From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Schwierzeck Date: Mon, 29 Apr 2019 16:27:12 +0200 Subject: [U-Boot] [PATCH] mips: rename arch mt7620 to mt7628 In-Reply-To: References: <1556444311-10203-1-git-send-email-weijie.gao@mediatek.com> <9d564739-9b8d-6b98-1d6b-f9bd8bcca222@denx.de> <1556530080.4849.43.camel@mcddlt001> Message-ID: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 8bit To: u-boot@lists.denx.de Am 29.04.19 um 11:40 schrieb Stefan Roese: > On 29.04.19 11:28, Weijie Gao wrote: >> On Mon, 2019-04-29 at 07:08 +0200, Stefan Roese wrote: >>> On 28.04.19 11:38, Weijie Gao wrote: >>>> The MediaTek MT7620 and MT7628 SoCs are different. >>>> Although they use the same memory controller, the lowlevel code (CPU >>>> PLL) >>>> and other peripherals they use are totally different. Which means they >>>> should use seperate mach directories. >>> >>> s/seperate/separate >>> >>>> Currently the mach mt7620 contains only architecture code of MT7628. >>>> In case we add real arch support of MT7620 in the future, the arch >>>> should >>>> be renamed to mt7628, including both Kconfig files and directories. >>>> Other files affected are also modified. >>> >>> Perhaps it would be possible to support both SoC's (MT7620 and >>> MT7628/88) >>> in one mach directory? Frankly I don't know the differences in >>> detail, so >>> its your call. > > > >> Dear Stefan, >> >> Thanks for pointing out the missing files. >> >> Here is the summary of HW components needed by u-boot for MT7620 and >> MT7628: >> >> L1 data cache:            MT7620 can not lock it. >>                            MT7628 uses it to do DDR calibration. >> CPU frequency (PLL):      MT7620 can change it. MT7628 can't. >>                            The PLL registers are different. >> DRAM controller:          Near the same. >>                            MT7620 can't do calibration. >>                            MT7628 has extra PAD configurations. > > These "devices / controllers" are handled in the mach-foo directory. > >> GPIO controller:          Not the same IP core. >> SPI controller:           Not the same IP core. >> Frame engine:             Similar IP core, different generation. >> Built-in Ethernet switch: Not the same IP core. > > And these controllers are handled in the drivers/foo directly. So any > different IP core (between MT7620 and MT7628) has no effect to the > mach directory. > > Please don't misunderstand me. I absolutely agree that we need to > differentiate between those two SoC's. So moving to CONFIG_SOC_MT7628 > instead of SOC_MT7620 makes perfect sense. I only want to avoid the > creation of another mach-foo directory, where code might be shared > between both SoC's. > >> >> So I insist to split them into two mach directory. > > I see. Okay, lets move forward then with your patch and lets finally > decide if and what can be shared between those SoC's, once (if) support > for the MT7620 arrives in mainline. > > BTW: Do you plan on adding support for the MT7620 anytime soon? > I agree with Stefan, there is no need to create separate mach- directories. With the power of Kconfig and Kbuild you can easily handle multiple SoCs within one mach- directory, for instance look at mach-bmips or mach-mscc. Could you rather rename to mach-mediatek or mach-mtmips so that we would have the Kconfig symbols ARCH_MEDIATEK and SOC_MT7628 (plus SOC_MT7620 in the future)? Thanks. BTW: it's good to see that another vendor in the MIPS area is stepping up to maintain its products in mainline ;) -- - Daniel