public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-14 18:29 Andries.Brouwer
  2001-08-14 19:35 ` Gunther Mayer
  0 siblings, 1 reply; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-14 18:29 UTC (permalink / raw)
  To: Gunther.Mayer, linux-kernel; +Cc: Andries.Brouwer

    From Gunther.Mayer@t-online.de Tue Aug 14 19:10:40 2001

    >     I can confirm what you suggest:
    >         My mouse (Logitech wheel USB/PS2) sends indeed AA 00.
    >     So, I extended my patch:
    >     psmouse_reconnect = 0: Do nothing (just pass all to userspace)
    >     psmouse_reconnect = 1: Flush Q & ping mouse on AA 00 (default)
    >     psmouse_reconnect = 2: Flush Q & ping mouse on AA (old behaviour)
    > 
    >     With reconnect 1 or 2: After reconnecting, mouse behaves strange
    >         (jumping around the screen)

    This is a serious bug in many user-space drivers. PS/2 mouse protocol
    was designed to easily re-synchronize (think about transmission errors/
    lost bytes).

The fragment of text you reply to is not about the 3-byte PS/2
protocol, but about the 4-byte wheelmouse protocol.

^ permalink raw reply	[flat|nested] 19+ messages in thread
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-14 21:26 Andries.Brouwer
  0 siblings, 0 replies; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-14 21:26 UTC (permalink / raw)
  To: Andries.Brouwer, Gunther.Mayer; +Cc: linux-kernel


> PS/2 mouse protocol was designed to easily re-synchronize

Not very easily. The Microsoft protocol synchronizes after
a single packet, but the PS/2 protocol has packets of three
bytes, two of which have completely arbitrary contents,
while the first byte has arbitrary contents except that
bit 3 is always set. No doubt it takes only a few packets
to resynchronize in practice, but in theory one could have
arbitrarily long strings of ..., 8, 8, 8, 8, 8, 12, 8, 8, 8, ...
and nobody knows whether the 12 was a mouse button.

> _This_ shows, user-mode must be notified of the re-connect!

Yes, we agree completely: this AUX_RECONNECT stuff should
be ripped out of the kernel, no sysctl, no ioctl, no boot param,
just out. Now gpm can handle the sequence AA 00 as it sees fit,
possibly by reinitializing the mouse, and if X gets its mouse
events via gpm it doesnt need the same code itself.

Andries

^ permalink raw reply	[flat|nested] 19+ messages in thread
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-14 11:12 Andries.Brouwer
  2001-08-14 17:10 ` Gunther Mayer
  0 siblings, 1 reply; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-14 11:12 UTC (permalink / raw)
  To: Andries.Brouwer, garloff; +Cc: alan, linux-kernel, mantel, rubini, torvalds

    From garloff@garloff.de Tue Aug 14 11:57:23 2001

    I can confirm what you suggest:
	My mouse (Logitech wheel USB/PS2) sends indeed AA 00.
    So, I extended my patch:
    psmouse_reconnect = 0: Do nothing (just pass all to userspace)
    psmouse_reconnect = 1: Flush Q & ping mouse on AA 00 (default)
    psmouse_reconnect = 2: Flush Q & ping mouse on AA (old behaviour)

    With reconnect 1 or 2: After reconnecting, mouse behaves strange
	(jumping around the screen)
    With reconnect 0:      Mouse is dead

    In both cases restarting gpm gets the mouse back to work again.
    It seems the imps2 driver does some initialization to the mouse.

    If I use the plain ps2 driver, then finally, I see the benefit of the
    reconnect code in the kernel:
    With reconnect = 1 or 2: It works after replugging
    With reconnect = 0:      Mouse is dead after replugging

    In the latter case restarting gpm helps.

Before having an opinion about what would be appropriate,
let me make sure that I understand the facts that you report.

You talk about reconnect, but what is your definition of reconnect?
Is it that the mouse sends AA or AA 00, or is it that you unplug
and replug the mouse?

Andries

^ permalink raw reply	[flat|nested] 19+ messages in thread
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-02 17:27 Andries.Brouwer
  2001-08-14  9:57 ` Kurt Garloff
  0 siblings, 1 reply; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-02 17:27 UTC (permalink / raw)
  To: Andries.Brouwer, garloff; +Cc: alan, linux-kernel, mantel, rubini, torvalds

    From: Kurt Garloff <garloff@suse.de>

    > Not precisely - it is a boot parameter "psaux-reconnect".
    > That is better than a sysctl.

    Why should that be better than a sysctl? Boot parameters are ugly. You
    need to reboot in order to change them ...

Of course I hope that we'll handle this correctly at some point,
without any options or parameters. In my eyes a sysctl is heavier
infrastructure than a boot parameter, so I prefer the latter
when a temporary fix is needed.

    Your other mail implies that we can fix the problem without manual
    intervention by parsing AA 00 instead of just AA. If it's true, I'd=20
    consider that the best solution.=20

Maybe precisely one person reported this, and his address
now bounces. If there exist people who need this "psaux-reconnect"
they can report on the codes they see. Note that just like AA is
a perfectly normal code, also the sequence AA 00 is perfectly
normal. Testing for that only diminishes the probability of
getting it by accident.

Instead of adding boot parameters or sysctls or heuristics,
probably we should just transfer the codes seen to user space,
e.g. to gpm.  Then it is up to gpm to recognize an AA 00 sequence
and decide whether that is something special.

Andries



^ permalink raw reply	[flat|nested] 19+ messages in thread
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-02 16:02 Andries.Brouwer
  2001-08-02 16:18 ` Kurt Garloff
  0 siblings, 1 reply; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-02 16:02 UTC (permalink / raw)
  To: Andries.Brouwer, alan
  Cc: brent, garloff, linux-kernel, mantel, rubini, torvalds

From: Alan Cox <alan@lxorguk.ukuu.org.uk>:

> 2.2 has had the sysctl for ages, and it defaults to off

Not precisely - it is a boot parameter "psaux-reconnect".
That is better than a sysctl.

^ permalink raw reply	[flat|nested] 19+ messages in thread
[parent not found: <no.id>]
* Re: [PATCH] make psaux reconnect adjustable
@ 2001-08-02 11:55 Andries.Brouwer
  0 siblings, 0 replies; 19+ messages in thread
From: Andries.Brouwer @ 2001-08-02 11:55 UTC (permalink / raw)
  To: alan, garloff, torvalds; +Cc: brent, linux-kernel, mantel, rubini

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2749 bytes --]

    From: Kurt Garloff <garloff@suse.de>

    working on notebooks I got used to the touchpads.
    Now, a lot of notebooks have a Synaptics touchpad.
    It offers a few additional features, such as tossing or the third mouse
    button (by a short click in the corner) ...

    Unfortunately, the synps2 generates nonstandard codes when in
    extended mode, amongst which the reconnect (170) token.
    The kernel (since 2.2.15) does interpret it as such and empties the queue.

    This seems to appropriate for a real plug event. For a synps2, it's not,
    but makes your mouse dead for a second. Instead the data should just
    be passed to userspace (gpm).

    So I made the behaviour switchable via a sysctl.=20

Hmm. I don't think there exists a "reconnect token" AA.
At power up or after a reset (FF) the mouse sends AA
(self test passed) followed by 00 (the ID: I am a PS/2 mouse).
During operation the value AA is entirely legal and has no special
significance.

Moreover, it seems that the specs say that the host should not react
to the AA but wait for the 00. The expected protocol at power up is:
Mouse: AA 00. Host: F4 (enable). Mouse: FA (ack).
Afterwards the host will send F3 to set the sample rate
(or to initialize a wheelmouse).

A quote [1]:
"At power-on, the PS/2 device performs a self-test and calibration,
then transmits the completion code $AA and ID code $00.  If the
device fails its self-test, it transmits error code $FC and ID code $00.
This processing also occurs when a software Reset ($FF) command is received.
The host should not attempt to send commands to the device until
the calibration/self-test is complete.
Power-on self-test and calibration takes 300­1000ms."

The reaction to FF (reset) from the host is an immediate FA (ack)
followed half a second later by AA 00.

Therefore, I think the kernel mouse handling was broken when this
strange AUX_RECONNECT stuff was introduced. It caused a tiny trickle
of complaints.
Bug 1: AA is not necessarily anything special
Bug 2: the sequence AA 00 from the mouse should not be interrupted

I don't think that it is a good idea to start building infrastructure
around it. By default this AUX_RECONNECT should be disabled, since
it is just plain wrong. Then the Synaptics touchpad will work.
In other words, no sysctl but #if 0 ... #endif.

For people who unplug and replug their PS/2 mouse with running
machine, or who use a KVM switch (I think it was only Brent Verner
who asked for this code): if what I say is correct you should
always see 00 following the AA. So, there may exist a more cautious
patch that will bite fewer people and does not react to AA but to
the sequence AA 00.

Andries

[1] See http://www.synaptics.com/decaf/utilities/tp-intf2-4.PDF

^ permalink raw reply	[flat|nested] 19+ messages in thread
* [PATCH] make psaux reconnect adjustable
@ 2001-08-02  2:21 Kurt Garloff
  0 siblings, 0 replies; 19+ messages in thread
From: Kurt Garloff @ 2001-08-02  2:21 UTC (permalink / raw)
  To: Linus Torvalds, Alan Cox
  Cc: Alessandro Rubini, Hubert Mantel, Linux kernel list


[-- Attachment #1.1: Type: text/plain, Size: 1453 bytes --]

Hi Linus Alan,

working on notebooks I got used to the touchpads.
Now, a lot of notebooks have a Synaptics touchpad.
It offers a few additional features, such as tossing or the third mouse
button (by a short click in the corner) ...

gpm -t synps2 does support those additional features and via the -R epeater
mode you also get it under X11.

Unfortunately, the synps2 generates nonstandard codes when in extended mode,
amongst which the reconnect (170) token.
The kernel (since 2.2.15) does interpret it as such and empties the queue.

This seems to appropriate for a real plug event. For a synps2, it's not, but
makes your mouse dead for a second. Instead the data should just be passed
to userspace (gpm).

So I made the behaviour switchable via a sysctl. 
/proc/sys/dev/ps2/psmouse_reconnect (defaults to 1 = the interpret behaviour)
Being at it, I also made the kbd error and unknown scancode reporting
switchable. (It used to be ifdefs.)

Please apply attached patch (against 2.4.7).

Allesandro, should I submit a patch for gpm to automatically handle this for
synps2 in case the kernel patch gets accepted?

Regards,
-- 
Kurt Garloff                   <kurt@garloff.de>         [Eindhoven, NL]
Physics: Plasma simulations  <K.Garloff@Phys.TUE.NL>  [TU Eindhoven, NL]
Linux: SCSI, Security          <garloff@suse.de>    [SuSE Nuernberg, DE]
 (See mail header or public key servers for PGP2 and GPG public keys.)

[-- Attachment #1.2: 247-psaux-reconnect-sysctl.diff --]
[-- Type: text/plain, Size: 5729 bytes --]

--- linux/include/linux/sysctl.h.orig	Tue Jul 31 23:49:42 2001
+++ linux/include/linux/sysctl.h	Thu Aug  2 03:41:34 2001
@@ -594,7 +594,8 @@
 	DEV_HWMON=2,
 	DEV_PARPORT=3,
 	DEV_RAID=4,
-	DEV_MAC_HID=5
+	DEV_MAC_HID=5,
+	DEV_PSAUX=6,
 };
 
 /* /proc/sys/dev/cdrom */
@@ -653,6 +654,13 @@
 	DEV_MAC_HID_MOUSE_BUTTON2_KEYCODE=4,
 	DEV_MAC_HID_MOUSE_BUTTON3_KEYCODE=5,
 	DEV_MAC_HID_ADB_MOUSE_SENDS_KEYCODES=6
+};
+
+/* /proc/sys/dev/psaux */
+enum {
+	DEV_PSMOUSE_RECONNECT=1,
+	DEV_KBD_REPORT_UNKN=2,
+	DEV_KBD_REPORT_TO=3,
 };
 
 #ifdef __KERNEL__
--- linux/drivers/char/pc_keyb.c.orig	Fri Apr  6 19:42:55 2001
+++ linux/drivers/char/pc_keyb.c	Thu Aug  2 04:01:15 2001
@@ -92,8 +92,83 @@
 #define AUX_INTS_ON  (KBD_MODE_KCC | KBD_MODE_SYS | KBD_MODE_MOUSE_INT | KBD_MODE_KBD_INT)
 
 #define MAX_RETRIES	60		/* some aux operations take long time*/
+
 #endif /* CONFIG_PSMOUSE */
 
+/* We want to be able to handle the psmouse reconnect token; unfortunately the
+ * Synaptics touchpads (and probably others too) use it for their extented
+ * functionality and produce them in extended mode (as set by gpm -t synps2).
+ * So we make this adjustable via a sysctl.  garloff@suse.de, 2001-08-01 */
+
+#ifdef CONFIG_SYSCTL
+#include <linux/sysctl.h>
+#ifdef CONFIG_PSMOUSE
+int sysctl_psmouse_reconnect = 1;
+#endif
+int sysctl_kbd_report_unkn = 0;
+int sysctl_kbd_report_to = 0;
+
+static int psaux_sysctl_handler (ctl_table *ctl, int write, struct file *filp,
+				 void *buffer, size_t *lenp)
+{
+	int *valp = ctl->data;
+	int ret = proc_dointvec(ctl, write, filp, buffer, lenp); 
+	if (write) {
+		if (*valp)
+			*valp = 1;
+	}
+	return ret;
+}
+			
+
+ctl_table psaux_table[] = {
+#ifdef CONFIG_PSMOUSE   
+        {DEV_PSMOUSE_RECONNECT, "psmouse_reconnect", &sysctl_psmouse_reconnect,
+                sizeof(int), 0644, NULL, &psaux_sysctl_handler},
+#endif   
+	{DEV_KBD_REPORT_UNKN, "kbd_report_unknown", &sysctl_kbd_report_unkn,
+                sizeof(int), 0644, NULL, &psaux_sysctl_handler},
+        {DEV_KBD_REPORT_TO, "kbd_report_timeout", &sysctl_kbd_report_to,
+                sizeof(int), 0644, NULL, &psaux_sysctl_handler},
+	{0}
+};
+
+ctl_table psaux_psaux_table[] = {
+        {DEV_CDROM, "ps2", NULL, 0, 0555, psaux_table},
+        {0}
+        };
+
+ctl_table psaux_root_table[] = {
+#ifdef CONFIG_PROC_FS
+        {CTL_DEV, "dev", NULL, 0, 0555, psaux_psaux_table},
+#endif /* CONFIG_PROC_FS */
+        {0}
+        };
+static struct ctl_table_header *psaux_sysctl_header;
+
+static void psaux_sysctl_register (void)
+{
+	static int initialized;
+	if (initialized) return;
+	
+	psaux_sysctl_header = register_sysctl_table (psaux_root_table, 1);
+	/*psaux_root_table->child->de->owner = THIS_MODULE;*/
+	initialized++;
+}
+
+/*
+static void psaux_sysctl_unregister (void)
+{
+	if (psaux_sysctl_header)
+		unregister_sysctl_table (psaux_sysctl_header);
+}
+ */
+#else /* CONFIG_SYSCTL */
+#define sysctl_psmouse_reconnect 1
+#define sysctl_kbd_report_unkn 0
+#define sysctl_kbd_report_to 0
+#endif /* CONFIG_SYSCTL */
+
 /*
  * Wait for keyboard controller input buffer to drain.
  *
@@ -123,9 +198,8 @@
 		mdelay(1);
 		timeout--;
 	} while (timeout);
-#ifdef KBD_REPORT_TIMEOUTS
-	printk(KERN_WARNING "Keyboard timed out[1]\n");
-#endif
+	if (sysctl_kbd_report_to)
+		printk(KERN_WARNING "Keyboard timed out[1]\n");
 }
 
 /*
@@ -320,10 +394,8 @@
 		  *keycode = E1_PAUSE;
 		  prev_scancode = 0;
 	      } else {
-#ifdef KBD_REPORT_UNKN
-		  if (!raw_mode)
+		  if (!raw_mode && sysctl_kbd_report_unkn)
 		    printk(KERN_INFO "keyboard: unknown e1 escape sequence\n");
-#endif
 		  prev_scancode = 0;
 		  return 0;
 	      }
@@ -348,11 +420,9 @@
 	      if (e0_keys[scancode])
 		*keycode = e0_keys[scancode];
 	      else {
-#ifdef KBD_REPORT_UNKN
-		  if (!raw_mode)
+		  if (!raw_mode && sysctl_kbd_report_unkn)
 		    printk(KERN_INFO "keyboard: unknown scancode e0 %02x\n",
 			   scancode);
-#endif
 		  return 0;
 	      }
 	  }
@@ -370,11 +440,9 @@
 	  *keycode = high_keys[scancode - SC_LIM];
 
 	  if (!*keycode) {
-	      if (!raw_mode) {
-#ifdef KBD_REPORT_UNKN
+	      if (!raw_mode && sysctl_kbd_report_unkn) {
 		  printk(KERN_INFO "keyboard: unrecognized scancode (%02x)"
 			 " - ignored\n", scancode);
-#endif
 	      }
 	      return 0;
 	  }
@@ -404,12 +472,15 @@
 		mouse_reply_expected = 0;
 	}
 	else if(scancode == AUX_RECONNECT){
-		queue->head = queue->tail = 0;  /* Flush input queue */
-		__aux_write_ack(AUX_ENABLE_DEV);  /* ping the mouse :) */
-		return;
+		if (sysctl_psmouse_reconnect) {
+			queue->head = queue->tail = 0;  /* Flush input queue */
+			__aux_write_ack(AUX_ENABLE_DEV);  /* ping the mouse :) */
+			return;
+		}
 	}
+	else
+		add_mouse_randomness(scancode);
 
-	add_mouse_randomness(scancode);
 	if (aux_count) {
 		int head = queue->head;
 
@@ -511,17 +582,14 @@
 			if (resend)
 				break;
 			mdelay(1);
-			if (!--timeout) {
-#ifdef KBD_REPORT_TIMEOUTS
+			if (!--timeout && sysctl_kbd_report_to) {
 				printk(KERN_WARNING "keyboard: Timeout - AT keyboard not present?\n");
-#endif
 				return 0;
 			}
 		}
 	} while (retries-- > 0);
-#ifdef KBD_REPORT_TIMEOUTS
-	printk(KERN_WARNING "keyboard: Too many NACKs -- noisy kbd cable?\n");
-#endif
+	if (sysctl_kbd_report_to)
+		printk(KERN_WARNING "keyboard: Too many NACKs -- noisy kbd cable?\n");
 	return 0;
 }
 
@@ -751,6 +819,7 @@
 
 	/* Ok, finally allocate the IRQ, and off we go.. */
 	kbd_request_irq(keyboard_interrupt);
+	psaux_sysctl_register ();
 }
 
 #if defined CONFIG_PSMOUSE

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

^ permalink raw reply	[flat|nested] 19+ messages in thread
end of thread, other threads:[~2001-08-15 15:49 UTC | newest]

Thread overview: 19+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2001-08-14 18:29 [PATCH] make psaux reconnect adjustable Andries.Brouwer
2001-08-14 19:35 ` Gunther Mayer
  -- strict thread matches above, loose matches on Subject: below --
2001-08-14 21:26 Andries.Brouwer
2001-08-14 11:12 Andries.Brouwer
2001-08-14 17:10 ` Gunther Mayer
2001-08-14 21:56   ` Kurt Garloff
2001-08-02 17:27 Andries.Brouwer
2001-08-14  9:57 ` Kurt Garloff
2001-08-14 15:03   ` Kurt Garloff
2001-08-14 16:58     ` Linus Torvalds
2001-08-14 17:35       ` Kurt Garloff
2001-08-14 21:29       ` Kurt Garloff
2001-08-15 15:50         ` Gunther Mayer
2001-08-02 16:02 Andries.Brouwer
2001-08-02 16:18 ` Kurt Garloff
     [not found] <no.id>
2001-08-02 15:03 ` Alan Cox
2001-08-02 17:34 ` Alan Cox
2001-08-02 11:55 Andries.Brouwer
2001-08-02  2:21 Kurt Garloff

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox