From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Hurley Subject: Re: [PATCH 2/4] of: DT quirks infrastructure Date: Fri, 20 Feb 2015 09:21:38 -0500 Message-ID: <54E742F2.80506@hurleysoftware.com> References: <1424271576-1952-1-git-send-email-pantelis.antoniou@konsulko.com> <1424271576-1952-3-git-send-email-pantelis.antoniou@konsulko.com> <20150218154106.GC29429@leverpostej> <20150218173115.GG29429@leverpostej> <76BD1B22-BAED-4205-9B34-186907CE0217@konsulko.com> <54E613E7.2020405@gmail.com> <670D0881-DBF0-45E8-A502-A6DB2B77A750@konsulko.com> <54E61DD2.3060002@gmail.com> <53F2F94C-0C43-4A54-B8CD-EEC454A0AC19@konsulko.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <53F2F94C-0C43-4A54-B8CD-EEC454A0AC19@konsulko.com> Sender: linux-kernel-owner@vger.kernel.org To: Pantelis Antoniou , frowand.list@gmail.com Cc: Mark Rutland , "devicetree@vger.kernel.org" , Tony Lindgren , Koen Kooi , Nicolas Ferre , "linux-kernel@vger.kernel.org" , Grant Likely , Ludovic Desroches , "linux-arm-kernel@lists.infradead.org" , Matt Porter , Guenter Roeck List-Id: devicetree@vger.kernel.org On 02/19/2015 12:38 PM, Pantelis Antoniou wrote: >=20 >> On Feb 19, 2015, at 19:30 , Frank Rowand wr= ote: >> >> On 2/19/2015 9:00 AM, Pantelis Antoniou wrote: >>> Hi Frank, >>> >>>> On Feb 19, 2015, at 18:48 , Frank Rowand = wrote: >>>> >>>> On 2/19/2015 6:29 AM, Pantelis Antoniou wrote: >>>>> Hi Mark, >>>>> >>>>>> On Feb 18, 2015, at 19:31 , Mark Rutland = wrote: >>>>>> >>>>>>>>> +While this may in theory work, in practice it is very cumber= some >>>>>>>>> +for the following reasons: >>>>>>>>> + >>>>>>>>> +1. The act of selecting a different boot device tree blob re= quires >>>>>>>>> +a reasonably advanced bootloader with some kind of configura= tion or >>>>>>>>> +scripting capabilities. Sadly this is not the case many time= s, the >>>>>>>>> +bootloader is extremely dumb and can only use a single dt bl= ob. >>>>>>>> >>>>>>>> You can have several bootloader builds, or even a single build= with >>>>>>>> something like appended DTB to get an appropriate DTB if the s= ame binary >>>>>>>> will otherwise work across all variants of a board. >>>>>>>> >>>>>>> >>>>>>> No, the same DTB will not work across all the variants of a boa= rd. >>>>>> >>>>>> I wasn't on about the DTB. I was on about the loader binary, in = the case >>>>>> the FW/bootloader could be common even if the DTB couldn't. >>>>>> >>>>>> To some extent there must be a DTB that will work across all var= iants >>>>>> (albeit with limited utility) or the quirk approach wouldn't wor= k=E2=80=A6 >>>>>> >>>>> >>>>> That=E2=80=99s not correct; the only part of the DTB that needs t= o be common >>>>> is the model property that would allow the quirk detection logic = to fire. >>>>> >>>>> So, there is a base DTB that will work on all variants, but that = only means >>>>> that it will work only up to the point that the quirk detector me= thod >>>>> can work. So while in recommended practice there are common subse= ts >>>>> of the DTB that might work, they might be unsafe. >>>>> >>>>> For instance on the beaglebone the regulator configuration is dif= ferent >>>>> between white and black, it is imperative you get them right othe= rwise >>>>> you risk board damage. >>>>> >>>>>>>> So it's not necessarily true that you need a complex bootloade= r. >>>>>>>> >>>>>>> >>>>>>>>> +2. On many instances boot time is extremely critical; in som= e cases >>>>>>>>> +there are hard requirements like having working video feeds = in under >>>>>>>>> +2 seconds from power-up. This leaves an extremely small time= budget for >>>>>>>>> +boot-up, as low as 500ms to kernel entry. The sanest way to = get there >>>>>>>>> +is by removing the standard bootloader from the normal boot = sequence >>>>>>>>> +altogether by having a very small boot shim that loads the k= ernel and >>>>>>>>> +immediately jumps to kernel, like falcon-boot mode in u-boot= does. >>>>>>>> >>>>>>>> Given my previous comments above I don't see why this is relev= ant. >>>>>>>> You're already passing _some_ DTB here, so if you can organise= for the >>>>>>>> board to statically provide a sane DTB that's fine, or you can= resort to >>>>>>>> appended DTB if it's not possible to update the board configur= ation. >>>>>>>> >>>>>>> >>>>>>> You=E2=80=99re missing the point. I can=E2=80=99t use the same = DTB for each revision of the >>>>>>> board. Each board is similar but it=E2=80=99s not identical. >>>>>> >>>>>> I think you've misunderstood my point. If you program the board = with the >>>>>> relevant DTB, or use appended DTB, then you will pass the correc= t DTB to >>>>>> the kernel without need for quirks. >>>>>> >>>>>> I understand that each variant is somewhat incompatible (and hen= ce needs >>>>>> its own DTB). >>>>> >>>>> In theory it might work, in practice this does not. Ludovic menti= oned that they >>>>> have 27 different DTBs in use at the moment. At a relatively comm= on 60k per DTB >>>>> that=E2=80=99s 27x60k =3D 1.6MB of DTBs, that need to be installe= d. >>>> >>>> < snip > >>>> >>>> Or you can install the correct DTB on the board. You trust your m= anufacturing line >>>> to install the correct resistors. You trust your manufacturing li= ne to install the >>>> correct kernel version (eg an updated version to resolve a securit= y issue). >>>> >>>> I thought the DT blob was supposed to follow the same standard tha= t other OS's or >>>> bootloaders understood. Are you willing to break that? (This is = one of those >>>> ripples I mentioned in my other emails.) >>>> >>> >>> Trust no-one. >>> >>> This is one of those things that the kernel community doesn=E2=80=99= t understand which makes people >>> who push product quite mad. >>> >>> Engineering a product is not only about meeting customer spec, in o= rder to turn a profit >>> the whole endeavor must be engineered as well for manufacturability= =2E >>> >>> Yes, you can always manually install files in the bootloader. For 1= board no problem. >>> For 10 doable. For 100 I guess you can hire an extra guy. For 1 mil= lion? Guess what, >>> instead of turning a profit you=E2=80=99re losing money if you only= have a few cents of profit >>> per unit. >> >> I'm not installing physical components manually. Why would I be ins= talling software >> manually? (rhetorical question) >> >=20 > Because on high volume product runs the flash comes preprogrammed and= is soldered as is. >=20 > Having a single binary to flash to every revision of the board makes = logistics considerably > easier. >=20 > Having to boot and tweak the bootloader settings to select the correc= t dtb (even if it=E2=80=99s present > on the flash medium) takes time and is error-prone. >=20 > Factory time =3D=3D money, errors =3D=3D money. >=20 >>> >>> No knobs to tweak means no knobs to break. And a broken knob can ha= ve pretty bad consequences >>> for a few million units.=20 >> >> And you produce a few million units before testing that the first on= e off the line works? >> >=20 > The first one off the line works. The rest will get some burn in and = functional testing if you=E2=80=99re > lucky. In many cases where the product is very cheap it might make fi= nancial sense to just ship > as is and deal with recalls, if you=E2=80=99re reasonably happy after= a little bit of statistical sampling. >=20 > Hardware is hard :) I'm failing to see how this series improves your manufacturing process = at all. 1. Won't you have to provide the factory with different eeprom images f= or the White and Black? You _trust_ them to get that right, or more likely= , you have process control procedures in place so that you don't get 1 mil= lion Blacks flashed with the White eeprom image. 2. The White and Black use different memory technology so it's not as i= f the eMMC from the Black will end up on the White SMT line (or vice versa= ). 3 For that matter, why wouldn't you worry that all the microSD cards i= ntended for the White were accidentally assembled with the first 50,000 Blac= ks; at that point you're losing a lot more than a few cents of profit. And = that has nothing to do with what image you provided. 3. The factory is just as likely to use some other customer's image by = accident, so you're just as likely to have the same failure rate if you have n= o test process at the factory. 4. If you're using offline programming, the image has to be tested afte= r reflow anyway. IOW, your QA process will not change at all =3D=3D same cost. Regards, Peter Hurley