On Mon, Oct 22, 2012 at 11:18:58AM +0800, Peter Chen wrote: > On Fri, Oct 19, 2012 at 07:20:01PM +0300, Felipe Balbi wrote: > > Hi, > > > > On Tue, Oct 16, 2012 at 09:36:46AM +0800, Peter Chen wrote: > > > The patch includes both API change and caller change. > > > The main changes like below: > > > > > > - add notify_suspend/notify_resume callback > > > > > > This let usb phy driver has the chance to change hw settings during > > > the controller suspend/resume procedure. > > > > > > Besides, old parameter "port" is useless for phy notify, as one usb > > > phy is only for one usb port. New parameter "speed" stands for > > > the device's speed which is on the port. > > > > > > - implement notify_suspend/notify_resume callback for mxs phy driver > > > These notify will be called during the bus suspend/resume procedure. > > > > > > - Add phy notify at suspend/resume procedure for chipidea host driver > > > > > > - refine phy notify operation during connection and disconnection > > > > > > The history of this problem like below: > > > At some i.mx SoCs, when controller works at host mode, the PHY > > > register needs to be changed at device connect, disconnect, bus > > > suspend and resume due to the SoC limitations. > > > > > > The phy notification should be added according to below rules: > > > > > > 1. Only set HW_USBPHY_CTRL.ENHOSTDISCONDETECT > > > during high speed host mode. > > > 2. Do not set HW_USBPHY_CTRL.ENHOSTDISCONDETECT > > > during the reset and speed negotiation period. > > > 3. Do not set HW_USBPHY_CTRL.ENHOSTDISCONDETECT > > > during host suspend/resume sequence. > > > > > > Please refer: i.mx23RM(page 413) for detail. > > > http://www.freescale.com/files/dsp/doc/ref_manual/IMX23RM.pdf > > > > > > Freescale i.MX SoC, i.mx23, i.mx28 and i.mx6(i.mx6SL does not > > > need to follow the 3rd rule) need to follow above rules. > > > > > > The correct notification setting method should be: > > > 1. Set connect notify after the second bus reset. > > > 2. Set disconnect notify after disconnection. > > > 3. Set suspend nofity after bus goes to suspend (portsc.suspendM=1). > > > 4. Set resume notify after resume (portsc.fpr=0). > > > > > > Signed-off-by: Peter Chen <peter.c...@freescale.com> > > > Tested-by: Mike Thompson <mpthomp...@gmail.com> > > > > sorry but you're doing too much in a single patch. Please split the > > patch before I review it any further. > Ok, I will. But I will put *.h and *.c at one patch to avoid git bitsec > error.
that's fine, it doesn't matter how many files you touch as long as the patch is a single, logical, self-contained change. Look at how many things you're doing in a single patch: you add new function pointers to a structure, you implement those for mxs phy driver, you make changes to connect/disconnect sequence of the phy and so on... all those should be split to their own patch. -- balbi
signature.asc
Description: Digital signature
