From mboxrd@z Thu Jan  1 00:00:00 1970
From: Marek Vasut <marek.vasut@gmail.com>
Date: Tue, 19 Jun 2012 23:16:52 +0200
Subject: [U-Boot] [PATCH v2 1/1] tegra: usb: Fix device enumeration
	problem of USB1
In-Reply-To: <4FE0B327.3060502@wwwdotorg.org>
References: <4B9C9637D5087840A465BDCB251780E9E2D629AF74@HKMAIL02.nvidia.com>
	<4FE0B327.3060502@wwwdotorg.org>
Message-ID: <201206192316.52907.marek.vasut@gmail.com>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

Dear Stephen Warren,

> On 06/19/2012 02:47 AM, Jim Lin wrote:
> > A known hardware issue of USB1 port where bit 1 (connect status
> > change) of PORTSC register will be set after issuing Port Reset
> > (like "usb reset" in u-boot command line).
> > This will be treated as an error and stops later device enumeration.
> > 
> > Therefore we add a definition in header file and a callback function
> > to clear that bit after Port Reset in order to proceed later device
> > enumeration.
> > 
> > CONFIG_USB_EHCI_SUBMIT_ROOT_POST_CALLBACK

We did such fix in the past by adding weak function. Check ehci-hcd.c

> 
> This style of fix (adding an optional callback from the USB core) seems
> like the idea to me. Thanks.
> 
> However, I'm a little worried that it implements the bug workaround at a
> different time to the Linux kernel. The kernel does:
> 
> * Assert reset to the port
> * Poll the port for enable (any time delays are here)
> * Clear the CSC bit
> 
> So, the CSC bit is cleared a long time after the port is reset.
> 
> However, this U-Boot patch does:
> 
> * Assert reset to the port
> * Clear the CSC bit
> * Wait a long time; 200mS
> * Check the port is enabled
> 
> So, the CSC bit is cleared immediately after the port is reset.
> 
> Is this U-Boot fix still guaranteed to solve the problem? In your
> testing, how many times did you test without this fix, and what was the
> failure percentage. How many times did you test with the fix, and what
> was the failure percentage?
> 
> I think it'd be safer to execute the callback from inside
> hub_port_reset() right before the call to usb_get_port_status(). That
> would make the time when CSC gets cleared much more equivalent between
> the Linux kernel and U-Boot. The implementation of the callback function
> could probably be a lot simpler to (e.g. doesn't need to check which
> feature is being set etc.)

Best regards,
Marek Vasut