Re: [PATCH v2 10/12] xenctx: change is_kernel_text() into is_kernel_addr().

[Don Slutz <dslutz@verizon.com>]

On 11/08/13 04:50, Ian Campbell wrote:
> On Fri, 2013-11-08 at 08:40 +0000, Jan Beulich wrote:
>>>>> On 08.11.13 at 01:51, Don Slutz <dslutz@verizon.com> wrote:
>>> On 11/07/13 03:35, Jan Beulich wrote:
>>>>>>> On 06.11.13 at 21:08, Don Slutz <dslutz@verizon.com> wrote:
>>>>> +        return KERNEL_TEXT_ADDR;
>>>>>        if (addr >= kernel_sinittext &&
>>>>>            addr <= kernel_einittext)
>>>>> -        return 1;
>>>>> -    return 0;
>>>>> +        return KERNEL_TEXT_ADDR;
>>>>> +    if (addr >= kernel_text &&
>>>>> +        addr <= kernel_end)
>>>>> +        return KERNEL_DATA_ADDR;
>>>> As you supposedly filtered out all text ranges before, did you really
>>>> mean to compare to kernel_text here (rather than kernel_start)?
>>> Yes, I did.  I think it is better to use the value that is in the system map
>>> over the default.  It has changed:
>>>
>>> dcs-xen-54:~/xen>grep " _text" /boot/System.map-*
>>> /boot/System.map-2.6.18-128.el5:ffffffff80000000 A _text
>>> /boot/System.map-3.6.11-5.fc17.x86_64:ffffffff81000000 T _text
>>> /boot/System.map-3.8.11-100.fc17.x86_64:ffffffff81000000 T _text
>>>
>>> But since it is a command line argument, if specified, should it be used
>>> instead?
>> I guess so.
> Yes, since the user may not have a symbol table for the given kernel and
> it may not even be Linux. I'm sure a Windows kernel guru debugging a
> Windows guest would know the right number for that version of Windows.
> (IIRC the option was added by just such a guru...)
If you do not specify a symbol file, kernel_start will be used. This is 
the case when both are specified.  If I am reading this right that is 2 
votes to use kernel_start if specified otherwise kernel_text.
    -Don Slutz
>>   In any event - after having split out all text pieces
>> (hopefully), the check here should cover the whole kernel image,
>> no matter how the ordering between text and data is.
>>
>> Jan
>>
>