| Message ID | 20190716050327.8866-1-bjorn.andersson@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | remoteproc: qcom: Move glink_ssr notification after stop | expand |
Reviewed-by: Sibi Sankar <sibis@codeaurora.org> Tested-by: Sibi Sankar <sibis@codeaurora.org> On 2019-07-16 10:33, Bjorn Andersson wrote: > glink_ssr is used to signal a remote processor "A" about the stopping > of > another remote processor "B", so that in the event that remote > processor > B is ever booted again the fifos of the glink channel between A and B > is > in a known state. > > But if remote processor A receives this notification before B is > actually stopped the newly reset fifo indices will be interpreted as > there being "data" on the channel and either side of the channel will > enter a fatal error handler. > > Move the glink_ssr notification to the "unprepare" state of the > rproc_subdev to avoid this issue. > > This has the side effect of us not notifying the dying remote processor > itself about its fate, which has been seen to block in certain resource > constraint scenarios. > > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> > --- > drivers/remoteproc/qcom_common.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/remoteproc/qcom_common.c > b/drivers/remoteproc/qcom_common.c > index 6a448429f223..60650bcc8c67 100644 > --- a/drivers/remoteproc/qcom_common.c > +++ b/drivers/remoteproc/qcom_common.c > @@ -200,7 +200,7 @@ void qcom_unregister_ssr_notifier(struct > notifier_block *nb) > } > EXPORT_SYMBOL_GPL(qcom_unregister_ssr_notifier); > > -static void ssr_notify_stop(struct rproc_subdev *subdev, bool crashed) > +static void ssr_notify_unprepare(struct rproc_subdev *subdev) > { > struct qcom_rproc_ssr *ssr = to_ssr_subdev(subdev); > > @@ -220,7 +220,7 @@ void qcom_add_ssr_subdev(struct rproc *rproc, > struct qcom_rproc_ssr *ssr, > const char *ssr_name) > { > ssr->name = ssr_name; > - ssr->subdev.stop = ssr_notify_stop; > + ssr->subdev.unprepare = ssr_notify_unprepare; > > rproc_add_subdev(rproc, &ssr->subdev); > }
diff --git a/drivers/remoteproc/qcom_common.c b/drivers/remoteproc/qcom_common.c index 6a448429f223..60650bcc8c67 100644 --- a/drivers/remoteproc/qcom_common.c +++ b/drivers/remoteproc/qcom_common.c @@ -200,7 +200,7 @@ void qcom_unregister_ssr_notifier(struct notifier_block *nb) } EXPORT_SYMBOL_GPL(qcom_unregister_ssr_notifier); -static void ssr_notify_stop(struct rproc_subdev *subdev, bool crashed) +static void ssr_notify_unprepare(struct rproc_subdev *subdev) { struct qcom_rproc_ssr *ssr = to_ssr_subdev(subdev); @@ -220,7 +220,7 @@ void qcom_add_ssr_subdev(struct rproc *rproc, struct qcom_rproc_ssr *ssr, const char *ssr_name) { ssr->name = ssr_name; - ssr->subdev.stop = ssr_notify_stop; + ssr->subdev.unprepare = ssr_notify_unprepare; rproc_add_subdev(rproc, &ssr->subdev); }
glink_ssr is used to signal a remote processor "A" about the stopping of another remote processor "B", so that in the event that remote processor B is ever booted again the fifos of the glink channel between A and B is in a known state. But if remote processor A receives this notification before B is actually stopped the newly reset fifo indices will be interpreted as there being "data" on the channel and either side of the channel will enter a fatal error handler. Move the glink_ssr notification to the "unprepare" state of the rproc_subdev to avoid this issue. This has the side effect of us not notifying the dying remote processor itself about its fate, which has been seen to block in certain resource constraint scenarios. Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> --- drivers/remoteproc/qcom_common.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)