[PATCH] ipmi: set si_trydefaults=0 for ARM64

[PATCH] ipmi: set si_trydefaults=0 for ARM64

[Tony Camuso]

Port I/O space does not exist in ARM64 and is not mapped. Attempts to
access it on ARM systems cause stack traces and worse.

Signed-off-by: Tony Camuso <tcamuso@redhat.com>
---
 drivers/char/ipmi/ipmi_si_intf.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
index 615abbf..85dcc86 100644
--- a/drivers/char/ipmi/ipmi_si_intf.c
+++ b/drivers/char/ipmi/ipmi_si_intf.c
@@ -3841,6 +3841,11 @@ static int init_ipmi_si(void)
 		spmi_find_bmc();
 #endif
 
+#ifdef CONFIG_ARM64
+	/* Don't touch port io space */
+	si_trydefaults = 0;
+#endif
+
 #ifdef CONFIG_PARISC
 	register_parisc_driver(&ipmi_parisc_driver);
 	parisc_registered = true;
-- 
1.8.3.1

Re: [PATCH] ipmi: set si_trydefaults=0 for ARM64

[Corey Minyard]

On 06/20/2016 01:26 PM, Tony Camuso wrote:
> Port I/O space does not exist in ARM64 and is not mapped. Attempts to
> access it on ARM systems cause stack traces and worse.

At this point, I think it is best to just completely pull out all concept
of "default addresses" in the IPMI driver.  The defaults were disabled
by default in 3.16, this is as good an impetus as any to just get rid
of them.

If you want, you can do a patch, or I can pull them out if you would
prefer that.

Thanks,

-corey

> Signed-off-by: Tony Camuso <tcamuso@redhat.com>
> ---
>   drivers/char/ipmi/ipmi_si_intf.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
> index 615abbf..85dcc86 100644
> --- a/drivers/char/ipmi/ipmi_si_intf.c
> +++ b/drivers/char/ipmi/ipmi_si_intf.c
> @@ -3841,6 +3841,11 @@ static int init_ipmi_si(void)
>   		spmi_find_bmc();
>   #endif
>   
> +#ifdef CONFIG_ARM64
> +	/* Don't touch port io space */
> +	si_trydefaults = 0;
> +#endif
> +
>   #ifdef CONFIG_PARISC
>   	register_parisc_driver(&ipmi_parisc_driver);
>   	parisc_registered = true;

Re: [PATCH] ipmi: set si_trydefaults=0 for ARM64

[Tony Camuso]

On 06/20/2016 04:57 PM, Corey Minyard wrote:
> On 06/20/2016 01:26 PM, Tony Camuso wrote:
>> Port I/O space does not exist in ARM64 and is not mapped. Attempts to
>> access it on ARM systems cause stack traces and worse.
>
> At this point, I think it is best to just completely pull out all concept
> of "default addresses" in the IPMI driver.  The defaults were disabled
> by default in 3.16, this is as good an impetus as any to just get rid
> of them.
>
> If you want, you can do a patch, or I can pull them out if you would
> prefer that.

I'll give it a go. Should have something later today.

Regards,
Tony


>
> Thanks,
>
> -corey
>
>> Signed-off-by: Tony Camuso <tcamuso@redhat.com>
>> ---
>>   drivers/char/ipmi/ipmi_si_intf.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
>> index 615abbf..85dcc86 100644
>> --- a/drivers/char/ipmi/ipmi_si_intf.c
>> +++ b/drivers/char/ipmi/ipmi_si_intf.c
>> @@ -3841,6 +3841,11 @@ static int init_ipmi_si(void)
>>           spmi_find_bmc();
>>   #endif
>>   +#ifdef CONFIG_ARM64
>> +    /* Don't touch port io space */
>> +    si_trydefaults = 0;
>> +#endif
>> +
>>   #ifdef CONFIG_PARISC
>>       register_parisc_driver(&ipmi_parisc_driver);
>>       parisc_registered = true;
>