Changeset 3265 in vbox

Timestamp:

Jun 25, 2007 8:50:42 AM (17 years ago)

Author:

vboxsync

Message:

NACK mouse commands we don't implement (not doing so slows down OS/2 boot up by 10+ seconds). Return 0 if anyone tries to read non-standard 8042 RAM (commands 0x21 thru 0x3f).

File:

• trunk/src/VBox/Devices/Input/DevPS2.cpp (modified) (3 diffs)

trunk/src/VBox/Devices/Input/DevPS2.cpp

@@ -166 +166 @@
 #define AUX_RESET               0xFF    /* Reset aux device */
 #define AUX_ACK                 0xFA    /* Command byte ACK. */
+#ifdef VBOX
+#define AUX_NACK                        0xFE    /* Command byte NACK. */
+#endif 
 
 #define MOUSE_STATUS_REMOTE     0x40
@@ -505 +508 @@
         /* ignore that - I don't know what is its use */
         break;
+#ifdef VBOX /* Make OS/2 happy. */
+    /* The 8042 RAM is readble using commands 0x20 thru 0x3f, and writable 
+       by 0x60 thru 0x7f. Now days only the firs byte, the mode, is used.
+       We'll ignore the writes (0x61..7f) and return 0 for all the reads
+       just to make some OS/2 debug stuff a bit happier. */
+    case 0x21: case 0x22: case 0x23: case 0x24: case 0x25: case 0x26: case 0x27:
+    case 0x28: case 0x29: case 0x2a: case 0x2b: case 0x2c: case 0x2d: case 0x2e: case 0x2f:
+    case 0x30: case 0x31: case 0x32: case 0x33: case 0x34: case 0x35: case 0x36: case 0x37:
+    case 0x38: case 0x39: case 0x3a: case 0x3b: case 0x3c: case 0x3d: case 0x3e: case 0x3f:
+        kbd_queue(s, 0, 0);
+        Log(("kbd: reading non-standard RAM addr %#x\n", val & 0x1f));
+        break;
+#endif 
     default:
         Log(("kbd: unsupported keyboard cmd=0x%02x\n", val));
@@ -898 +914 @@
             break;
         default:
+#ifdef VBOX
+            /* NACK all commands we don't know. 
+
+               The usecase for this is the OS/2 mouse driver which will try 
+               read 0xE2 in order to figure out if it's a trackpoint device 
+               or not. If it doesn't get a NACK (or ACK) on the command it'll
+               do several hundred thousand status reads before giving up. This 
+               is slows down the OS/2 boot up considerably. (It also seems that
+               the code is somehow vulnerable while polling like this and that
+               mouse or keyboard input at this point might screw things up badly.)
+
+               From http://www.win.tue.nl/~aeb/linux/kbd/scancodes-13.html:
+
+               Every command or data byte sent to the mouse (except for the 
+               resend command fe) is ACKed with fa. If the command or data 
+               is invalid, it is NACKed with fe. If the next byte is again 
+               invalid, the reply is ERROR: fc. */
+            /** @todo send error if we NACKed the previous command? */
+            kbd_queue(s, AUX_NACK, 1);
+#endif
             break;
         }