Hi,

Peter Griffin [off-list ref] writes:

Hi Felipe,

On Fri, 25 Mar 2016, Felipe Balbi wrote:

quoted

Hi,

Gregory CLEMENT [off-list ref] writes:

quoted

quoted

Peter Griffin [off-list ref] writes:

quoted

Otherwise generic-xhci and xhci-platform which have no data get wrongly
detected as XHCI_PLAT_TYPE_MARVELL_ARMADA by xhci_plat_type_is().

This fixes a regression in v4.5 for STiH407 family SoC's which use the
synopsis dwc3 IP, whereby the disable_clk error path gets taken due to
wrongly being detected as XHCI_PLAT_TYPE_MARVELL_ARMADA and the hcd never
gets added.

I suspect this will also fix other dwc3 DT platforms such as Exynos,
although I've only tested on STih410 SoC.

Fixes: 4efb2f694114 ("usb: host: xhci-plat: add struct xhci_plat_priv")
Cc: stable at vger.kernel.org
Cc: gregory.clement at free-electrons.com
Cc: yoshihiro.shimoda.uh at renesas.com
Signed-off-by: Peter Griffin <peter.griffin@linaro.org>
---
 drivers/usb/host/xhci-plat.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/usb/host/xhci-plat.h b/drivers/usb/host/xhci-plat.h
index 5a2e2e3..529c3c4 100644
--- a/drivers/usb/host/xhci-plat.h
+++ b/drivers/usb/host/xhci-plat.h

@@ -14,7 +14,7 @@
 #include "xhci.h"	/* for hcd_to_xhci() */
 
 enum xhci_plat_type {
-	XHCI_PLAT_TYPE_MARVELL_ARMADA,
+	XHCI_PLAT_TYPE_MARVELL_ARMADA = 1,
 	XHCI_PLAT_TYPE_RENESAS_RCAR_GEN2,
 	XHCI_PLAT_TYPE_RENESAS_RCAR_GEN3,

aren't these platforms using device tree ? Why aren't these just
different compatible strings ?

According to 4efb2f69411456d35051e9047c15157c9a5ba217 "usb: host:
xhci-plat: add struct xhci_plat_priv" :

This patch adds struct xhci_plat_priv to simplify the code to match
platform specific variables. For now, this patch adds a member "type" in
the structure

that's fine but the answer doesn't exactly match my question ;-)

My point is that this enum shouldn't be necessary at all. We have
compatible flags to make these checks instead. How about below ?
(untested, uncompiled, yada yada yada). Note that we DON'T need this
xhci_plat_type trickery, just need to be a little bit smarter about how
we use driver_data:

Your solution certainly looks more elegant.

cool thanks. Now that I think about this more carefully, we might wanna
take $subject anyway for the -rc and get my version applied for v4.7
merge window. What do you think ?

quoted

diff --git a/drivers/usb/host/xhci-mvebu.c b/drivers/usb/host/xhci-mvebu.c
index 1eefc988192d..1ea6c18b74f3 100644
--- a/drivers/usb/host/xhci-mvebu.c
+++ b/drivers/usb/host/xhci-mvebu.c

@@ -41,8 +41,9 @@ static void xhci_mvebu_mbus_config(void __iomem *base,
 	}
 }
 
-int xhci_mvebu_mbus_init_quirk(struct platform_device *pdev)
+int xhci_mvebu_mbus_init_quirk(struct usb_hcd *hcd)
 {
+	struct platform_device *pdev = to_platform_device(hcd->self.controller);
 	struct resource	*res;
 	void __iomem *base;
 	const struct mbus_dram_target_info *dram;

diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c
index 5c15e9bc5f7a..adb77c60a9ae 100644
--- a/drivers/usb/host/xhci-plat.c
+++ b/drivers/usb/host/xhci-plat.c

@@ -47,43 +47,56 @@ static void xhci_plat_quirks(struct device *dev, struct xhci_hcd *xhci)
 	xhci->quirks |= XHCI_PLAT;
 }
 
+static void xhci_priv_plat_start(struct usb_hcd *hcd)
+{
+	struct xhci_plat_priv *priv = hcd_to_xhci_priv(hcd);
+
+	if (priv->plat_start)
+		priv->plat_start(hcd);
+}
+
+static int xhci_priv_init_quirk(struct usb_hcd *hcd)
+{
+	struct xhci_plat_priv *priv = hcd_to_xhci_priv(hcd);
+
+	if (!priv->init_quirk)
+		return 0;
+
+	return priv->init_quirk(hcd);
+}
+
 /* called during probe() after chip reset completes */
 static int xhci_plat_setup(struct usb_hcd *hcd)
 {
 	int ret;
 
-	if (xhci_plat_type_is(hcd, XHCI_PLAT_TYPE_RENESAS_RCAR_GEN2) ||
-	    xhci_plat_type_is(hcd, XHCI_PLAT_TYPE_RENESAS_RCAR_GEN3)) {
-		ret = xhci_rcar_init_quirk(hcd);
-		if (ret)
-			return ret;
-	}
+	ret = xhci_priv_init_quirk(hcd);
+	if (ret)
+		return ret;
 
 	return xhci_gen_setup(hcd, xhci_plat_quirks);
 }
 
 static int xhci_plat_start(struct usb_hcd *hcd)
 {
-	if (xhci_plat_type_is(hcd, XHCI_PLAT_TYPE_RENESAS_RCAR_GEN2) ||
-	    xhci_plat_type_is(hcd, XHCI_PLAT_TYPE_RENESAS_RCAR_GEN3))
-		xhci_rcar_start(hcd);
-
+	xhci_priv_plat_start(hcd);
 	return xhci_run(hcd);
 }
 
 #ifdef CONFIG_OF
 static const struct xhci_plat_priv xhci_plat_marvell_armada = {
-	.type = XHCI_PLAT_TYPE_MARVELL_ARMADA,
+	.init_quirk = xhci_mvebu_mbus_init_quirk,
 };
 
 static const struct xhci_plat_priv xhci_plat_renesas_rcar_gen2 = {
-	.type = XHCI_PLAT_TYPE_RENESAS_RCAR_GEN2,
 	.firmware_name = XHCI_RCAR_FIRMWARE_NAME_V1,
+	.init_quirk = xhci_rcar_init_quirk,
 };
 
 static const struct xhci_plat_priv xhci_plat_renesas_rcar_gen3 = {
-	.type = XHCI_PLAT_TYPE_RENESAS_RCAR_GEN3,
 	.firmware_name = XHCI_RCAR_FIRMWARE_NAME_V2,
+	.init_quirk = xhci_rcar_init_quirk,
+	.plat_start = xhci_rcar_start,
 };
 
 static const struct of_device_id usb_xhci_of_match[] = {

diff --git a/drivers/usb/host/xhci-plat.h b/drivers/usb/host/xhci-plat.h
index 5a2e2e3936c4..c4d565980832 100644
--- a/drivers/usb/host/xhci-plat.h
+++ b/drivers/usb/host/xhci-plat.h

@@ -13,27 +13,13 @@
 
 #include "xhci.h"	/* for hcd_to_xhci() */
 
-enum xhci_plat_type {
-	XHCI_PLAT_TYPE_MARVELL_ARMADA,
-	XHCI_PLAT_TYPE_RENESAS_RCAR_GEN2,
-	XHCI_PLAT_TYPE_RENESAS_RCAR_GEN3,
-};
-
 struct xhci_plat_priv {
 	enum xhci_plat_type type;
 	const char *firmware_name;
+	void (*plat_start)(struct usb_hcd *);
+	int (*init_quirk)(struct usb_hcd *);
 };
 
 #define hcd_to_xhci_priv(h) ((struct xhci_plat_priv *)hcd_to_xhci(h)->priv)
 
-static inline bool xhci_plat_type_is(struct usb_hcd *hcd,
-				     enum xhci_plat_type type)
-{
-	struct xhci_plat_priv *priv = hcd_to_xhci_priv(hcd);
-
-	if (priv && priv->type == type)
-		return true;
-	else
-		return false;
-}
 #endif	/* _XHCI_PLAT_H */

ps: there might be bugs there, but it's a holiday and I really shouldn't
be spending time on this right now ;-)

I'm also off on holiday now until Sunday 10th April... yay :-)

heh, cool :-)

quoted

Anyway, have fun testing. Let me know if it doesn't work.

I only have access to STi platforms which were broken by this change.
Not any of the platforms which rely on the functionality which
was introduced (although I can't see any reason why your patch wouldn't work).

Maybe Yoshihiro (on CC) could test this on the Renesas platforms and
confirm?

sure, that would be great; then we avoid further regressions ;-)

-- 
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 818 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20160326/5f3fc385/attachment-0001.sig>