* Re: [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags
2010-10-19 1:43 [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags Nobuhiro Iwamatsu
@ 2010-10-19 1:50 ` Nobuhiro Iwamatsu
2010-10-19 2:47 ` Paul Mundt
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Nobuhiro Iwamatsu @ 2010-10-19 1:50 UTC (permalink / raw)
To: linux-sh
Oh, sorry. Subject is wrong.
s/cou/cpu
Nobuhiro
2010/10/19 Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>:
> There is not method to confirm which of 32bit mode or 29bit mode Linux
> kernel works.
> This patch adds a mode to cpu_flags and can confirm it from auxv and proc
> filesystem.
>
> Signed-off-by: Nobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
> ---
> arch/sh/include/asm/cpu-features.h | 1 +
> arch/sh/kernel/cpu/sh4/probe.c | 4 ++++
> arch/sh/kernel/setup.c | 8 +++++++-
> 3 files changed, 12 insertions(+), 1 deletions(-)
>
> diff --git a/arch/sh/include/asm/cpu-features.h b/arch/sh/include/asm/cpu-features.h
> index 694abe4..dc1265b 100644
> --- a/arch/sh/include/asm/cpu-features.h
> +++ b/arch/sh/include/asm/cpu-features.h
> @@ -22,5 +22,6 @@
> #define CPU_HAS_L2_CACHE 0x0080 /* Secondary cache / URAM */
> #define CPU_HAS_OP32 0x0100 /* 32-bit instruction support */
> #define CPU_HAS_PTEAEX 0x0200 /* PTE ASID Extension support */
> +#define CPU_IS_32BIT_MODE 0x0400 /* linux kernel works 32bit mode */
>
> #endif /* __ASM_SH_CPU_FEATURES_H */
> diff --git a/arch/sh/kernel/cpu/sh4/probe.c b/arch/sh/kernel/cpu/sh4/probe.c
> index b93458f..4bda234 100644
> --- a/arch/sh/kernel/cpu/sh4/probe.c
> +++ b/arch/sh/kernel/cpu/sh4/probe.c
> @@ -189,6 +189,10 @@ void __cpuinit cpu_probe(void)
> break;
> }
>
> +#ifdef CONFIG_32BIT
> + boot_cpu_data.flags |= CPU_IS_32BIT_MODE;
> +#endif
> +
> /*
> * On anything that's not a direct-mapped cache, look to the CVR
> * for I/D-cache specifics.
> diff --git a/arch/sh/kernel/setup.c b/arch/sh/kernel/setup.c
> index 4e27846..460ac72 100644
> --- a/arch/sh/kernel/setup.c
> +++ b/arch/sh/kernel/setup.c
> @@ -352,7 +352,8 @@ EXPORT_SYMBOL(get_cpu_subtype);
> /* Symbolic CPU flags, keep in sync with asm/cpu-features.h */
> static const char *cpu_flags[] = {
> "none", "fpu", "p2flush", "mmuassoc", "dsp", "perfctr",
> - "ptea", "llsc", "l2", "op32", "pteaex", NULL
> + "ptea", "llsc", "l2", "op32", "pteaex", "32bit",
> + NULL
> };
>
> static void show_cpuflags(struct seq_file *m, struct sh_cpuinfo *c)
> @@ -410,6 +411,11 @@ static int show_cpuinfo(struct seq_file *m, void *v)
> else
> seq_printf(m, "cut\t\t: %d.%d\n", c->cut_major, c->cut_minor);
>
> + if (c->flags & CPU_IS_32BIT_MODE)
> + seq_printf(m, "cpu mode\t: 32bit\n");
> + else
> + seq_printf(m, "cpu mode\t: 29bit\n");
> +
> show_cpuflags(m, c);
>
> seq_printf(m, "cache type\t: ");
> --
> 1.7.1
>
>
--
Nobuhiro Iwamatsu
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags
2010-10-19 1:43 [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags Nobuhiro Iwamatsu
2010-10-19 1:50 ` Nobuhiro Iwamatsu
@ 2010-10-19 2:47 ` Paul Mundt
2010-10-20 2:13 ` Nobuhiro Iwamatsu
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2010-10-19 2:47 UTC (permalink / raw)
To: linux-sh
On Tue, Oct 19, 2010 at 10:43:29AM +0900, Nobuhiro Iwamatsu wrote:
> There is not method to confirm which of 32bit mode or 29bit mode Linux
> kernel works.
> This patch adds a mode to cpu_flags and can confirm it from auxv and proc
> filesystem.
>
Is there some use you had in mind for this?
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags
2010-10-19 1:43 [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags Nobuhiro Iwamatsu
2010-10-19 1:50 ` Nobuhiro Iwamatsu
2010-10-19 2:47 ` Paul Mundt
@ 2010-10-20 2:13 ` Nobuhiro Iwamatsu
2010-10-25 6:14 ` Paul Mundt
2010-10-26 0:19 ` Nobuhiro Iwamatsu
4 siblings, 0 replies; 6+ messages in thread
From: Nobuhiro Iwamatsu @ 2010-10-20 2:13 UTC (permalink / raw)
To: linux-sh
2010/10/19 Paul Mundt <lethal@linux-sh.org>:
> On Tue, Oct 19, 2010 at 10:43:29AM +0900, Nobuhiro Iwamatsu wrote:
>> There is not method to confirm which of 32bit mode or 29bit mode Linux
>> kernel works.
>> This patch adds a mode to cpu_flags and can confirm it from auxv and proc
>> filesystem.
>>
> Is there some use you had in mind for this?
First , I wanted these as information simply.
And I am debuging kexec-toos for 32bit mode.
I need this information to use as __pa in kernel.
If there is a good method, please teach it.
Best regards,
Nobuhiro
--
Nobuhiro Iwamatsu
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags
2010-10-19 1:43 [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags Nobuhiro Iwamatsu
` (2 preceding siblings ...)
2010-10-20 2:13 ` Nobuhiro Iwamatsu
@ 2010-10-25 6:14 ` Paul Mundt
2010-10-26 0:19 ` Nobuhiro Iwamatsu
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2010-10-25 6:14 UTC (permalink / raw)
To: linux-sh
On Wed, Oct 20, 2010 at 11:13:09AM +0900, Nobuhiro Iwamatsu wrote:
> 2010/10/19 Paul Mundt <lethal@linux-sh.org>:
> > On Tue, Oct 19, 2010 at 10:43:29AM +0900, Nobuhiro Iwamatsu wrote:
> >> There is not method to confirm which of 32bit mode or 29bit mode Linux
> >> kernel works.
> >> This patch adds a mode to cpu_flags and can confirm it from auxv and proc
> >> filesystem.
> >>
> > Is there some use you had in mind for this?
>
> First , I wanted these as information simply.
>
> And I am debuging kexec-toos for 32bit mode.
> I need this information to use as __pa in kernel.
> If there is a good method, please teach it.
>
While I agree that having this information exposed is useful, I don't
think it really falls under a CPU capability as such. 32-bit capable CPUs
can be either in 29 or 32-bit physical mode, so the existence of the PMB
is not sufficient for determining configuration.
The x86 approach for reporting the physical address space size seems like
a pretty good fit for this though, so how about this instead?
---
diff --git a/arch/sh/include/asm/processor.h b/arch/sh/include/asm/processor.h
index 0a58cb2..c9e7cbc 100644
--- a/arch/sh/include/asm/processor.h
+++ b/arch/sh/include/asm/processor.h
@@ -89,6 +89,7 @@ struct sh_cpuinfo {
struct task_struct *idle;
#endif
+ unsigned int phys_bits;
unsigned long flags;
} __attribute__ ((aligned(L1_CACHE_BYTES)));
diff --git a/arch/sh/kernel/cpu/init.c b/arch/sh/kernel/cpu/init.c
index 9766106..fac742e 100644
--- a/arch/sh/kernel/cpu/init.c
+++ b/arch/sh/kernel/cpu/init.c
@@ -340,6 +340,8 @@ asmlinkage void __cpuinit cpu_init(void)
*/
current_cpu_data.asid_cache = NO_CONTEXT;
+ current_cpu_data.phys_bits = __in_29bit_mode() ? 29 : 32;
+
speculative_execution_init();
expmask_init();
diff --git a/arch/sh/kernel/setup.c b/arch/sh/kernel/setup.c
index 4e27846..0af25e0 100644
--- a/arch/sh/kernel/setup.c
+++ b/arch/sh/kernel/setup.c
@@ -432,6 +432,8 @@ static int show_cpuinfo(struct seq_file *m, void *v)
if (c->flags & CPU_HAS_L2_CACHE)
show_cacheinfo(m, "scache", c->scache);
+ seq_printf(m, "address sizes\t: %u bits physical\n", c->phys_bits);
+
seq_printf(m, "bogomips\t: %lu.%02lu\n",
c->loops_per_jiffy/(500000/HZ),
(c->loops_per_jiffy/(5000/HZ)) % 100);
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags
2010-10-19 1:43 [RFC/PATCH] sh: Add 32bit cou mode to cpu_flags Nobuhiro Iwamatsu
` (3 preceding siblings ...)
2010-10-25 6:14 ` Paul Mundt
@ 2010-10-26 0:19 ` Nobuhiro Iwamatsu
4 siblings, 0 replies; 6+ messages in thread
From: Nobuhiro Iwamatsu @ 2010-10-26 0:19 UTC (permalink / raw)
To: linux-sh
Hi, Paul.
2010/10/25 Paul Mundt <lethal@linux-sh.org>:
> On Wed, Oct 20, 2010 at 11:13:09AM +0900, Nobuhiro Iwamatsu wrote:
>> 2010/10/19 Paul Mundt <lethal@linux-sh.org>:
>> > On Tue, Oct 19, 2010 at 10:43:29AM +0900, Nobuhiro Iwamatsu wrote:
>> >> There is not method to confirm which of 32bit mode or 29bit mode Linux
>> >> kernel works.
>> >> This patch adds a mode to cpu_flags and can confirm it from auxv and proc
>> >> filesystem.
>> >>
>> > Is there some use you had in mind for this?
>>
>> First , I wanted these as information simply.
>>
>> And I am debuging kexec-toos for 32bit mode.
>> I need this information to use as __pa in kernel.
>> If there is a good method, please teach it.
>>
> While I agree that having this information exposed is useful, I don't
> think it really falls under a CPU as such. 32-bit capable CPUs
> can be either in 29 or 32-bit physical mode, so the existence of the PMB
> is not sufficient for determining configuration.
>
I thought whether I might be able to be in auxv, because the mode of
the CPU was changed,
But I agree.
> The x86 approach for reporting the physical address space size seems like
> a pretty good fit for this though, so how about this instead?
Looks good to me, thanks for fixing it.
Best regards,
Nobuhiro
--
Nobuhiro Iwamatsu
^ permalink raw reply [flat|nested] 6+ messages in thread