Re: [RFC PATCH v2 2/3] usb: xhci-mtk: modify the SOF/ITP interval for mt8195

[Chunfeng Yun <chunfeng.yun@mediatek.com>]

On Mon, 2021-02-08 at 13:43 +0200, Mathias Nyman wrote:
> On 7.2.2021 4.27, Chunfeng Yun wrote:
> > Hi Mathias,
> > 
> > On Wed, 2021-02-03 at 18:26 +0800, Chunfeng Yun wrote:
> >> There are 4 USB controllers on MT8195, the controllers (IP1~IP3,
> >> exclude IP0) have a wrong default SOF/ITP interval which is
> >> calculated from the frame counter clock 24Mhz by default, but
> >> in fact, the frame counter clock is 48Mhz, so we should set
> >> the accurate interval according to 48Mhz for those controllers.
> >> Note: the first controller no need set it.
> >>
> >> Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
> >> ---
> >> v2: fix typo of comaptible
> >> ---
> >>  drivers/usb/host/xhci-mtk.c | 63 +++++++++++++++++++++++++++++++++++++
> >>  1 file changed, 63 insertions(+)
> >>
> >> diff --git a/drivers/usb/host/xhci-mtk.c b/drivers/usb/host/xhci-mtk.c
> >> index 8f321f39ab96..0a68c4ac8b48 100644
> >> --- a/drivers/usb/host/xhci-mtk.c
> >> +++ b/drivers/usb/host/xhci-mtk.c
> >> @@ -68,11 +68,71 @@
> >>  #define SSC_IP_SLEEP_EN	BIT(4)
> >>  #define SSC_SPM_INT_EN		BIT(1)
> >>  
> > Can I Read/Write the following xHCI controller's registers  in
> > xhci-mtk.c?
> > 
> > Ideally, xhci-mtk.c should not access them, because xhci-mtk is only a
> > glue driver used to initialize clocks/power and IPPC registers which
> > don't belong to xHCI controller.
> > 
> 
> These *_EOF registers look like they are Mediatek vendor specific registers
> and not part of public xHCI register-level spec. 
> So I think accessing them from xhci-mtk.c makes sense.
> 
> If those register offsets are hardcoded like this in the Mediatek spec then
> this is fine, 
Check it with our DE, it's this case.
> but if those offsets are found from a vendor specific xHCI
> extended capability entry (see xhci spec section 7) then we should dig them out
> from there. 
> >> +/* xHCI csr */
> >> +#define LS_EOF			0x930
> >> +#define LS_EOF_OFFSET		0x89
> >> +
> >> +#define FS_EOF			0x934
> >> +#define FS_EOF_OFFSET		0x2e
> >> +
> >> +#define SS_GEN1_EOF		0x93c
> >> +#define SS_GEN1_EOF_OFFSET	0x78
> >> +
> >> +#define HFCNTR_CFG		0x944
> >> +#define ITP_DELTA_CLK		(0xa << 1)
> >> +#define ITP_DELTA_CLK_MASK	GENMASK(5, 1)
> >> +#define FRMCNT_LEV1_RANG	(0x12b << 8)
> >> +#define FRMCNT_LEV1_RANG_MASK	GENMASK(19, 8)
> >> +
> >> +#define SS_GEN2_EOF		0x990
> >> +#define SS_GEN2_EOF_OFFSET	0x3c
> >> +#define EOF_OFFSET_MASK		GENMASK(11, 0)
> >> +
> >>  enum ssusb_uwk_vers {
> >>  	SSUSB_UWK_V1 = 1,
> >>  	SSUSB_UWK_V2,
> >>  };
> >>  
> >> +/*
> >> + * MT8195 has 4 controllers, the controller1~3's default SOF/ITP interval
> >> + * is calculated from the frame counter clock 24M, but in fact, the clock
> >> + * is 48M, so need change the interval.
> >> + */
> >> +static void xhci_mtk_set_frame_interval(struct xhci_hcd_mtk *mtk)
> >> +{
> >> +	struct device *dev = mtk->dev;
> >> +	struct usb_hcd *hcd = mtk->hcd;
> >> +	u32 value;
> >> +
> >> +	if (!of_device_is_compatible(dev->of_node, "mediatek,mt8195-xhci"))
> >> +		return;
> >> +
> >> +	value = readl(hcd->regs + HFCNTR_CFG);
> >> +	value &= ~(ITP_DELTA_CLK_MASK | FRMCNT_LEV1_RANG_MASK);
> >> +	value |= (ITP_DELTA_CLK | FRMCNT_LEV1_RANG);
> >> +	writel(value, hcd->regs + HFCNTR_CFG);
> >> +
> >> +	value = readl(hcd->regs + LS_EOF);
> >> +	value &= ~EOF_OFFSET_MASK;
> >> +	value |= LS_EOF_OFFSET;
> >> +	writel(value, hcd->regs + LS_EOF);
> >> +
> >> +	value = readl(hcd->regs + FS_EOF);
> >> +	value &= ~EOF_OFFSET_MASK;
> >> +	value |= FS_EOF_OFFSET;
> >> +	writel(value, hcd->regs + FS_EOF);
> >> +
> >> +	value = readl(hcd->regs + SS_GEN1_EOF);
> >> +	value &= ~EOF_OFFSET_MASK;
> >> +	value |= SS_GEN1_EOF_OFFSET;
> >> +	writel(value, hcd->regs + SS_GEN1_EOF);
> >> +
> >> +	value = readl(hcd->regs + SS_GEN2_EOF);
> >> +	value &= ~EOF_OFFSET_MASK;
> >> +	value |= SS_GEN2_EOF_OFFSET;
> >> +	writel(value, hcd->regs + SS_GEN2_EOF);
> 
> Minor nit about names,
> Register offsets from MMIO start are named *_EOF while clock multipliers? are named *_EOF_OFFSET.
> This was a bit confusing
Good point, the names come from register map docs, I'll modify it,
thanks a lot
> 
> Thanks
> -Mathias