bug: usb hub enumeration hangs

bug: usb hub enumeration hangs

[Nick Owens]

[-- Attachment #1: Type: text/plain, Size: 1094 bytes --]

on my new machine xhci/usbd lock up when my IBM ultranav with an
internal hub is plugged in.

attached is a patch that fixes it for me. if it looks ok, i will commit it.

if anyone has hubs, whether just a hub or inside something else like a
keyboard, please test this.

i have traced this down to usbd attempting to set the 'port enable'
feature on one of the ports. when it attempts to enable the port, a
usb request is made but there is no reply, making usbd hang while
doing the endpoint write.

from the USB 2 spec:

11.24.2.7.1.2 PORT_ENABLE
...
This bit may be set only as a result of a SetPortFeature(PORT_RESET) request.
...
The hub response to a SetPortFeature(PORT_ENABLE) request is not
specified. The preferred behavior is
that the hub respond with a Request Error. This may not be used by the
USB System Software. The
ClearPortFeature(PORT_ENABLE) request is supported as specified in
Section 11.5.1.4.

it's pretty clear to me we shouldn't fiddle with this bit. the patch
fixes use of it, and additionally replaces port enabling with
unsuspend when we check for port suspension.

[-- Attachment #2: usbd.patch --]
[-- Type: text/x-patch, Size: 2255 bytes --]

diff --git a/sys/src/cmd/nusb/usbd/dat.h b/sys/src/cmd/nusb/usbd/dat.h
--- a/sys/src/cmd/nusb/usbd/dat.h
+++ b/sys/src/cmd/nusb/usbd/dat.h
@@ -55,7 +55,6 @@ enum
 
 	/* Delays, timeouts (ms) */
 	Spawndelay	= 250,		/* how often may we re-spawn a driver */
-	Connectdelay	= 500,		/* how much to wait after a connect */
 	Resetdelay	= 20,		/* how much to wait after a reset */
 	Enabledelay	= 20,		/* how much to wait after an enable */
 	Powerdelay	= 100,		/* after powering up ports */
diff --git a/sys/src/cmd/nusb/usbd/hub.c b/sys/src/cmd/nusb/usbd/hub.c
--- a/sys/src/cmd/nusb/usbd/hub.c
+++ b/sys/src/cmd/nusb/usbd/hub.c
@@ -362,7 +362,7 @@ portattach(Hub *h, int p, u32int sts)
 	pp->state = Pattached;
 	dprint(2, "%s: %s: port %d attach sts %#ux\n", argv0, d->dir, p, sts);
 	if(h->dev->isusb3){
-		sleep(Connectdelay);
+		sleep(Enabledelay);
 		sts = portstatus(h, p);
 		if(sts == -1)
 			goto Fail;
@@ -372,9 +372,6 @@ portattach(Hub *h, int p, u32int sts)
 		}
 		sp = "super";
 	} else {
-		sleep(Connectdelay);
-		if(hubfeature(h, p, Fportenable, 1) < 0)
-			dprint(2, "%s: %s: port %d: enable: %r\n", argv0, d->dir, p);
 		sleep(Enabledelay);
 		if(hubfeature(h, p, Fportreset, 1) < 0){
 			dprint(2, "%s: %s: port %d: reset: %r\n", argv0, d->dir, p);
@@ -558,12 +555,7 @@ portreset(Hub *h, int p)
 		goto Fail;
 	if((sts & PSenable) == 0){
 		dprint(2, "%s: %s: port %d: not enabled?\n", argv0, d->dir, p);
-		if(h->dev->isusb3)
-			goto Fail;
-		hubfeature(h, p, Fportenable, 1);
-		sts = portstatus(h, p);
-		if((sts & PSenable) == 0)
-			goto Fail;
+		goto Fail;
 	}
 	nd = pp->dev;
 	opendevdata(nd, ORDWR);
@@ -635,11 +627,11 @@ enumhub(Hub *h, int p)
 	onhubs = nhubs;
 	if(!h->dev->isusb3){
 		if((sts & PSsuspend) != 0){
-			if(hubfeature(h, p, Fportenable, 1) < 0)
-				dprint(2, "%s: %s: port %d: enable: %r\n", argv0, d->dir, p);
+			if(hubfeature(h, p, Fportsuspend, 0) < 0)
+				dprint(2, "%s: %s: port %d: unsuspend: %r\n", argv0, d->dir, p);
 			sleep(Enabledelay);
 			sts = portstatus(h, p);
-			fprint(2, "%s: %s: port %d: resumed (sts %#ux)\n", argv0, d->dir, p, sts);
+			fprint(2, "%s: %s: port %d: unsuspended (sts %#ux)\n", argv0, d->dir, p, sts);
 		}
 	}
 	if((pp->sts & PSpresent) == 0 && (sts & PSpresent) != 0){

Re: [9front] bug: usb hub enumeration hangs

[Ethan Gardener]

On Tue, Apr 16, 2019, at 11:23 AM, Nick Owens wrote:
> 
> if anyone has hubs, whether just a hub or inside something else like a
> keyboard, please test this.

I used a keyboard with a hub years ago. The hub was okay. The keyboard wasn't quite all right, but it wasn't quite all right with other OSs too. 

Re: [9front] bug: usb hub enumeration hangs

[Ethan Gardener]

On Tue, Apr 16, 2019, at 11:27 AM, Ethan Gardener wrote:
> 
> I used a keyboard with a hub years ago. The hub was okay. The keyboard 
> wasn't quite all right, but it wasn't quite all right with other OSs 
> too. 

Just tried it again: keyboard + 2 mice all working simultaneously. (All ports filled.) A different mouse didn't work at all. The keyboardhub and not-working mouse produce messages, "required 4 attempts (no data from device)", but no messages were produced for the working mice. Then I tried a joystick in place of one of the mice. It also works, its buttons appearing as mouse buttons. No messages on kprint.

Re: [9front] bug: usb hub enumeration hangs

[cinap_lenrek]

good catch! i'll try it.

--
cinap

Re: [9front] bug: usb hub enumeration hangs

[cinap_lenrek]

works for me!

--
cinap

Re: [9front] bug: usb hub enumeration hangs

[Nick Owens]

committed in f7b606b2be21.