On Wed, Nov 29, 2017 at 12:02:44PM +0900, Masahiro Yamada wrote:
2017-11-29 11:30 GMT+09:00 Tom Rini trini@konsulko.com:
On Wed, Nov 29, 2017 at 11:01:02AM +0900, Masahiro Yamada wrote:
2017-11-29 10:42 GMT+09:00 Tom Rini trini@konsulko.com:
On Wed, Nov 29, 2017 at 12:17:35AM +0900, Masahiro Yamada wrote:
2017-11-28 23:52 GMT+09:00 Tom Rini trini@konsulko.com:
On Tue, Nov 28, 2017 at 09:23:18PM +0900, Masahiro Yamada wrote:
> Buildman test passed. > > > Masahiro Yamada (5): > Move CONFIG_PANIC_HANG to Kconfig > ARM: openrd: set CONFIG_LOGLEVEL to 2 > Enable CONFIG_PANIC_HANG for boards without do_reset() > treewide: convert assert() to BUG_ON() > Remove assert()
Have you done a size check on this series? If not, I'll fire one up, thanks!
I think I did it correctly.
At least, I needed the following patch for openrd boards: http://patchwork.ozlabs.org/patch/842106/
Looks like the NAND core pull-request needs it, too.
So, in my testing (which is gcc-6 for everything thanks to Debian/9): 06: Remove assert() arm: + clearfog peach-pi snow smdk5250 smdk5420 turris_omnia spring omap3_evm peach-pit -(am335x_hs_evm) arm-linux-gnueabihf-ld.bfd: region `.sram' overflowed by 760 bytes +(clearfog,turris_omnia) arm-linux-gnueabihf-ld.bfd: SPL image too big +(am335x_hs_evm) arm-linux-gnueabihf-ld.bfd: region `.sram' overflowed by 1080 bytes +(snow,smdk5250,peach-pi,smdk5420,spring,peach-pit) arch/arm/mach-exynos/built-in.o: In function `clock_calc_best_scalar': +(snow,smdk5250,peach-pi,smdk5420,spring,peach-pit) build/../arch/arm/mach-exynos/clock.c:1408: undefined reference to `panic' +(omap3_evm) arm-linux-gnueabihf-ld.bfd: region `.sram' overflowed by 28 bytes
And then a lot of platforms grow in size. Taking a harder peek, BUG_ON is always set. However, assert() is only set when DEBUG is set. Now, I can certainly see an argument along the lines of that being fairly unexpected, and I don't know that I would disagree.
After a very quick look over what git grep -l says, perhaps we want to introduce a new (default off) option to preserve the current behavior of assert() in BUG_ON, as it looks like almost only kernel imported code is using BUG_ON.
Do you have an idea for the option name?
ENABLE_BUG_ON_CHECKS ?
Do you mean CONFIG option? i.e. CONFIG_ENABLE_BUG_ON_CHECKS so we can turn it on/off globally?
Yes, a CONFIG option to turn it on/off.