Re: [PATCH] arm64: dts: rockchip: Change serial baud rate for Pinephone Pro to 1.5 MB

[Maya Matuszczyk]

Hi Heiko

pt., 28 lip 2023 o 21:00 Heiko Stuebner <heiko@xxxxxxxxx> napisał(a):
>
> Hi Javier,
>
> Am Samstag, 22. Juli 2023, 01:06:54 CEST schrieb Javier Martinez Canillas:
> > Heiko Stübner <heiko@xxxxxxxxx> writes:
> > > Am Dienstag, 18. April 2023, 14:11:53 CEST schrieb Javier Martinez Canillas:
> > >> Heiko Stübner <heiko@xxxxxxxxx> writes:
> > >> > Am Dienstag, 4. April 2023, 14:52:02 CEST schrieb Peter Geis:
> > >> >> On Tue, Apr 4, 2023 at 3:55 AM Heiko Stübner <heiko@xxxxxxxxx> wrote:
> > >> >> > Am Montag, 3. April 2023, 19:59:37 CEST schrieb Javier Martinez Canillas:
> > >> >> > > This baud rate is set for the device by mainline u-boot and is also what
> > >> >> > > is set in the Pinebook Pro Device Tree, which is a device similar to the
> > >> >> > > PinePhone Pro but with a different form factor.
> > >> >> > >
> > >> >> > > Otherwise, the baud rate of the firmware and Linux don't match by default
> > >> >> > > and a 'console=ttyS2,1500000n8' kernel command line parameter is required
> > >> >> > > to have proper output for both.
> > >> >> >
> > >> >> > The interesting question is always if this will break someone else's setup.
> > >> >> > I've never really understood the strange setting of 1.5MBps, but on the
> > >> >> > other hand it _is_ a reality on most boards.
> > >> >
> > >> >> The 1.5M baud is default because the clock structure on rockchip
> > >> >> devices does not allow a clean 115200 baud. By attempting to force
> > >> >> 115200, it will always be slightly off (either low or high depending
> > >> >> on how the driver decided to round). If this actually causes any
> > >> >> problems is the subject of much debate.
> > >> >
> > >> > thanks so much for this piece of clock-detail. As I wrote, I never really
> > >> > understood the why _before_ but also never cared that much to dive
> > >> > into it and find out.
> > >> >
> > >> > So your explanation closes one knowledge gap in my head.
> > >> >
> > >> > Thanks a lot :-)
> > >>
> > >> Did you make a decision about this? I guess the clock explanation is yet
> > >> another argument in favour of switching the PPP to a 1.5 Mbps baud rate ?
> > >
> > > Sorry, but no decision made here. Either way it's breaking for someone,
> > > which makes this quite hard.
> > >
> >
> > Another ping on this patch.
> >
> > > The rate accuracy is the one side, the two-boot issue is the other side.
> > > And mainline u-boot (and levinboot - whatever that is) provides a 3rd side.
> > >
> > > People starting with the phone probably won't replace the bootloader
> > > in a first step but instead might play with a system image or newer kernel.
> > > So if the uart will break for everyone using the default bootloader from
> > > the factory that is somewhat bad.
> > >
> >
> > Probably won't replace the DTB shipped with the firmware either? If one is
> > replacing the firmware provided DTB witch the one in the mainline kernel,
> > probably such person is also using mainline u-boot?
>
> Not necessarily.
>
> I.e. putting an extlinux.conf on an sd-card with a kernel-image and dtb
> is not rocket science ;-)
>
>
> > > I don't have a Pinephone Pro myself, so I really hoped for some Acks
> > > or similar to appear in the meantime.
> > >
> >
> > For someone like me who is only using mainline u-boot, linux, etc then
> > having a consistent uart baud rate across all components is really useful.
> >
> > Otherwise I either have serial console for u-boot or the kernel, but can't
> > have both working so is annoying.
> >
> > It would be good to have a definite answer on this. Since every time that
> > I try to hack on my PPP, I end changing my DTS and remember this patch :)
>
> So far people only reported "breaks my setup". I'm in a pickle here ;-) .
> Without anybody saying "I want to also move into this direction" I really
> feel I should not merge a patch that breaks other peoples setups.

Well, I'd prefer 1.5M baud rate as it is more consistent with other Rockchip
boards and it makes for a much more usable terminal experience when
logged in, it also doesn't affect boot times when serial is enabled with a
high loglevel and console on serial as 115200 does.

Though I'm just fine with using kernel's cmdline to set a baud rate.

*shrugs*

Best Regards,
Maya Matuszczyk

pt., 28 lip 2023 o 21:00 Heiko Stuebner <heiko@xxxxxxxxx> napisał(a):
>
> Hi Javier,
>
> Am Samstag, 22. Juli 2023, 01:06:54 CEST schrieb Javier Martinez Canillas:
> > Heiko Stübner <heiko@xxxxxxxxx> writes:
> > > Am Dienstag, 18. April 2023, 14:11:53 CEST schrieb Javier Martinez Canillas:
> > >> Heiko Stübner <heiko@xxxxxxxxx> writes:
> > >> > Am Dienstag, 4. April 2023, 14:52:02 CEST schrieb Peter Geis:
> > >> >> On Tue, Apr 4, 2023 at 3:55 AM Heiko Stübner <heiko@xxxxxxxxx> wrote:
> > >> >> > Am Montag, 3. April 2023, 19:59:37 CEST schrieb Javier Martinez Canillas:
> > >> >> > > This baud rate is set for the device by mainline u-boot and is also what
> > >> >> > > is set in the Pinebook Pro Device Tree, which is a device similar to the
> > >> >> > > PinePhone Pro but with a different form factor.
> > >> >> > >
> > >> >> > > Otherwise, the baud rate of the firmware and Linux don't match by default
> > >> >> > > and a 'console=ttyS2,1500000n8' kernel command line parameter is required
> > >> >> > > to have proper output for both.
> > >> >> >
> > >> >> > The interesting question is always if this will break someone else's setup.
> > >> >> > I've never really understood the strange setting of 1.5MBps, but on the
> > >> >> > other hand it _is_ a reality on most boards.
> > >> >
> > >> >> The 1.5M baud is default because the clock structure on rockchip
> > >> >> devices does not allow a clean 115200 baud. By attempting to force
> > >> >> 115200, it will always be slightly off (either low or high depending
> > >> >> on how the driver decided to round). If this actually causes any
> > >> >> problems is the subject of much debate.
> > >> >
> > >> > thanks so much for this piece of clock-detail. As I wrote, I never really
> > >> > understood the why _before_ but also never cared that much to dive
> > >> > into it and find out.
> > >> >
> > >> > So your explanation closes one knowledge gap in my head.
> > >> >
> > >> > Thanks a lot :-)
> > >>
> > >> Did you make a decision about this? I guess the clock explanation is yet
> > >> another argument in favour of switching the PPP to a 1.5 Mbps baud rate ?
> > >
> > > Sorry, but no decision made here. Either way it's breaking for someone,
> > > which makes this quite hard.
> > >
> >
> > Another ping on this patch.
> >
> > > The rate accuracy is the one side, the two-boot issue is the other side.
> > > And mainline u-boot (and levinboot - whatever that is) provides a 3rd side.
> > >
> > > People starting with the phone probably won't replace the bootloader
> > > in a first step but instead might play with a system image or newer kernel.
> > > So if the uart will break for everyone using the default bootloader from
> > > the factory that is somewhat bad.
> > >
> >
> > Probably won't replace the DTB shipped with the firmware either? If one is
> > replacing the firmware provided DTB witch the one in the mainline kernel,
> > probably such person is also using mainline u-boot?
>
> Not necessarily.
>
> I.e. putting an extlinux.conf on an sd-card with a kernel-image and dtb
> is not rocket science ;-)
>
>
> > > I don't have a Pinephone Pro myself, so I really hoped for some Acks
> > > or similar to appear in the meantime.
> > >
> >
> > For someone like me who is only using mainline u-boot, linux, etc then
> > having a consistent uart baud rate across all components is really useful.
> >
> > Otherwise I either have serial console for u-boot or the kernel, but can't
> > have both working so is annoying.
> >
> > It would be good to have a definite answer on this. Since every time that
> > I try to hack on my PPP, I end changing my DTS and remember this patch :)
>
> So far people only reported "breaks my setup". I'm in a pickle here ;-) .
> Without anybody saying "I want to also move into this direction" I really
> feel I should not merge a patch that breaks other peoples setups.
>
>
> Heiko
>
>
>
> _______________________________________________
> Linux-rockchip mailing list
> Linux-rockchip@xxxxxxxxxxxxxxxxxxx
> http://lists.infradead.org/mailman/listinfo/linux-rockchip