diff mbox series

[v3] ehci: fix EHCI host controller initialization sequence

Message ID 20210109181828.GA7585@himera.home
State Superseded
Headers show
Series [v3] ehci: fix EHCI host controller initialization sequence | expand

Commit Message

Eugene Korenevsky Jan. 9, 2021, 6:18 p.m. UTC 
According to EHCI spec, EHCI HC clears USBSTS.HCHalted whenever USBCMD.RS=1.
However, it is a good practice to wait some time after setting USBCMD.RS
(approximately 100ms) until USBSTS.HCHalted become zero.

Instead, previous version of ehci_run() wrote 1 to USBCMD.RS, issued read to
USBCMD and waited for 5 ms.
That worked incorrectly at least for VirtualBox's EHCI virtual HC and caused
accidental hangs.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=211095 
Reviewed-by: Alan Stern <stern@rowland.harvard.edu>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Eugene Korenevsky <ekorenevsky@astralinux.ru>
---
 drivers/usb/host/ehci-hcd.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

Comments

Eugene Korenevsky Jan. 10, 2021, 7:37 a.m. UTC | #1 
> > -	ehci_readl(ehci, &ehci->regs->command);	/* unblock posted writes */


> You should not remove the ehci_readl call above.  With that line gone, 

> the PCI bus might not send the new value of configured_flag to the 

> controller until after the 5-ms sleep has ended, which makes the msleep 

> useless.


Could not find reads from USBCMD in similar drivers (for
example here: https://github.com/NetBSD/src/blob/trunk/sys/dev/usb/ehci.c#L625).
Is this feature (reading from USBCMD for unblocking posted writes)
documented anywhere or it is found empirically?

--
Eugene
Alan Stern Jan. 10, 2021, 4:44 p.m. UTC | #2 
On Sun, Jan 10, 2021 at 10:37:37AM +0300, Eugene Korenevsky wrote:
> > > -	ehci_readl(ehci, &ehci->regs->command);	/* unblock posted writes */

> 

> > You should not remove the ehci_readl call above.  With that line gone, 

> > the PCI bus might not send the new value of configured_flag to the 

> > controller until after the 5-ms sleep has ended, which makes the msleep 

> > useless.

> 

> Could not find reads from USBCMD in similar drivers (for

> example here: https://github.com/NetBSD/src/blob/trunk/sys/dev/usb/ehci.c#L625).


I don't know why NetBSD doesn't do this.  However, note that the 
following poll loop does a read every millisecond, so the first time 
through the loop, the earlier writes will be unblocked.

> Is this feature (reading from USBCMD for unblocking posted writes)

> documented anywhere or it is found empirically?


USBCMD is nothing special; it's just a convenient register to read.  Any 
other register in the controller would do just as well.

This is part of the PCI spec.  Writes are posted (meaning they aren't 
delivered to the device right away), but a read will cause all 
previously posted writes for any address on the same device to be 
delivered before the read finishes.

Alan Stern
diff mbox series

Patch

diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
index e358ae17d51e..f3b73b5ab6e3 100644
--- a/drivers/usb/host/ehci-hcd.c
+++ b/drivers/usb/host/ehci-hcd.c
@@ -574,6 +574,7 @@  static int ehci_run (struct usb_hcd *hcd)
 	struct ehci_hcd		*ehci = hcd_to_ehci (hcd);
 	u32			temp;
 	u32			hcc_params;
+	int			rc;
 
 	hcd->uses_new_polling = 1;
 
@@ -629,9 +630,19 @@  static int ehci_run (struct usb_hcd *hcd)
 	down_write(&ehci_cf_port_reset_rwsem);
 	ehci->rh_state = EHCI_RH_RUNNING;
 	ehci_writel(ehci, FLAG_CF, &ehci->regs->configured_flag);
-	ehci_readl(ehci, &ehci->regs->command);	/* unblock posted writes */
+
+	/* Wait until HC become operational */
 	msleep(5);
+	rc = ehci_handshake(ehci, &ehci->regs->status, STS_HALT, 0, 100 * 1000);
+
 	up_write(&ehci_cf_port_reset_rwsem);
+
+	if (rc) {
+		ehci_err(ehci, "USB %x.%x, controller refused to start: %d\n",
+			 ((ehci->sbrn & 0xf0)>>4), (ehci->sbrn & 0x0f), rc);
+		return rc;
+	}
+
 	ehci->last_periodic_enable = ktime_get_real();
 
 	temp = HC_VERSION(ehci, ehci_readl(ehci, &ehci->caps->hc_capbase));