On 7/23/21 10:18 AM, Kyle Tso wrote:
On Fri, Jul 16, 2021 at 5:39 PM Kyle Tso <kyletso@xxxxxxxxxx> wrote:
tcpm.c could work well without PD capabilities. Do not block the probe
"could" is a bit vague. What is the use case ?
if capabilities are not defined in fwnode and skip the PD power
negotiation in the state machine.
Signed-off-by: Kyle Tso <kyletso@xxxxxxxxxx>
---
Hi, any comments about this patch?
First question would be if this is documented/standardized. More comments below.
thanks,
Kyle
drivers/usb/typec/tcpm/tcpm.c | 50 ++++++++++++++++++++---------------
1 file changed, 29 insertions(+), 21 deletions(-)
diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c
index 5b22a1c931a9..a42de5e17d24 100644
--- a/drivers/usb/typec/tcpm/tcpm.c
+++ b/drivers/usb/typec/tcpm/tcpm.c
@@ -3914,6 +3914,8 @@ static void run_state_machine(struct tcpm_port *port)
if (port->ams == POWER_ROLE_SWAP ||
port->ams == FAST_ROLE_SWAP)
tcpm_ams_finish(port);
+ if (!port->nr_src_pdo)
+ tcpm_set_state(port, SRC_READY, 0);
port->upcoming_state = SRC_SEND_CAPABILITIES;
tcpm_ams_start(port, POWER_NEGOTIATION);
break;
@@ -4161,7 +4163,10 @@ static void run_state_machine(struct tcpm_port *port)
current_lim = PD_P_SNK_STDBY_MW / 5;
tcpm_set_current_limit(port, current_lim, 5000);
tcpm_set_charge(port, true);
- tcpm_set_state(port, SNK_WAIT_CAPABILITIES, 0);
+ if (!port->nr_snk_pdo)
+ tcpm_set_state(port, SNK_READY, 0);
+ else
+ tcpm_set_state(port, SNK_WAIT_CAPABILITIES, 0);
break;
}
/*
@@ -5939,15 +5944,17 @@ static int tcpm_fw_get_caps(struct tcpm_port *port,
/* Get source pdos */
ret = fwnode_property_count_u32(fwnode, "source-pdos");
- if (ret <= 0)
- return -EINVAL;
+ if (ret < 0)
+ ret = 0;
This makes the capability properties optional. I think that would have
to be documented.
port->nr_src_pdo = min(ret, PDO_MAX_OBJECTS);
- ret = fwnode_property_read_u32_array(fwnode, "source-pdos",
- port->src_pdo, port->nr_src_pdo);
- if ((ret < 0) || tcpm_validate_caps(port, port->src_pdo,
- port->nr_src_pdo))
- return -EINVAL;
+ if (port->nr_src_pdo) {
+ ret = fwnode_property_read_u32_array(fwnode, "source-pdos",
+ port->src_pdo, port->nr_src_pdo);
+ if ((ret < 0) || tcpm_validate_caps(port, port->src_pdo,
+ port->nr_src_pdo))
+ return -EINVAL;
+ }
if (port->port_type == TYPEC_PORT_SRC)
return 0;
@@ -5963,19 +5970,21 @@ static int tcpm_fw_get_caps(struct tcpm_port *port,
sink:
/* Get sink pdos */
ret = fwnode_property_count_u32(fwnode, "sink-pdos");
- if (ret <= 0)
- return -EINVAL;
+ if (ret < 0)
+ ret = 0;
port->nr_snk_pdo = min(ret, PDO_MAX_OBJECTS);
- ret = fwnode_property_read_u32_array(fwnode, "sink-pdos",
- port->snk_pdo, port->nr_snk_pdo);
- if ((ret < 0) || tcpm_validate_caps(port, port->snk_pdo,
- port->nr_snk_pdo))
- return -EINVAL;
+ if (port->nr_snk_pdo) {
+ ret = fwnode_property_read_u32_array(fwnode, "sink-pdos",
+ port->snk_pdo, port->nr_snk_pdo);
+ if ((ret < 0) || tcpm_validate_caps(port, port->snk_pdo,
+ port->nr_snk_pdo))
+ return -EINVAL;
- if (fwnode_property_read_u32(fwnode, "op-sink-microwatt", &mw) < 0)
- return -EINVAL;
- port->operating_snk_mw = mw / 1000;
+ if (fwnode_property_read_u32(fwnode, "op-sink-microwatt", &mw) < 0)
+ return -EINVAL;
+ port->operating_snk_mw = mw / 1000;
+ }
port->self_powered = fwnode_property_read_bool(fwnode, "self-powered");
@@ -6283,9 +6292,8 @@ struct tcpm_port *tcpm_register_port(struct device *dev, struct tcpc_dev *tcpc)
int err;
if (!dev || !tcpc ||
- !tcpc->get_vbus || !tcpc->set_cc || !tcpc->get_cc ||
- !tcpc->set_polarity || !tcpc->set_vconn || !tcpc->set_vbus ||
- !tcpc->set_pd_rx || !tcpc->set_roles || !tcpc->pd_transmit)
+ !tcpc->get_vbus || !tcpc->set_cc || !tcpc->get_cc || !tcpc->set_polarity ||
+ !tcpc->set_vconn || !tcpc->set_vbus || !tcpc->set_roles)
With this change, if a driver does not define the necessary pd callbacks
(set_pd_rx, pd_transmit), but its devicetree data does provide pdo properties,
we'll get a nice crash.
On top of that, I am quite sure that the set_pd_rx() callback is still called
from various places even if neither sink-pdos nor source-pdos properties
are defined.
I think this really tries to handle two conditions: A low level driver that
doesn't support PD, and a system where the low level driver does support PD
but the system itself doesn't. And then there is the odd case where the system
only supports either source or sink PD while claiming to support the other.
Maybe it would make sense to separate both conditions, for example by introducing
a new flag such as pd_supported to indicate that the system doesn't support the
PD protocol.
Guenter
return ERR_PTR(-EINVAL);
port = devm_kzalloc(dev, sizeof(*port), GFP_KERNEL);
--
2.32.0.402.g57bb445576-goog
