutes waiting for response from the device
// if the user doesn't click any button within 5 minutes you
// get a signing error and the device sort of locks up on the signing screen
self.signing_loop(handle, chain_id, data)
}
MessageType::MessageType_EthereumTxRequest => {
let resp: EthereumTxRequest = protobuf::core::parse_from_bytes(&bytes)?;
if resp.has_data_length() {
let mut msg = EthereumTxAck::new();
let len = resp.get_data_length() as usize;
msg.set_data_chunk(data[..len].to_vec());
self.send_device_message(handle, &MessageType::MessageType_EthereumTxAck, &msg)?;
self.signing_loop(handle, chain_id, &data[len..])
} else {
let v = resp.get_signature_v();
let r = H256::from_slice(resp.get_signature_r());
let s = H256::from_slice(resp.get_signature_s());
if let Some(c_id) = *chain_id {
// If there is a chain_id supplied, Trezor will return a v
// part of the signature that is already adjusted for EIP-155,
// so v' = v + 2 * chain_id + 35, but code further down the
// pipeline will already do this transformation, so remove it here
let adjustment = 35 + 2 * c_id as u32;
Ok(Signature::from_rsv(&r, &s, (max(v, adjustment) - adjustment) as u8))
} else {
// If there isn't a chain_id, v will be returned as v + 27
let adjusted_v = if v < 27 { v } else { v - 27 };
Ok(Signature::from_rsv(&r, &s, adjusted_v as u8))
}
}
}
MessageType::MessageType_Failure => Err(Error::Protocol("Last message sent to Trezor failed")),
_ => Err(Error::Protocol("Unexpected response from Trezor device.")),
}
}
fn send_device_message(&self, device: &hidapi::HidDevice, msg_type: &MessageType, msg: &Message) -> Result<usize, Error> {
let msg_id = *msg_type as u16;
let mut message = msg.write_to_bytes()?;
let msg_size = message.len();
let mut data = Vec::new();
let hid_version = self.probe_hid_version(device)?;
// Magic constants
data.push('#' as u8);
data.push('#' as u8);
// Convert msg_id to BE and split into bytes
data.push(((msg_id >> 8) & 0xFF) as u8);
data.push((msg_id & 0xFF) as u8);
// Convert msg_size to BE and split into bytes
data.push(((msg_size >> 24) & 0xFF) as u8);
data.push(((msg_size >> 16) & 0xFF) as u8);
data.push(((msg_size >> 8) & 0xFF) as u8);
data.push((msg_size & 0xFF) as u8);
data.append(&mut message);
while data.len() % 63 > 0 {
data.push(0);
}
let mut total_written = 0;
for chunk in data.chunks(63) {
let mut padded_chunk = match hid_version {
HidVersion::V1 => vec!['?' as u8],
HidVersion::V2 => vec![0, '?' as u8],
};
padded_chunk.extend_from_slice(&chunk);
total_written += device.write(&padded_chunk)?;
}
Ok(total_written)
}
fn probe_hid_version(&self, device: &hidapi::HidDevice) -> Result<HidVersion, Error> {
let mut buf2 = [0xFFu8; 65];
buf2[0] = 0;
buf2[1] = 63;
let mut buf1 = [0xFFu8; 64];
buf1[0] = 63;
if device.write(&buf2)? == 65 {
Ok(HidVersion::V2)
} else if device.write(&buf1)? == 64 {
Ok(HidVersion::V1)
} else {
Err(Error::Usb("Unable to determine HID Version"))
}
}
fn read_device_response(&self, device: &hidapi::HidDevice) -> Result<(MessageType, Vec<u8>), Error> {
let protocol_err = Error::Protocol(&"Unexpected wire response from Trezor Device");
let mut buf = vec![0; 64];
let first_chunk = device.read_timeout(&mut buf, 300_000)?;
if first_chunk < 9 || buf[0] != '?' as u8 || buf[1] != '#' as u8 || buf[2] != '#' as u8 {
return Err(protocol_err);
}
let msg_type = MessageType::from_i32(((buf[3] as i32 & 0xFF) << 8) + (buf[4] as i32 & 0xFF)).ok_or(protocol_err)?;
let msg_size = ((buf[5] as u32 & 0xFF) << 24) + ((buf[6] as u32 & 0xFF) << 16) + ((buf[7] as u32 & 0xFF) << 8) + (buf[8] as u32 & 0xFF);
let mut data = Vec::new();
data.extend_from_slice(&buf[9..]);
while data.len() < (msg_size as usize) {
device.read_timeout(&mut buf, 10_000)?;
data.extend_from_slice(&buf[1..]);
}
Ok((msg_type, data[..msg_size as usize].to_vec()))
}
}
// Try to connect to the device using polling in at most the time specified by the `timeout`
fn try_connect_polling(trezor: Arc<Manager>, duration: Duration) -> bool {
let start_time = Instant::now();
while start_time.elapsed() <= duration {
if let Ok(_) = trezor.update_devices() {
return true
}
}
false
}
/// Trezor event handler
/// A separate thread is handeling incoming events
///
/// Note, that this run to completion and race-conditions can't occur but this can
/// therefore starve other events for being process with a spinlock or similar
pub struct EventHandler {
trezor: Weak<Manager>,
}
impl EventHandler {
/// Trezor event handler constructor
pub fn new(trezor: Weak<Manager>) -> Self {
Self { trezor: trezor }
}
}
impl libusb::Hotplug for EventHandler {
fn device_arrived(&mut self, _device: libusb::Device) {
debug!(target: "hw", "Trezor V1 arrived");
if let Some(trezor) = self.trezor.upgrade() {
if try_connect_polling(trezor, Duration::from_millis(500)) != true {
debug!(target: "hw", "Ledger connect timeout");
}
}
}
fn device_left(&mut self, _device: libusb::Device) {
debug!(target: "hw", "Trezor V1 left");
if let Some(trezor) = self.trezor.upgrade() {
if try_connect_polling(trezor, Duration::from_millis(500)) != true {
debug!(target: "hw", "Ledger disconnect timeout");
}
}
}
}
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
#[test]
#[ignore]
/// This test can't be run without an actual trezor device connected
/// (and unlocked) attached to the machine that's running the test
fn test_signature() {
use ethereum_types::{H160, H256, U256};
Trezor Support (#6403) * Copy modal from keepkey branch and generalize The keepkey PinMatrix modal needs to be the same for Trezor, but we should probably try to keep it general since it can be used for both. * Add trezor communication code This is a result of much trial-and-error and a couple of dead-ends in how to communicate and wire everything up. Code here is still a bit WIP with lots of debug prints and stuff. The test works though, it is possible to sign a transaction. * Extend the basic lib to allow Trezor This is kind of ugly and needs some cleanup and generalization. I’ve just copy-pasted some things to bring in the trezor wallets. I’ve also had to add a lock to the USB API so that only one thing talks to the USB at once. * Add RPC plumbing needed We need to be able to get “locked” devices from the frontend to figure out if we’re going to display the PinMatrix or not. Then we need to be able to send a pin to a device. * Add logic to query backend for Trezor and display PinMatrix There’s a bug somewhere here because signing a transaction fails if you take too long to press the confirm button on the device. * Change back to paritytech branch As my fork has been merged in. * Converting spaces to tabs, as it should be * Incorporate correct handling of EIP-155 Turns out the Trezor was adjusting the v part of the signature, and we’re already doing that so it was done twice. * Some circular logic here that was incorrect BE-encoded U256 is almost the same as RLP encoded without the size-byte, except for <u8 sized values. What’s really done is BE-encoded U256 and then left-trimmed to the smallest size. Kind of obvious in hindsight. * Resolve issue where not clicking fast enough fails The device will not repeat a ButtonRequest when you read from it, so you need to have a blocking `read` for whatever amount of time that you want to give the user to click. You could also have a shorter timeout but keep retrying for some amount of time, but it would amount to the same thing. * Scan after pin entry to make accepting it faster * Remove ability to cancel pin request * Some slight cleanup * Probe for the correct HID Version to determine padding * Move the PinMatrix from Accounts to Application * Removing unused dependencies * Mistake in copying over stuff from keepkey branch * Simplify FormattedMessage * Move generated code to external crate * Remove ethcore-util dependency * Fix broken import in test This test is useless without a connected Trezor, not sure how to make it useful without one. * Merge branch 'master' into fh-4500-trezor-support # Conflicts: # rpc/src/v1/helpers/dispatch.rs * Ignore test that can't be run without trezor device * Fixing grumbles * Avoiding owning data in RPC method * Checking for overflow in v part of signature * s/network_id/chain_id * Propagating an error from the HID Api * Condensing code a little bit * Fixing UI. * Debugging trezor. * Minor styling tweak * Make message type into an actual type This makes the message type that the RPC message accepts into an actual type as opposed to just a string, based on feedback. Although I’m not 100% sure this has actually improved the situation. Overall I think the hardware wallet interface needs some refactoring love. * Split the trezor RPC endpoint It’s split into two more generic endpoints that should be suitable for any hardware wallets with the same behavior to sit behind. * Reflect RPC method split in javascript * Fix bug with pin entry * Fix deadlock for Ledger * Avoid having a USB lock in just listing locked wallets * Fix javascript issue (see #6509) * Replace Mutex with RwLock * Update Ledger test * Fix typo causing faulty signatures (sometimes) * *Actually* fix tests * Update git submodule Needed to make tests pass * Swap line orders to prevent possible deadlock * Make setPinMatrixRequest an @action
2017-09-14 19:28:43 +02:00
let hidapi = Arc::new(Mutex::new(hidapi::HidApi::new().unwrap()));
let manager = Manager::new(hidapi.clone());
let addr: Address = H160::from("some_addr");
manager.update_devices().unwrap();
let t_info = TransactionInfo {
nonce: U256::from(1),
gas_price: U256::from(100),
gas_limit: U256::from(21_000),
to: Some(H160::from("some_other_addr")),
chain_id: Some(17),
value: U256::from(1_000_000),
data: (&[1u8; 3000]).to_vec(),
};
let signature = manager.sign_transaction(&addr, &t_info).unwrap();
let expected = Signature::from_rsv(
&H256::from("device_specific_r"),
&H256::from("device_specific_s"),
0x01
);
assert_eq!(signature, expected)
}