Hi, the preempt_disable/preempt_enable sequence in do_cpu() [traps.c] exists quite long (patch submitted in Oct. 2004), so it should be nothing wrong there. Can somebody please comment why use of preempt_disable/enable in do_cpu will not result in "scheduling while atomic" for fpu-less cpu (with enabled preemption). The sequence looks like do_cpu() | preempt_disable() | fpu_emulator_cop1Handler() | | cond_reshed() | | | schedule() <------ scheduling while atomic The proposed patch was tested for Sibyte, but it has fpu (AFAIK) and has no fpu_emulator_cop1Handler called. -- Thank you, Pavel Kiryukhin mailto:vksavl@xxxxxxxxxxx Friday, October 22, 2004, 10:06:43 PM, you wrote: ML> Hello ! ML> The attached patch incorporates preemption enable/disable in some parts ML> of the kernel. I have tested this on the Broadcom Sibyte. Please review ML> ... ML> Thanks ML> Manish Lachwani <skip> ML> Index: linux-2.6.8.1/arch/mips/kernel/traps.c ML> =================================================================== ML> --- linux-2.6.8.1.orig/arch/mips/kernel/traps.c ML> +++ linux-2.6.8.1/arch/mips/kernel/traps.c <skip> ML> case 1: ML> +preempt_disable(); ML> + ML> own_fpu(); ML> if (current->used_math) { /* Using the FPU again. */ ML> restore_fp(current); ML> @@ -674,6 +690,8 @@ ML> force_sig(sig, current); ML> } ML> +preempt_enable(); ML> + ML> return; ML> case 2: <skip>
