On Fri, Jul 8, 2011 at 12:16 PM, Greg Ungerer <g...@snapgear.com> wrote:
> Hi Erjiang, > > > On 05/07/11 13:08, Erjiang Li wrote: > >> Hi All, >> I have been working on uClinux kernel on ARM platform recently, and I >> want to decrease the kernel stack size from 8K to 4K, i.e. change the >> following code (in include/asm-arm/thread_info.h)**: >> #define THREAD_SIZE_ORDER 1 >> #define THREAD_SIZE 8192 >> to >> #define THREAD_SIZE_ORDER 0 >> #define THREAD_SIZE 4096 >> >> But after this modification, it fails to boot up linux on my ARM >> development board now. It stops at the position of pdflush_init(), and >> the functions calling chain is as following: >> ... >> ->pdflush_init()->start_one_**pdflush_thread()->kthread_run(** >> )->kthread_create()->wait_for_**completion()->wait_for_common(** >> )->do_wait_for_common()->**schedule() >> >> In schedule() function, below code is executed repeatedly: >> if (unlikely(test_thread_flag(**TIF_NEED_RESCHED))) >> goto need_resched; >> >> So the function schedule() can't return. >> > > Have you checked that the stack at this point isn't larger than > the 4k you have allowed for? > > I know the above call chain doesn't look too bad, but double check > the current stack. And of course the real problem is that interrupts > will come in and need more of the current kernel stack. > > I am sorry for that my original findings were not correct. schedule() function can return, the code "goto need_resched" is not executed repeatedly. It's ret_to_user/ret_slow_syscall entry in arch/arm/kernel/entry-common.S calls schedule() function repeatedly. I have checked for stack, and don't find any overflows at this point. I did want to use a separate IRQ stack for IRQ handling, like x86. But I don't know how much & how difficult the work will be. > > > BTW, my uclinux version is 2.6.25, and my ARM cpu is MMU-less. >> > > That is kind of old. > > > > Can anyone give me some advice on this problem. I wonder if we can >> decrease THREAD_SIZE to 4K on ARM platform. Some one told me that ARM >> uClinux needs separate interrupt stack for 4K THREAD_SIZE. Is that true? >> > > I haven't heard that. I use 4k stacks on uClinux/ColdFire builds and > that seems ok on the systems I have used it on. > > Regards > Greg > > > ------------------------------**------------------------------** > ------------ > Greg Ungerer -- Principal Engineer EMAIL: g...@snapgear.com > SnapGear Group, McAfee PHONE: +61 7 3435 2888 > 8 Gardner Close FAX: +61 7 3217 5323 > Milton, QLD, 4064, Australia WEB: http://www.SnapGear.com > -- Best Regards Erjian
_______________________________________________ uClinux-dev mailing list uClinux-dev@uclinux.org http://mailman.uclinux.org/mailman/listinfo/uclinux-dev This message was resent by uclinux-dev@uclinux.org To unsubscribe see: http://mailman.uclinux.org/mailman/options/uclinux-dev
