Skip to contentSkip to conversation

USB-OHCI + USB broken in 2.4.14/15pre2?

DISCUSSION Thomas Winischhofer 1 participant 2 messages Nov 18, 2001 Kbuild

lore
Share & tools
lore
#0 lore raw
Quoted 7 lines · Linus Torvalds

In article <[email protected]> you write:

Quoted 5 lines

I tried to "recover" this behavior by temporarily patching ohci_pci_resume() so that it does a brutal hc_restart(ohci) instead of nothing when detecting this "odd PCI resume" situation - without any success.

Quoted 5 lines

I would suggest trying to do a "pci_enable_device(dev);" at the very top of ohci_pci_resume(). It sounds like your suspend/resume doesn't re-enable PCI interrupt routing, and doing the device enable will make the kernel re-route the interrupt for you. If that helps, please send me the tested patch, and forward it to the appropriate USB people too.

Kiitos/tack; however, to make it actually work on my particular machine I needed to

1) call hc_restart(ohci) in default section of switch statement AND

2) like you said, insert "pci_enable_device(dev);" right after the declarations in ohci_pci_resume(). (I think this should be done after the check for concurrent resumes, anyway).

This is what the log says this time:

----

Nov 18 13:40:46 oland cardmgr[189]: executing: './network suspend eth1'
Nov 18 13:40:46 oland kernel: usb-ohci.c: USB suspend: usb-00:01.2
Nov 18 13:40:46 oland kernel: NETDEV WATCHDOG: eth1: transmit timed out
Nov 18 13:40:46 oland kernel: eth1: Transmit timeout.
Nov 18 13:40:46 oland kernel: usb-ohci.c: Bus suspended
Nov 18 13:40:46 oland kernel: usb-ohci.c: USB suspend: usb-00:01.3
Nov 18 13:40:47 oland kernel: usb-ohci.c: Bus suspended
Nov 18 13:40:51 oland logger: Storing ALSA mixer settings...done.
Nov 18 13:40:52 oland logger: Shutting down ALSA sound driver (version
0.9.0beta9): done.
Nov 18 13:40:53 oland apmd[350]: User Suspend
Nov 18 13:40:57 oland kernel: PCI: Found IRQ 11 for device 00:01.2
Nov 18 13:40:57 oland kernel: PCI: Sharing IRQ 11 with 00:01.3
Nov 18 13:40:57 oland kernel: usb-ohci.c: odd PCI resume for usb-00:01.2
Nov 18 13:40:57 oland kernel: usb.c: USB disconnect on device 1
Nov 18 13:40:58 oland kernel: hub.c: USB hub found
Nov 18 13:40:58 oland kernel: hub.c: 3 ports detected
Nov 18 13:40:58 oland kernel: PCI: Found IRQ 11 for device 00:01.3
Nov 18 13:40:58 oland kernel: PCI: Sharing IRQ 11 with 00:01.2
Nov 18 13:40:58 oland kernel: usb-ohci.c: odd PCI resume for usb-00:01.3
Nov 18 13:40:58 oland kernel: usb.c: USB disconnect on device 1
Nov 18 13:40:58 oland kernel: usb.c: USB disconnect on device 2
Nov 18 13:40:58 oland kernel: usb-ohci.c: TD leak, 1
Nov 18 13:40:58 oland kernel: hub.c: USB hub found
Nov 18 13:40:58 oland kernel: hub.c: 3 ports detected
Nov 18 13:40:58 oland kernel: APIC error on CPU0: 00(40)
Nov 18 13:40:59 oland cardmgr[189]: executing: './network resume eth1'
Nov 18 13:41:00 oland kernel: hub.c: USB new device connect on bus2/1,
assigned device number 4
Nov 18 13:41:00 oland kernel: input0: USB HID v1.10 Mouse [Logitech USB
Mouse] on usb2:4.0
Nov 18 13:41:00 oland logger: ALSA driver (version 0.9.0beta9) is
already running.
Nov 18 13:41:01 oland kernel: PCI: Found IRQ 11 for device 00:01.4
Nov 18 13:41:01 oland kernel: PCI: Sharing IRQ 11 with 00:03.0
Nov 18 13:41:03 oland apmd[350]: Normal Resume after 00:00:10 (99%
unknown) AC power
Nov 18 13:41:04 oland logger: ALSA driver (version 0.9.0beta9) is
already running.
Nov 18 13:41:05 oland apmd[350]: Normal Resume after 00:00:12 (99%

unknown) AC power
----

As you can see, the APIC error still shows up in the log, although AFTER hub.c says that it found a hub a the ports (ie. after re-initialization).

Wondering what "TD leak, 1" means...?

I think this is more a dirty work-around than an appropriate solution; or is it ok to assume that the control flags are corrupted (or reset to OHCI_USB_OPER) after pci_enable_device() leading into "odd PCI resume"...? (In this case I would be glad to submit a patch...)

Thomas

Signature · 3 lines
Thomas Winischhofer
Vienna/Austria                  Check it out:         
mailto:[email protected]              *** http://www.webit.com/tw
#1 lore raw
Quoted 9 lines · Thomas Winischhofer

Kiitos/tack; however, to make it actually work on my particular machine I needed to

1) call hc_restart(ohci) in default section of switch statement AND

2) like you said, insert "pci_enable_device(dev);" right after the declarations in ohci_pci_resume(). (I think this should be done after the check for concurrent resumes, anyway).

WRONG. This does as a matter of fact _not_ work:

When a device is connected to the bus at the time of suspend, using my somewhat brutal method recovers that device at resume from suspend. So far so good.

But: When there's _no_ device connected to the usb at the time of suspend, the bus does NOT recover properly and produces errors after resume whenever I try to connect a device.

Here's what I did:

----

usr/src/linux/drivers/usb/usb-ohci.c renamed from root/usb-ohci.c +7 -4
@@ -2678,16 +2678,16 @@
 	mdelay (500); /* No schedule here ! */
 	switch (readl (&ohci->regs->control) & OHCI_CTRL_HCFS) {
 		case OHCI_USB_RESET: /* dbg statt info!/TW */
-			dbg("Bus in reset phase ???");
+			info("Bus in reset phase ???");
 			break;
 		case OHCI_USB_RESUME:
-			dbg("Bus in resume phase ???");
+			info("Bus in resume phase ???");
 			break;
 		case OHCI_USB_OPER:
-			dbg("Bus in operational phase ???");
+			info("Bus in operational phase ???");
 			break;
 		case OHCI_USB_SUSPEND:
-			dbg("Bus suspended");
+			info("Bus suspended");
 			break;
 	}
 	/* In some rare situations, Apple's OHCI have happily trashed
@@ -2719,6 +2719,8 @@
 	int		temp;
 	unsigned long	flags;
 
+	pci_enable_device(dev);
+	
 	/* guard against multiple resumes */
 	atomic_inc (&ohci->resume_count);
 	if (atomic_read (&ohci->resume_count) != 1) {
@@ -2812,6 +2814,7 @@
 
 	default:
 		warn ("odd PCI resume for usb-%s", dev->slot_name);
+		hc_restart (ohci);
 	}
 
 	/* controller is operational, extra resumes are harmless */

----

The log:

---- suspending:

Nov 18 19:45:00 oland cardmgr[189]: executing: './network suspend eth1'
Nov 18 19:45:00 oland kernel: NETDEV WATCHDOG: eth1: transmit timed out
Nov 18 19:45:00 oland kernel: eth1: Transmit timeout.
Nov 18 19:45:00 oland kernel: usb-ohci.c: USB suspend: usb-00:01.2
Nov 18 19:45:01 oland kernel: usb-ohci.c: Bus suspended
Nov 18 19:45:01 oland kernel: usb-ohci.c: USB suspend: usb-00:01.3
Nov 18 19:45:01 oland kernel: usb-ohci.c: Bus suspended
Nov 18 19:45:06 oland apmd[350]: User Suspend

----

---- resuming:

Nov 18 19:45:13 oland kernel: PCI: Found IRQ 11 for device 00:01.2
Nov 18 19:45:13 oland kernel: PCI: Sharing IRQ 11 with 00:01.3
Nov 18 19:45:13 oland kernel: usb-ohci.c: odd PCI resume for usb-00:01.2
Nov 18 19:45:13 oland kernel: usb.c: USB disconnect on device 1
Nov 18 19:45:14 oland kernel: hub.c: USB hub found
Nov 18 19:45:14 oland kernel: hub.c: 3 ports detected
Nov 18 19:45:14 oland kernel: PCI: Found IRQ 11 for device 00:01.3
Nov 18 19:45:14 oland kernel: PCI: Sharing IRQ 11 with 00:01.2
Nov 18 19:45:14 oland kernel: usb-ohci.c: odd PCI resume for usb-00:01.3
Nov 18 19:45:14 oland kernel: usb.c: USB disconnect on device 1
Nov 18 19:45:14 oland kernel: hub.c: USB hub found
Nov 18 19:45:14 oland kernel: hub.c: 3 ports detected
Nov 18 19:45:14 oland kernel: APIC error on CPU0: 00(40)
Nov 18 19:45:15 oland cardmgr[189]: executing: './network resume eth1'
Nov 18 19:45:16 oland apmd[350]: Normal Resume after 00:00:10 (99%
unknown) AC power
Nov 18 19:45:17 oland apmd[350]: Normal Resume after 00:00:11 (99%

unknown) AC power
----

(Why resume twice?)

---- connecting device (mouse)

Nov 18 19:45:29 oland kernel: hub.c: USB new device connect on bus1/3,
assigned device number 3
Nov 18 19:45:32 oland kernel: usb_control/bulk_msg: timeout
Nov 18 19:45:32 oland kernel: usb.c: USB device not accepting new
address=3 (error=-110)
Nov 18 19:45:33 oland kernel: hub.c: USB new device connect on bus1/3,
assigned device number 4
Nov 18 19:45:36 oland kernel: usb_control/bulk_msg: timeout
Nov 18 19:45:36 oland kernel: usb.c: USB device not accepting new

address=4 (error=-110)
----

There we are again. What do the ohci gurus say?

Thomas

Signature · 3 lines
Thomas Winischhofer
Vienna/Austria                  Check it out:         
mailto:[email protected]              *** http://www.webit.com/tw