[PATCH v2] usb: typec: tipd: Don't block probing of consumer of "connector" nodes

Similar as with tcpm this patch lets fw_devlink know not to wait on the
fwnode to be populated as a struct device.

Without this patch, USB functionality can be broken on some previously
supported boards.

Fixes: 28ec344bb891 ("usb: typec: tcpm: Don't block probing of consumers of "connector" nodes")
Signed-off-by: Martin Kepplinger <redacted>
---

revision history
----------------
v2: (thank you Saravana)
* add a code-comment why the call is needed.

v1:
https://lore.kernel.org/linux-usb/20210713073946.102501-1-martin.kepplinger@puri.sm/



 drivers/usb/typec/tipd/core.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c
index 938219bc1b4b..21b3ae25c76d 100644
--- a/drivers/usb/typec/tipd/core.c
+++ b/drivers/usb/typec/tipd/core.c

@@ -629,6 +629,15 @@ static int tps6598x_probe(struct i2c_client *client)
 	if (!fwnode)
 		return -ENODEV;
 
+	/*
+	 * This fwnode has a "compatible" property, but is never populated as a
+	 * struct device. Instead we simply parse it to read the properties.
+	 * This breaks fw_devlink=on. To maintain backward compatibility
+	 * with existing DT files, we work around this by deleting any
+	 * fwnode_links to/from this fwnode.
+	 */
+	fw_devlink_purge_absent_suppliers(fwnode);
+
 	tps->role_sw = fwnode_usb_role_switch_get(fwnode);
 	if (IS_ERR(tps->role_sw)) {
 		ret = PTR_ERR(tps->role_sw);

-- 
2.30.2

Wed, Jul 14, 2021 at 08:18:07AM +0200, Martin Kepplinger kirjoitti:

Similar as with tcpm this patch lets fw_devlink know not to wait on the
fwnode to be populated as a struct device.

Without this patch, USB functionality can be broken on some previously
supported boards.

Fixes: 28ec344bb891 ("usb: typec: tcpm: Don't block probing of consumers of "connector" nodes")
Signed-off-by: Martin Kepplinger <redacted>

Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>

-- 
heikki

On Tue, Jul 13, 2021 at 11:18 PM Martin Kepplinger
[off-list ref] wrote:

Hey Martin,

As part of a series I'm working on, I'm looking into deleting
fw_devlink_purge_absent_suppliers() and having the driver core figure
this out automatically.

So I was making sure all the current uses of
fw_devlink_purge_absent_suppliers() would automatically be handled by
driver core. But when I looked at this usage of
fw_devlink_purge_absent_suppliers(), I didn't see any "compatible"
property in the "connector" child node. So, I'm confused how you had a
problem without this patch.

fw_devlink would have created fwnode links
between typec_pd and usb_dwc3_0, figured out it's a cycle and have
stopped enforcing the "remote-endpoint" dependencies.

Can you give me more details please on why you were having a problem
without this patch?

I'm looking at arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi as an
example because you listed it in your v1 patch.

-Saravana

        tps->role_sw = fwnode_usb_role_switch_get(fwnode);
        if (IS_ERR(tps->role_sw)) {
                ret = PTR_ERR(tps->role_sw);
--
2.30.2

Am Donnerstag, dem 21.07.2022 um 16:06 -0700 schrieb Saravana Kannan:

On Tue, Jul 13, 2021 at 11:18 PM Martin Kepplinger
[off-list ref] wrote:

quoted

Similar as with tcpm this patch lets fw_devlink know not to wait on
the
fwnode to be populated as a struct device.

Without this patch, USB functionality can be broken on some
previously
supported boards.

Fixes: 28ec344bb891 ("usb: typec: tcpm: Don't block probing of
consumers of "connector" nodes")
Signed-off-by: Martin Kepplinger <redacted>
---

revision history
----------------
v2: (thank you Saravana)
* add a code-comment why the call is needed.

v1:
https://lore.kernel.org/linux-usb/20210713073946.102501-1-martin.kepplinger@puri.sm/



 drivers/usb/typec/tipd/core.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/usb/typec/tipd/core.c

b/drivers/usb/typec/tipd/core.c
index 938219bc1b4b..21b3ae25c76d 100644

--- a/drivers/usb/typec/tipd/core.c
+++ b/drivers/usb/typec/tipd/core.c

@@ -629,6 +629,15 @@ static int tps6598x_probe(struct i2c_client

*client)
        if (!fwnode)
                return -ENODEV;

+       /*
+        * This fwnode has a "compatible" property, but is never
populated as a
+        * struct device. Instead we simply parse it to read the
properties.
+        * This breaks fw_devlink=on. To maintain backward
compatibility
+        * with existing DT files, we work around this by deleting
any
+        * fwnode_links to/from this fwnode.
+        */
+       fw_devlink_purge_absent_suppliers(fwnode);
+

Hey Martin,

As part of a series I'm working on, I'm looking into deleting
fw_devlink_purge_absent_suppliers() and having the driver core figure
this out automatically.

So I was making sure all the current uses of
fw_devlink_purge_absent_suppliers() would automatically be handled by
driver core. But when I looked at this usage of
fw_devlink_purge_absent_suppliers(), I didn't see any "compatible"
property in the "connector" child node. So, I'm confused how you had
a
problem without this patch.

fw_devlink would have created fwnode links
between typec_pd and usb_dwc3_0, figured out it's a cycle and have
stopped enforcing the "remote-endpoint" dependencies.

Can you give me more details please on why you were having a problem
without this patch?

I'm looking at arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi as
an
example because you listed it in your v1 patch.

-Saravana

It's downstream. I'm very sorry about that. The patch that adds the
compatible string (
https://source.puri.sm/martin.kepplinger/linux-next/-/commit/3d887ea76158851ae50f192094d7865cc3f4da9d
) sits in our tree (
https://source.puri.sm/martin.kepplinger/linux-next/-/commits/5.19-rc7/librem5__integration
) and I'll make sure it'll be sent out after I'm back from vacations in
a few weeks.

Sorry for the confusion!

                             martin

quoted

        tps->role_sw = fwnode_usb_role_switch_get(fwnode);
        if (IS_ERR(tps->role_sw)) {
                ret = PTR_ERR(tps->role_sw);
--
2.30.2