Re: [Xen-devel] [PATCH] xen/apic: implement io apic read with hypercall

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 20/04/12 13:38, Ian Campbell wrote:
> On Fri, 2012-04-20 at 12:13 +0100, Lin Ming wrote:
>> On Fri, 2012-04-20 at 10:58 +0100, Andrew Cooper wrote:
>>> On 20/04/12 10:25, Lin Ming wrote:
>>>> Implements xen_io_apic_read with hypercall, so it returns proper IO-APIC
>>>> information instead of fabricated one.
>>>>
>>>> Signed-off-by: Lin Ming <mlin@xxxxxxxxxxxxx>
>>>> ---
>>>>  arch/x86/xen/apic.c |   16 +++++++++++-----
>>>>  1 files changed, 11 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/arch/x86/xen/apic.c b/arch/x86/xen/apic.c
>>>> index aee16ab..f1f392d 100644
>>>> --- a/arch/x86/xen/apic.c
>>>> +++ b/arch/x86/xen/apic.c
>>>> @@ -1,14 +1,20 @@
>>>>  #include <linux/init.h>
>>>>  #include <asm/x86_init.h>
>>>> +#include <asm/apic.h>
>>>> +#include <xen/interface/physdev.h>
>>>> +#include <asm/xen/hypercall.h>
>>>>  
>>>>  unsigned int xen_io_apic_read(unsigned apic, unsigned reg)
>>>>  {
>>>> -  if (reg == 0x1)
>>>> -          return 0x00170020;
>>>> -  else if (reg == 0x0)
>>>> -          return apic << 24;
>>>> +  struct physdev_apic apic_op;
>>>> +  int ret;
>>>>  
>>>> -  return 0xff;
>>>> +  apic_op.apic_physbase = mpc_ioapic_addr(apic);
>>>> +  apic_op.reg = reg;
>>>> +  ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op);
>>>> +  if (ret)
>>>> +          return ret;
>>>> +  return apic_op.value;
>>> Hypercall ret errors are negative, yet this function is unsigned.  Given
>>> that the previous function had no possible way to fail, perhaps on error
>>> you should fake up the values as before.
>> How about return -1 on error?
>> The calling function can check -1 for error. 
> Isn't -1 potentially (at least theoretically) a valid value to read from
> one of these registers?

Theoretically yes, but practically it would only be buggy hardware
returning -1.

>
> Under what circumstances can these hypercalls fail? Would a BUG_ON be
> appropriate/

-EFAULT, -EPERM, anything xsm_apic() could return (which looks only to
be -EPERM).  The call into Xen itself will return 0 as a value if an
invalid physbase is passed in the hypercall.

So a BUG_ON() is not safe/sensible for domU.

>
>> unsigned int ret = apic_read(...);
>> if (ret == -1)
>>      //handle error.
>>
>> Thanks,
>> Lin Ming
>>
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxx
>> http://lists.xen.org/xen-devel
>

-- 
Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer
T: +44 (0)1223 225 900, http://www.citrix.com


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel