Message ID | 20180615095939.1331-1-srinivas.kandagatla@linaro.org (mailing list archive) |
---|---|
State | Accepted |
Headers | show |
Hi Srinivas, On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: > Some of the rpmsg devices need to switch on power domains to communicate > with remote processor. For example on Qualcomm DB820c platform LPASS > power domain needs to switched on for any kind of audio services. > This patch adds the missing power domain support in rpmsg core. > > Without this patch attempting to play audio via QDSP on DB820c would > reboot the system. > > Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> > --- > drivers/rpmsg/rpmsg_core.c | 7 +++++++ > 1 file changed, 7 insertions(+) > > diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c > index b714a543a91d..8122807db380 100644 > --- a/drivers/rpmsg/rpmsg_core.c > +++ b/drivers/rpmsg/rpmsg_core.c > @@ -15,6 +15,7 @@ > #include <linux/module.h> > #include <linux/rpmsg.h> > #include <linux/of_device.h> > +#include <linux/pm_domain.h> > #include <linux/slab.h> > > #include "rpmsg_internal.h" > @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) > struct rpmsg_endpoint *ept = NULL; > int err; > > + err = dev_pm_domain_attach(dev, true); > + if (err) > + goto out; This patch has broken the virtio-rpmsg stack based rpmsg devices. These devices are non-DT and the rpmsg_dev_probe() is now failing with -19 (-ENODEV) error code. Loic, Arnaud, Can one of you double-confirm this behavior on ST platforms as well? The patch came through 4.14.71 stable release and broke our downstream kernels. regards Suman > + > if (rpdrv->callback) { > strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); > chinfo.src = rpdev->src; > @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) > > rpdrv->remove(rpdev); > > + dev_pm_domain_detach(dev, true); > + > if (rpdev->ept) > rpmsg_destroy_ept(rpdev->ept); > >
Hi Suman, On 25/09/18 01:25, Suman Anna wrote: > Hi Srinivas, > > On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: >> Some of the rpmsg devices need to switch on power domains to communicate >> with remote processor. For example on Qualcomm DB820c platform LPASS >> power domain needs to switched on for any kind of audio services. >> This patch adds the missing power domain support in rpmsg core. >> >> Without this patch attempting to play audio via QDSP on DB820c would >> reboot the system. >> >> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> >> --- >> drivers/rpmsg/rpmsg_core.c | 7 +++++++ >> 1 file changed, 7 insertions(+) >> >> diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c >> index b714a543a91d..8122807db380 100644 >> --- a/drivers/rpmsg/rpmsg_core.c >> +++ b/drivers/rpmsg/rpmsg_core.c >> @@ -15,6 +15,7 @@ >> #include <linux/module.h> >> #include <linux/rpmsg.h> >> #include <linux/of_device.h> >> +#include <linux/pm_domain.h> >> #include <linux/slab.h> >> >> #include "rpmsg_internal.h" >> @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) >> struct rpmsg_endpoint *ept = NULL; >> int err; >> >> + err = dev_pm_domain_attach(dev, true); >> + if (err) >> + goto out; > > This patch has broken the virtio-rpmsg stack based rpmsg devices. These > devices are non-DT and the rpmsg_dev_probe() is now failing with -19 > (-ENODEV) error code. > There seems to be some dependency on this patch ( https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v4.19-rc5&id=919b7308fcc452cd4e282bab389c33384a9f3790 ) which seems to rework the return values from acpi_dev_pm_attach() and genpd_dev_pm_attach() --srini > Loic, Arnaud, > Can one of you double-confirm this behavior on ST platforms as well? > The patch came through 4.14.71 stable release and broke our downstream > kernels. > > regards > Suman > >> + >> if (rpdrv->callback) { >> strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); >> chinfo.src = rpdev->src; >> @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) >> >> rpdrv->remove(rpdev); >> >> + dev_pm_domain_detach(dev, true); >> + >> if (rpdev->ept) >> rpmsg_destroy_ept(rpdev->ept); >> >> >
Hi Srini, On 09/25/2018 03:29 AM, Srinivas Kandagatla wrote: > Hi Suman, > > On 25/09/18 01:25, Suman Anna wrote: >> Hi Srinivas, >> >> On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: >>> Some of the rpmsg devices need to switch on power domains to communicate >>> with remote processor. For example on Qualcomm DB820c platform LPASS >>> power domain needs to switched on for any kind of audio services. >>> This patch adds the missing power domain support in rpmsg core. >>> >>> Without this patch attempting to play audio via QDSP on DB820c would >>> reboot the system. >>> >>> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> >>> --- >>> drivers/rpmsg/rpmsg_core.c | 7 +++++++ >>> 1 file changed, 7 insertions(+) >>> >>> diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c >>> index b714a543a91d..8122807db380 100644 >>> --- a/drivers/rpmsg/rpmsg_core.c >>> +++ b/drivers/rpmsg/rpmsg_core.c >>> @@ -15,6 +15,7 @@ >>> #include <linux/module.h> >>> #include <linux/rpmsg.h> >>> #include <linux/of_device.h> >>> +#include <linux/pm_domain.h> >>> #include <linux/slab.h> >>> #include "rpmsg_internal.h" >>> @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) >>> struct rpmsg_endpoint *ept = NULL; >>> int err; >>> + err = dev_pm_domain_attach(dev, true); >>> + if (err) >>> + goto out; >> >> This patch has broken the virtio-rpmsg stack based rpmsg devices. These >> devices are non-DT and the rpmsg_dev_probe() is now failing with -19 >> (-ENODEV) error code. >> > > There seems to be some dependency on this patch ( > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v4.19-rc5&id=919b7308fcc452cd4e282bab389c33384a9f3790 > ) which seems to rework the return values from acpi_dev_pm_attach() and > genpd_dev_pm_attach() > OK, I see that dependent commit is part of the 4.18 Linux kernel. So mainline kernel has no issues, but this patch was picked up in 4.14.71 (don't know why) as commit 1ed3a9307230, so 4.14 stable is broken. So, how do we proceed here - don't think we can pick all those dependent patches against stable. regards Suman > > > --srini > > >> Loic, Arnaud, >> Can one of you double-confirm this behavior on ST platforms as well? >> The patch came through 4.14.71 stable release and broke our downstream >> kernels. >> >> regards >> Suman >> >>> + >>> if (rpdrv->callback) { >>> strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); >>> chinfo.src = rpdev->src; >>> @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) >>> rpdrv->remove(rpdev); >>> + dev_pm_domain_detach(dev, true); >>> + >>> if (rpdev->ept) >>> rpmsg_destroy_ept(rpdev->ept); >>> >>
Hi Suman On 09/25/2018 02:25 AM, Suman Anna wrote: > Hi Srinivas, > > On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: >> Some of the rpmsg devices need to switch on power domains to communicate >> with remote processor. For example on Qualcomm DB820c platform LPASS >> power domain needs to switched on for any kind of audio services. >> This patch adds the missing power domain support in rpmsg core. >> >> Without this patch attempting to play audio via QDSP on DB820c would >> reboot the system. >> >> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> >> --- >> drivers/rpmsg/rpmsg_core.c | 7 +++++++ >> 1 file changed, 7 insertions(+) >> >> diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c >> index b714a543a91d..8122807db380 100644 >> --- a/drivers/rpmsg/rpmsg_core.c >> +++ b/drivers/rpmsg/rpmsg_core.c >> @@ -15,6 +15,7 @@ >> #include <linux/module.h> >> #include <linux/rpmsg.h> >> #include <linux/of_device.h> >> +#include <linux/pm_domain.h> >> #include <linux/slab.h> >> >> #include "rpmsg_internal.h" >> @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) >> struct rpmsg_endpoint *ept = NULL; >> int err; >> >> + err = dev_pm_domain_attach(dev, true); >> + if (err) >> + goto out; > > This patch has broken the virtio-rpmsg stack based rpmsg devices. These > devices are non-DT and the rpmsg_dev_probe() is now failing with -19 > (-ENODEV) error code. > > Loic, Arnaud, > Can one of you double-confirm this behavior on ST platforms as well? > The patch came through 4.14.71 stable release and broke our downstream > kernels. I do not reproduce issue on 4.14.0 + your patch, on my ST platform (not able to test on the LTS branch). By looking the source code, seems that your issue is related to genpd_dev_pm_attach that returns -ENODEV because you have no remoteproc device node... Do you have a rproc node defined in your DT? we have one. Regards Arnaud > > regards > Suman > >> + >> if (rpdrv->callback) { >> strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); >> chinfo.src = rpdev->src; >> @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) >> >> rpdrv->remove(rpdev); >> >> + dev_pm_domain_detach(dev, true); >> + >> if (rpdev->ept) >> rpmsg_destroy_ept(rpdev->ept); >> >> >
Hi Arnaud, > On 09/25/2018 02:25 AM, Suman Anna wrote: >> Hi Srinivas, >> >> On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: >>> Some of the rpmsg devices need to switch on power domains to communicate >>> with remote processor. For example on Qualcomm DB820c platform LPASS >>> power domain needs to switched on for any kind of audio services. >>> This patch adds the missing power domain support in rpmsg core. >>> >>> Without this patch attempting to play audio via QDSP on DB820c would >>> reboot the system. >>> >>> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> >>> --- >>> drivers/rpmsg/rpmsg_core.c | 7 +++++++ >>> 1 file changed, 7 insertions(+) >>> >>> diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c >>> index b714a543a91d..8122807db380 100644 >>> --- a/drivers/rpmsg/rpmsg_core.c >>> +++ b/drivers/rpmsg/rpmsg_core.c >>> @@ -15,6 +15,7 @@ >>> #include <linux/module.h> >>> #include <linux/rpmsg.h> >>> #include <linux/of_device.h> >>> +#include <linux/pm_domain.h> >>> #include <linux/slab.h> >>> >>> #include "rpmsg_internal.h" >>> @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) >>> struct rpmsg_endpoint *ept = NULL; >>> int err; >>> >>> + err = dev_pm_domain_attach(dev, true); >>> + if (err) >>> + goto out; >> >> This patch has broken the virtio-rpmsg stack based rpmsg devices. These >> devices are non-DT and the rpmsg_dev_probe() is now failing with -19 >> (-ENODEV) error code. >> >> Loic, Arnaud, >> Can one of you double-confirm this behavior on ST platforms as well? >> The patch came through 4.14.71 stable release and broke our downstream >> kernels. > I do not reproduce issue on 4.14.0 + your patch, on my ST platform (not > able to test on the LTS branch). > By looking the source code, seems that your issue is related to > genpd_dev_pm_attach that returns -ENODEV because you have > no remoteproc device node... > Do you have a rproc node defined in your DT? we have one. The dev here is not the remoteproc dev but the rpmsg device which does not have an OF node, and so I get the ENODEV due to the check on dev->of_node in genpd_dev_pm_attach. I do have DT remoteproc nodes. Are you publishing an rpmsg device from the remote side and are they being probed? Issue is seen only during the probe phase, you will not see any errors if the rpmsg devices are just being published and created. regards Suman > > Regards > Arnaud > >> >> regards >> Suman >> >>> + >>> if (rpdrv->callback) { >>> strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); >>> chinfo.src = rpdev->src; >>> @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) >>> >>> rpdrv->remove(rpdev); >>> >>> + dev_pm_domain_detach(dev, true); >>> + >>> if (rpdev->ept) >>> rpmsg_destroy_ept(rpdev->ept); >>> >>> >>
Hi Suman, On 09/25/2018 09:02 PM, Suman Anna wrote: > Hi Arnaud, > >> On 09/25/2018 02:25 AM, Suman Anna wrote: >>> Hi Srinivas, >>> >>> On 06/15/2018 04:59 AM, Srinivas Kandagatla wrote: >>>> Some of the rpmsg devices need to switch on power domains to communicate >>>> with remote processor. For example on Qualcomm DB820c platform LPASS >>>> power domain needs to switched on for any kind of audio services. >>>> This patch adds the missing power domain support in rpmsg core. >>>> >>>> Without this patch attempting to play audio via QDSP on DB820c would >>>> reboot the system. >>>> >>>> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> >>>> --- >>>> drivers/rpmsg/rpmsg_core.c | 7 +++++++ >>>> 1 file changed, 7 insertions(+) >>>> >>>> diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c >>>> index b714a543a91d..8122807db380 100644 >>>> --- a/drivers/rpmsg/rpmsg_core.c >>>> +++ b/drivers/rpmsg/rpmsg_core.c >>>> @@ -15,6 +15,7 @@ >>>> #include <linux/module.h> >>>> #include <linux/rpmsg.h> >>>> #include <linux/of_device.h> >>>> +#include <linux/pm_domain.h> >>>> #include <linux/slab.h> >>>> >>>> #include "rpmsg_internal.h" >>>> @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) >>>> struct rpmsg_endpoint *ept = NULL; >>>> int err; >>>> >>>> + err = dev_pm_domain_attach(dev, true); >>>> + if (err) >>>> + goto out; >>> >>> This patch has broken the virtio-rpmsg stack based rpmsg devices. These >>> devices are non-DT and the rpmsg_dev_probe() is now failing with -19 >>> (-ENODEV) error code. >>> >>> Loic, Arnaud, >>> Can one of you double-confirm this behavior on ST platforms as well? >>> The patch came through 4.14.71 stable release and broke our downstream >>> kernels. > >> I do not reproduce issue on 4.14.0 + your patch, on my ST platform (not >> able to test on the LTS branch). >> By looking the source code, seems that your issue is related to >> genpd_dev_pm_attach that returns -ENODEV because you have >> no remoteproc device node... >> Do you have a rproc node defined in your DT? we have one. > > The dev here is not the remoteproc dev but the rpmsg device which does > not have an OF node, and so I get the ENODEV due to the check on > dev->of_node in genpd_dev_pm_attach. I do have DT remoteproc nodes. > > Are you publishing an rpmsg device from the remote side and are they > being probed? Issue is seen only during the probe phase, you will not > see any errors if the rpmsg devices are just being published and created. Sorry something was wrong in my fw used yesterday for testing...so forget my previous status. I re-tested it and I have the same issue (-ENODEV) with rpmsg device probed on ns announcement rpmsg from remote side. So we confirm the issue. Regards Arnaud > > regards > Suman > >> >> Regards >> Arnaud >> >>> >>> regards >>> Suman >>> >>>> + >>>> if (rpdrv->callback) { >>>> strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); >>>> chinfo.src = rpdev->src; >>>> @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) >>>> >>>> rpdrv->remove(rpdev); >>>> >>>> + dev_pm_domain_detach(dev, true); >>>> + >>>> if (rpdev->ept) >>>> rpmsg_destroy_ept(rpdev->ept); >>>> >>>> >>> >
On 26/09/18 13:48, Arnaud Pouliquen wrote: > Sorry something was wrong in my fw used yesterday for testing...so > forget my previous status. > > I re-tested it and I have the same issue (-ENODEV) with rpmsg device > probed on ns announcement rpmsg from remote side. > So we confirm the issue. A revert patch is in the stable queue now for 4.14 and 4.18 stable. http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary Issue reported should be fixed soon!! --srini
On Wed 26 Sep 06:06 PDT 2018, Srinivas Kandagatla wrote: > On 26/09/18 13:48, Arnaud Pouliquen wrote: > > Sorry something was wrong in my fw used yesterday for testing...so > > forget my previous status. > > > > I re-tested it and I have the same issue (-ENODEV) with rpmsg device > > probed on ns announcement rpmsg from remote side. > > So we confirm the issue. > > > A revert patch is in the stable queue now for 4.14 and 4.18 stable. > http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary > > Issue reported should be fixed soon!! > Thanks guys! Regards, Bjorn
diff --git a/drivers/rpmsg/rpmsg_core.c b/drivers/rpmsg/rpmsg_core.c index b714a543a91d..8122807db380 100644 --- a/drivers/rpmsg/rpmsg_core.c +++ b/drivers/rpmsg/rpmsg_core.c @@ -15,6 +15,7 @@ #include <linux/module.h> #include <linux/rpmsg.h> #include <linux/of_device.h> +#include <linux/pm_domain.h> #include <linux/slab.h> #include "rpmsg_internal.h" @@ -449,6 +450,10 @@ static int rpmsg_dev_probe(struct device *dev) struct rpmsg_endpoint *ept = NULL; int err; + err = dev_pm_domain_attach(dev, true); + if (err) + goto out; + if (rpdrv->callback) { strncpy(chinfo.name, rpdev->id.name, RPMSG_NAME_SIZE); chinfo.src = rpdev->src; @@ -490,6 +495,8 @@ static int rpmsg_dev_remove(struct device *dev) rpdrv->remove(rpdev); + dev_pm_domain_detach(dev, true); + if (rpdev->ept) rpmsg_destroy_ept(rpdev->ept);
Some of the rpmsg devices need to switch on power domains to communicate with remote processor. For example on Qualcomm DB820c platform LPASS power domain needs to switched on for any kind of audio services. This patch adds the missing power domain support in rpmsg core. Without this patch attempting to play audio via QDSP on DB820c would reboot the system. Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> --- drivers/rpmsg/rpmsg_core.c | 7 +++++++ 1 file changed, 7 insertions(+)