| Message ID | 1476298306-9138-1-git-send-email-scott.branden@broadcom.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden <scott.branden@broadcom.com> wrote: > Enable support for on board SPI EEPROM by turning on > CONFIG_EEPROM_AT25. This needs to be on in order to > boot and test the kernel with a static rootfs image > that is not rebuilt everytime the kernel is rebuilt. If we did this for every kernel option we'd get a huge kernel. In general, we've said that static options for what's needed to boot to rootfs (i.e. storage and network drivers for nfsroot) are fine to enable statically. I doubt you need the EEPROM driver to boot to rootfs on your system, so please enable it as a module instead. Look into using config fragments in case you need to modify the options for local builds, it should be a convenient way to have a small delta to apply to fit your internal needs, instead of completely forking the config file. -Olof
Hi Olof, On 16-10-17 02:58 PM, Olof Johansson wrote: > Hi, > > On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden > <scott.branden@broadcom.com> wrote: >> Enable support for on board SPI EEPROM by turning on >> CONFIG_EEPROM_AT25. This needs to be on in order to >> boot and test the kernel with a static rootfs image >> that is not rebuilt everytime the kernel is rebuilt. > > If we did this for every kernel option we'd get a huge kernel. > > In general, we've said that static options for what's needed to boot > to rootfs (i.e. storage and network drivers for nfsroot) are fine to > enable statically. > > I doubt you need the EEPROM driver to boot to rootfs on your system, > so please enable it as a module instead. > > Look into using config fragments in case you need to modify the > options for local builds, it should be a convenient way to have a > small delta to apply to fit your internal needs, instead of completely > forking the config file. Do you allow such config fragments to be upstreamed or do we need to maintain these in our tree? > > > -Olof > Thanks, Scott
On Mon, Oct 17, 2016 at 4:24 PM, Scott Branden <scott.branden@broadcom.com> wrote: > Hi Olof, > > On 16-10-17 02:58 PM, Olof Johansson wrote: >> >> Hi, >> >> On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden >> <scott.branden@broadcom.com> wrote: >>> >>> Enable support for on board SPI EEPROM by turning on >>> CONFIG_EEPROM_AT25. This needs to be on in order to >>> boot and test the kernel with a static rootfs image >>> that is not rebuilt everytime the kernel is rebuilt. >> >> >> If we did this for every kernel option we'd get a huge kernel. >> >> In general, we've said that static options for what's needed to boot >> to rootfs (i.e. storage and network drivers for nfsroot) are fine to >> enable statically. >> >> I doubt you need the EEPROM driver to boot to rootfs on your system, >> so please enable it as a module instead. >> >> Look into using config fragments in case you need to modify the >> options for local builds, it should be a convenient way to have a >> small delta to apply to fit your internal needs, instead of completely >> forking the config file. > > > Do you allow such config fragments to be upstreamed or do we need to > maintain these in our tree? There's no place for them upstream. Maintain locally or in a separate repo. -Olof
Hi Olof, On 16-10-17 05:04 PM, Olof Johansson wrote: > On Mon, Oct 17, 2016 at 4:24 PM, Scott Branden > <scott.branden@broadcom.com> wrote: >> Hi Olof, >> >> On 16-10-17 02:58 PM, Olof Johansson wrote: >>> >>> Hi, >>> >>> On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden >>> <scott.branden@broadcom.com> wrote: >>>> >>>> Enable support for on board SPI EEPROM by turning on >>>> CONFIG_EEPROM_AT25. This needs to be on in order to >>>> boot and test the kernel with a static rootfs image >>>> that is not rebuilt everytime the kernel is rebuilt. >>> >>> >>> If we did this for every kernel option we'd get a huge kernel. >>> >>> In general, we've said that static options for what's needed to boot >>> to rootfs (i.e. storage and network drivers for nfsroot) are fine to >>> enable statically. >>> >>> I doubt you need the EEPROM driver to boot to rootfs on your system, >>> so please enable it as a module instead. OK, I will upstream as module and need config fragments maintained locally in order to test defconfig on our test setup. >>> >>> Look into using config fragments in case you need to modify the >>> options for local builds, it should be a convenient way to have a >>> small delta to apply to fit your internal needs, instead of completely >>> forking the config file. >> >> >> Do you allow such config fragments to be upstreamed or do we need to >> maintain these in our tree? > > There's no place for them upstream. Maintain locally or in a separate repo. If that is the case - shall we cleanup arch/arm/configs and delete dram_0x00000000.config that was introduced in the 4.4 kernel? > > > -Olof > Thanks, Scott
On Tue, Oct 18, 2016 at 1:38 PM, Scott Branden <scott.branden@broadcom.com> wrote: > Hi Olof, > > On 16-10-17 05:04 PM, Olof Johansson wrote: >> >> On Mon, Oct 17, 2016 at 4:24 PM, Scott Branden >> <scott.branden@broadcom.com> wrote: >>> >>> Hi Olof, >>> >>> On 16-10-17 02:58 PM, Olof Johansson wrote: >>>> >>>> >>>> Hi, >>>> >>>> On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden >>>> <scott.branden@broadcom.com> wrote: >>>>> >>>>> >>>>> Enable support for on board SPI EEPROM by turning on >>>>> CONFIG_EEPROM_AT25. This needs to be on in order to >>>>> boot and test the kernel with a static rootfs image >>>>> that is not rebuilt everytime the kernel is rebuilt. >>>> >>>> >>>> >>>> If we did this for every kernel option we'd get a huge kernel. >>>> >>>> In general, we've said that static options for what's needed to boot >>>> to rootfs (i.e. storage and network drivers for nfsroot) are fine to >>>> enable statically. >>>> >>>> I doubt you need the EEPROM driver to boot to rootfs on your system, >>>> so please enable it as a module instead. > > OK, I will upstream as module and need config fragments maintained locally > in order to test defconfig on our test setup. Great. >>>> >>>> >>>> Look into using config fragments in case you need to modify the >>>> options for local builds, it should be a convenient way to have a >>>> small delta to apply to fit your internal needs, instead of completely >>>> forking the config file. >>> >>> >>> >>> Do you allow such config fragments to be upstreamed or do we need to >>> maintain these in our tree? >> >> >> There's no place for them upstream. Maintain locally or in a separate >> repo. > > If that is the case - shall we cleanup arch/arm/configs and delete > dram_0x00000000.config that was introduced in the 4.4 kernel? That one is a bit different, in that it allows us to do defconfig consolidation, and as such keeps the number of defconfigs needed down. In particular, see the email from Arnd here: http://marc.info/?l=linux-arm-kernel&m=145700132713703 -Olof
On 16-10-18 01:48 PM, Olof Johansson wrote: > On Tue, Oct 18, 2016 at 1:38 PM, Scott Branden > <scott.branden@broadcom.com> wrote: >> Hi Olof, >> >> On 16-10-17 05:04 PM, Olof Johansson wrote: >>> >>> On Mon, Oct 17, 2016 at 4:24 PM, Scott Branden >>> <scott.branden@broadcom.com> wrote: >>>> >>>> Hi Olof, >>>> >>>> On 16-10-17 02:58 PM, Olof Johansson wrote: >>>>> >>>>> >>>>> Hi, >>>>> >>>>> On Wed, Oct 12, 2016 at 11:51 AM, Scott Branden >>>>> <scott.branden@broadcom.com> wrote: >>>>>> >>>>>> >>>>>> Enable support for on board SPI EEPROM by turning on >>>>>> CONFIG_EEPROM_AT25. This needs to be on in order to >>>>>> boot and test the kernel with a static rootfs image >>>>>> that is not rebuilt everytime the kernel is rebuilt. >>>>> >>>>> >>>>> >>>>> If we did this for every kernel option we'd get a huge kernel. >>>>> >>>>> In general, we've said that static options for what's needed to boot >>>>> to rootfs (i.e. storage and network drivers for nfsroot) are fine to >>>>> enable statically. >>>>> >>>>> I doubt you need the EEPROM driver to boot to rootfs on your system, >>>>> so please enable it as a module instead. >> >> OK, I will upstream as module and need config fragments maintained locally >> in order to test defconfig on our test setup. > > Great. > >>>>> >>>>> >>>>> Look into using config fragments in case you need to modify the >>>>> options for local builds, it should be a convenient way to have a >>>>> small delta to apply to fit your internal needs, instead of completely >>>>> forking the config file. >>>> >>>> >>>> >>>> Do you allow such config fragments to be upstreamed or do we need to >>>> maintain these in our tree? >>> >>> >>> There's no place for them upstream. Maintain locally or in a separate >>> repo. >> >> If that is the case - shall we cleanup arch/arm/configs and delete >> dram_0x00000000.config that was introduced in the 4.4 kernel? > > That one is a bit different, in that it allows us to do defconfig > consolidation, and as such keeps the number of defconfigs needed down. > > In particular, see the email from Arnd here: > > http://marc.info/?l=linux-arm-kernel&m=145700132713703 Thanks for link. I have be.config and le.config that allow you to switch the defconfig between big and little endian. Does this make sense to upstream to arm/configs if you have accepted dram_0x00000000.config? Would you also accept this to arm64/configs? We actually use big and little endian on the same SoC more on arm64 platforms. But, in order to boot big endian we need to maintain this outside the kernel right now. > > > -Olof >
On Tuesday, October 18, 2016 3:23:26 PM CEST Scott Branden wrote: > I have be.config and le.config that allow you to switch the defconfig > between big and little endian. Does this make sense to upstream to > arm/configs if you have accepted dram_0x00000000.config? Yes, they clearly fall into the same category, let's merge those as well. > Would you also accept this to arm64/configs? We actually use > big and little endian on the same SoC more on arm64 platforms. But, in > order to boot big endian we need to maintain this outside the kernel > right now. I'm in favor of that, but let's see what the arm64 maintainers think. Arnd
On Wed, Oct 19, 2016 at 1:04 AM, Arnd Bergmann <arnd@arndb.de> wrote: > On Tuesday, October 18, 2016 3:23:26 PM CEST Scott Branden wrote: >> I have be.config and le.config that allow you to switch the defconfig >> between big and little endian. Does this make sense to upstream to >> arm/configs if you have accepted dram_0x00000000.config? > > Yes, they clearly fall into the same category, let's merge those as well. > >> Would you also accept this to arm64/configs? We actually use >> big and little endian on the same SoC more on arm64 platforms. But, in >> order to boot big endian we need to maintain this outside the kernel >> right now. > > I'm in favor of that, but let's see what the arm64 maintainers think. Single-line fragments aren't really all that valuable, IMHO. Flipping just one option is trivial to do without fragments, or when they're not simple Y/N flips (i.e. like the ram base). Fragments are mostly useful when you need to flip several options together to enable some set of functionality. -Olof
diff --git a/arch/arm64/configs/defconfig b/arch/arm64/configs/defconfig index eadf485..9955ee1 100644 --- a/arch/arm64/configs/defconfig +++ b/arch/arm64/configs/defconfig @@ -136,6 +136,7 @@ CONFIG_MTD_SPI_NOR=y CONFIG_BLK_DEV_LOOP=y CONFIG_BLK_DEV_NBD=m CONFIG_VIRTIO_BLK=y +CONFIG_EEPROM_AT25=y CONFIG_SRAM=y # CONFIG_SCSI_PROC_FS is not set CONFIG_BLK_DEV_SD=y
Enable support for on board SPI EEPROM by turning on CONFIG_EEPROM_AT25. This needs to be on in order to boot and test the kernel with a static rootfs image that is not rebuilt everytime the kernel is rebuilt. Signed-off-by: Scott Branden <scott.branden@broadcom.com> --- arch/arm64/configs/defconfig | 1 + 1 file changed, 1 insertion(+)