| Message ID | 1649759407-24049-4-git-send-email-quic_linyyuan@quicinc.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | usb: typec: ucsi: allow retry to find role switch | expand |
On Tue, Apr 12, 2022 at 06:30:07PM +0800, Linyu Yuan wrote: > When role switch enabled, return -EAGAIN if fail to find it due to > module load ordering issue, then restart ucsi init work to find > it again every 100ms. > > It also means change ucsi init work to delayed_work. > > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> > --- > drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ > drivers/usb/typec/ucsi/ucsi.h | 2 +- > 2 files changed, 17 insertions(+), 13 deletions(-) > > diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c > index 684fc4f..fabd4df 100644 > --- a/drivers/usb/typec/ucsi/ucsi.c > +++ b/drivers/usb/typec/ucsi/ucsi.c > @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, int index) > con->num = index + 1; > con->ucsi = ucsi; > > + cap->fwnode = ucsi_find_fwnode(con); > + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > + if (IS_ERR(con->usb_role_sw)) { > + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > + index + 1); > + return -EAGAIN; > + } No, this is wrong. If there is a role switch assigned in the firmware for the port, then fwnode_usb_role_switch_get() will return -EPROBE_DEFER until the switch has been registered. So I think what you want to do is just make the failure to get the role switch fatal in this driver. Just return PTR_ERR(con->usb_role_sw). thanks,
> From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > Sent: Wednesday, April 13, 2022 3:46 PM > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > module load late > > On Tue, Apr 12, 2022 at 06:30:07PM +0800, Linyu Yuan wrote: > > When role switch enabled, return -EAGAIN if fail to find it due to > > module load ordering issue, then restart ucsi init work to find > > it again every 100ms. > > > > It also means change ucsi init work to delayed_work. > > > > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> > > --- > > drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ > > drivers/usb/typec/ucsi/ucsi.h | 2 +- > > 2 files changed, 17 insertions(+), 13 deletions(-) > > > > diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c > > index 684fc4f..fabd4df 100644 > > --- a/drivers/usb/typec/ucsi/ucsi.c > > +++ b/drivers/usb/typec/ucsi/ucsi.c > > @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, int > index) > > con->num = index + 1; > > con->ucsi = ucsi; > > > > + cap->fwnode = ucsi_find_fwnode(con); > > + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > > + if (IS_ERR(con->usb_role_sw)) { > > + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > > + index + 1); > > + return -EAGAIN; > > + } > > No, this is wrong. If there is a role switch assigned in the firmware > for the port, then fwnode_usb_role_switch_get() will return > -EPROBE_DEFER until the switch has been registered. > > So I think what you want to do is just make the failure to get the > role switch fatal in this driver. Just return PTR_ERR(con->usb_role_sw). Thanks, will change it to check -EPROBE_DEFER. > > thanks, > > -- > heikki
On Wed, Apr 13, 2022 at 08:00:35AM +0000, Linyu Yuan (QUIC) wrote: > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > Sent: Wednesday, April 13, 2022 3:46 PM > > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > > module load late > > > > On Tue, Apr 12, 2022 at 06:30:07PM +0800, Linyu Yuan wrote: > > > When role switch enabled, return -EAGAIN if fail to find it due to > > > module load ordering issue, then restart ucsi init work to find > > > it again every 100ms. > > > > > > It also means change ucsi init work to delayed_work. > > > > > > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> > > > --- > > > drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ > > > drivers/usb/typec/ucsi/ucsi.h | 2 +- > > > 2 files changed, 17 insertions(+), 13 deletions(-) > > > > > > diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c > > > index 684fc4f..fabd4df 100644 > > > --- a/drivers/usb/typec/ucsi/ucsi.c > > > +++ b/drivers/usb/typec/ucsi/ucsi.c > > > @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, int > > index) > > > con->num = index + 1; > > > con->ucsi = ucsi; > > > > > > + cap->fwnode = ucsi_find_fwnode(con); > > > + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > > > + if (IS_ERR(con->usb_role_sw)) { > > > + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > > > + index + 1); > > > + return -EAGAIN; > > > + } > > > > No, this is wrong. If there is a role switch assigned in the firmware > > for the port, then fwnode_usb_role_switch_get() will return > > -EPROBE_DEFER until the switch has been registered. > > > > So I think what you want to do is just make the failure to get the > > role switch fatal in this driver. Just return PTR_ERR(con->usb_role_sw). > Thanks, will change it to check -EPROBE_DEFER. No, you don't check that, you just return that. So you don't need that crazy polling mechanism where you queue a work until you get the resource. Just let the driver core handle the situation. thanks,
> From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > Sent: Wednesday, April 13, 2022 4:10 PM > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > module load late > > On Wed, Apr 13, 2022 at 08:00:35AM +0000, Linyu Yuan (QUIC) wrote: > > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > > Sent: Wednesday, April 13, 2022 3:46 PM > > > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > > > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > > > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > > > module load late > > > > > > On Tue, Apr 12, 2022 at 06:30:07PM +0800, Linyu Yuan wrote: > > > > When role switch enabled, return -EAGAIN if fail to find it due to > > > > module load ordering issue, then restart ucsi init work to find > > > > it again every 100ms. > > > > > > > > It also means change ucsi init work to delayed_work. > > > > > > > > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> > > > > --- > > > > drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ > > > > drivers/usb/typec/ucsi/ucsi.h | 2 +- > > > > 2 files changed, 17 insertions(+), 13 deletions(-) > > > > > > > > diff --git a/drivers/usb/typec/ucsi/ucsi.c > b/drivers/usb/typec/ucsi/ucsi.c > > > > index 684fc4f..fabd4df 100644 > > > > --- a/drivers/usb/typec/ucsi/ucsi.c > > > > +++ b/drivers/usb/typec/ucsi/ucsi.c > > > > @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, > int > > > index) > > > > con->num = index + 1; > > > > con->ucsi = ucsi; > > > > > > > > + cap->fwnode = ucsi_find_fwnode(con); > > > > + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > > > > + if (IS_ERR(con->usb_role_sw)) { > > > > + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > > > > + index + 1); > > > > + return -EAGAIN; > > > > + } > > > > > > No, this is wrong. If there is a role switch assigned in the firmware > > > for the port, then fwnode_usb_role_switch_get() will return > > > -EPROBE_DEFER until the switch has been registered. > > > > > > So I think what you want to do is just make the failure to get the > > > role switch fatal in this driver. Just return PTR_ERR(con->usb_role_sw). > > Thanks, will change it to check -EPROBE_DEFER. > > No, you don't check that, you just return that. Yes, I do like it. > > So you don't need that crazy polling mechanism where you queue a work > until you get the resource. Just let the driver core handle the > situation. The issue is when a UCSI implementation driver probe, it call ucsi_register(), But this function will not return -EPROBE_DEFER, It just queue a work to start connector discovery. > > thanks, > > -- > heikki
On Wed, Apr 13, 2022 at 08:16:53AM +0000, Linyu Yuan (QUIC) wrote: > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > Sent: Wednesday, April 13, 2022 4:10 PM > > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > > module load late > > > > On Wed, Apr 13, 2022 at 08:00:35AM +0000, Linyu Yuan (QUIC) wrote: > > > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > > > Sent: Wednesday, April 13, 2022 3:46 PM > > > > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com> > > > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; linux- > > > > usb@vger.kernel.org; Jack Pham (QUIC) <quic_jackp@quicinc.com> > > > > Subject: Re: [PATCH 3/3] usb: typec: ucsi: retry find role swithch when > > > > module load late > > > > > > > > On Tue, Apr 12, 2022 at 06:30:07PM +0800, Linyu Yuan wrote: > > > > > When role switch enabled, return -EAGAIN if fail to find it due to > > > > > module load ordering issue, then restart ucsi init work to find > > > > > it again every 100ms. > > > > > > > > > > It also means change ucsi init work to delayed_work. > > > > > > > > > > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> > > > > > --- > > > > > drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ > > > > > drivers/usb/typec/ucsi/ucsi.h | 2 +- > > > > > 2 files changed, 17 insertions(+), 13 deletions(-) > > > > > > > > > > diff --git a/drivers/usb/typec/ucsi/ucsi.c > > b/drivers/usb/typec/ucsi/ucsi.c > > > > > index 684fc4f..fabd4df 100644 > > > > > --- a/drivers/usb/typec/ucsi/ucsi.c > > > > > +++ b/drivers/usb/typec/ucsi/ucsi.c > > > > > @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, > > int > > > > index) > > > > > con->num = index + 1; > > > > > con->ucsi = ucsi; > > > > > > > > > > + cap->fwnode = ucsi_find_fwnode(con); > > > > > + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > > > > > + if (IS_ERR(con->usb_role_sw)) { > > > > > + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > > > > > + index + 1); > > > > > + return -EAGAIN; > > > > > + } > > > > > > > > No, this is wrong. If there is a role switch assigned in the firmware > > > > for the port, then fwnode_usb_role_switch_get() will return > > > > -EPROBE_DEFER until the switch has been registered. > > > > > > > > So I think what you want to do is just make the failure to get the > > > > role switch fatal in this driver. Just return PTR_ERR(con->usb_role_sw). > > > Thanks, will change it to check -EPROBE_DEFER. > > > > No, you don't check that, you just return that. > Yes, I do like it. > > > > So you don't need that crazy polling mechanism where you queue a work > > until you get the resource. Just let the driver core handle the > > situation. > The issue is when a UCSI implementation driver probe, it call ucsi_register(), > But this function will not return -EPROBE_DEFER, > It just queue a work to start connector discovery. Ah, right right. This is a library. But we should be able use wait_for_device_probe(). So if fwnode_usb_role_switch_get() returns -EPROBE_DEFER, you call wait_for_device_probe() and try again. Br,
> From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > Sent: Wednesday, April 13, 2022 6:24 PM > > > No, you don't check that, you just return that. > > Yes, I do like it. > > > > > > So you don't need that crazy polling mechanism where you queue a work > > > until you get the resource. Just let the driver core handle the > > > situation. > > The issue is when a UCSI implementation driver probe, it call ucsi_register(), > > But this function will not return -EPROBE_DEFER, > > It just queue a work to start connector discovery. > > Ah, right right. This is a library. But we should be able use > wait_for_device_probe(). > > So if fwnode_usb_role_switch_get() returns -EPROBE_DEFER, you call > wait_for_device_probe() and try again. Do you mean do as below, find_role_switch: con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); if (IS_ERR(con->usb_role_sw)) { dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", con->num); if (con->usb_role_sw == -EPROBE_DEFER) { wait_for_device_probe(); goto find_role_switch; } return PTR_ERR(con->usb_role_sw); } Seem wait_for_device_probe() will wait all drivers probe, Can we accept it ? > > Br, > > -- > heikki
On Wed, Apr 13, 2022 at 11:01:23AM +0000, Linyu Yuan (QUIC) wrote: > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > Sent: Wednesday, April 13, 2022 6:24 PM > > > > No, you don't check that, you just return that. > > > Yes, I do like it. > > > > > > > > So you don't need that crazy polling mechanism where you queue a work > > > > until you get the resource. Just let the driver core handle the > > > > situation. > > > The issue is when a UCSI implementation driver probe, it call ucsi_register(), > > > But this function will not return -EPROBE_DEFER, > > > It just queue a work to start connector discovery. > > > > Ah, right right. This is a library. But we should be able use > > wait_for_device_probe(). > > > > So if fwnode_usb_role_switch_get() returns -EPROBE_DEFER, you call > > wait_for_device_probe() and try again. > Do you mean do as below, > > find_role_switch: > con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > if (IS_ERR(con->usb_role_sw)) { > dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > con->num); > if (con->usb_role_sw == -EPROBE_DEFER) { > wait_for_device_probe(); > goto find_role_switch; > } > > return PTR_ERR(con->usb_role_sw); > } Yes, something like that. Perhaps you could just do that in a loop, and you should have a delay there. > Seem wait_for_device_probe() will wait all drivers probe, > Can we accept it ? That part is not a problem, but I'm not sure if it appropriate to call the function from drivers. Let me check if there is something else that we can do... thanks,
On Wed, Apr 13, 2022 at 02:19:49PM +0300, Heikki Krogerus wrote: > On Wed, Apr 13, 2022 at 11:01:23AM +0000, Linyu Yuan (QUIC) wrote: > > > From: Heikki Krogerus <heikki.krogerus@linux.intel.com> > > > Sent: Wednesday, April 13, 2022 6:24 PM > > > > > No, you don't check that, you just return that. > > > > Yes, I do like it. > > > > > > > > > > So you don't need that crazy polling mechanism where you queue a work > > > > > until you get the resource. Just let the driver core handle the > > > > > situation. > > > > The issue is when a UCSI implementation driver probe, it call ucsi_register(), > > > > But this function will not return -EPROBE_DEFER, > > > > It just queue a work to start connector discovery. > > > > > > Ah, right right. This is a library. But we should be able use > > > wait_for_device_probe(). > > > > > > So if fwnode_usb_role_switch_get() returns -EPROBE_DEFER, you call > > > wait_for_device_probe() and try again. > > Do you mean do as below, > > > > find_role_switch: > > con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); > > if (IS_ERR(con->usb_role_sw)) { > > dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", > > con->num); > > if (con->usb_role_sw == -EPROBE_DEFER) { > > wait_for_device_probe(); > > goto find_role_switch; > > } > > > > return PTR_ERR(con->usb_role_sw); > > } > > Yes, something like that. Perhaps you could just do that in a loop, > and you should have a delay there. > > > Seem wait_for_device_probe() will wait all drivers probe, > > Can we accept it ? > > That part is not a problem, but I'm not sure if it appropriate to call > the function from drivers. Let me check if there is something else > that we can do... OK, I don't think we should use that in drivers, and I don't have any better ideas. Sorry for the noise. I notised that you have already send a couple of new versions of this series. I'll review the last one of those, but in the future please don't send the new versions so fast. thanks,
diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c index 684fc4f..fabd4df 100644 --- a/drivers/usb/typec/ucsi/ucsi.c +++ b/drivers/usb/typec/ucsi/ucsi.c @@ -1053,6 +1053,14 @@ static int ucsi_register_port(struct ucsi *ucsi, int index) con->num = index + 1; con->ucsi = ucsi; + cap->fwnode = ucsi_find_fwnode(con); + con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); + if (IS_ERR(con->usb_role_sw)) { + dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", + index + 1); + return -EAGAIN; + } + /* Delay other interactions with the con until registration is complete */ mutex_lock(&con->lock); @@ -1088,7 +1096,6 @@ static int ucsi_register_port(struct ucsi *ucsi, int index) if (con->cap.op_mode & UCSI_CONCAP_OPMODE_DEBUG_ACCESSORY) *accessory = TYPEC_ACCESSORY_DEBUG; - cap->fwnode = ucsi_find_fwnode(con); cap->driver_data = con; cap->ops = &ucsi_ops; @@ -1147,13 +1154,6 @@ static int ucsi_register_port(struct ucsi *ucsi, int index) ucsi_port_psy_changed(con); } - con->usb_role_sw = fwnode_usb_role_switch_get(cap->fwnode); - if (IS_ERR(con->usb_role_sw)) { - dev_err(ucsi->dev, "con%d: failed to get usb role switch\n", - con->num); - con->usb_role_sw = NULL; - } - /* Only notify USB controller if partner supports USB data */ if (!(UCSI_CONSTAT_PARTNER_FLAGS(con->status.flags) & UCSI_CONSTAT_PARTNER_FLAG_USB)) u_role = USB_ROLE_NONE; @@ -1281,12 +1281,16 @@ static int ucsi_init(struct ucsi *ucsi) static void ucsi_init_work(struct work_struct *work) { - struct ucsi *ucsi = container_of(work, struct ucsi, work); + struct ucsi *ucsi = container_of(work, struct ucsi, work.work); int ret; ret = ucsi_init(ucsi); if (ret) dev_err(ucsi->dev, "PPM init failed (%d)\n", ret); + + + if (ret == -EAGAIN) + queue_delayed_work(system_long_wq, &ucsi->work, HZ/10); } /** @@ -1326,7 +1330,7 @@ struct ucsi *ucsi_create(struct device *dev, const struct ucsi_operations *ops) if (!ucsi) return ERR_PTR(-ENOMEM); - INIT_WORK(&ucsi->work, ucsi_init_work); + INIT_DELAYED_WORK(&ucsi->work, ucsi_init_work); mutex_init(&ucsi->ppm_lock); ucsi->dev = dev; ucsi->ops = ops; @@ -1361,7 +1365,7 @@ int ucsi_register(struct ucsi *ucsi) if (!ucsi->version) return -ENODEV; - queue_work(system_long_wq, &ucsi->work); + queue_delayed_work(system_long_wq, &ucsi->work, 0); return 0; } @@ -1378,7 +1382,7 @@ void ucsi_unregister(struct ucsi *ucsi) u64 cmd = UCSI_SET_NOTIFICATION_ENABLE; /* Make sure that we are not in the middle of driver initialization */ - cancel_work_sync(&ucsi->work); + cancel_delayed_work_sync(&ucsi->work); /* Disable notifications */ ucsi->ops->async_write(ucsi, UCSI_CONTROL, &cmd, sizeof(cmd)); diff --git a/drivers/usb/typec/ucsi/ucsi.h b/drivers/usb/typec/ucsi/ucsi.h index 280f1e1..3812017 100644 --- a/drivers/usb/typec/ucsi/ucsi.h +++ b/drivers/usb/typec/ucsi/ucsi.h @@ -287,7 +287,7 @@ struct ucsi { struct ucsi_capability cap; struct ucsi_connector *connector; - struct work_struct work; + struct delayed_work work; /* PPM Communication lock */ struct mutex ppm_lock;
When role switch enabled, return -EAGAIN if fail to find it due to module load ordering issue, then restart ucsi init work to find it again every 100ms. It also means change ucsi init work to delayed_work. Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com> --- drivers/usb/typec/ucsi/ucsi.c | 28 ++++++++++++++++------------ drivers/usb/typec/ucsi/ucsi.h | 2 +- 2 files changed, 17 insertions(+), 13 deletions(-)