feat: send all paste keystrokes to backend

This commit is contained in:
Siyuan Miao 2025-09-09 23:56:23 +02:00
parent bcc307b147
commit 4b0818502c
7 changed files with 224 additions and 89 deletions

View File

@ -475,6 +475,10 @@ func handleSessionRequest(
cloudLogger.Info().Interface("session", session).Msg("new session accepted")
cloudLogger.Trace().Interface("session", session).Msg("new session accepted")
// Cancel any ongoing keyboard report multi when session changes
cancelKeyboardReportMulti()
currentSession = session
_ = wsjson.Write(context.Background(), c, gin.H{"type": "answer", "data": sd})
return nil

View File

@ -1,7 +1,9 @@
package kvm
import (
"errors"
"fmt"
"io"
"time"
"github.com/jetkvm/kvm/internal/hidrpc"
@ -143,6 +145,10 @@ func reportHidRPC(params any, session *Session) {
}
if err := session.HidChannel.Send(message); err != nil {
if errors.Is(err, io.ErrClosedPipe) {
logger.Debug().Err(err).Msg("HID RPC channel closed, skipping reportHidRPC")
return
}
logger.Warn().Err(err).Msg("failed to send HID RPC message")
}
}

View File

@ -10,6 +10,7 @@ import (
"path/filepath"
"reflect"
"strconv"
"sync"
"time"
"github.com/pion/webrtc/v4"
@ -1049,6 +1050,101 @@ func rpcSetLocalLoopbackOnly(enabled bool) error {
return nil
}
// cancelKeyboardReportMulti cancels any ongoing keyboard report multi execution
func cancelKeyboardReportMulti() {
keyboardReportMultiLock.Lock()
defer keyboardReportMultiLock.Unlock()
if keyboardReportMultiCancel != nil {
keyboardReportMultiCancel()
logger.Info().Msg("canceled keyboard report multi")
keyboardReportMultiCancel = nil
}
}
func rpcKeyboardReportMultiWrapper(macro []map[string]any) (usbgadget.KeysDownState, error) {
keyboardReportMultiLock.Lock()
defer keyboardReportMultiLock.Unlock()
if keyboardReportMultiCancel != nil {
keyboardReportMultiCancel()
logger.Info().Msg("canceled previous keyboard report multi")
}
ctx, cancel := context.WithCancel(context.Background())
keyboardReportMultiCancel = cancel
result, err := rpcKeyboardReportMulti(ctx, macro)
keyboardReportMultiCancel = nil
return result, err
}
var (
keyboardReportMultiCancel context.CancelFunc
keyboardReportMultiLock sync.Mutex
)
func rpcKeyboardReportMulti(ctx context.Context, macro []map[string]any) (usbgadget.KeysDownState, error) {
var last usbgadget.KeysDownState
var err error
logger.Debug().Interface("macro", macro).Msg("Executing keyboard report multi")
for i, step := range macro {
// Check for cancellation before each step
select {
case <-ctx.Done():
logger.Debug().Msg("Keyboard report multi context cancelled")
return last, ctx.Err()
default:
}
var modifier byte
if m, ok := step["modifier"].(float64); ok {
modifier = byte(int(m))
} else if mi, ok := step["modifier"].(int); ok {
modifier = byte(mi)
} else if mb, ok := step["modifier"].(uint8); ok {
modifier = mb
}
var keys []byte
if arr, ok := step["keys"].([]any); ok {
keys = make([]byte, 0, len(arr))
for _, v := range arr {
if f, ok := v.(float64); ok {
keys = append(keys, byte(int(f)))
} else if i, ok := v.(int); ok {
keys = append(keys, byte(i))
} else if b, ok := v.(uint8); ok {
keys = append(keys, b)
}
}
} else if bs, ok := step["keys"].([]byte); ok {
keys = bs
}
// Use context-aware sleep that can be cancelled
select {
case <-time.After(100 * time.Millisecond):
// Sleep completed normally
case <-ctx.Done():
logger.Debug().Int("step", i).Msg("Keyboard report multi cancelled during sleep")
return last, ctx.Err()
}
last, err = rpcKeyboardReport(modifier, keys)
if err != nil {
logger.Warn().Err(err).Msg("failed to execute keyboard report multi")
return last, err
}
}
return last, nil
}
var rpcHandlers = map[string]RPCHandler{
"ping": {Func: rpcPing},
"reboot": {Func: rpcReboot, Params: []string{"force"}},
@ -1060,6 +1156,7 @@ var rpcHandlers = map[string]RPCHandler{
"setNetworkSettings": {Func: rpcSetNetworkSettings, Params: []string{"settings"}},
"renewDHCPLease": {Func: rpcRenewDHCPLease},
"keyboardReport": {Func: rpcKeyboardReport, Params: []string{"modifier", "keys"}},
"keyboardReportMulti": {Func: rpcKeyboardReportMultiWrapper, Params: []string{"macro"}},
"getKeyboardLedState": {Func: rpcGetKeyboardLedState},
"keypressReport": {Func: rpcKeypressReport, Params: []string{"key", "press"}},
"getKeyDownState": {Func: rpcGetKeysDownState},

View File

@ -8,35 +8,24 @@ import { GridCard } from "@components/Card";
import { TextAreaWithLabel } from "@components/TextArea";
import { SettingsPageHeader } from "@components/SettingsPageheader";
import { JsonRpcResponse, useJsonRpc } from "@/hooks/useJsonRpc";
import { useHidStore, useRTCStore, useUiStore, useSettingsStore } from "@/hooks/stores";
import { keys, modifiers } from "@/keyboardMappings";
import { KeyStroke } from "@/keyboardLayouts";
import { useHidStore, useSettingsStore, useUiStore } from "@/hooks/stores";
import useKeyboard from "@/hooks/useKeyboard";
import useKeyboardLayout from "@/hooks/useKeyboardLayout";
import notifications from "@/notifications";
const hidKeyboardPayload = (modifier: number, keys: number[]) => {
return { modifier, keys };
};
const modifierCode = (shift?: boolean, altRight?: boolean) => {
return (shift ? modifiers.ShiftLeft : 0)
| (altRight ? modifiers.AltRight : 0)
}
const noModifier = 0
export default function PasteModal() {
const TextAreaRef = useRef<HTMLTextAreaElement>(null);
const { setPasteModeEnabled } = useHidStore();
const { setDisableVideoFocusTrap } = useUiStore();
const { send } = useJsonRpc();
const { rpcDataChannel } = useRTCStore();
const { executeMacro } = useKeyboard();
const [invalidChars, setInvalidChars] = useState<string[]>([]);
const close = useClose();
const { setKeyboardLayout } = useSettingsStore();
const { selectedKeyboard } = useKeyboardLayout();
const { selectedKeyboard } = useKeyboardLayout();
useEffect(() => {
send("getKeyboardLayout", {}, (resp: JsonRpcResponse) => {
@ -52,15 +41,20 @@ export default function PasteModal() {
}, [setDisableVideoFocusTrap, setPasteModeEnabled]);
const onConfirmPaste = useCallback(async () => {
setPasteModeEnabled(false);
setDisableVideoFocusTrap(false);
// setPasteModeEnabled(false);
// setDisableVideoFocusTrap(false);
if (rpcDataChannel?.readyState !== "open" || !TextAreaRef.current) return;
if (!selectedKeyboard) return;
if (!TextAreaRef.current || !selectedKeyboard) return;
const text = TextAreaRef.current.value;
try {
const macroSteps: {
keys: string[] | null;
modifiers: string[] | null;
delay: number;
}[] = [];
for (const char of text) {
const keyprops = selectedKeyboard.chars[char];
if (!keyprops) continue;
@ -70,39 +64,41 @@ export default function PasteModal() {
// if this is an accented character, we need to send that accent FIRST
if (accentKey) {
await sendKeystroke({modifier: modifierCode(accentKey.shift, accentKey.altRight), keys: [ keys[accentKey.key] ] })
const accentModifiers: string[] = [];
if (accentKey.shift) accentModifiers.push("ShiftLeft");
if (accentKey.altRight) accentModifiers.push("AltRight");
macroSteps.push({
keys: [String(accentKey.key)],
modifiers: accentModifiers.length > 0 ? accentModifiers : null,
delay: 100,
});
}
// now send the actual key
await sendKeystroke({ modifier: modifierCode(shift, altRight), keys: [ keys[key] ]});
const modifiers: string[] = [];
if (shift) modifiers.push("ShiftLeft");
if (altRight) modifiers.push("AltRight");
macroSteps.push({
keys: [String(key)],
modifiers: modifiers.length > 0 ? modifiers : null,
delay: 100,
});
// if what was requested was a dead key, we need to send an unmodified space to emit
// just the accent character
if (deadKey) {
await sendKeystroke({ modifier: noModifier, keys: [ keys["Space"] ] });
}
if (deadKey) macroSteps.push({ keys: ["Space"], modifiers: null, delay: 100 });
}
// now send a message with no keys down to "release" the keys
await sendKeystroke({ modifier: 0, keys: [] });
if (macroSteps.length > 0) {
await executeMacro(macroSteps);
}
} catch (error) {
console.error("Failed to paste text:", error);
notifications.error("Failed to paste text");
}
async function sendKeystroke(stroke: KeyStroke) {
await new Promise<void>((resolve, reject) => {
send(
"keyboardReport",
hidKeyboardPayload(stroke.modifier, stroke.keys),
params => {
if ("error" in params) return reject(params.error);
resolve();
}
);
});
}
}, [selectedKeyboard, rpcDataChannel?.readyState, send, setDisableVideoFocusTrap, setPasteModeEnabled]);
}, [selectedKeyboard, executeMacro]);
useEffect(() => {
if (TextAreaRef.current) {
@ -122,14 +118,18 @@ export default function PasteModal() {
/>
<div
className="animate-fadeIn opacity-0 space-y-2"
className="animate-fadeIn space-y-2 opacity-0"
style={{
animationDuration: "0.7s",
animationDelay: "0.1s",
}}
>
<div>
<div className="w-full" onKeyUp={e => e.stopPropagation()} onKeyDown={e => e.stopPropagation()}>
<div
className="w-full"
onKeyUp={e => e.stopPropagation()}
onKeyDown={e => e.stopPropagation()}
>
<TextAreaWithLabel
ref={TextAreaRef}
label="Paste from host"
@ -173,7 +173,8 @@ export default function PasteModal() {
</div>
<div className="space-y-4">
<p className="text-xs text-slate-600 dark:text-slate-400">
Sending text using keyboard layout: {selectedKeyboard.isoCode}-{selectedKeyboard.name}
Sending text using keyboard layout: {selectedKeyboard.isoCode}-
{selectedKeyboard.name}
</p>
</div>
</div>
@ -181,7 +182,7 @@ export default function PasteModal() {
</div>
</div>
<div
className="flex animate-fadeIn opacity-0 items-center justify-end gap-x-2"
className="flex animate-fadeIn items-center justify-end gap-x-2 opacity-0"
style={{
animationDuration: "0.7s",
animationDelay: "0.2s",

View File

@ -1,6 +1,12 @@
import { useCallback } from "react";
import { hidErrorRollOver, hidKeyBufferSize, KeysDownState, useHidStore, useRTCStore } from "@/hooks/stores";
import {
hidErrorRollOver,
hidKeyBufferSize,
KeysDownState,
useHidStore,
useRTCStore,
} from "@/hooks/stores";
import { JsonRpcResponse, useJsonRpc } from "@/hooks/useJsonRpc";
import { useHidRpc } from "@/hooks/useHidRpc";
import { KeyboardLedStateMessage, KeysDownStateMessage } from "@/hooks/hidRpc";
@ -17,9 +23,9 @@ export default function useKeyboard() {
// is running on the cloud against a device that has not been updated yet and thus does not
// support the keyPressReport API. In that case, we need to handle the key presses locally
// and send the full state to the device, so it can behave like a real USB HID keyboard.
// This flag indicates whether the keyPressReport API is available on the device which is
// This flag indicates whether the keyPressReport API is available on the device which is
// dynamically set when the device responds to the first key press event or reports its
// keysDownState when queried since the keyPressReport was introduced together with the
// keysDownState when queried since the keyPressReport was introduced together with the
// getKeysDownState API.
// HidRPC is a binary format for exchanging keyboard and mouse events
@ -27,7 +33,7 @@ export default function useKeyboard() {
reportKeyboardEvent: sendKeyboardEventHidRpc,
reportKeypressEvent: sendKeypressEventHidRpc,
rpcHidReady,
} = useHidRpc((message) => {
} = useHidRpc(message => {
switch (message.constructor) {
case KeysDownStateMessage:
setKeysDownState((message as KeysDownStateMessage).keysDownState);
@ -48,7 +54,9 @@ export default function useKeyboard() {
async (state: KeysDownState) => {
if (rpcDataChannel?.readyState !== "open" && !rpcHidReady) return;
console.debug(`Send keyboardReport keys: ${state.keys}, modifier: ${state.modifier}`);
console.debug(
`Send keyboardReport keys: ${state.keys}, modifier: ${state.modifier}`,
);
if (rpcHidReady) {
console.debug("Sending keyboard report via HidRPC");
@ -56,31 +64,29 @@ export default function useKeyboard() {
return;
}
send("keyboardReport", { keys: state.keys, modifier: state.modifier }, (resp: JsonRpcResponse) => {
if ("error" in resp) {
console.error(`Failed to send keyboard report ${state}`, resp.error);
}
});
send(
"keyboardReport",
{ keys: state.keys, modifier: state.modifier },
(resp: JsonRpcResponse) => {
if ("error" in resp) {
console.error(`Failed to send keyboard report ${state}`, resp.error);
}
},
);
},
[
rpcDataChannel?.readyState,
rpcHidReady,
send,
sendKeyboardEventHidRpc,
],
[rpcDataChannel?.readyState, rpcHidReady, send, sendKeyboardEventHidRpc],
);
// resetKeyboardState is used to reset the keyboard state to no keys pressed and no modifiers.
// This is useful for macros and when the browser loses focus to ensure that the keyboard state
// is clean.
const resetKeyboardState = useCallback(
async () => {
// Reset the keys buffer to zeros and the modifier state to zero
keysDownState.keys.length = hidKeyBufferSize;
keysDownState.keys.fill(0);
keysDownState.modifier = 0;
sendKeyboardEvent(keysDownState);
}, [keysDownState, sendKeyboardEvent]);
const resetKeyboardState = useCallback(async () => {
// Reset the keys buffer to zeros and the modifier state to zero
keysDownState.keys.length = hidKeyBufferSize;
keysDownState.keys.fill(0);
keysDownState.modifier = 0;
sendKeyboardEvent(keysDownState);
}, [keysDownState, sendKeyboardEvent]);
// executeMacro is used to execute a macro consisting of multiple steps.
// Each step can have multiple keys, multiple modifiers and a delay.
@ -88,27 +94,32 @@ export default function useKeyboard() {
// After the delay, the keys and modifiers are released and the next step is executed.
// If a step has no keys or modifiers, it is treated as a delay-only step.
// A small pause is added between steps to ensure that the device can process the events.
const executeMacro = async (steps: { keys: string[] | null; modifiers: string[] | null; delay: number }[]) => {
for (const [index, step] of steps.entries()) {
const executeMacro = async (
steps: { keys: string[] | null; modifiers: string[] | null; delay: number }[],
) => {
const macro: KeysDownState[] = [];
for (const [_, step] of steps.entries()) {
const keyValues = (step.keys || []).map(key => keys[key]).filter(Boolean);
const modifierMask: number = (step.modifiers || []).map(mod => modifiers[mod]).reduce((acc, val) => acc + val, 0);
const modifierMask: number = (step.modifiers || [])
.map(mod => modifiers[mod])
.reduce((acc, val) => acc + val, 0);
// If the step has keys and/or modifiers, press them and hold for the delay
if (keyValues.length > 0 || modifierMask > 0) {
sendKeyboardEvent({ keys: keyValues, modifier: modifierMask });
await new Promise(resolve => setTimeout(resolve, step.delay || 50));
resetKeyboardState();
} else {
// This is a delay-only step, just wait for the delay amount
await new Promise(resolve => setTimeout(resolve, step.delay || 50));
}
// Add a small pause between steps if not the last step
if (index < steps.length - 1) {
await new Promise(resolve => setTimeout(resolve, 10));
macro.push({ keys: keyValues, modifier: modifierMask });
keysDownState.keys.length = hidKeyBufferSize;
keysDownState.keys.fill(0);
keysDownState.modifier = 0;
macro.push(keysDownState);
}
}
// KeyboardReportMessage
send("keyboardReportMulti", { macro }, (resp: JsonRpcResponse) => {
if ("error" in resp) {
console.error(`Failed to send keyboard report ${macro}`, resp.error);
}
});
};
// handleKeyPress is used to handle a key press or release event.
@ -132,7 +143,11 @@ export default function useKeyboard() {
sendKeypressEventHidRpc(key, press);
} else {
// if the keyPress api is not available, we need to handle the key locally
const downState = simulateDeviceSideKeyHandlingForLegacyDevices(keysDownState, key, press);
const downState = simulateDeviceSideKeyHandlingForLegacyDevices(
keysDownState,
key,
press,
);
sendKeyboardEvent(downState); // then we send the full state
// if we just sent ErrorRollOver, reset to empty state
@ -152,7 +167,11 @@ export default function useKeyboard() {
);
// IMPORTANT: See the keyPressReportApiAvailable comment above for the reason this exists
function simulateDeviceSideKeyHandlingForLegacyDevices(state: KeysDownState, key: number, press: boolean): KeysDownState {
function simulateDeviceSideKeyHandlingForLegacyDevices(
state: KeysDownState,
key: number,
press: boolean,
): KeysDownState {
// IMPORTANT: This code parallels the logic in the kernel's hid-gadget driver
// for handling key presses and releases. It ensures that the USB gadget
// behaves similarly to a real USB HID keyboard. This logic is paralleled
@ -164,7 +183,7 @@ export default function useKeyboard() {
if (modifierMask !== 0) {
// If the key is a modifier key, we update the keyboardModifier state
// by setting or clearing the corresponding bit in the modifier byte.
// This allows us to track the state of dynamic modifier keys like
// This allows us to track the state of dynamic modifier keys like
// Shift, Control, Alt, and Super.
if (press) {
modifiers |= modifierMask;
@ -181,7 +200,7 @@ export default function useKeyboard() {
// and if we find a zero byte, we can place the key there (if press is true)
if (keys[i] === key || keys[i] === 0) {
if (press) {
keys[i] = key // overwrites the zero byte or the same key if already pressed
keys[i] = key; // overwrites the zero byte or the same key if already pressed
} else {
// we are releasing the key, remove it from the buffer
if (keys[i] !== 0) {
@ -197,13 +216,15 @@ export default function useKeyboard() {
// If we reach here it means we didn't find an empty slot or the key in the buffer
if (overrun) {
if (press) {
console.warn(`keyboard buffer overflow current keys ${keys}, key: ${key} not added`);
console.warn(
`keyboard buffer overflow current keys ${keys}, key: ${key} not added`,
);
// Fill all key slots with ErrorRollOver (0x01) to indicate overflow
keys.length = hidKeyBufferSize;
keys.fill(hidErrorRollOver);
} else {
// If we are releasing a key, and we didn't find it in a slot, who cares?
console.debug(`key ${key} not found in buffer, nothing to release`)
console.debug(`key ${key} not found in buffer, nothing to release`);
}
}
}

4
web.go
View File

@ -198,6 +198,10 @@ func handleWebRTCSession(c *gin.Context) {
_ = peerConn.Close()
}()
}
// Cancel any ongoing keyboard report multi when session changes
cancelKeyboardReportMulti()
currentSession = session
c.JSON(http.StatusOK, gin.H{"sd": sd})
}

View File

@ -266,6 +266,8 @@ func newSession(config SessionConfig) (*Session, error) {
if connectionState == webrtc.ICEConnectionStateClosed {
scopedLogger.Debug().Msg("ICE Connection State is closed, unmounting virtual media")
if session == currentSession {
// Cancel any ongoing keyboard report multi when session closes
cancelKeyboardReportMulti()
currentSession = nil
}
// Stop RPC processor