Re: OHCI_QUIRK_INITRESET (was: 2.6.9-rc2-mm2 ohci_hcd doesn't work)

[David Brownell]

On Friday 24 September 2004 9:16 am, Bjorn Helgaas wrote:
> 
> The attached patch (which applies on top of Rui's patch for
> ALI M5237) fixes the problem for my DL360.  

Hmm, I'd rather avoid needing a quirk table ... especially
when I've always suspected this is some subtle bug in the
way Linux initializes!  Does this patch behave too?

- Dave
Cope differently with hardware that doesn't act like we expect:
kick in the "initreset" quirk automatically, so we won't need
to maintain a quirk table.

Signed-off-by: David Brownell <dbrownell@xxxxxxxxxxxxxxxxxxxxx>

--- 1.82/drivers/usb/host/ohci-hcd.c	Mon Sep 20 10:31:27 2004
+++ edited/drivers/usb/host/ohci-hcd.c	Sat Sep 25 15:53:19 2004
@@ -545,6 +545,7 @@
 	/* 2msec timelimit here means no irqs/preempt */
 	spin_lock_irq (&ohci->lock);
 
+retry:
 	/* HC Reset requires max 10 us delay */
 	writel (OHCI_HCR,  &ohci->regs->cmdstatus);
 	temp = 30;	/* ... allow extra time */
@@ -563,6 +564,8 @@
 	 * ... but some hardware won't init fmInterval "by the book"
 	 * (SiS, OPTi ...), so reset again instead.  SiS doesn't need
 	 * this if we write fmInterval after we're OPERATIONAL.
+	 * Unclear about ALi, ServerWorks, and others ... this could
+	 * easily be a longstanding bug in chip init on Linux.
 	 */
 	if (ohci->flags & OHCI_QUIRK_INITRESET) {
 		writel (ohci->hc_control, &ohci->regs->control);
@@ -586,6 +589,11 @@
 	 */
 	if ((ohci_readl (&ohci->regs->fminterval) & 0x3fff0000) == 0
 			|| !ohci_readl (&ohci->regs->periodicstart)) {
+		if (!(ohci->flags & OHCI_QUIRK_INITRESET)) {
+			ohci->flags |= OHCI_QUIRK_INITRESET;
+			ohci_dbg (ohci, "enabling initreset quirk\n");
+			goto retry;
+		}
 		spin_unlock_irq (&ohci->lock);
 		ohci_err (ohci, "init err (%08x %04x)\n",
 			ohci_readl (&ohci->regs->fminterval),
--- 1.38/drivers/usb/host/ohci-pci.c	Fri Sep 10 17:57:52 2004
+++ edited/drivers/usb/host/ohci-pci.c	Sat Sep 25 15:51:41 2004
@@ -69,8 +69,6 @@
 				&& pdev->device == 0xc861) {
 			ohci_info (ohci,
 				"WARNING: OPTi workarounds unavailable\n");
-			/* OPTi sometimes acts wierd during init */
-			ohci->flags = OHCI_QUIRK_INITRESET;
 		}
 
 		/* Check for NSC87560. We have to look at the bridge (fn1) to
@@ -88,13 +86,6 @@
 				ohci_info (ohci, "Using NSC SuperIO setup\n");
 			}
 		}
-
-		/* SiS sometimes acts wierd during init */
-		else if (pdev->vendor == PCI_VENDOR_ID_SI) {
-			ohci->flags = OHCI_QUIRK_INITRESET;
-			ohci_info(ohci, "SiS init quirk\n");
-		}
-	
 	}
 
 	/* NOTE: there may have already been a first reset, to