| Message ID | 20160721151134.5803-1-paul.gortmaker@windriver.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Paul Gortmaker <paul.gortmaker@windriver.com> writes: > Note the output from the following: > > $ git grep STACKPROTECTOR arch/arm/configs/ > arch/arm/configs/aspeed_g4_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > arch/arm/configs/aspeed_g5_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > arch/arm/configs/bcm2835_defconfig:CONFIG_CC_STACKPROTECTOR_REGULAR=y > $ I don't see why rpi1 shoud be special, so for bcm2835: Acked-by: Eric Anholt <eric@anholt.net>
Hi Paul, On Fri, Jul 22, 2016 at 12:41 AM, Paul Gortmaker <paul.gortmaker@windriver.com> wrote: > Note the output from the following: > > $ git grep STACKPROTECTOR arch/arm/configs/ > arch/arm/configs/aspeed_g4_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > arch/arm/configs/aspeed_g5_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > arch/arm/configs/bcm2835_defconfig:CONFIG_CC_STACKPROTECTOR_REGULAR=y > $ > > Only three defconfigs specify a value. And two of the three ask for > the strong variant, which isn't supported by older toolchains. > > Due to the nature of ARM having more platform specific code than say > x86, the allyesconfig and allmodconfig aren't as effective for build > coverage. So, in addition, I like to use a trivial script to walk all > the defconfigs and build each one. > > However I will get false positives on unsupported stackprotector values > with an older toolchain like gcc-4.6.3. As in this instance I am just > using the compiler as a glorified syntax checker on a machine where I > build a bunch of other arch for the same reason, there is no real > motivation to get a newer toolchain for improved optimization etc. I'm happy to remove it from the Aspeed configurations as I'm not sure why it was enabled in the first place. However, I do not agree with the reasoning here. If you're building to check syntax a modern GCC will certainly pick up on more than one from four years ago. > Since there are only three of them, and there is nothing about these > settings that are board/platform specific, I propose we just eliminate > the three existing instances and take the default. This makes sense to me. Acked-by: Joel Stanley <joel@jms.id.au>
[Re: [PATCH] ARM: dont specify STACKPROTECTOR in defconfigs] On 22/07/2016 (Fri 01:40) Joel Stanley wrote: > Hi Paul, > > On Fri, Jul 22, 2016 at 12:41 AM, Paul Gortmaker > <paul.gortmaker@windriver.com> wrote: > > Note the output from the following: > > > > $ git grep STACKPROTECTOR arch/arm/configs/ > > arch/arm/configs/aspeed_g4_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > > arch/arm/configs/aspeed_g5_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y > > arch/arm/configs/bcm2835_defconfig:CONFIG_CC_STACKPROTECTOR_REGULAR=y > > $ > > > > Only three defconfigs specify a value. And two of the three ask for > > the strong variant, which isn't supported by older toolchains. > > > > Due to the nature of ARM having more platform specific code than say > > x86, the allyesconfig and allmodconfig aren't as effective for build > > coverage. So, in addition, I like to use a trivial script to walk all > > the defconfigs and build each one. > > > > However I will get false positives on unsupported stackprotector values > > with an older toolchain like gcc-4.6.3. As in this instance I am just > > using the compiler as a glorified syntax checker on a machine where I > > build a bunch of other arch for the same reason, there is no real > > motivation to get a newer toolchain for improved optimization etc. > > I'm happy to remove it from the Aspeed configurations as I'm not sure > why it was enabled in the first place. > > However, I do not agree with the reasoning here. If you're building to > check syntax a modern GCC will certainly pick up on more than one from > four years ago. Just to clarify, syntax in this case is just for fat fingered typos and ensuring functions resolve with the appropriate header includes. If I was coding new stuff specifically for ARM, then that would be different. > > > Since there are only three of them, and there is nothing about these > > settings that are board/platform specific, I propose we just eliminate > > the three existing instances and take the default. > > This makes sense to me. > > Acked-by: Joel Stanley <joel@jms.id.au> Thanks, Paul. --
diff --git a/arch/arm/configs/aspeed_g4_defconfig b/arch/arm/configs/aspeed_g4_defconfig index b6e54ee9bdbd..78dae22c56de 100644 --- a/arch/arm/configs/aspeed_g4_defconfig +++ b/arch/arm/configs/aspeed_g4_defconfig @@ -18,7 +18,6 @@ CONFIG_BPF_SYSCALL=y CONFIG_EMBEDDED=y # CONFIG_COMPAT_BRK is not set CONFIG_SLAB=y -CONFIG_CC_STACKPROTECTOR_STRONG=y CONFIG_MODULES=y CONFIG_MODULE_UNLOAD=y # CONFIG_BLOCK is not set diff --git a/arch/arm/configs/aspeed_g5_defconfig b/arch/arm/configs/aspeed_g5_defconfig index 892605167357..2253a09cc3c2 100644 --- a/arch/arm/configs/aspeed_g5_defconfig +++ b/arch/arm/configs/aspeed_g5_defconfig @@ -18,7 +18,6 @@ CONFIG_BPF_SYSCALL=y CONFIG_EMBEDDED=y # CONFIG_COMPAT_BRK is not set CONFIG_SLAB=y -CONFIG_CC_STACKPROTECTOR_STRONG=y CONFIG_MODULES=y CONFIG_MODULE_UNLOAD=y # CONFIG_BLOCK is not set diff --git a/arch/arm/configs/bcm2835_defconfig b/arch/arm/configs/bcm2835_defconfig index cc55e4252fda..8c06b380c561 100644 --- a/arch/arm/configs/bcm2835_defconfig +++ b/arch/arm/configs/bcm2835_defconfig @@ -24,7 +24,6 @@ CONFIG_EMBEDDED=y CONFIG_PROFILING=y CONFIG_OPROFILE=y CONFIG_JUMP_LABEL=y -CONFIG_CC_STACKPROTECTOR_REGULAR=y CONFIG_ARCH_MULTI_V6=y CONFIG_ARCH_BCM=y CONFIG_ARCH_BCM2835=y
Note the output from the following: $ git grep STACKPROTECTOR arch/arm/configs/ arch/arm/configs/aspeed_g4_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y arch/arm/configs/aspeed_g5_defconfig:CONFIG_CC_STACKPROTECTOR_STRONG=y arch/arm/configs/bcm2835_defconfig:CONFIG_CC_STACKPROTECTOR_REGULAR=y $ Only three defconfigs specify a value. And two of the three ask for the strong variant, which isn't supported by older toolchains. Due to the nature of ARM having more platform specific code than say x86, the allyesconfig and allmodconfig aren't as effective for build coverage. So, in addition, I like to use a trivial script to walk all the defconfigs and build each one. However I will get false positives on unsupported stackprotector values with an older toolchain like gcc-4.6.3. As in this instance I am just using the compiler as a glorified syntax checker on a machine where I build a bunch of other arch for the same reason, there is no real motivation to get a newer toolchain for improved optimization etc. Since there are only three of them, and there is nothing about these settings that are board/platform specific, I propose we just eliminate the three existing instances and take the default. Cc: Russell King <linux@armlinux.org.uk> Cc: Florian Fainelli <f.fainelli@gmail.com> Cc: Ray Jui <rjui@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Stephen Warren <swarren@wwwdotorg.org> Cc: Lee Jones <lee@kernel.org> Cc: Eric Anholt <eric@anholt.net> Cc: Joel Stanley <joel@jms.id.au> Cc: linux-arm-kernel@lists.infradead.org Cc: bcm-kernel-feedback-list@broadcom.com Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> --- arch/arm/configs/aspeed_g4_defconfig | 1 - arch/arm/configs/aspeed_g5_defconfig | 1 - arch/arm/configs/bcm2835_defconfig | 1 - 3 files changed, 3 deletions(-)