From: Adam Ford <aford173@gmail.com>
To: "H. Nikolaus Schaller" <hns@goldelico.com>
Cc: "André Roth" <neolynx@gmail.com>,
Linux-OMAP <linux-omap@vger.kernel.org>,
"Discussions about the Letux Kernel"
<letux-kernel@openphoenux.org>,
"Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
"Andreas Kemnade" <andreas@kemnade.info>,
"Tony Lindgren" <tony@atomide.com>
Subject: Re: [Letux-kernel] [RFC PATCH 0/3] Enable 1GHz support on omap36xx
Date: Mon, 9 Sep 2019 11:20:44 -0500 [thread overview]
Message-ID: <CAHCN7xJe+sBVZHfXN0ZEaepxAR8RGuMTVGA4GMR5Nqn5LG2TeA@mail.gmail.com> (raw)
In-Reply-To: <87420DBD-770F-4C32-9499-A3AEA5876E8A@goldelico.com>
On Mon, Sep 9, 2019 at 9:57 AM H. Nikolaus Schaller <hns@goldelico.com> wrote:
>
> Hi Adam,
>
> > Am 09.09.2019 um 16:26 schrieb Adam Ford <aford173@gmail.com>:
> >
> > On Sat, Sep 7, 2019 at 2:38 AM H. Nikolaus Schaller <hns@goldelico.com> wrote:
> >>
> >> Hi Adam,
> >>
> >>> Am 02.09.2019 um 23:10 schrieb Adam Ford <aford173@gmail.com>:
> >>>
> >>> On Mon, Sep 2, 2019 at 10:46 AM H. Nikolaus Schaller <hns@goldelico.com> wrote:
> >>>>
> >>>>
> >>>>
> >>>> But my tests show that decoding works now. So you already might give it a try.
> >>>
> >>> I am traveling all this week, but I have an omap3530, DM3730
> >>> (omap3630), and an AM3517 that I use for testing.
> >>
> >> now as the omap3430 and omap3630 opp-v2 tables are installed,
> >> we could add am35x7 as well.
> >>
> >> What needs to be done:
> >>
> >> 1. add OPP-v2 table to am3517.dtsi
> >>
> >> for example copy skeleton from omap36xx.dtsi
> >>
> >> and define reasonable clock speeds. I would think about
> >> 150 MHz, 300 MHz, 600MHz.
> >
> > This might be more of a question for TI, but can we match the 3430
> > list of frequencies?
> >
> > Something like:
> > 125000 1200000
> > 250000 1200000
> > 500000 1200000
> > 550000 1200000
> > 600000 1200000
>
> And another question: is it more derived from omap3430 or omap3630?
>
> >
> >
> >>
> >> Debatable is if we need a clock-latency definition.
> >>
> >> 2. change all voltages to 1.2V
> >>
> >> opp-microvolt = <1200000 1200000 1200000>;
> >>
> >> There is no point to specify 3 voltages <target min max> here since we
> >> will never need a min and a max value.
> >>
> >> opp-microvolt = <1200000>;
> >>
> >> should also be ok (AFAIK, parser handles single-value records).
> >>
> >> 3. AFAIK there is no speed binned eFuse...
> >>
> >> But the ti-cpufreq driver always wants to read some eFuse register.
> >>
> >> So please check if you can read 0x4800244C and 0x4830A204
> >> like on omap36xx and if they produce stable values (and not
> >> random noise).
> >
> > For the AM3517,
> >
> > 0x4800244C = 0000 0cc0
>
> If it behaves like an dm3730 (Table 1-6) this would be read as 800/600MHz
> and some reserved code in bit 7:6.
>
> If it behaves like an omap3530 (Table 1-6) this would bean 600MHz but reserved
> value for IVA Frequency.
>
>
> > 0x4830A204 = 1b86 802f
>
> would be decoded (Table 1-7) as "AM/DM37x ES1.1"
>
> omap35xx would have a different code (Table 1-9). Most similar is "OMAP35x ES2.0" with 0x1B7A E02F
>
> So this seems to answer that the am3517 is indeed a derivative of the am/dm37xx.
>
> Therefore the only OPPs would be OPP50 (300 MHz) and OPP100 (600 MHz).
>
> Only tests or TI internal documentations could show if the am3517 still
> runs stable at newly invented "OPP25" (150MHz).
>
> >
> > The AM3517 shows these are valid addresses and I read them multiple
> > times and they yielded the same results even after power cycling.
> >
> >
> >>
> >> If yes, we simply assume that am3517 is similar enough to omap3630,
> >> ignore that there is no eFuse, but read the register anyways and
> >> then ignore the bit if it is 0 or 1.
> >>
> >> This means that all OPPs can get
> >>
> >> opp-supported-hw = <0xffffffff 0xffffffff>;
> >>
> >> There could also be a default handler if this property is missing,
> >> but I have not researched this.
> >>
> > Like this?
> >
> > opp1-125000000 {
> > opp-hz = /bits/ 64 <125000000>;
> > opp-microvolt = <1200000>;
> > opp-supported-hw = <0xffffffff 0xffffffff>;
> > };
> >
> > opp2-250000000 {
> > opp-hz = /bits/ 64 <250000000>;
> > opp-microvolt = <1200000>;
> > opp-supported-hw = <0xffffffff 0xffffffff>;
> > opp-suspend;
> > };
> >
> > opp3-500000000 {
> > opp-hz = /bits/ 64 <500000000>;
> > opp-microvolt = <1200000>;
> > opp-supported-hw = <0xffffffff 0xffffffff>;
> > };
> >
> > opp4-550000000 {
> > opp-hz = /bits/ 64 <550000000>;
> > opp-microvolt = <1200000>;
> > opp-supported-hw = <0xffffffff 0xffffffff>;
> > };
> >
> > opp5-600000000 {
> > opp-hz = /bits/ 64 <600000000>;
> > opp-microvolt = <1200000>;
> > opp-supported-hw = <0xffffffff 0xffffffff>;
> > };
>
> Yes.
>
> >
> > What does opp-suspend do? I noticed it in the 34xx.dtsi
>
> Good question. I think it is the OPP to be chosen before suspend:
>
> https://www.kernel.org/doc/Documentation/devicetree/bindings/opp/opp.txt
> says
>
> - opp-suspend: Marks the OPP to be used during device suspend. Only one OPP in
> the table should have this.
>
> But that doesn't mean the drivers make use of this marker.
>
> This makes me also wonder if we should tag the OPP1G and OPP6 as "turbo-mode"...
>
> Another question that came up by private mail from André was if we
> should better disable the turbo OPPs of omap34xx and 36xx by default
> (status = "disabled";) because there are concerns about overheating
> the chips and we have no thermal regulation like for omap4 & 5.
>
> But this would mean that every board DTS would have to set it explicitly
> to "enabled".
>
> And another concern is if the 1GHz OPP doesn't also need to switch the
> ABB bias LDO to a different mode. This is not done by the ti-cpufreq driver.
> Maybe it is done by some driver in mach-omap but I have not searched for.
ti-omap5-opp-supply.txt shows
Required Properties for Device Node:
- vdd-supply: phandle to regulator controlling VDD supply
- vbb-supply: phandle to regulator controlling Body Bias supply
(Usually Adaptive Body Bias regulator)
Looking at the ti-cpufreq.c driver, it appears as if there is both a
vdd and a vbb listed in the table (line 302). AFAICT, we should be
able to just set it up pointing VDD to the PMIC and VBB at the ABB
unless I am missing something.
Currently, I have the follwing in my device tree:
/* Reroute power feeding the CPU to come from the external PMIC */
®_arm
{
vin-supply = <&sw1a_reg>;
};
®_soc
{
vin-supply = <&sw1c_reg>;
};
Is this still correct with the new cpufreq driver? I am wondering if
we need VDD and VBB references under the cpu entry. The
ti-omap5-opp-supply ready me also reads:
/* Device Node (CPU) */
cpus {
cpu0: cpu@0 {
device_type = "cpu";
...
vdd-supply = <&vcc>;
vbb-supply = <&abb_mpu>;
};
};
This isn't in the ti-cpufreq.txt, however it might actually be
utilized based on a quick skim of the driver.
adam
>
> So the concern is that we will run the turbo modes outside of the TI specs
> while before applying the patch set this would be a lesser problem (OPP130
> should also be thermally limited to 90°C).
>
> I.e. users of 1GHz capable boards will not only see 25% more speed but
> suddenly higher SoC temperatures than the years before.
>
> >
> >> 4. add compatible to ti-cpufreq
> >> and share the register offsets, bit masks etc. with omap3630:
> >>
> >> { .compatible = "ti,am33xx", .data = &am3x_soc_data, },
> >> { .compatible = "ti,am3517", .data = &omap36xx_soc_data, },
> >> { .compatible = "ti,am43", .data = &am4x_soc_data, },
> >> { .compatible = "ti,dra7", .data = &dra7_soc_data },
> >> { .compatible = "ti,omap3430", .data = &omap34xx_soc_data, },
> >> { .compatible = "ti,omap3630", .data = &omap36xx_soc_data, },
> >>
> >> 5. configure for CONFIG_ARM_TI_CPUFREQ=y
> >>
> >> This should IMHO suffice.
> >
> > If you're OK with what I am proposing, I'll run some tests and submit
> > a patch. I won't promise to fully understand what's happening. :-)
>
> Same for me :)
>
> BR and thanks,
> Nikolaus
>
next prev parent reply other threads:[~2019-09-09 16:20 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20190801012823.28730-1-neolynx@gmail.com>
[not found] ` <CAHCN7x+nD0J6KZYtfH+0ApQTPO5byO2obMkUwc9Uf4WubyRbTw@mail.gmail.com>
[not found] ` <C04F49BA-1229-4E96-9FCF-4FC662D1DB11@goldelico.com>
[not found] ` <CAHCN7x+Ye6sB_YqO0sAX1OJDw64B-qGS3pL545v3Xk5z914cwQ@mail.gmail.com>
[not found] ` <0C1EF64E-B33C-4BFA-A7D3-471DD1B9EE86@goldelico.com>
[not found] ` <515048DE-138D-4400-8168-F2B7D61F1005@goldelico.com>
[not found] ` <CAHCN7xLPCX9rZ0+7KVBiA_bgZ6tg6VeCXqD-UXu+6iwpFMPVrA@mail.gmail.com>
[not found] ` <7B3D1D77-3E8C-444F-90B9-6DF2641178B8@goldelico.com>
[not found] ` <CAHCN7xLW58ggx3CpVL=HdCVHWo6D-MCTB91A_9rtSRoZQ+xJuQ@mail.gmail.com>
2019-09-07 7:37 ` [Letux-kernel] [RFC PATCH 0/3] Enable 1GHz support on omap36xx H. Nikolaus Schaller
2019-09-09 14:26 ` Adam Ford
2019-09-09 14:56 ` H. Nikolaus Schaller
2019-09-09 16:20 ` Adam Ford [this message]
2019-09-09 16:32 ` Adam Ford
2019-09-09 16:32 ` Tony Lindgren
2019-09-09 16:38 ` Adam Ford
2019-09-09 17:03 ` H. Nikolaus Schaller
2019-09-09 16:54 ` H. Nikolaus Schaller
2019-09-09 18:11 ` H. Nikolaus Schaller
2019-09-09 19:13 ` Adam Ford
2019-09-10 16:59 ` H. Nikolaus Schaller
2019-09-10 18:30 ` Adam Ford
2019-09-10 18:51 ` H. Nikolaus Schaller
2019-09-10 19:26 ` H. Nikolaus Schaller
2019-09-10 19:36 ` Adam Ford
2019-09-10 19:55 ` H. Nikolaus Schaller
2019-09-10 20:06 ` Adam Ford
2019-09-11 0:24 ` Adam Ford
2019-09-11 0:41 ` Adam Ford
2019-09-11 5:13 ` H. Nikolaus Schaller
2019-09-11 6:03 ` H. Nikolaus Schaller
2019-09-11 6:49 ` H. Nikolaus Schaller
2019-09-11 12:43 ` Adam Ford
2019-09-11 15:46 ` H. Nikolaus Schaller
2019-09-11 15:56 ` Adam Ford
2019-09-11 16:01 ` H. Nikolaus Schaller
2019-09-11 17:43 ` H. Nikolaus Schaller
2019-09-11 17:49 ` Adam Ford
2019-09-12 13:58 ` Adam Ford
2019-09-12 18:52 ` Adam Ford
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAHCN7xJe+sBVZHfXN0ZEaepxAR8RGuMTVGA4GMR5Nqn5LG2TeA@mail.gmail.com \
--to=aford173@gmail.com \
--cc=andreas@kemnade.info \
--cc=hns@goldelico.com \
--cc=letux-kernel@openphoenux.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-omap@vger.kernel.org \
--cc=neolynx@gmail.com \
--cc=tony@atomide.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).