| Message ID | 20210621161616.77524-7-manivannan.sadhasivam@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | MHI patches for v5.14 | expand |
On Mon, Jun 21, 2021 at 09:46:14PM +0530, Manivannan Sadhasivam wrote: > From: Hemant Kumar <hemantk@codeaurora.org> > > Event ring priorities are currently set to 1 and are unused. > Default processing priority for event rings is set to regular > tasklet. Controllers can choose to use high priority tasklet > scheduling for certain event rings critical for processing such > as ones transporting control information if they wish to avoid > system scheduling delays for those packets. In order to support > these use cases, allow controllers to set event ring priority to > high. > > Signed-off-by: Hemant Kumar <hemantk@codeaurora.org> > Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org> > Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > Link: https://lore.kernel.org/r/1624053903-24653-2-git-send-email-bbhatt@codeaurora.org > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > --- > drivers/bus/mhi/core/init.c | 3 +-- > drivers/bus/mhi/core/main.c | 9 +++++++-- > include/linux/mhi.h | 2 +- > 3 files changed, 9 insertions(+), 5 deletions(-) > > diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c > index c81b377fca8f..444676034bf0 100644 > --- a/drivers/bus/mhi/core/init.c > +++ b/drivers/bus/mhi/core/init.c > @@ -673,8 +673,7 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl, > &mhi_cntrl->mhi_chan[mhi_event->chan]; > } > > - /* Priority is fixed to 1 for now */ > - mhi_event->priority = 1; > + mhi_event->priority = event_cfg->priority; > > mhi_event->db_cfg.brstmode = event_cfg->mode; > if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode)) > diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c > index 8ac73f9e92a6..3775c77dec63 100644 > --- a/drivers/bus/mhi/core/main.c > +++ b/drivers/bus/mhi/core/main.c > @@ -454,10 +454,15 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev) > > if (mhi_dev) > mhi_notify(mhi_dev, MHI_CB_PENDING_DATA); > - } else { > - tasklet_schedule(&mhi_event->task); > + > + return IRQ_HANDLED; > } > > + if (!mhi_event->priority) > + tasklet_hi_schedule(&mhi_event->task); > + else > + tasklet_schedule(&mhi_event->task); > + > return IRQ_HANDLED; > } > > diff --git a/include/linux/mhi.h b/include/linux/mhi.h > index 86cea5256e3c..bf23c213429c 100644 > --- a/include/linux/mhi.h > +++ b/include/linux/mhi.h > @@ -250,7 +250,7 @@ struct mhi_channel_config { > * @irq_moderation_ms: Delay irq for additional events to be aggregated > * @irq: IRQ associated with this ring > * @channel: Dedicated channel number. U32_MAX indicates a non-dedicated ring > - * @priority: Priority of this ring. Use 1 for now > + * @priority: Processing priority of this ring. 0 is high and 1 is regular Why is 0 high and 1 low? Does that feel backwards? Shouldn't this be a boolean, or if not, an enumerated type so that people can read the code over time? thanks, greg k-h
On Thu, Jun 24, 2021 at 03:53:33PM +0200, Greg KH wrote: > On Mon, Jun 21, 2021 at 09:46:14PM +0530, Manivannan Sadhasivam wrote: > > From: Hemant Kumar <hemantk@codeaurora.org> > > > > Event ring priorities are currently set to 1 and are unused. > > Default processing priority for event rings is set to regular > > tasklet. Controllers can choose to use high priority tasklet > > scheduling for certain event rings critical for processing such > > as ones transporting control information if they wish to avoid > > system scheduling delays for those packets. In order to support > > these use cases, allow controllers to set event ring priority to > > high. > > > > Signed-off-by: Hemant Kumar <hemantk@codeaurora.org> > > Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org> > > Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > Link: https://lore.kernel.org/r/1624053903-24653-2-git-send-email-bbhatt@codeaurora.org > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > --- > > drivers/bus/mhi/core/init.c | 3 +-- > > drivers/bus/mhi/core/main.c | 9 +++++++-- > > include/linux/mhi.h | 2 +- > > 3 files changed, 9 insertions(+), 5 deletions(-) > > > > diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c > > index c81b377fca8f..444676034bf0 100644 > > --- a/drivers/bus/mhi/core/init.c > > +++ b/drivers/bus/mhi/core/init.c > > @@ -673,8 +673,7 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl, > > &mhi_cntrl->mhi_chan[mhi_event->chan]; > > } > > > > - /* Priority is fixed to 1 for now */ > > - mhi_event->priority = 1; > > + mhi_event->priority = event_cfg->priority; > > > > mhi_event->db_cfg.brstmode = event_cfg->mode; > > if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode)) > > diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c > > index 8ac73f9e92a6..3775c77dec63 100644 > > --- a/drivers/bus/mhi/core/main.c > > +++ b/drivers/bus/mhi/core/main.c > > @@ -454,10 +454,15 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev) > > > > if (mhi_dev) > > mhi_notify(mhi_dev, MHI_CB_PENDING_DATA); > > - } else { > > - tasklet_schedule(&mhi_event->task); > > + > > + return IRQ_HANDLED; > > } > > > > + if (!mhi_event->priority) > > + tasklet_hi_schedule(&mhi_event->task); > > + else > > + tasklet_schedule(&mhi_event->task); > > + > > return IRQ_HANDLED; > > } > > > > diff --git a/include/linux/mhi.h b/include/linux/mhi.h > > index 86cea5256e3c..bf23c213429c 100644 > > --- a/include/linux/mhi.h > > +++ b/include/linux/mhi.h > > @@ -250,7 +250,7 @@ struct mhi_channel_config { > > * @irq_moderation_ms: Delay irq for additional events to be aggregated > > * @irq: IRQ associated with this ring > > * @channel: Dedicated channel number. U32_MAX indicates a non-dedicated ring > > - * @priority: Priority of this ring. Use 1 for now > > + * @priority: Processing priority of this ring. 0 is high and 1 is regular > > Why is 0 high and 1 low? Does that feel backwards? > That's because, "1" was used from the beginning by the controller drivers as the regular priority. And I thought of using "0" as high priority so that we can leave the controller drivers unmodified. > Shouldn't this be a boolean, or if not, an enumerated type so that > people can read the code over time? > Bhaumik proposed an enum but I wanted 0/1 so that the controller drivers can be untouched. Also, I don't see any immediate requirement for other priorities. Will make it a bool then. Thanks, Mani > thanks, > > greg k-h
On Thu, Jun 24, 2021 at 07:54:53PM +0530, Manivannan Sadhasivam wrote: > On Thu, Jun 24, 2021 at 03:53:33PM +0200, Greg KH wrote: > > On Mon, Jun 21, 2021 at 09:46:14PM +0530, Manivannan Sadhasivam wrote: > > > From: Hemant Kumar <hemantk@codeaurora.org> > > > > > > Event ring priorities are currently set to 1 and are unused. > > > Default processing priority for event rings is set to regular > > > tasklet. Controllers can choose to use high priority tasklet > > > scheduling for certain event rings critical for processing such > > > as ones transporting control information if they wish to avoid > > > system scheduling delays for those packets. In order to support > > > these use cases, allow controllers to set event ring priority to > > > high. > > > > > > Signed-off-by: Hemant Kumar <hemantk@codeaurora.org> > > > Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org> > > > Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > > Link: https://lore.kernel.org/r/1624053903-24653-2-git-send-email-bbhatt@codeaurora.org > > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > > --- > > > drivers/bus/mhi/core/init.c | 3 +-- > > > drivers/bus/mhi/core/main.c | 9 +++++++-- > > > include/linux/mhi.h | 2 +- > > > 3 files changed, 9 insertions(+), 5 deletions(-) > > > > > > diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c > > > index c81b377fca8f..444676034bf0 100644 > > > --- a/drivers/bus/mhi/core/init.c > > > +++ b/drivers/bus/mhi/core/init.c > > > @@ -673,8 +673,7 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl, > > > &mhi_cntrl->mhi_chan[mhi_event->chan]; > > > } > > > > > > - /* Priority is fixed to 1 for now */ > > > - mhi_event->priority = 1; > > > + mhi_event->priority = event_cfg->priority; > > > > > > mhi_event->db_cfg.brstmode = event_cfg->mode; > > > if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode)) > > > diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c > > > index 8ac73f9e92a6..3775c77dec63 100644 > > > --- a/drivers/bus/mhi/core/main.c > > > +++ b/drivers/bus/mhi/core/main.c > > > @@ -454,10 +454,15 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev) > > > > > > if (mhi_dev) > > > mhi_notify(mhi_dev, MHI_CB_PENDING_DATA); > > > - } else { > > > - tasklet_schedule(&mhi_event->task); > > > + > > > + return IRQ_HANDLED; > > > } > > > > > > + if (!mhi_event->priority) > > > + tasklet_hi_schedule(&mhi_event->task); > > > + else > > > + tasklet_schedule(&mhi_event->task); > > > + > > > return IRQ_HANDLED; > > > } > > > > > > diff --git a/include/linux/mhi.h b/include/linux/mhi.h > > > index 86cea5256e3c..bf23c213429c 100644 > > > --- a/include/linux/mhi.h > > > +++ b/include/linux/mhi.h > > > @@ -250,7 +250,7 @@ struct mhi_channel_config { > > > * @irq_moderation_ms: Delay irq for additional events to be aggregated > > > * @irq: IRQ associated with this ring > > > * @channel: Dedicated channel number. U32_MAX indicates a non-dedicated ring > > > - * @priority: Priority of this ring. Use 1 for now > > > + * @priority: Processing priority of this ring. 0 is high and 1 is regular > > > > Why is 0 high and 1 low? Does that feel backwards? > > > > That's because, "1" was used from the beginning by the controller drivers > as the regular priority. And I thought of using "0" as high priority so > that we can leave the controller drivers unmodified. There's no problem modifying everyone, how much work is that? > > Shouldn't this be a boolean, or if not, an enumerated type so that > > people can read the code over time? > > > > Bhaumik proposed an enum but I wanted 0/1 so that the controller drivers > can be untouched. Also, I don't see any immediate requirement for other > priorities. > > Will make it a bool then. Rename it when you change it so that you know you catch all existing users. thanks, greg k-h
On Thu, Jun 24, 2021 at 04:40:38PM +0200, Greg KH wrote: > On Thu, Jun 24, 2021 at 07:54:53PM +0530, Manivannan Sadhasivam wrote: > > On Thu, Jun 24, 2021 at 03:53:33PM +0200, Greg KH wrote: > > > On Mon, Jun 21, 2021 at 09:46:14PM +0530, Manivannan Sadhasivam wrote: > > > > From: Hemant Kumar <hemantk@codeaurora.org> > > > > > > > > Event ring priorities are currently set to 1 and are unused. > > > > Default processing priority for event rings is set to regular > > > > tasklet. Controllers can choose to use high priority tasklet > > > > scheduling for certain event rings critical for processing such > > > > as ones transporting control information if they wish to avoid > > > > system scheduling delays for those packets. In order to support > > > > these use cases, allow controllers to set event ring priority to > > > > high. > > > > > > > > Signed-off-by: Hemant Kumar <hemantk@codeaurora.org> > > > > Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org> > > > > Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > > > Link: https://lore.kernel.org/r/1624053903-24653-2-git-send-email-bbhatt@codeaurora.org > > > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> > > > > --- > > > > drivers/bus/mhi/core/init.c | 3 +-- > > > > drivers/bus/mhi/core/main.c | 9 +++++++-- > > > > include/linux/mhi.h | 2 +- > > > > 3 files changed, 9 insertions(+), 5 deletions(-) > > > > > > > > diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c > > > > index c81b377fca8f..444676034bf0 100644 > > > > --- a/drivers/bus/mhi/core/init.c > > > > +++ b/drivers/bus/mhi/core/init.c [...] > > That's because, "1" was used from the beginning by the controller drivers > > as the regular priority. And I thought of using "0" as high priority so > > that we can leave the controller drivers unmodified. > > There's no problem modifying everyone, how much work is that? > I thought of minimizing the diff if we can avoid...err > > > Shouldn't this be a boolean, or if not, an enumerated type so that > > > people can read the code over time? > > > > > > > Bhaumik proposed an enum but I wanted 0/1 so that the controller drivers > > can be untouched. Also, I don't see any immediate requirement for other > > priorities. > > > > Will make it a bool then. > > Rename it when you change it so that you know you catch all existing > users. > Okay. Bhaumik, can you please use the enum (as you did)? Thanks, Mani > thanks, > > greg k-h
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c index c81b377fca8f..444676034bf0 100644 --- a/drivers/bus/mhi/core/init.c +++ b/drivers/bus/mhi/core/init.c @@ -673,8 +673,7 @@ static int parse_ev_cfg(struct mhi_controller *mhi_cntrl, &mhi_cntrl->mhi_chan[mhi_event->chan]; } - /* Priority is fixed to 1 for now */ - mhi_event->priority = 1; + mhi_event->priority = event_cfg->priority; mhi_event->db_cfg.brstmode = event_cfg->mode; if (MHI_INVALID_BRSTMODE(mhi_event->db_cfg.brstmode)) diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c index 8ac73f9e92a6..3775c77dec63 100644 --- a/drivers/bus/mhi/core/main.c +++ b/drivers/bus/mhi/core/main.c @@ -454,10 +454,15 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev) if (mhi_dev) mhi_notify(mhi_dev, MHI_CB_PENDING_DATA); - } else { - tasklet_schedule(&mhi_event->task); + + return IRQ_HANDLED; } + if (!mhi_event->priority) + tasklet_hi_schedule(&mhi_event->task); + else + tasklet_schedule(&mhi_event->task); + return IRQ_HANDLED; } diff --git a/include/linux/mhi.h b/include/linux/mhi.h index 86cea5256e3c..bf23c213429c 100644 --- a/include/linux/mhi.h +++ b/include/linux/mhi.h @@ -250,7 +250,7 @@ struct mhi_channel_config { * @irq_moderation_ms: Delay irq for additional events to be aggregated * @irq: IRQ associated with this ring * @channel: Dedicated channel number. U32_MAX indicates a non-dedicated ring - * @priority: Priority of this ring. Use 1 for now + * @priority: Processing priority of this ring. 0 is high and 1 is regular * @mode: Doorbell mode * @data_type: Type of data this ring will process * @hardware_event: This ring is associated with hardware channels