[PATCH RESEND v3 04/11] usb: xhci-rcar: convert to readl_poll_timeout_atomic()

[Daniel Thompson]

On Fri, Sep 11, 2020 at 10:33:21AM +0800, Chunfeng Yun wrote:
> On Thu, 2020-09-10 at 14:12 +0100, Daniel Thompson wrote:
> > On Thu, Sep 10, 2020 at 04:21:45PM +0800, Chunfeng Yun wrote:
> > > Use readl_poll_timeout_atomic() to simplify code
> > > 
> > > Cc: Mathias Nyman <mathias.nyman at linux.intel.com>
> > > Cc: Yoshihiro Shimoda <yoshihiro.shimoda.uh at renesas.com>
> > > Signed-off-by: Chunfeng Yun <chunfeng.yun at mediatek.com>
> > > ---
> > > v2~v3: no changes
> > > ---
> > >  drivers/usb/host/xhci-rcar.c | 43 ++++++++++++-------------------------------
> > >  1 file changed, 12 insertions(+), 31 deletions(-)
> > > 
> > > diff --git a/drivers/usb/host/xhci-rcar.c b/drivers/usb/host/xhci-rcar.c
> > > index c1025d3..74f836f 100644
> > > --- a/drivers/usb/host/xhci-rcar.c
> > > +++ b/drivers/usb/host/xhci-rcar.c
> > > @@ -6,6 +6,7 @@
> > >   */
> > >  
> > >  #include <linux/firmware.h>
> > > +#include <linux/iopoll.h>
> > >  #include <linux/module.h>
> > >  #include <linux/platform_device.h>
> > >  #include <linux/of.h>
> > > @@ -127,8 +128,7 @@ static int xhci_rcar_download_firmware(struct usb_hcd *hcd)
> > >  	void __iomem *regs = hcd->regs;
> > >  	struct xhci_plat_priv *priv = hcd_to_xhci_priv(hcd);
> > >  	const struct firmware *fw;
> > > -	int retval, index, j, time;
> > > -	int timeout = 10000;
> > > +	int retval, index, j;
> > >  	u32 data, val, temp;
> > >  	u32 quirks = 0;
> > >  	const struct soc_device_attribute *attr;
> > > @@ -166,32 +166,19 @@ static int xhci_rcar_download_firmware(struct usb_hcd *hcd)
> > >  		temp |= RCAR_USB3_DL_CTRL_FW_SET_DATA0;
> > >  		writel(temp, regs + RCAR_USB3_DL_CTRL);
> > >  
> > > -		for (time = 0; time < timeout; time++) {
> > > -			val = readl(regs + RCAR_USB3_DL_CTRL);
> > > -			if ((val & RCAR_USB3_DL_CTRL_FW_SET_DATA0) == 0)
> > > -				break;
> > > -			udelay(1);
> > > -		}
> > > -		if (time == timeout) {
> > > -			retval = -ETIMEDOUT;
> > > +		retval = readl_poll_timeout_atomic(regs + RCAR_USB3_DL_CTRL,
> > > +				val, !(val & RCAR_USB3_DL_CTRL_FW_SET_DATA0),
> > > +				1, 10000);
> > > +		if (retval < 0)
> > >  			break;
> > > -		}
> > >  	}
> > >  
> > >  	temp = readl(regs + RCAR_USB3_DL_CTRL);
> > >  	temp &= ~RCAR_USB3_DL_CTRL_ENABLE;
> > >  	writel(temp, regs + RCAR_USB3_DL_CTRL);
> > >  
> > > -	for (time = 0; time < timeout; time++) {
> > > -		val = readl(regs + RCAR_USB3_DL_CTRL);
> > > -		if (val & RCAR_USB3_DL_CTRL_FW_SUCCESS) {
> > > -			retval = 0;
> > > -			break;
> > > -		}
> > > -		udelay(1);
> > > -	}
> > > -	if (time == timeout)
> > > -		retval = -ETIMEDOUT;
> > > +	retval = readl_poll_timeout_atomic((regs + RCAR_USB3_DL_CTRL),
> > > +			val, (val & RCAR_USB3_DL_CTRL_FW_SUCCESS), 1, 10000);
> > 
> > Directly assigning to retval at this point will clobber a previous
> > -ETIMEDOUT error.
> > 
> > In other words if there is a timeout waiting for FW_SET_DATA0, but not for
> > DW_SUCCESS, then we will return the wrong return value.
>
> Yes, agree with you, but seems I keep its original logic unchanged.

I disagree.

Your patch does not preserve the original logic. Your patch explicitly
sets retval to zero if the second loop succeeds. The original code does
not do this. As a result there is a change of return code for one of the
error paths.


Daniel.