Re: [PATCH] x86/traps: Remove redundant trapnr parameter from fatal_trap()

[Don Slutz <dslutz@verizon.com>]

On 09/01/14 12:27, Andrew Cooper wrote:
> On 01/09/14 17:10, Don Slutz wrote:
>> On 09/01/14 06:06, Andrew Cooper wrote:
>>> It is always available via regs->entry_vector.
>>>
>>> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
>>> CC: Jan Beulich <JBeulich@suse.com>
>>> ---

>>> -void fatal_trap(int trapnr, const struct cpu_user_regs *regs)
>>> +void fatal_trap(const struct cpu_user_regs *regs)
>>>    {
>>>        static DEFINE_PER_CPU(char, depth);
>>> +    unsigned int trapnr = regs->entry_vector;
>> Should this be:
>>
>> unsigned int trapnr = regs->entry_vector &
>> ~(TRAP_regs_partial|TRAP_syscall);
>>
>>
>> To get rid of the extra bits?
>>
>>    -Don Slutz
> No; I don't think so.  There should be no paths into fatal_trap() which
> would set these bits, as these bits are only set as a result of guest
> sys/hypercall interaction, while fatal_trap() is a terminal error condition.
>
> Furthermore, if a path is discovered then a) it will be more obvious
> from the error message and b) we probably have a security problem to fix.
>
> ~Andrew

Ok.

Reviewed-by: Don Slutz <dslutz@verizon.com>