Stefan Wahren reports a problem with a warning fix that was merged for v4.15: we had lots of device nodes with a 'phys' property pointing to a device node that is not compliant with the binding documented in Documentation/devicetree/bindings/phy/phy-bindings.txt
This generally works because USB HCD drivers that support both the generic phy subsystem and the older usb-phy subsystem ignore most errors from phy_get() and related calls and then use the usb-phy driver instead.
However, usb_add_hcd() (along with the respective functions in dwc2 and dwc3) propagate the EPROBE_DEFER return code so we can try again whenever the driver gets loaded. In case the driver is written for the usb-phy subsystem (like usb-generic-phy aka usb-nop-xceiv), we will never load a generic-phy driver for it, and keep failing here.
There is only a small number of remaining usb-phy drivers that support device tree, so this adds a workaround by providing a full list of the potentially affected drivers, and always failing the probe with -ENODEV here, which is the same behavior that we used to get with incorrect device tree files. Since we generally want older kernels to also want to work with the fixed devicetree files, it would be good to backport the patch into stable kernels as well (3.13+ are possibly affected). Reverting back to the DTS sources that work would in theory fix USB support for now, but in the long run we'd run into the same problem again when the drivers get ported from usb-phy to generic-phy.
Fixes: 014d6da6cb25 ("ARM: dts: bcm283x: Fix DTC warnings about missing phy-cells") Link: https://marc.info/?l=linux-usb&m=151518314314753&w=2 Cc: stable@vger.kernel.org Cc: Stefan Wahren stefan.wahren@i2se.com Cc: Felipe Balbi balbi@kernel.org Signed-off-by: Arnd Bergmann arnd@arndb.de --- This obviously needs to be tested, I wrote this up as a reply to Stefan's bug report. I'm fairly sure that I covered all usb-phy driver strings here. My goal is to have a fix merged into 4.15 rather than reverting all the DT fixes. --- drivers/phy/phy-core.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c index b4964b067aec..bb4dd2a2de2d 100644 --- a/drivers/phy/phy-core.c +++ b/drivers/phy/phy-core.c @@ -387,6 +387,24 @@ int phy_calibrate(struct phy *phy) } EXPORT_SYMBOL_GPL(phy_calibrate);
+static struct of_device_id __maybe_unused legacy_usbphy[] = { + { .compatible = "fsl,imx23-usbphy" }, + { .compatible = "fsl,imx6q-usbphy" }, + { .compatible = "fsl,imx6sl-usbphy" }, + { .compatible = "fsl,imx6sx-usbphy" }, + { .compatible = "fsl,imx6ul-usbphy" }, + { .compatible = "fsl,vf610-usbphy" }, + { .compatible = "nvidia,tegra20-usb-phy" }, + { .compatible = "nvidia,tegra30-usb-phy" }, + { .compatible = "nxp,isp1301" }, + { .compatible = "ti,am335x-usb-ctrl-module" }, + { .compatible = "ti,am335x-usb-phy" }, + { .compatible = "ti,keystone-usbphy" }, + { .compatible = "ti,twl6030-usb" }, + { .compatible = "usb-nop-xceiv" }, + {}, +}; + /** * _of_phy_get() - lookup and obtain a reference to a phy by phandle * @np: device_node for which to get the phy @@ -410,6 +428,15 @@ static struct phy *_of_phy_get(struct device_node *np, int index) if (ret) return ERR_PTR(-ENODEV);
+ /* + * Some USB host controllers use a "phys" property to refer to + * a device that does not have a generic phy driver but that + * has a driver for the older usb-phy framework. + * We must not return -EPROBE_DEFER for those, so bail out early. + */ + if (of_match_node(legacy_usbphy, args.np)) + return ERR_PTR(-ENODEV); + mutex_lock(&phy_provider_mutex); phy_provider = of_phy_provider_lookup(args.np); if (IS_ERR(phy_provider) || !try_module_get(phy_provider->owner)) {
Hi Arnd,
On Monday 08 January 2018 06:31 PM, Arnd Bergmann wrote:
Stefan Wahren reports a problem with a warning fix that was merged for v4.15: we had lots of device nodes with a 'phys' property pointing to a device node that is not compliant with the binding documented in Documentation/devicetree/bindings/phy/phy-bindings.txt
This generally works because USB HCD drivers that support both the generic phy subsystem and the older usb-phy subsystem ignore most errors from phy_get() and related calls and then use the usb-phy driver instead.
However, usb_add_hcd() (along with the respective functions in dwc2 and dwc3) propagate the EPROBE_DEFER return code so we can try again whenever the driver gets loaded. In case the driver is written for the usb-phy subsystem (like usb-generic-phy aka usb-nop-xceiv), we will never load a generic-phy driver for it, and keep failing here.
There is only a small number of remaining usb-phy drivers that support device tree, so this adds a workaround by providing a full list of the potentially affected drivers, and always failing the probe with -ENODEV here, which is the same behavior that we used to get with incorrect device tree files. Since we generally want older kernels to also want to work with the fixed devicetree files, it would be good to backport the patch into stable kernels as well (3.13+ are possibly affected). Reverting back to the DTS sources that work would in theory fix USB support for now, but in the long run we'd run into the same problem again when the drivers get ported from usb-phy to generic-phy.
Fixes: 014d6da6cb25 ("ARM: dts: bcm283x: Fix DTC warnings about missing phy-cells") Link: https://marc.info/?l=linux-usb&m=151518314314753&w=2 Cc: stable@vger.kernel.org Cc: Stefan Wahren stefan.wahren@i2se.com Cc: Felipe Balbi balbi@kernel.org Signed-off-by: Arnd Bergmann arnd@arndb.de
This obviously needs to be tested, I wrote this up as a reply to Stefan's bug report. I'm fairly sure that I covered all usb-phy driver strings here. My goal is to have a fix merged into 4.15 rather than reverting all the DT fixes.
Shouldn't the fix be in phy consumer drivers to not return error if it's able to find the phy either using usb-phy or generic phy?
drivers/phy/phy-core.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c index b4964b067aec..bb4dd2a2de2d 100644 --- a/drivers/phy/phy-core.c +++ b/drivers/phy/phy-core.c @@ -387,6 +387,24 @@ int phy_calibrate(struct phy *phy) } EXPORT_SYMBOL_GPL(phy_calibrate); +static struct of_device_id __maybe_unused legacy_usbphy[] = {
- { .compatible = "fsl,imx23-usbphy" },
- { .compatible = "fsl,imx6q-usbphy" },
- { .compatible = "fsl,imx6sl-usbphy" },
- { .compatible = "fsl,imx6sx-usbphy" },
- { .compatible = "fsl,imx6ul-usbphy" },
- { .compatible = "fsl,vf610-usbphy" },
- { .compatible = "nvidia,tegra20-usb-phy" },
- { .compatible = "nvidia,tegra30-usb-phy" },
- { .compatible = "nxp,isp1301" },
- { .compatible = "ti,am335x-usb-ctrl-module" },
- { .compatible = "ti,am335x-usb-phy" },
- { .compatible = "ti,keystone-usbphy" },
- { .compatible = "ti,twl6030-usb" },
- { .compatible = "usb-nop-xceiv" },
- {},
"ti,am335x-usb-ctrl-module" and "ti,twl6030-usb" are not phys.
Thanks Kishon
On Mon, Jan 8, 2018 at 7:32 PM, Kishon Vijay Abraham I kishon@ti.com wrote:
Hi Arnd,
On Monday 08 January 2018 06:31 PM, Arnd Bergmann wrote:
Stefan Wahren reports a problem with a warning fix that was merged for v4.15: we had lots of device nodes with a 'phys' property pointing to a device node that is not compliant with the binding documented in Documentation/devicetree/bindings/phy/phy-bindings.txt
This generally works because USB HCD drivers that support both the generic phy subsystem and the older usb-phy subsystem ignore most errors from phy_get() and related calls and then use the usb-phy driver instead.
However, usb_add_hcd() (along with the respective functions in dwc2 and dwc3) propagate the EPROBE_DEFER return code so we can try again whenever the driver gets loaded. In case the driver is written for the usb-phy subsystem (like usb-generic-phy aka usb-nop-xceiv), we will never load a generic-phy driver for it, and keep failing here.
There is only a small number of remaining usb-phy drivers that support device tree, so this adds a workaround by providing a full list of the potentially affected drivers, and always failing the probe with -ENODEV here, which is the same behavior that we used to get with incorrect device tree files. Since we generally want older kernels to also want to work with the fixed devicetree files, it would be good to backport the patch into stable kernels as well (3.13+ are possibly affected). Reverting back to the DTS sources that work would in theory fix USB support for now, but in the long run we'd run into the same problem again when the drivers get ported from usb-phy to generic-phy.
Fixes: 014d6da6cb25 ("ARM: dts: bcm283x: Fix DTC warnings about missing phy-cells") Link: https://marc.info/?l=linux-usb&m=151518314314753&w=2 Cc: stable@vger.kernel.org Cc: Stefan Wahren stefan.wahren@i2se.com Cc: Felipe Balbi balbi@kernel.org Signed-off-by: Arnd Bergmann arnd@arndb.de
This obviously needs to be tested, I wrote this up as a reply to Stefan's bug report. I'm fairly sure that I covered all usb-phy driver strings here. My goal is to have a fix merged into 4.15 rather than reverting all the DT fixes.
Shouldn't the fix be in phy consumer drivers to not return error if it's able to find the phy either using usb-phy or generic phy?
Stefan has posted a patch to that effect now, but I fear that might be a little fragile, in particular this short before the release with the regression in place.
The main problem is that we'd have to change the generic usb_add_hcd() function in addition to dwc2 and dwc3 to ignore -EPROBE_DEFER from phy_get() whenever usb_get_phy_dev() has already succeeded.
If there is any HCD that relies on usb_add_hcd() to get both the usb_phy and the phy structures, and it may need to defer probing when the latter one isn't ready yet, that fix would break another driver.
diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c index b4964b067aec..bb4dd2a2de2d 100644 --- a/drivers/phy/phy-core.c +++ b/drivers/phy/phy-core.c @@ -387,6 +387,24 @@ int phy_calibrate(struct phy *phy) } EXPORT_SYMBOL_GPL(phy_calibrate);
+static struct of_device_id __maybe_unused legacy_usbphy[] = {
{ .compatible = "fsl,imx23-usbphy" },
{ .compatible = "fsl,imx6q-usbphy" },
{ .compatible = "fsl,imx6sl-usbphy" },
{ .compatible = "fsl,imx6sx-usbphy" },
{ .compatible = "fsl,imx6ul-usbphy" },
{ .compatible = "fsl,vf610-usbphy" },
{ .compatible = "nvidia,tegra20-usb-phy" },
{ .compatible = "nvidia,tegra30-usb-phy" },
{ .compatible = "nxp,isp1301" },
{ .compatible = "ti,am335x-usb-ctrl-module" },
{ .compatible = "ti,am335x-usb-phy" },
{ .compatible = "ti,keystone-usbphy" },
{ .compatible = "ti,twl6030-usb" },
{ .compatible = "usb-nop-xceiv" },
{},
"ti,am335x-usb-ctrl-module" and "ti,twl6030-usb" are not phys.
Ok, I see.
Arnd
Hi Arnd,
On Thursday 11 January 2018 02:27 AM, Arnd Bergmann wrote:
On Mon, Jan 8, 2018 at 7:32 PM, Kishon Vijay Abraham I kishon@ti.com wrote:
Hi Arnd,
On Monday 08 January 2018 06:31 PM, Arnd Bergmann wrote:
Stefan Wahren reports a problem with a warning fix that was merged for v4.15: we had lots of device nodes with a 'phys' property pointing to a device node that is not compliant with the binding documented in Documentation/devicetree/bindings/phy/phy-bindings.txt
This generally works because USB HCD drivers that support both the generic phy subsystem and the older usb-phy subsystem ignore most errors from phy_get() and related calls and then use the usb-phy driver instead.
However, usb_add_hcd() (along with the respective functions in dwc2 and dwc3) propagate the EPROBE_DEFER return code so we can try again whenever the driver gets loaded. In case the driver is written for the usb-phy subsystem (like usb-generic-phy aka usb-nop-xceiv), we will never load a generic-phy driver for it, and keep failing here.
There is only a small number of remaining usb-phy drivers that support device tree, so this adds a workaround by providing a full list of the potentially affected drivers, and always failing the probe with -ENODEV here, which is the same behavior that we used to get with incorrect device tree files. Since we generally want older kernels to also want to work with the fixed devicetree files, it would be good to backport the patch into stable kernels as well (3.13+ are possibly affected). Reverting back to the DTS sources that work would in theory fix USB support for now, but in the long run we'd run into the same problem again when the drivers get ported from usb-phy to generic-phy.
Fixes: 014d6da6cb25 ("ARM: dts: bcm283x: Fix DTC warnings about missing phy-cells") Link: https://marc.info/?l=linux-usb&m=151518314314753&w=2 Cc: stable@vger.kernel.org Cc: Stefan Wahren stefan.wahren@i2se.com Cc: Felipe Balbi balbi@kernel.org Signed-off-by: Arnd Bergmann arnd@arndb.de
This obviously needs to be tested, I wrote this up as a reply to Stefan's bug report. I'm fairly sure that I covered all usb-phy driver strings here. My goal is to have a fix merged into 4.15 rather than reverting all the DT fixes.
Shouldn't the fix be in phy consumer drivers to not return error if it's able to find the phy either using usb-phy or generic phy?
Stefan has posted a patch to that effect now, but I fear that might be a little fragile, in particular this short before the release with the regression in place.
The main problem is that we'd have to change the generic usb_add_hcd() function in addition to dwc2 and dwc3 to ignore -EPROBE_DEFER from phy_get() whenever usb_get_phy_dev() has already succeeded.
If there is any HCD that relies on usb_add_hcd() to get both the usb_phy and the phy structures, and it may need to defer probing when the latter one isn't ready yet, that fix would break another driver.
hmm.. IMO the better thing right now would be to revert the dt patch which adds #phy-cells. We have to see if there are better fixes in order to add #phy-cells warning fix in stable tree.
Thanks Kishon
On Thu, Jan 11, 2018 at 2:30 PM, Kishon Vijay Abraham I kishon@ti.com wrote:
On Thursday 11 January 2018 02:27 AM, Arnd Bergmann wrote:
On Mon, Jan 8, 2018 at 7:32 PM, Kishon Vijay Abraham I kishon@ti.com wrote:
On Monday 08 January 2018 06:31 PM, Arnd Bergmann wrote:
Stefan Wahren reports a problem with a warning fix that was merged
This obviously needs to be tested, I wrote this up as a reply to Stefan's bug report. I'm fairly sure that I covered all usb-phy driver strings here. My goal is to have a fix merged into 4.15 rather than reverting all the DT fixes.
Shouldn't the fix be in phy consumer drivers to not return error if it's able to find the phy either using usb-phy or generic phy?
Stefan has posted a patch to that effect now, but I fear that might be a little fragile, in particular this short before the release with the regression in place.
The main problem is that we'd have to change the generic usb_add_hcd() function in addition to dwc2 and dwc3 to ignore -EPROBE_DEFER from phy_get() whenever usb_get_phy_dev() has already succeeded.
If there is any HCD that relies on usb_add_hcd() to get both the usb_phy and the phy structures, and it may need to defer probing when the latter one isn't ready yet, that fix would break another driver.
hmm.. IMO the better thing right now would be to revert the dt patch which adds #phy-cells. We have to see if there are better fixes in order to add #phy-cells warning fix in stable tree.
Let's see which patches that would be, I think this is the full list of nodes that got an extra #phy-cells:
c22fe696157d ARM: dts: Fix dm814x missing phy-cells property f0e11ff8ff65 ARM: dts: am33xx: Add missing #phy-cells to ti,am335x-usb-phy c5bbf358b790 arm: dts: nspire: Add missing #phy-cells to usb-nop-xceiv 44e5dced2ef6 arm: dts: marvell: Add missing #phy-cells to usb-nop-xceiv 014d6da6cb25 ARM: dts: bcm283x: Fix DTC warnings about missing phy-cells f568f6f554b8 ARM: dts: omap: Add missing #phy-cells to usb-nop-xceiv
plus a couple in linux-next:
d745d5f277bf ARM: dts: imx51-zii-rdu1: Add missing #phy-cells to usb-nop-xceiv 915fbe59cbf2 ARM: dts: imx: Add missing #phy-cells to usb-nop-xceiv
It's a lot of patches to revert, and I guess it would get us back to hundreds of warnings in an allmodconfig build, so I'd first try to come up with ways to prove that at least some of them can stay.
Almost all the warnings are about "usb-nop-xceiv" phys, the only exceptions I could find are the OMAP ones (the first two patches), which use "ti,am335x-usb-phy" and are referenced from a "ti,musb-am33xx". That particular driver is not affected by the bug, so we can leave that in.
To deal with all the "usb-nop-xceiv" references including the one that Stefan reported, we could use a much simpler version of my earlier patch, do you think this is any better?
Signed-off-by: Arnd Bergmann arnd@arndb.de
diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c index b4964b067aec..f056d8fb3921 100644 --- a/drivers/phy/phy-core.c +++ b/drivers/phy/phy-core.c @@ -410,6 +410,10 @@ static struct phy *_of_phy_get(struct device_node *np, int index) if (ret) return ERR_PTR(-ENODEV);
+ /* This phy type handled by the usb-phy subsystem for now */ + if (of_device_is_compatible("usb-nop-xceiv")) + return ERR_PTR(-ENODEV); + mutex_lock(&phy_provider_mutex); phy_provider = of_phy_provider_lookup(args.np); if (IS_ERR(phy_provider) || !try_module_get(phy_provider->owner)) {
linux-stable-mirror@lists.linaro.org