Re: [xen-unstable test] 104131: regressions - FAIL

["Jan Beulich" <JBeulich@suse.com>]

>>> On 18.01.17 at 11:23, <kevin.tian@intel.com> wrote:
>>  From: Jan Beulich [mailto:JBeulich@suse.com]
>> Sent: Wednesday, January 18, 2017 5:38 PM
>> 
>> >>> On 18.01.17 at 05:57, <kevin.tian@intel.com> wrote:
>> > Attached was my earlier comment:
>> >
>> > --
>> >> >>> On 20.12.16 at 06:37, <kevin.tian@intel.com> wrote:
>> >> >>  From: Xuquan (Quan Xu) [mailto:xuquan8@huawei.com]
>> >> >> Sent: Friday, December 16, 2016 5:40 PM
>> >> >> -        if (pt_vector != -1)
>> >> >> -            vmx_set_eoi_exit_bitmap(v, pt_vector);
>> >> >> +        if ( pt_vector != -1 ) {
>> >> >> +            if ( intack.vector > pt_vector )
>> >> >> +                vmx_set_eoi_exit_bitmap(v, intack.vector);
>> >> >> +            else
>> >> >> +                vmx_set_eoi_exit_bitmap(v, pt_vector);
>> >> >> +        }
>> >> >
>> >> > Above can be simplified as one line change:
>> >> > 	if ( pt_vector != -1 )
>> >> > 		vmx_set_eoi_exit_bitmap(v, intack.vector);
>> >>
>> >> Hmm, I don't understand. Did you mean to use max() here? Or
>> >> else how is this an equivalent of the originally proposed code?
>> >>
>> >
>> > Original code is not 100% correct. The purpose is to set EOI exit
>> > bitmap for any vector which may block injection of pt_vector -
>> > give chance to recognize pt_vector in future intack and then do pt
>> > intr post. The simplified code achieves this effect same as original
>> > code if intack.vector >= vector. I cannot come up a case why
>> > intack.vector might be smaller than vector. If this case happens,
>> > we still need enable exit bitmap for intack.vector instead of
>> > pt_vector for said purpose while original code did it wrong.
>> >
>> > Thanks
>> > Kevin
>> > --
>> >
>> > Using intack.vector is always expected here regardless of the
>> > comparison result between intack.vector and pt_vector. The
>> > reason why I was OK adding an ASSERT was simply to test
>> > whether intack.vecor<pt_vector does happen which is
>> > orthogonal to the fix itself.
>> 
>> Well, a vector lower than pt_vector can't block delivery. Or wait:
> 
> There are two points here:
> 
> a) We need enable EOI exit bitmap when pt_vector is blocked.
> 
> b) As you said, ideally a vector lower than pt_vecotr cannot block
> 
> The patch fixed a) and then added an ASSERT to verify b). Strictly
> speaking they are separate issues.

Okay, I think I finally understand your argumentation here.

>> Don't we need to consider vector classes here, i.e.
>> 
>>             ASSERT((intack.vector >> 4) >= (pt_vector >> 4));
>> 
>> ?
>> 
> 
> However it still doesn't explain why original ASSERT is triggered.
> vlapic_find_highest_vector actually finds the highest vector, instead
> of highest class...
> 
> static int vlapic_find_highest_vector(const void *bitmap)
> {
>     const uint32_t *word = bitmap;
>     unsigned int word_offset = NR_VECTORS / 32;
> 
>     /* Work backwards through the bitmap (first 32-bit word in every four). */
>     while ( (word_offset != 0) && (word[(--word_offset)*4] == 0) )
>         continue;
> 
>     return (fls(word[word_offset*4]) - 1) + (word_offset * 32);
> }

Well, perhaps a PIR -> IRR syncing issue then (I in particular note
the early bailing from vmx_sync_pir_to_irr())? I guess we'd want
to see the entire IRR array (and perhaps also PI state) if the check
in the assertion fails.

Jan


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