Re: [v5][PATCH 1/2] xen:x86:mm:p2m: introduce set_identity_p2m_entry

["Chen, Tiejun" <tiejun.chen@intel.com>]

On 2014/7/29 16:19, Jan Beulich wrote:
>>>> On 29.07.14 at 09:35, <tiejun.chen@intel.com> wrote:
>> On 2014/7/29 15:05, Jan Beulich wrote:
>>>>>> On 29.07.14 at 08:40, <tiejun.chen@intel.com> wrote:
>>>> +    struct p2m_domain *p2m = p2m_get_hostp2m(d);
>>>> +    int ret = -EBUSY;
>>>> +
>>>> +    gfn_lock(p2m, gfn, 0);
>>>> +
>>>> +    tmp_mfn = p2m->get_entry(p2m, gfn, &p2mt, &a, 0, NULL);
>>>> +
>>>> +    if ( mfn_valid(tmp_mfn) )
>>>> +    {
>>>> +        gdprintk(XENLOG_ERR,
>>>> +                 "Overlapping RMRRs at %"PRIx64".\n", (paddr_t)gfn);
>>>
>>> Pointless cast: Just use %lx in the format string. Additionally I don't
>>> think the message text is correct: You don't really know whether
>>> what's there is another RMRR (or that the context you're being
>>> called in refers to an RMRR at all). On the contrary - if it was an
>>> RMRR (or to be precise, a previously established identity mapping),
>>> you'd want to report success. And generally we have no stop at
>>> the end of log messages.
>>
>> So just print this,
>>
>> +        gdprintk(XENLOG_ERR,
>> +                 "Overlapping at %lx.\n", (paddr_t)gfn);
>
> Of course not - such a message is really meaningless.
>
> printk(XENLOG_G_WARNING "Cannot identity map %d:%lx, already mapped to %lx\n",
>         d->domain_id, gfn, mfn_x(mfn));
>

Thanks your further help.

> would be a message conveying all information necessary to
> gain initial understanding of what the issue is.
>
>>>
>>>> +        goto out;
>>>
>>> Once again, when error handling is that simple please avoid using
>>> "goto".
>>>
>>
>> Its make no sense to me.
>>
>> Did you see this function in this same file,
>
> Referring to existing bad examples is never going to help.

But sometimes someone who are not familiar with this case may have no 
choice because something is really tricky.

>
>> Yes, previously I really can't understand what's that code style in xen.
>> So as I remember I ask you guy if xen has checkpatch.pl like Linux, qemu
>> or other stuff, but you didn't reply this point.
>
> We just have none, and for the specific case of using or not using
> "goto" it wouldn't help you anyway.
>
>> So I have to try
>> following existing codes. Now I'm curious what we should abide.
>
> Where is the problem with just writing

I doesn't mean I have a problem to write such codes but I think we need 
an unified coding style.

Anyway I can do as you show.

>
> int k(struct domain *d, unsigned long gfn)
> {
>      p2m_type_t p2mt;
>      p2m_access_t a;
>      mfn_t mfn;
>      struct p2m_domain *p2m = p2m_get_hostp2m(d);
>      int ret;
>
>      gfn_lock(p2m, gfn, 0);
>
>      mfn = p2m->get_entry(p2m, gfn, &p2mt, &a, 0, NULL);
>      if ( !mfn_valid(mfn) )
>          ret = p2m_set_entry(p2m, gfn, _mfn(gfn), PAGE_ORDER_4K, p2m_mmio_direct,
>                              p2m_access_rw);
>      else if ( mfn_x(mfn) == gfn )
>         ret = 0;
>      else
>      {
>          printk(XENLOG_G_WARNING
>                 "Cannot identity map %d:%lx, already mapped to %lx\n",
>                 d->domain_id, gfn, mfn_x(mfn));
>          ret = -EBUSY;
>      }
>
>      gfn_unlock(p2m, gfn, 0);
>
>      return ret;
> }
>
> ? Of course it may still be necessary to also inspect the obtained p2mt
> and a.
>

Are you saying this?

	if ( !p2m_is_valid(p2mt) ||
	     !mfn_valid(mfn) ||
	     (a != p2m_access_rw) )


Thanks
Tiejun