[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v7 2/5] sysctl: Add sysctl interface for querying PCI topology



On 21/04/15 14:13, Boris Ostrovsky wrote:
>
> On 04/21/2015 09:14 AM, Andrew Cooper wrote:
>> On 21/04/15 13:56, Boris Ostrovsky wrote:
>>> On 04/21/2015 03:01 AM, Jan Beulich wrote:
>>>>>>> --- a/docs/misc/xsm-flask.txt
>>>>>>> +++ b/docs/misc/xsm-flask.txt
>>>>>>> @@ -121,6 +121,7 @@ __HYPERVISOR_sysctl
>>>>>>> (xen/include/public/sysctl.h)
>>>>>>>     * XEN_SYSCTL_cpupool_op
>>>>>>>     * XEN_SYSCTL_scheduler_op
>>>>>>>     * XEN_SYSCTL_coverage_op
>>>>>>> + * XEN_SYSCTL_pcitopoinfo
>>>> No additions to this list are permitted. Either the new sub-op is
>>>> disaggregation safe (which it looks to be), or it can't be accepted.
>>> True, it *is* safe, but why then cputopoinfo and numainfo are in this
>>> list? They look to be safe as well.
>> This list includes "not yet audited".  It is quite likely that some
>> entries in the list are safe.
>>
>> fwiw, neither cputopo nor numainfo are currently safe for very large
>> systems.
>
> Why would safety of an operation depend on system size? And how are
> these two unsafe? (Because maybe then PCI topology query is unsafe in
> the same manner)

PCI topology has continuations in it.

cpu/numa info is bounded by host values, but does attempt to cover all
cpus/nodes in a single pass, which will get progressively longer as the
server gets larger.

~Andrew

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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.