From: Marek Vasut [mailto:marex@xxxxxxx] Sent: Thursday, April 6, 2023 4:10 PM > On 4/6/23 10:37, Christoph Niedermaier wrote: >> From: Marek Vasut [mailto:marex@xxxxxxx] >> Sent: Wednesday, April 5, 2023 8:47 PM >>> On 4/5/23 20:24, Christoph Niedermaier wrote: >>>> From: Marek Vasut [mailto:marex@xxxxxxx] >>>> Sent: Wednesday, April 5, 2023 6:25 PM >>>>> On 4/5/23 18:02, Christoph Niedermaier wrote: >>>>> >>>>> [...] >>>>> >>>>>> +/ { >>>>>> + model = "DH electronics i.MX6ULL DHCOR on maveo box"; >>>>>> + compatible = "dh,imx6ull-dhcor-maveo-box", "dh,imx6ull-dhcor-som", >>>>>> + "fsl,imx6ull"; >>>>>> + >>>>>> + aliases { >>>>>> + /delete-property/ mmc0; /* Avoid double definitions */ >>>>>> + /delete-property/ mmc1; >>>>>> + mmc2 = &usdhc2; /* eMMC should be mmc2 */ >>>>> >>>>> Why not mmc0 ? >>>>> >>>>> Use root=PARTUUID= when booting to avoid any dependency on >>>>> root=/dev/mmcblk2pN enumeration. >>>> >>>> This is due to software interchangeability with the DHCOM >>>> i.MX6ULL, where the eMMC is always mmc2. >>> >>> +CC Ulf , I vaguely recall some discussion about this enumeration and I >>> am not sure one can really depend on that. >> >> That why I think it good to have a defined number for mmcblk devices >> on an embedded system. An excerpt from [1]: > > I might be misremembering this, but could it be that, if any non-OF > SDMMC controller probes early and hogs the /dev/mmcblk2 before the OF > ones have a chance to probe, then the OF ones would fail to probe ? > >> Alternative solutions like PARTUUIDs do not cover the case where multiple >> mmcblk devices contain the same image. > > I agree, this is indeed a downside of PARTUUID . > >> This is a common issue on devices >> that can boot both from eMMC (for regular boot) and SD cards (as a >> temporary boot medium for development). When a firmware image is >> installed to eMMC after a test boot via SD card, there will be no >> reliable way to refer to a specific device using (PART)UUIDs oder >> LABELs > > This can be solved by the installer updating the PARTUUID on the eMMC > however. > >> [1] https://patchwork.kernel.org/project/linux-mmc/patch/20200825134441.17537-2-matthias.schiffer@xxxxxxxxxxxxxxx/ >> >> So far I have never had a problem with numbering mmcblk devices via aliases. > > Based on the above, I don't think either the aliases or PARTUUID is a > perfect solution, but the aliases should be fine for mx6ull at least? > So I think we can conclude this discussion thread ? Yes, I will send a new version with the changes on the first patch. Regards Christoph
