Date: Thu, 21 Jan 2016 21:43:32 -0700 From: Warner Losh <imp@bsdimp.com> To: Adrian Chadd <adrian.chadd@gmail.com> Cc: Stanislav Galabov <sgalabov@gmail.com>, "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org>, "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org> Subject: Re: SPI geom_flashmap/fdt_slicer support, FDT 'resets=' support and a move of ohci_fdt.c Message-ID: <CANCZdfr4frLMAfRiaHoVFPhi8hDa8rXSr5xjTyMQpUqoS-RT7g@mail.gmail.com> In-Reply-To: <CAJ-Vmo=DcfSXsuLUpcMvogjWNmNrniNsPxshhmHcGoWOxWM=CQ@mail.gmail.com> References: <B4B24B7D-B3EE-4F37-9E89-24FF17294C70@gmail.com> <CANCZdfrt-EdzFuXFSP2sTAX2vxP_RKAamKgUeL0u7hZYxeOF_A@mail.gmail.com> <2AB9D6E1-BFF8-4EEE-B366-C980B72C4779@gmail.com> <CAJ-Vmo=DcfSXsuLUpcMvogjWNmNrniNsPxshhmHcGoWOxWM=CQ@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Jan 21, 2016 at 8:35 PM, Adrian Chadd <adrian.chadd@gmail.com> wrote: > On 18 January 2016 at 06:49, Stanislav Galabov <sgalabov@gmail.com> wrote= : > > Hi Warner, > > > > I was thinking resets could help in general, not specifically in the > case of trying to implement a generic ohci_fdt driver. > > > > As I already mentioned to you off-list (and in order for this message t= o > possibly make some more sense), I saw that Linux makes use of the =E2=80= =98resets=E2=80=99 > property and looked at the documentation for it: > > > https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bi= ndings/reset/reset.txt > < > https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bi= ndings/reset/reset.txt > > > > > > For example, with the work I am currently doing on Ralink/Mediatek > support, I have over 10 different chips in the same family, that have ver= y > similar peripheral blocks, but their clocks and resets are controlled via > different bits in one of the SysCtl registers (the register itself is at > the same offset within the SysCtl block of each chip). So I may have chip= X > which has its USB (for example) reset controlled by bit 10 and chip Y wit= h > USB reset controlled by bit 12. > > > > So being able to write something like: > > resets =3D <&sysctl 10>; > > or > > resets =3D <&sysctl 11>; > > > > in my dts/dtsi files helps me immensely, instead of having to check wha= t > chip I am running on and based on that use a different register layout=E2= =80=A6 > > Then, if I wanted to (de)assert reset for a peripheral block that has > this property defined I=E2=80=99d just do fdt_reset_(de)assert_all(dev), = where dev > is the device_t for the peripheral in question. This would (de)assert all > reset pins associated with the peripheral. > > > > The same is the case for clock control (gating) in the Ralink/Mediatek > SoCs and this is the main reason I used fdt_clock that is already in > sys/dev/fdt and then thought about implementing the fdt_reset based on it= . > > > > I hope this clarifies a bit my reason for submitting the fdt_reset patc= h. > > This seems fine to me. Hm. Ian? Any comments? > I want to check on a few things before we head down this path. I've been traveling so haven't had a chance to look through it to see if Atmel uses this, and who else does... Warner > -a > > > Best wishes > > Stanislav > > > >> On Jan 18, 2016, at 02:04, Warner Losh <imp@bsdimp.com> wrote: > >> > >> I don't see how resets help. Maybe I missed where it was documented, > could you send that to me? > >> > >> Even with that, it seems that a generic ohci_fdt driver isn't possible= . > >> > >> Warner > >> > >> On Thu, Jan 14, 2016 at 2:01 AM, Stanislav Galabov <sgalabov@gmail.com > <mailto:sgalabov@gmail.com>> wrote: > >> Hi all, > >> > >> First off, sorry for the cross-post, I wasn=E2=80=99t very sure where = this > should go=E2=80=A6 > >> > >> I=E2=80=99ve created 3 PRs, which enable some functionality that my wo= rk on > Ralink/Mediatek SoCs would benefit from. > >> > >> 1. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206227 < > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206227> > >> - This enables geom_flashmap and fdt_slicer support for SPI flash chip= s > supported by the mx25l driver (sys/dev/flash/mx25l.c) > >> > >> 2. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206228 < > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206228> > >> - This adds support for FDT =E2=80=98resets=3D=E2=80=99 property in mu= ch the same way as > ian@=E2=80=99s sys/dev/fdt/fdt_clock* supports FDT =E2=80=98clocks=3D=E2= =80=98 property. In fact > this work is basically a modified version of fdt_clock* :-) > >> > >> 3. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206229 < > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D206229> > >> - This simply moves the at91 specific sys/dev/usb/controller/ohci_fdt.= c > to sys/dev/usb/controller/at91ohci_fdt.c (and changes the filename in > sys/arm/at91/files.at91 as well). The current naming is misleading IMHO a= nd > also, I have some (vague-ish) plans to see if I can implement generic > ohci_fdt and ehci_fdt based on dwc_otg_fdt, so that systems with standard > ehci/ohci controllers can reuse these. > >> > >> Patches are attached to the PRs. > >> > >> I would appreciate any feedback on the PRs and would also appreciate i= t > if someone could commit these if the proposed changes are appropriate. > >> > >> Best wishes, > >> Stanislav > >> _______________________________________________ > >> freebsd-arm@freebsd.org <mailto:freebsd-arm@freebsd.org> mailing list > >> https://lists.freebsd.org/mailman/listinfo/freebsd-arm < > https://lists.freebsd.org/mailman/listinfo/freebsd-arm> > >> To unsubscribe, send any mail to "freebsd-arm-unsubscribe@freebsd.org > <mailto:freebsd-arm-unsubscribe@freebsd.org>" > >> > > > > _______________________________________________ > > freebsd-arm@freebsd.org mailing list > > https://lists.freebsd.org/mailman/listinfo/freebsd-arm > > To unsubscribe, send any mail to "freebsd-arm-unsubscribe@freebsd.org" >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfr4frLMAfRiaHoVFPhi8hDa8rXSr5xjTyMQpUqoS-RT7g>