| Message ID | 1585371871-29406-2-git-send-email-sunil@amarulasolutions.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | Fix ATF v2.2 build for rk3399, add roc-rk3399-pc board | expand |
Hello, I've applied your patch, but after doing a number of changes, see below. On Sat, 28 Mar 2020 10:34:29 +0530 sunil@amarulasolutions.com wrote: > DEVELOPERS | 3 +++ > package/Config.in.host | 1 + > package/arm-gnu-a-toolchain/Config.in.host | 9 ++++++++ I've entirely dropped the visible Config.in.host option for this package, it is merely a build dependency of ATF, so we don't need (at this point at least) a visible option. > +N: Suniel Mahesh <sunil@amarulasolutions.com> > +F: package/arm-gnu-a-toolchain Directories end with a / in DEVELOPERS. > +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN > + bool "host arm-gnu-a-toolchain" > + help > + arm trusted firmware requires a bare metal toolchain to build some platforms > + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. > + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for > + the host machine. Lines where way too long: you forgot to run "make check-package", which would have pointed out this coding style issue. Of course, since I dropped the Config.in.host, it was not a big issue. > --- /dev/null > +++ b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk > @@ -0,0 +1,27 @@ > +################################################################################ > +# > +# arm-gnu-a-toolchain > +# > +################################################################################ > + > +ARM_GNU_A_TOOLCHAIN_SITE = https://developer.arm.com/-/media/Files/downloads/gnu-a/9.2-2019.12/binrel You should have used the <pkg>_VERSION variable here instead of hardcoding 9.2-2019.12. > +ARM_GNU_A_TOOLCHAIN_VERSION = 9.2-2019.12 > +ARM_GNU_A_TOOLCHAIN_SOURCE = gcc-arm-$(ARM_GNU_A_TOOLCHAIN_VERSION)-x86_64-arm-none-eabi.tar.xz > +ARM_GNU_A_TOOLCHAIN_LICENSE = GPL-3.0+ > +ARM_GNU_A_TOOLCHAIN_LICENSE_FILES = I've dropped this last variable, since it's empty. > +HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR = $(HOST_DIR)/opt/gcc-arm-none-eabi > + > +define HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_CMDS > + rm -rf $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > + mkdir -p $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > + cp -rf $(@D)/* $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > + > + cd $(HOST_DIR)/bin && \ You need to create $(HOST_DIR)/bin before cd-ing into it, as this directory may not exist. > + for i in $(HOST_DIR)/opt/gcc-arm-none-eabi/bin/*; do \ > + ln -sf $$(echo $$i | sed 's%^$(HOST_DIR)%..%') .; \ Simplified as: cd $(HOST_DIR)/bin && \ for i in ../opt/gcc-arm-none-eabi/bin/*; do \ ln -sf $$i; \ done I was also not sure these symlinks where really useful, it would have been possible to pass an explicit path to the toolchain when building ATF. But ok, I've kept the symlinks, we'll see how it goes. Thanks! Thomas
Hi On Sun, Mar 29, 2020 at 3:14 PM Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: > > Hello, > > I've applied your patch, but after doing a number of changes, see below. > > On Sat, 28 Mar 2020 10:34:29 +0530 > sunil@amarulasolutions.com wrote: > > > DEVELOPERS | 3 +++ > > package/Config.in.host | 1 + > > package/arm-gnu-a-toolchain/Config.in.host | 9 ++++++++ > > I've entirely dropped the visible Config.in.host option for this > package, it is merely a build dependency of ATF, so we don't need (at > this point at least) a visible option. > > > +N: Suniel Mahesh <sunil@amarulasolutions.com> > > +F: package/arm-gnu-a-toolchain > > Directories end with a / in DEVELOPERS. > > > +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN > > + bool "host arm-gnu-a-toolchain" > > + help > > + arm trusted firmware requires a bare metal toolchain to build some platforms > > + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. > > + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for > > + the host machine. > > Lines where way too long: you forgot to run "make check-package", which > would have pointed out this coding style issue. Of course, since I > dropped the Config.in.host, it was not a big issue. Does it mean that with this we can enable same for nanopi NEO4 (which was dropped in 8af7b11bd8a95bc41e4de46f727f3b0afde465a4) with comment: This defconfig tries to build an ARM Trusted Firmware version that needs an ARM32 toolchain, which is not available as the platform is an ARM64 one. The correct solution for this is to have a package in Buildroot for an ARM32 bare-metal toolchain, but this wasn't done in time for the 2019.08 release. but when I just reverted mentioned commit NEO4 boots fine without issue. So I'm bit confused if I need to add support for NEO4 with this changes to NEO4 and resurrect it? Thanks. > > > --- /dev/null > > +++ b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk > > @@ -0,0 +1,27 @@ > > +################################################################################ > > +# > > +# arm-gnu-a-toolchain > > +# > > +################################################################################ > > + > > +ARM_GNU_A_TOOLCHAIN_SITE = https://developer.arm.com/-/media/Files/downloads/gnu-a/9.2-2019.12/binrel > > You should have used the <pkg>_VERSION variable here instead of hardcoding 9.2-2019.12. > > > +ARM_GNU_A_TOOLCHAIN_VERSION = 9.2-2019.12 > > +ARM_GNU_A_TOOLCHAIN_SOURCE = gcc-arm-$(ARM_GNU_A_TOOLCHAIN_VERSION)-x86_64-arm-none-eabi.tar.xz > > +ARM_GNU_A_TOOLCHAIN_LICENSE = GPL-3.0+ > > +ARM_GNU_A_TOOLCHAIN_LICENSE_FILES = > > I've dropped this last variable, since it's empty. > > > +HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR = $(HOST_DIR)/opt/gcc-arm-none-eabi > > + > > +define HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_CMDS > > + rm -rf $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + mkdir -p $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + cp -rf $(@D)/* $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + > > + cd $(HOST_DIR)/bin && \ > > You need to create $(HOST_DIR)/bin before cd-ing into it, as this > directory may not exist. > > > + for i in $(HOST_DIR)/opt/gcc-arm-none-eabi/bin/*; do \ > > + ln -sf $$(echo $$i | sed 's%^$(HOST_DIR)%..%') .; \ > > Simplified as: > > cd $(HOST_DIR)/bin && \ > for i in ../opt/gcc-arm-none-eabi/bin/*; do \ > ln -sf $$i; \ > done > > I was also not sure these symlinks where really useful, it would have > been possible to pass an explicit path to the toolchain when building > ATF. But ok, I've kept the symlinks, we'll see how it goes. > > Thanks! > > Thomas > -- > Thomas Petazzoni, CTO, Bootlin > Embedded Linux and Kernel engineering > https://bootlin.com > _______________________________________________ > buildroot mailing list > buildroot@busybox.net > http://lists.busybox.net/mailman/listinfo/buildroot BR, marek
> Hi > > On Sun, Mar 29, 2020 at 3:14 PM Thomas Petazzoni > <thomas.petazzoni@bootlin.com> wrote: > > > > Hello, > > > > I've applied your patch, but after doing a number of changes, see below. > > > > On Sat, 28 Mar 2020 10:34:29 +0530 > > sunil@amarulasolutions.com wrote: > > > > > DEVELOPERS | 3 +++ > > > package/Config.in.host | 1 + > > > package/arm-gnu-a-toolchain/Config.in.host | 9 ++++++++ > > > > I've entirely dropped the visible Config.in.host option for this > > package, it is merely a build dependency of ATF, so we don't need (at > > this point at least) a visible option. > > > > > +N: Suniel Mahesh <sunil@amarulasolutions.com> > > > +F: package/arm-gnu-a-toolchain > > > > Directories end with a / in DEVELOPERS. > > > > > +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN > > > + bool "host arm-gnu-a-toolchain" > > > + help > > > + arm trusted firmware requires a bare metal toolchain to build some platforms > > > + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. > > > + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for > > > + the host machine. > > > > Lines where way too long: you forgot to run "make check-package", which > > would have pointed out this coding style issue. Of course, since I > > dropped the Config.in.host, it was not a big issue. > Does it mean that with this we can enable same for nanopi NEO4 (which > was dropped in 8af7b11bd8a95bc41e4de46f727f3b0afde465a4) > with comment: > > This defconfig tries to build an ARM Trusted Firmware version that > needs an ARM32 toolchain, which is not available as the platform is an > ARM64 one. The correct solution for this is to have a package in > Buildroot for an ARM32 bare-metal toolchain, but this wasn't done in > time for the 2019.08 release. > > but when I just reverted mentioned commit NEO4 boots fine without > issue. So I'm bit confused if I need to add support for NEO4 with this > changes to NEO4 and resurrect it? Thanks. You mean if you revert that commit w/o any changes, NEO4 builds and boots fine ? Probably you have already installed arm32 toolchain on your build server, that is why it works fine for you. Looking at the patches from Sunil, you have to add BR2_TARGET_ARM_TRUSTED_FIRMWARE_NEEDS_ARM32_TOOLCHAIN=y into ATF section of NEO4 board defconfig. Regards, Sergey
On Mon, Mar 30, 2020 at 9:47 AM Sergey Matyukevich <geomatsi@gmail.com> wrote: > > > Hi > > > > On Sun, Mar 29, 2020 at 3:14 PM Thomas Petazzoni > > <thomas.petazzoni@bootlin.com> wrote: > > > > > > Hello, > > > > > > I've applied your patch, but after doing a number of changes, see below. > > > > > > On Sat, 28 Mar 2020 10:34:29 +0530 > > > sunil@amarulasolutions.com wrote: > > > > > > > DEVELOPERS | 3 +++ > > > > package/Config.in.host | 1 + > > > > package/arm-gnu-a-toolchain/Config.in.host | 9 ++++++++ > > > > > > I've entirely dropped the visible Config.in.host option for this > > > package, it is merely a build dependency of ATF, so we don't need (at > > > this point at least) a visible option. > > > > > > > +N: Suniel Mahesh <sunil@amarulasolutions.com> > > > > +F: package/arm-gnu-a-toolchain > > > > > > Directories end with a / in DEVELOPERS. > > > > > > > +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN > > > > + bool "host arm-gnu-a-toolchain" > > > > + help > > > > + arm trusted firmware requires a bare metal toolchain to build some platforms > > > > + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. > > > > + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for > > > > + the host machine. > > > > > > Lines where way too long: you forgot to run "make check-package", which > > > would have pointed out this coding style issue. Of course, since I > > > dropped the Config.in.host, it was not a big issue. > > Does it mean that with this we can enable same for nanopi NEO4 (which > > was dropped in 8af7b11bd8a95bc41e4de46f727f3b0afde465a4) > > with comment: > > > > This defconfig tries to build an ARM Trusted Firmware version that > > needs an ARM32 toolchain, which is not available as the platform is an > > ARM64 one. The correct solution for this is to have a package in > > Buildroot for an ARM32 bare-metal toolchain, but this wasn't done in > > time for the 2019.08 release. > > > > but when I just reverted mentioned commit NEO4 boots fine without > > issue. So I'm bit confused if I need to add support for NEO4 with this > > changes to NEO4 and resurrect it? Thanks. > > You mean if you revert that commit w/o any changes, NEO4 builds and > boots fine ? Probably you have already installed arm32 toolchain on > your build server, that is why it works fine for you. OK taht could be an option as I have arm-none-eabi installed on my laptop and that is the reason why it works ;). > > Looking at the patches from Sunil, you have to add > BR2_TARGET_ARM_TRUSTED_FIRMWARE_NEEDS_ARM32_TOOLCHAIN=y > into ATF section of NEO4 board defconfig. OK I'll revert patch which remove NEO4 board and add this option to defconfig. Thanks a lot. > > Regards, > Sergey BR, marek
On Sun, Mar 29, 2020 at 6:44 PM Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: > > Hello, > > I've applied your patch, but after doing a number of changes, see below. Thanks for all the changes and suggestions. Next time I will make sure that I do make check-package before submitting. Suniel > > On Sat, 28 Mar 2020 10:34:29 +0530 > sunil@amarulasolutions.com wrote: > > > DEVELOPERS | 3 +++ > > package/Config.in.host | 1 + > > package/arm-gnu-a-toolchain/Config.in.host | 9 ++++++++ > > I've entirely dropped the visible Config.in.host option for this > package, it is merely a build dependency of ATF, so we don't need (at > this point at least) a visible option. > > > +N: Suniel Mahesh <sunil@amarulasolutions.com> > > +F: package/arm-gnu-a-toolchain > > Directories end with a / in DEVELOPERS. > > > +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN > > + bool "host arm-gnu-a-toolchain" > > + help > > + arm trusted firmware requires a bare metal toolchain to build some platforms > > + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. > > + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for > > + the host machine. > > Lines where way too long: you forgot to run "make check-package", which > would have pointed out this coding style issue. Of course, since I > dropped the Config.in.host, it was not a big issue. > > > --- /dev/null > > +++ b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk > > @@ -0,0 +1,27 @@ > > +################################################################################ > > +# > > +# arm-gnu-a-toolchain > > +# > > +################################################################################ > > + > > +ARM_GNU_A_TOOLCHAIN_SITE = https://developer.arm.com/-/media/Files/downloads/gnu-a/9.2-2019.12/binrel > > You should have used the <pkg>_VERSION variable here instead of hardcoding 9.2-2019.12. > > > +ARM_GNU_A_TOOLCHAIN_VERSION = 9.2-2019.12 > > +ARM_GNU_A_TOOLCHAIN_SOURCE = gcc-arm-$(ARM_GNU_A_TOOLCHAIN_VERSION)-x86_64-arm-none-eabi.tar.xz > > +ARM_GNU_A_TOOLCHAIN_LICENSE = GPL-3.0+ > > +ARM_GNU_A_TOOLCHAIN_LICENSE_FILES = > > I've dropped this last variable, since it's empty. > > > +HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR = $(HOST_DIR)/opt/gcc-arm-none-eabi > > + > > +define HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_CMDS > > + rm -rf $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + mkdir -p $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + cp -rf $(@D)/* $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) > > + > > + cd $(HOST_DIR)/bin && \ > > You need to create $(HOST_DIR)/bin before cd-ing into it, as this > directory may not exist. > > > + for i in $(HOST_DIR)/opt/gcc-arm-none-eabi/bin/*; do \ > > + ln -sf $$(echo $$i | sed 's%^$(HOST_DIR)%..%') .; \ > > Simplified as: > > cd $(HOST_DIR)/bin && \ > for i in ../opt/gcc-arm-none-eabi/bin/*; do \ > ln -sf $$i; \ > done > > I was also not sure these symlinks where really useful, it would have > been possible to pass an explicit path to the toolchain when building > ATF. But ok, I've kept the symlinks, we'll see how it goes. > > Thanks! > > Thomas > -- > Thomas Petazzoni, CTO, Bootlin > Embedded Linux and Kernel engineering > https://bootlin.com
diff --git a/DEVELOPERS b/DEVELOPERS index 1fb4e65..ce1a761 100644 --- a/DEVELOPERS +++ b/DEVELOPERS @@ -2389,6 +2389,9 @@ N: Steven Noonan <steven@uplinklabs.net> F: package/hwloc/ F: package/powertop/ +N: Suniel Mahesh <sunil@amarulasolutions.com> +F: package/arm-gnu-a-toolchain + N: Sven Haardiek <sven.haardiek@iotec-gmbh.de> F: package/lcdproc/ F: package/python-influxdb/ diff --git a/package/Config.in.host b/package/Config.in.host index eb8f63f..19dcab6 100644 --- a/package/Config.in.host +++ b/package/Config.in.host @@ -2,6 +2,7 @@ menu "Host utilities" source "package/aespipe/Config.in.host" source "package/android-tools/Config.in.host" + source "package/arm-gnu-a-toolchain/Config.in.host" source "package/btrfs-progs/Config.in.host" source "package/cargo/Config.in.host" source "package/cbootimage/Config.in.host" diff --git a/package/arm-gnu-a-toolchain/Config.in.host b/package/arm-gnu-a-toolchain/Config.in.host new file mode 100644 index 0000000..07b73a7 --- /dev/null +++ b/package/arm-gnu-a-toolchain/Config.in.host @@ -0,0 +1,9 @@ +config BR2_PACKAGE_HOST_ARM_GNU_A_TOOLCHAIN + bool "host arm-gnu-a-toolchain" + help + arm trusted firmware requires a bare metal toolchain to build some platforms + which host cortex-m series core, for instance rockchip rk3399 has a cortex-m0 core. + This package installs a pre-built cross-compilation ARM-A bare metal toolchain for + the host machine. + + https://developer.arm.com/-/media/Files/downloads/gnu-a/9.2-2019.12/binrel/gcc-arm-9.2-2019.12-x86_64-arm-none-eabi.tar.xz diff --git a/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.hash b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.hash new file mode 100644 index 0000000..b18e56c --- /dev/null +++ b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.hash @@ -0,0 +1,2 @@ +# Locally calculated +sha256 ac952d89ae0fc3543e81099e7d34917efc621f5def112eee843fd1ce755eca8c gcc-arm-9.2-2019.12-x86_64-arm-none-eabi.tar.xz diff --git a/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk new file mode 100644 index 0000000..cbe8aba --- /dev/null +++ b/package/arm-gnu-a-toolchain/arm-gnu-a-toolchain.mk @@ -0,0 +1,27 @@ +################################################################################ +# +# arm-gnu-a-toolchain +# +################################################################################ + +ARM_GNU_A_TOOLCHAIN_SITE = https://developer.arm.com/-/media/Files/downloads/gnu-a/9.2-2019.12/binrel +ARM_GNU_A_TOOLCHAIN_VERSION = 9.2-2019.12 +ARM_GNU_A_TOOLCHAIN_SOURCE = gcc-arm-$(ARM_GNU_A_TOOLCHAIN_VERSION)-x86_64-arm-none-eabi.tar.xz +ARM_GNU_A_TOOLCHAIN_LICENSE = GPL-3.0+ +ARM_GNU_A_TOOLCHAIN_LICENSE_FILES = + +HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR = $(HOST_DIR)/opt/gcc-arm-none-eabi + +define HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_CMDS + rm -rf $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) + mkdir -p $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) + cp -rf $(@D)/* $(HOST_ARM_GNU_A_TOOLCHAIN_INSTALL_DIR) + + cd $(HOST_DIR)/bin && \ + for i in $(HOST_DIR)/opt/gcc-arm-none-eabi/bin/*; do \ + ln -sf $$(echo $$i | sed 's%^$(HOST_DIR)%..%') .; \ + done + +endef + +$(eval $(host-generic-package))