| Message ID | 20221102112451.128110-1-peng.fan@oss.nxp.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | remoteproc: imx: add start up delay | expand |
On Wed, 2 Nov 2022 at 05:23, Peng Fan (OSS) <peng.fan@oss.nxp.com> wrote: > > From: Peng Fan <peng.fan@nxp.com> > > V2: > Rebased on linux-next > > V1: > https://lore.kernel.org/lkml/20220609123500.3492475-1-peng.fan@oss.nxp.com/ > > There is case that after remoteproc start remote processor[M4], the M4 > runs slow and before M4 finish its own rpmsg framework initialization, > linux sends out vring kick message, then M4 firmware drops the kick > message. Some NXP released Cortex-M[x] images has such limitation that > it requires linux sends out vring kick message after M4 firmware finish > its rpmsg framework initialization. > > The best case is to use a method to let M4 notify Linux that M4 has > finished initialization, but we could not patch released firmware, > then update driver to detect notification. > > So add delay before linux send out vring kick message. It is not good to > use a fixed time delay in driver, so I choose to get that from device > tree. > From where I stand this is a hack to hide the lack of motivation to enact the real solution that is outlined above. I also wonder how these problems were not caught during the testing phase. Either find a way to upgrade your firmware or keep this in your internal tree. > Peng Fan (2): > dt-bindings: remoteproc: imx_rproc: add fsl,startup-delay-ms > remoteproc: imx_rproc: delay after kick remote processor > > .../devicetree/bindings/remoteproc/fsl,imx-rproc.yaml | 4 ++++ > drivers/remoteproc/imx_rproc.c | 9 +++++++++ > 2 files changed, 13 insertions(+) > > -- > 2.37.1 >
> Subject: Re: [PATCH V2 0/2] remoteproc: imx: add start up delay > > On Wed, 2 Nov 2022 at 05:23, Peng Fan (OSS) <peng.fan@oss.nxp.com> > wrote: > > > > From: Peng Fan <peng.fan@nxp.com> > > > > V2: > > Rebased on linux-next > > > > V1: > > > > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore > > .kernel.org%2Flkml%2F20220609123500.3492475-1- > peng.fan%40oss.nxp.com%2 > > > F&data=05%7C01%7Cpeng.fan%40nxp.com%7Cf14584bdef9349e744ca > 08dabcf7 > > > e1be%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C63803007007 > 2451509%7 > > > CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ > BTiI6Ik1 > > > haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=Ck73gkOPpNVT > kcZL8olCZP8B > > %2BrIpZHV48uZBl2R8HdM%3D&reserved=0 > > > > There is case that after remoteproc start remote processor[M4], the M4 > > runs slow and before M4 finish its own rpmsg framework initialization, > > linux sends out vring kick message, then M4 firmware drops the kick > > message. Some NXP released Cortex-M[x] images has such limitation that > > it requires linux sends out vring kick message after M4 firmware > > finish its rpmsg framework initialization. > > > > The best case is to use a method to let M4 notify Linux that M4 has > > finished initialization, but we could not patch released firmware, > > then update driver to detect notification. > > > > So add delay before linux send out vring kick message. It is not good > > to use a fixed time delay in driver, so I choose to get that from > > device tree. > > > > From where I stand this is a hack to hide the lack of motivation to enact the > real solution that is outlined above. I also wonder how these problems > were not caught during the testing phase. Either find a way to upgrade your > firmware or keep this in your internal tree. In the beginning, i.mx not migrated to use remoteproc, i.MX release only support uboot kick Cortex-M core and use downstream imx_rpmsg driver to communicate with remote core. There is no runtime stop/start. After migrated to remoteproc, we do see some issue, but old firmware was there. Thanks, Peng. > > > Peng Fan (2): > > dt-bindings: remoteproc: imx_rproc: add fsl,startup-delay-ms > > remoteproc: imx_rproc: delay after kick remote processor > > > > .../devicetree/bindings/remoteproc/fsl,imx-rproc.yaml | 4 ++++ > > drivers/remoteproc/imx_rproc.c | 9 +++++++++ > > 2 files changed, 13 insertions(+) > > > > -- > > 2.37.1 > >
On Thu, Nov 03, 2022 at 05:08:15AM +0000, Peng Fan wrote: > > Subject: Re: [PATCH V2 0/2] remoteproc: imx: add start up delay > > > > On Wed, 2 Nov 2022 at 05:23, Peng Fan (OSS) <peng.fan@oss.nxp.com> > > wrote: > > > > > > From: Peng Fan <peng.fan@nxp.com> > > > > > > V2: > > > Rebased on linux-next > > > > > > V1: > > > > > > > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore > > > .kernel.org%2Flkml%2F20220609123500.3492475-1- > > peng.fan%40oss.nxp.com%2 > > > > > F&data=05%7C01%7Cpeng.fan%40nxp.com%7Cf14584bdef9349e744ca > > 08dabcf7 > > > > > e1be%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C63803007007 > > 2451509%7 > > > > > CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ > > BTiI6Ik1 > > > > > haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=Ck73gkOPpNVT > > kcZL8olCZP8B > > > %2BrIpZHV48uZBl2R8HdM%3D&reserved=0 > > > > > > There is case that after remoteproc start remote processor[M4], the M4 > > > runs slow and before M4 finish its own rpmsg framework initialization, > > > linux sends out vring kick message, then M4 firmware drops the kick > > > message. Some NXP released Cortex-M[x] images has such limitation that > > > it requires linux sends out vring kick message after M4 firmware > > > finish its rpmsg framework initialization. > > > > > > The best case is to use a method to let M4 notify Linux that M4 has > > > finished initialization, but we could not patch released firmware, > > > then update driver to detect notification. > > > > > > So add delay before linux send out vring kick message. It is not good > > > to use a fixed time delay in driver, so I choose to get that from > > > device tree. > > > > > > > From where I stand this is a hack to hide the lack of motivation to enact the > > real solution that is outlined above. I also wonder how these problems > > were not caught during the testing phase. Either find a way to upgrade your > > firmware or keep this in your internal tree. > In the beginning, i.mx not migrated to use remoteproc, i.MX release only support > uboot kick Cortex-M core and use downstream imx_rpmsg driver to > communicate with remote core. There is no runtime stop/start. > > After migrated to remoteproc, we do see some issue, but old firmware was > there. You could also attach() to that remote processor, if the HW supports it. But my position remains the same - either find a way to work with this situation that is not a hack or keep this patch in your internal tree. > > Thanks, > Peng. > > > > > > Peng Fan (2): > > > dt-bindings: remoteproc: imx_rproc: add fsl,startup-delay-ms > > > remoteproc: imx_rproc: delay after kick remote processor > > > > > > .../devicetree/bindings/remoteproc/fsl,imx-rproc.yaml | 4 ++++ > > > drivers/remoteproc/imx_rproc.c | 9 +++++++++ > > > 2 files changed, 13 insertions(+) > > > > > > -- > > > 2.37.1 > > >
From: Peng Fan <peng.fan@nxp.com> V2: Rebased on linux-next V1: https://lore.kernel.org/lkml/20220609123500.3492475-1-peng.fan@oss.nxp.com/ There is case that after remoteproc start remote processor[M4], the M4 runs slow and before M4 finish its own rpmsg framework initialization, linux sends out vring kick message, then M4 firmware drops the kick message. Some NXP released Cortex-M[x] images has such limitation that it requires linux sends out vring kick message after M4 firmware finish its rpmsg framework initialization. The best case is to use a method to let M4 notify Linux that M4 has finished initialization, but we could not patch released firmware, then update driver to detect notification. So add delay before linux send out vring kick message. It is not good to use a fixed time delay in driver, so I choose to get that from device tree. Peng Fan (2): dt-bindings: remoteproc: imx_rproc: add fsl,startup-delay-ms remoteproc: imx_rproc: delay after kick remote processor .../devicetree/bindings/remoteproc/fsl,imx-rproc.yaml | 4 ++++ drivers/remoteproc/imx_rproc.c | 9 +++++++++ 2 files changed, 13 insertions(+)