| Message ID | 20240616-b4-mips-ipi-improvements-v1-4-e332687f1692@flygoat.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | MIPS: IPI Improvements | expand |
Hi, Jiaxun On Mon, Jun 17, 2024 at 5:03 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: > > This will give platform code a genuine chance to setup > IPI IRQ in prepare_cpus. > > This is the best place for platforms to setup IPI as smp_setup > is too early for IRQ subsystem. mips_smp_ipi_init() is an early_initcall() function, why do you say it is in smp_setup()? Huacai > > Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> > --- > arch/mips/kernel/smp.c | 11 ++++++++++- > 1 file changed, 10 insertions(+), 1 deletion(-) > > diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c > index fe053fe52147..ddf96c28e2f0 100644 > --- a/arch/mips/kernel/smp.c > +++ b/arch/mips/kernel/smp.c > @@ -375,7 +375,6 @@ static int __init mips_smp_ipi_init(void) > > return 0; > } > -early_initcall(mips_smp_ipi_init); > #endif > > /* > @@ -460,12 +459,22 @@ void __init smp_cpus_done(unsigned int max_cpus) > /* called from main before smp_init() */ > void __init smp_prepare_cpus(unsigned int max_cpus) > { > + int rc; > + > init_new_context(current, &init_mm); > current_thread_info()->cpu = 0; > mp_ops->prepare_cpus(max_cpus); > set_cpu_sibling_map(0); > set_cpu_core_map(0); > calculate_cpu_foreign_map(); > +#ifdef CONFIG_GENERIC_IRQ_IPI > + rc = mips_smp_ipi_init(); > + if (rc) { > + pr_err("Failed to initialize IPI - disabling SMP"); > + init_cpu_present(cpumask_of(0)); > + return; > + } > +#endif > #ifndef CONFIG_HOTPLUG_CPU > init_cpu_present(cpu_possible_mask); > #endif > > -- > 2.43.0 >
在2024年6月17日六月 下午2:53，Huacai Chen写道： > Hi, Jiaxun > > On Mon, Jun 17, 2024 at 5:03 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: >> >> This will give platform code a genuine chance to setup >> IPI IRQ in prepare_cpus. >> >> This is the best place for platforms to setup IPI as smp_setup >> is too early for IRQ subsystem. > mips_smp_ipi_init() is an early_initcall() function, why do you say it > is in smp_setup()? Sorry, I was trying to say that smp_setup is not a good point so we should go prepare_cpus. The intention of this patch is to move mips_smp_ipi_init to a certain point so platform would gain control over it. Thanks - Jiaxun > > > Huacai >> >> Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> >> --- >> arch/mips/kernel/smp.c | 11 ++++++++++- >> 1 file changed, 10 insertions(+), 1 deletion(-) >> >> diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c >> index fe053fe52147..ddf96c28e2f0 100644 >> --- a/arch/mips/kernel/smp.c >> +++ b/arch/mips/kernel/smp.c >> @@ -375,7 +375,6 @@ static int __init mips_smp_ipi_init(void) >> >> return 0; >> } >> -early_initcall(mips_smp_ipi_init); >> #endif >> >> /* >> @@ -460,12 +459,22 @@ void __init smp_cpus_done(unsigned int max_cpus) >> /* called from main before smp_init() */ >> void __init smp_prepare_cpus(unsigned int max_cpus) >> { >> + int rc; >> + >> init_new_context(current, &init_mm); >> current_thread_info()->cpu = 0; >> mp_ops->prepare_cpus(max_cpus); >> set_cpu_sibling_map(0); >> set_cpu_core_map(0); >> calculate_cpu_foreign_map(); >> +#ifdef CONFIG_GENERIC_IRQ_IPI >> + rc = mips_smp_ipi_init(); >> + if (rc) { >> + pr_err("Failed to initialize IPI - disabling SMP"); >> + init_cpu_present(cpumask_of(0)); >> + return; >> + } >> +#endif >> #ifndef CONFIG_HOTPLUG_CPU >> init_cpu_present(cpu_possible_mask); >> #endif >> >> -- >> 2.43.0 >>
On Tue, Jun 18, 2024 at 6:10 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: > > > > 在2024年6月17日六月 下午2:53，Huacai Chen写道： > > Hi, Jiaxun > > > > On Mon, Jun 17, 2024 at 5:03 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: > >> > >> This will give platform code a genuine chance to setup > >> IPI IRQ in prepare_cpus. > >> > >> This is the best place for platforms to setup IPI as smp_setup > >> is too early for IRQ subsystem. > > mips_smp_ipi_init() is an early_initcall() function, why do you say it > > is in smp_setup()? > > Sorry, I was trying to say that smp_setup is not a good point so we should > go prepare_cpus. It is not in smp_setup() now, then how do you move it from smp_setup()? Huacai > > The intention of this patch is to move mips_smp_ipi_init to a certain point > so platform would gain control over it. > > Thanks > - Jiaxun > > > > > > > Huacai > >> > >> Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> > >> --- > >> arch/mips/kernel/smp.c | 11 ++++++++++- > >> 1 file changed, 10 insertions(+), 1 deletion(-) > >> > >> diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c > >> index fe053fe52147..ddf96c28e2f0 100644 > >> --- a/arch/mips/kernel/smp.c > >> +++ b/arch/mips/kernel/smp.c > >> @@ -375,7 +375,6 @@ static int __init mips_smp_ipi_init(void) > >> > >> return 0; > >> } > >> -early_initcall(mips_smp_ipi_init); > >> #endif > >> > >> /* > >> @@ -460,12 +459,22 @@ void __init smp_cpus_done(unsigned int max_cpus) > >> /* called from main before smp_init() */ > >> void __init smp_prepare_cpus(unsigned int max_cpus) > >> { > >> + int rc; > >> + > >> init_new_context(current, &init_mm); > >> current_thread_info()->cpu = 0; > >> mp_ops->prepare_cpus(max_cpus); > >> set_cpu_sibling_map(0); > >> set_cpu_core_map(0); > >> calculate_cpu_foreign_map(); > >> +#ifdef CONFIG_GENERIC_IRQ_IPI > >> + rc = mips_smp_ipi_init(); > >> + if (rc) { > >> + pr_err("Failed to initialize IPI - disabling SMP"); > >> + init_cpu_present(cpumask_of(0)); > >> + return; > >> + } > >> +#endif > >> #ifndef CONFIG_HOTPLUG_CPU > >> init_cpu_present(cpu_possible_mask); > >> #endif > >> > >> -- > >> 2.43.0 > >> > > -- > - Jiaxun
在2024年6月18日六月 上午4:51，Huacai Chen写道： > On Tue, Jun 18, 2024 at 6:10 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: >> >> >> >> 在2024年6月17日六月 下午2:53，Huacai Chen写道： >> > Hi, Jiaxun >> > >> > On Mon, Jun 17, 2024 at 5:03 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: >> >> >> >> This will give platform code a genuine chance to setup >> >> IPI IRQ in prepare_cpus. >> >> >> >> This is the best place for platforms to setup IPI as smp_setup >> >> is too early for IRQ subsystem. >> > mips_smp_ipi_init() is an early_initcall() function, why do you say it >> > is in smp_setup()? >> >> Sorry, I was trying to say that smp_setup is not a good point so we should >> go prepare_cpus. > It is not in smp_setup() now, then how do you move it from smp_setup()? [...] Well I was trying to justify where should I move it to, not where does it from. I should improve the commit message to justify the reason first. Thanks - Jiaxun > > Huacai > >> >> The intention of this patch is to move mips_smp_ipi_init to a certain point >> so platform would gain control over it. >> >> Thanks >> - Jiaxun >> >> > >> > >> > Huacai >> >> >> >> Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> >> >> --- >> >> arch/mips/kernel/smp.c | 11 ++++++++++- >> >> 1 file changed, 10 insertions(+), 1 deletion(-) >> >> >> >> diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c >> >> index fe053fe52147..ddf96c28e2f0 100644 >> >> --- a/arch/mips/kernel/smp.c >> >> +++ b/arch/mips/kernel/smp.c >> >> @@ -375,7 +375,6 @@ static int __init mips_smp_ipi_init(void) >> >> >> >> return 0; >> >> } >> >> -early_initcall(mips_smp_ipi_init); >> >> #endif >> >> >> >> /* >> >> @@ -460,12 +459,22 @@ void __init smp_cpus_done(unsigned int max_cpus) >> >> /* called from main before smp_init() */ >> >> void __init smp_prepare_cpus(unsigned int max_cpus) >> >> { >> >> + int rc; >> >> + >> >> init_new_context(current, &init_mm); >> >> current_thread_info()->cpu = 0; >> >> mp_ops->prepare_cpus(max_cpus); >> >> set_cpu_sibling_map(0); >> >> set_cpu_core_map(0); >> >> calculate_cpu_foreign_map(); >> >> +#ifdef CONFIG_GENERIC_IRQ_IPI >> >> + rc = mips_smp_ipi_init(); >> >> + if (rc) { >> >> + pr_err("Failed to initialize IPI - disabling SMP"); >> >> + init_cpu_present(cpumask_of(0)); >> >> + return; >> >> + } >> >> +#endif >> >> #ifndef CONFIG_HOTPLUG_CPU >> >> init_cpu_present(cpu_possible_mask); >> >> #endif >> >> >> >> -- >> >> 2.43.0 >> >> >> >> -- >> - Jiaxun
On Tue, Jun 18, 2024 at 03:50:33PM +0100, Jiaxun Yang wrote: > > > 在2024年6月18日六月 上午4:51，Huacai Chen写道： > > On Tue, Jun 18, 2024 at 6:10 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: > >> > >> > >> > >> 在2024年6月17日六月 下午2:53，Huacai Chen写道： > >> > Hi, Jiaxun > >> > > >> > On Mon, Jun 17, 2024 at 5:03 AM Jiaxun Yang <jiaxun.yang@flygoat.com> wrote: > >> >> > >> >> This will give platform code a genuine chance to setup > >> >> IPI IRQ in prepare_cpus. > >> >> > >> >> This is the best place for platforms to setup IPI as smp_setup > >> >> is too early for IRQ subsystem. > >> > mips_smp_ipi_init() is an early_initcall() function, why do you say it > >> > is in smp_setup()? > >> > >> Sorry, I was trying to say that smp_setup is not a good point so we should > >> go prepare_cpus. > > It is not in smp_setup() now, then how do you move it from smp_setup()? > [...] > > Well I was trying to justify where should I move it to, not where > does it from. > > I should improve the commit message to justify the reason first. please do Thomas.
diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c index fe053fe52147..ddf96c28e2f0 100644 --- a/arch/mips/kernel/smp.c +++ b/arch/mips/kernel/smp.c @@ -375,7 +375,6 @@ static int __init mips_smp_ipi_init(void) return 0; } -early_initcall(mips_smp_ipi_init); #endif /* @@ -460,12 +459,22 @@ void __init smp_cpus_done(unsigned int max_cpus) /* called from main before smp_init() */ void __init smp_prepare_cpus(unsigned int max_cpus) { + int rc; + init_new_context(current, &init_mm); current_thread_info()->cpu = 0; mp_ops->prepare_cpus(max_cpus); set_cpu_sibling_map(0); set_cpu_core_map(0); calculate_cpu_foreign_map(); +#ifdef CONFIG_GENERIC_IRQ_IPI + rc = mips_smp_ipi_init(); + if (rc) { + pr_err("Failed to initialize IPI - disabling SMP"); + init_cpu_present(cpumask_of(0)); + return; + } +#endif #ifndef CONFIG_HOTPLUG_CPU init_cpu_present(cpu_possible_mask); #endif
This will give platform code a genuine chance to setup IPI IRQ in prepare_cpus. This is the best place for platforms to setup IPI as smp_setup is too early for IRQ subsystem. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> --- arch/mips/kernel/smp.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)