Using c/s 20189, I can''t boot my (only) system without APIC anymore. I didn''t have time to look into the issue in more detail so far, but since it seems likely that this is connected to the recent per-CPU-IRQ-vector changes, I wanted to ask whether these had been tested on APIC-less systems. Thanks, Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Zhang, Xiantao
2009-Sep-12 01:58 UTC
[Xen-devel] RE: regression from IRQ handling changes?
Jan,
Frankly speaking, we haven''t done the testings which can cover
non-APIC case. Does anybody really care about such case now ? I can''t
figure out the reason why modern system doesn''t leverage apic. If
non-APIC case is necessary, maybe we need to fix it. As I know, Xen is
becoming more and more dependent on the components (APIC, ACPI....), so non-APIC
or non-ACPI cases maybe becoming less and less important.
Xiantao
Jan Beulich wrote:> Using c/s 20189, I can''t boot my (only) system without APIC
anymore. I
> didn''t have time to look into the issue in more detail so far, but
> since it seems likely that this is connected to the recent
> per-CPU-IRQ-vector changes, I wanted to ask whether these had been
> tested on APIC-less systems.
>
> Thanks, Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
While agreeing from a technical perspective, customer perspective on our products is different: Whenever something like proper UP (no-APIC) support breaks in our products, we almost immediately get reports of this, since there will always be someone running Xen e.g. on an old laptop (no matter whether Xen is meant to be run there). Less frequent for ACPI, but quite recently we had a report there too (and I''m under the impression that no-ACPI support has been broken for quite a while) - luckily, a BIOS update fixed the issue for them, so I didn''t have to look into the issue in more detail. Otoh, if we really wanted to switch over to APIC and ACPI being a requirement, I wouldn''t mind that step - it just should be that it''s clear whether this is intended to work. And if it isn''t, quite a bit of code should be eliminated/cleaned up. Jan>>> "Zhang, Xiantao" <xiantao.zhang@intel.com> 12.09.09 03:58 >>>Jan, Frankly speaking, we haven''t done the testings which can cover non-APIC case. Does anybody really care about such case now ? I can''t figure out the reason why modern system doesn''t leverage apic. If non-APIC case is necessary, maybe we need to fix it. As I know, Xen is becoming more and more dependent on the components (APIC, ACPI....), so non-APIC or non-ACPI cases maybe becoming less and less important. Xiantao Jan Beulich wrote:> Using c/s 20189, I can''t boot my (only) system without APIC anymore. I > didn''t have time to look into the issue in more detail so far, but > since it seems likely that this is connected to the recent > per-CPU-IRQ-vector changes, I wanted to ask whether these had been > tested on APIC-less systems. > > Thanks, Jan_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2009-Sep-14 14:30 UTC
Re: [Xen-devel] RE: regression from IRQ handling changes?
It isn''t hard to boot UP and with APIC disabled to test this case. I would prefer to contineu to support it for a while longer, so please do take a look at it Xiantao. Thanks, Keir On 14/09/2009 07:59, "Jan Beulich" <JBeulich@novell.com> wrote:> While agreeing from a technical perspective, customer perspective on our > products is different: Whenever something like proper UP (no-APIC) > support breaks in our products, we almost immediately get reports of this, > since there will always be someone running Xen e.g. on an old laptop (no > matter whether Xen is meant to be run there). > > Less frequent for ACPI, but quite recently we had a report there too (and > I''m under the impression that no-ACPI support has been broken for quite > a while) - luckily, a BIOS update fixed the issue for them, so I didn''t have > to look into the issue in more detail. > > Otoh, if we really wanted to switch over to APIC and ACPI being a > requirement, I wouldn''t mind that step - it just should be that it''s clear > whether this is intended to work. And if it isn''t, quite a bit of code should > be eliminated/cleaned up. > > Jan > >>>> "Zhang, Xiantao" <xiantao.zhang@intel.com> 12.09.09 03:58 >>> > Jan, > Frankly speaking, we haven''t done the testings which can cover non-APIC > case. Does anybody really care about such case now ? I can''t figure out the > reason why modern system doesn''t leverage apic. If non-APIC case is necessary, > maybe we need to fix it. As I know, Xen is becoming more and more dependent > on the components (APIC, ACPI....), so non-APIC or non-ACPI cases maybe > becoming less and less important. > Xiantao > > Jan Beulich wrote: >> Using c/s 20189, I can''t boot my (only) system without APIC anymore. I >> didn''t have time to look into the issue in more detail so far, but >> since it seems likely that this is connected to the recent >> per-CPU-IRQ-vector changes, I wanted to ask whether these had been >> tested on APIC-less systems. >> >> Thanks, Jan > > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Zhang, Xiantao
2009-Sep-14 15:13 UTC
RE: [Xen-devel] RE: regression from IRQ handling changes?
Okay, I will have a look about it. As you said, it should be easy to fix it. Xiantao Keir Fraser wrote:> It isn''t hard to boot UP and with APIC disabled to test this case. I > would prefer to contineu to support it for a while longer, so please > do take a look at it Xiantao. > > Thanks, > Keir > > On 14/09/2009 07:59, "Jan Beulich" <JBeulich@novell.com> wrote: > >> While agreeing from a technical perspective, customer perspective on >> our products is different: Whenever something like proper UP >> (no-APIC) >> support breaks in our products, we almost immediately get reports of >> this, since there will always be someone running Xen e.g. on an old >> laptop (no matter whether Xen is meant to be run there). >> >> Less frequent for ACPI, but quite recently we had a report there too >> (and >> I''m under the impression that no-ACPI support has been broken for >> quite >> a while) - luckily, a BIOS update fixed the issue for them, so I >> didn''t have to look into the issue in more detail. >> >> Otoh, if we really wanted to switch over to APIC and ACPI being a >> requirement, I wouldn''t mind that step - it just should be that it''s >> clear whether this is intended to work. And if it isn''t, quite a bit >> of code should be eliminated/cleaned up. >> >> Jan >> >>>>> "Zhang, Xiantao" <xiantao.zhang@intel.com> 12.09.09 03:58 >>> >> Jan, >> Frankly speaking, we haven''t done the testings which can cover >> non-APIC case. Does anybody really care about such case now ? I >> can''t figure out the reason why modern system doesn''t leverage apic. >> If non-APIC case is necessary, maybe we need to fix it. As I know, >> Xen is becoming more and more dependent on the components (APIC, >> ACPI....), so non-APIC or non-ACPI cases maybe becoming less and >> less important. >> Xiantao >> >> Jan Beulich wrote: >>> Using c/s 20189, I can''t boot my (only) system without APIC >>> anymore. I didn''t have time to look into the issue in more detail >>> so far, but since it seems likely that this is connected to the >>> recent per-CPU-IRQ-vector changes, I wanted to ask whether these >>> had been tested on APIC-less systems. >>> >>> Thanks, Jan >> >> >> >> _______________________________________________ >> Xen-devel mailing list >> Xen-devel@lists.xensource.com >> http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel