[PATCH 4/5] qemu: Prefer -numa cpu over -numa node,cpus=

Igor Mammedov imammedo at redhat.com
Fri Jun 5 11:35:24 UTC 2020


On Thu, 4 Jun 2020 10:58:01 +0200
Michal Privoznik <mprivozn at redhat.com> wrote:

> On 5/27/20 3:58 PM, Igor Mammedov wrote:
> > On Tue, 26 May 2020 17:31:09 +0200
> > Michal Privoznik <mprivozn at redhat.com> wrote:
> >   
> >> On 5/26/20 4:51 PM, Igor Mammedov wrote:  
> >>> On Mon, 25 May 2020 10:05:08 +0200
> >>> Michal Privoznik <mprivozn at redhat.com> wrote:
> >>>      
> >>>>
> >>>> This is a problem. The domain XML that is provided can't be changed,
> >>>> mostly because mgmt apps construct it on the fly and then just pass it
> >>>> as a RO string to libvirt. While libvirt could create a separate cache,
> >>>> there has to be a better way.
> >>>>
> >>>> I mean, I can add some more code that once the guest is running
> >>>> preserves the mapping during migration. But that assumes a running QEMU.
> >>>> When starting a domain from scratch, is it acceptable it vCPU topology
> >>>> changes? I suspect it is not.  
> >>> I'm not sure I got you but
> >>> vCPU topology isn't changnig but when starting QEMU, user has to map
> >>> 'concrete vCPUs' to spencific numa nodes. The issue here is that
> >>> to specify concrete vCPUs user needs to get layout from QEMU first
> >>> as it's a function of target/machine/-smp and possibly cpu type.  
> >>
> >> Assume the following config: 4 vCPUs (2 sockets, 2 cores, 1 thread
> >> topology) and 2 NUMA nodes and the following assignment to NUMA:
> >>
> >> node 0: cpus=0-1
> >> node 1: cpus=2-3
> >>
> >> With old libvirt & qemu (and assuming x86_64 - not EPYC), I assume the
> >> following topology is going to be used:
> >>
> >> node 0: socket=0,core=0,thread=0 (vCPU0)  socket=0,core=1,thread=0 (vCPU1)
> >> node 1: socket=1,core=0,thread=0 (vCPU2)  socket=1,core=1,thread=0 (vCPU3)
> >>
> >> Now, user upgrades libvirt & qemu but doesn't change the config. And on
> >> a fresh new start (no migration), they might get a different topology:
> >>
> >> node 0: socket=0,core=0,thread=0 (vCPU0)  socket=1,core=0,thread=0 (vCPU1)
> >> node 1: socket=0,core=1,thread=0 (vCPU2)  socket=1,core=1,thread=0 (vCPU3)  
> > 
> > that shouldn't happen at least for as long as machine version stays the same  
> 
> Shouldn't as in it's bad if it happens or as in QEMU won't change 
> topology for released machine types?
it's the second

> Well, we are talking about libvirt 
> generating the topology.
> 
> >> The problem here is not how to assign vCPUs to NUMA nodes, the problem
> >> is how to translate vCPU IDs to socket=,core=,thread=.  
> > if you are talking about libvirt's vCPU IDs, then it's separate issue
> > as it's user facing API, I think it should not rely on cpu_index.
> > Instead it should map vCPU IDs to ([socket,]core[,thread]) tuple
> > or maybe drop notion of vCPU IDs and expose ([socket,]core[,thread])
> > to users if they ask for numa aware config.  
> 
> And this is the thing I am asking. How to map vCPU IDs to 
> socket,core,thread and how to do it reliably.
vCPU ID has the same drawbacks as cpu_index in QEMU, it provides zero
information about topology. Which is fine in non NUMA case since user
doesn't care about topology at all (I'm assuming it's libvirt who does
pinning and it would use topology info to pin vcpus correctly). 
But for NUMA case, as a user I'd like to see/use topology instead of
vCPU ID, especially if user is in charge of assigning vCPUs to nodes.

I'd drop vCPU IDs concept altogether and use ([socket,]core[,thread])
tuple to describe vCPUs instead. It should work fine for both usecases
and you wouldn't have to do mapping to vCPU IDs. (I'm talking here about
new configs that use new machine types and ignore compatibility. 
More on the later see below)
 
> > 
> > PS:
> > I'm curious how libvirt currently implements numa mapping and
> > how it's correlated with pinnig to host nodes?
> > Does it have any sort of code to calculate topology based on cpu_index
> > so it could properly assign vCPUs to nodes or all the pain of
> > assigning vCPU IDs to nodes is on the user shoulders?  
> 
> It's on users. In the domain XML they specify number of vCPUs, and then 
> they can assign individual IDs to NUMA nodes. For instance:
> 
>    <vcpu>8</vcpu>
> 
>    <cpu>
>       <numa>
>         <cell id='0' cpus='0-3' memory='2097152' unit='KiB'/>
>         <cell id='1' cpus='4-7' memory='2097152' unit='KiB'/>
>       </numa>
>    </cpu>
> 
> translates to:
> 
>    -smp 8,sockets=8,cores=1,threads=1
>    -numa node,nodeid=0,cpus=0-3,mem=...
>    -numa node,nodeid=1,cpus=4-7,mem=...
> 
> 
> The sockets=,cores=,threads= is formatted every time, even if no 
> topology was specified in the domain XML. If no topology was specified 
> then every vCPU is in its own socket and has 1 core and 1 thread.
> 
> If topology is specified then the -smp looks accordingly. But all that 
> libvirt uses to assing vCPUs to NUMA nodes is vCPU ID. If it has to use 
> sockets,cores,threads then so be it, but that means libvirt needs to 
> learn the mapping of vCPU IDs to sockets=,cores=,threads=; because if it 
> doesn't and generates the mapping differently to QEMU then for the above 
> snippet vCPUs might move between NUMA nodes. I mean, if there is a 
> domain with the above config it has some topology (that QEMU came up 
> with). Now, after we change it and user updates libvirt & QEMU, libvirt 
> might (in general) come with a different topology and if the VM is 
> booted again it will see say CPU1 move to NUMA#1 (for example).
> 
> This happened because libvirt came up with vCPU ID -> socket,core,thread 
> mapping itself. I mean, in this patch the algorithm is copied from 
> x86_topo_ids_from_idx(), but I bet there are different mappings (I can 
> see x86_topo_ids_from_idx_epyc() and other architectures might have 
> completely different mapping - powe9 perhaps?).

* in case of x86 if I recall correctly there are 3 mappings now,
  - x86_topo_ids_from_idx() original can produce 2 different tologies
    depending on compat_apic_id_mode
  - x86_topo_ids_from_idx_epyc - newish and probably now final yet,
    so there is a chance that we will have several variants if we decide
    to keep bugs on per machine version.
* Power (spapr) advertises (core_id) in query_hotpluggbale_cpus but internally
  it might have more than 1 thread per core so mapping to cpu_index is not
  1:1 over there
* arm/virt - should advertise (thread_id) in query_hotpluggbale_cpus
  and mapping to cpu_index happens to be 1:1 but a subject to change in future
* s390 - advertises (core_id) and mapping to cpu_index is 1:1 so far

Like I said before, for compatibility sake QEMU can introduce cpu_index 
in the output of query_hotpluggbale_cpus for released machine types, to help
libvirt to map old configs to '-numa cpus='. For new configs it would be better
move away from vCPU ID configs.


> Maybe I'm misunderstanding cpu_index and vCPU ID? I thought it is the 
> same thing.
>From your examples it looks like it's, and it should be decoupled from each
other for new configs (or dropped altogeter).

> 
> >   
> >>> that applies not only '-numa cpu' but also to -device cpufoo,
> >>> that's why query-hotpluggable-cpus was introduced to let
> >>> user get the list of possible CPUs (including topo properties needed to
> >>> create them) for a given set of CLI options.
> >>>
> >>> If I recall right libvirt uses topo properies during cpu hotplug but
> >>> treats it mainly as opaqueue info so it could feed it back to QEMU.
> >>>
> >>>      
> >>>>>> tries to avoid that as much as it can.
> >>>>>>        
> >>>>>>>
> >>>>>>> How to present it to libvirt user I'm not sure (give them that list perhaps
> >>>>>>> and let select from it???)  
> >>>>>>
> >>>>>> This is what I am trying to figure out in the cover letter. Maybe we
> >>>>>> need to let users configure the topology (well, vCPU id to [socket, die,
> >>>>>> core, thread] mapping), but then again, in my testing the guest ignored
> >>>>>> that and displayed different topology (true, I was testing with -cpu
> >>>>>> host, so maybe that's why).  
> >>>>> there is ongiong issue with EPYC VCPUs topology, but I otherwise it should work.
> >>>>> Just report bug to qemu-devel, if it's broken.
> >>>>>           
> >>>>>>        
> >>>>>>> But it's irrelevant, to the patch, magical IDs for socket/core/...whatever
> >>>>>>> should not be generated by libvirt anymore, but rather taken from QEMU for given
> >>>>>>> machine + -smp combination.  
> >>>>>>
> >>>>>> Taken when? We can do this for running machines, but not for freshly
> >>>>>> started ones, can we?  
> >>>>>
> >>>>> it can be used for freshly started as well,
> >>>>> QEMU -S -preconfig -M pc -smp ...
> >>>>> (QMP) query-hotpluggable-cpus
> >>>>> (QMP) set-numa-node ...
> >>>>> ...
> >>>>> (QMP) exit-preconfig
> >>>>> (QMP) other stuff libvirt does (like hot-plugging CPUs , ...)
> >>>>> (QMP) cont  
> >>>>
> >>>> I'm not sure this works. query-hotpluggable-cpus does not map vCPU ID
> >>>> <-> socket/core/thread, For '-smp 2,sockets=2,cores=1,threads=1' the
> >>>> 'query-hotpluggable-cpus' returns:
> >>>>
> >>>> {"return": [{"props": {"core-id": 0, "thread-id": 0, "socket-id": 1},
> >>>> "vcpus-count": 1, "type": "qemu64-x86_64-cpu"}, {"props": {"core-id": 0,
> >>>> "thread-id": 0, "socket-id": 0}, "vcpus-count": 1, "type":
> >>>> "qemu64-x86_64-cpu"}]}  
> >>>
> >>> that's the list I was taling about, which is implicitly ordered by cpu_index  
> >>
> >> Aha! So in this case it would be:
> >>
> >> vCPU0 -> socket=1,core=0,thread=0
> >> vCPU1 -> socket=0,core=0,thread=0
> >>
> >> But that doesn't feel right. Is the cpu_index increasing or decreasing
> >> as I go through the array?  
> > it's array with decreasing order and index in it currently == cpu_index for
> > present and possible CPUs. Content of array is immutable for given
> > -M/-smp combination, to keep migration working. We can try to add
> > x-cpu-index to cpu entries, so you won't have to rely on order to help with
> > migrating from old CLI (but only for old machine types where old CLI actually
> > worked worked).  
> 
> That might help. So we won't hardcode any mapping in libvirt rather than 
> ask QEMU what it thinks the topology is. Cool.
> 
> So it would work like this:
> 
> 1) libvirt starts:
>     qemu -preconfig -S -smp 8,sockets=2,cores=2,threads=2
> 
> 2) libvirt uses "query-hotpluggable-cpus" to learn what topology it came 
> up with, IOW what is the vCPU ID <-> socket,core,thread mapping
For power/arm/s390 provided tuples will be different (described above)

> 
> 3) libvirt configures NUMA nodes, assigns vCPUs to them using 
> [socket,core,thread] based on the mapping it learned in step 2)
> 
> 4) preconfig is exited, machine resumed
> 
> Very well. What I don't understand is why we need to have steps 2 and 3. 
> Because in step 2, QEMU needs to report the mapping. Therefore it has to 
> have some internal code that handles the mapping. Having said that, we 
> can have new set of steps:
 in step 2 
   """cpu_index""" <-> socket,core,thread mapping, could be provided only
   for old machine types to help with mapping old vCPU IDs based configs
   to 'new' [socket,core,thread] scheme. New machines should not
   provide cpu_index as part of "query-hotpluggable-cpus"

whole point of exercise is in getting rid of old ambiguous cpu_index
interface /-numa node,cpus/ which duplicates topology aware '-numa cpu'
one and dropping NUMA code that supports there /that's quite a bit/

That libvirt happens to have 1:1 mapping between cpu_index and vCPU IDs
is unfortunate and should be decoupled. It still can keep using
vCPU IDs as interface to upper layers but remap it internally to
socket,core,thread using its own arch-independed deterministic algorithm
for new configs/machine types.
However if I were a user, I'd like to describe CPUs (in config) using
topology attributes (doubly so if I'm the one who later assigns them
to NUMA nodes).

> 
> 1) libvirt starts:
>     qemu -preconfig -S -smp 8,sockets=2,cores=2,threads=2
> 
> 2) libvirt configures NUMA nodes, assigns vCPUs to them using vCPU IDs, 
> QEMU will use the internal code to map IDs to [socket,core,thread]
I really can't parse this sentence (in preconfig context).


> 3) preconfig is exited, machine resumed
> 
> 
> And since there is no need to preconfig anymore, we can have one step 
> actually:
> 
> 1) libvirt starts:
>    qemu -S -smp 8,sockets=2,cores=2,threads=2 -numa node -numa cpu,cpus= 
> -numa node -numa cpu,cpus=
you've probably meant here the old way /-numa node,cpus=/ which uses cpu_index

> Or, we can move the mapping into libvirt (that's what I tried to do in 
> this patch). I'm not against it, but we will need to do it exactly like 
> QEMU is doing now.
as shortcut it will work untill something changes on QEMU side,
and then it becomes silently broken [for new machine types].
That's the reason for introducing preconfig, so that topo info
could be introspected (per -M/-smp combination) and libvirt doesn't have
to duplicate and maintain complex architecture depended topology code
from QEMU.

> Then we can do plain
> 
> 1) qemu -S -smp 8,sockets=2,cores=2,threads=2 -numa node -numa 
> cpu,socket=,core=,thread= -numa node -numa cpu,socket=,core=,thread=
> 
> 
> >   
> >> Also, how is this able to express holes? E.g.
> >> there might be some CPUs that don't have linear topology, and for
> >> instance while socket=0,core=0,thread=0 and socket=0,core=0,thread=2
> >> exist, socket=0,core=0,thread=1 does not. How am I supposed to know that
> >> by just looking at the array?  
> > speaking of x86, QEMU curently does not implement topologies with holes
> > in [socket/core/thread] tuple but if it were it shouldn't matter as all
> > CPUs and their realations with each other are described within that array.
> > 
> >     
> >>>> And 'query-cpus' or 'query-cpus-fast' which map vCPU ID onto
> >>>> socket/core/thread are not allowed in preconfig state.  
> >>> these 2 commands apply to present cpu only, if I'm not mistaken.
> >>> query-hotpluggable-cpus shows not only present but also CPUs that
> >>> could be hotplugged with device_add or used with -device.  
> >>
> >> Fair enough. I haven't looked into the code that much.
> >>  
> >>>
> >>>      
> >>>> But if I take a step back, the whole point of deprecating -numa
> >>>> node,cpus= is that QEMU no longer wants to do vCPU ID <->
> >>>> socket/core/thread mapping because it's ambiguous. So it feels a bit
> >>>> weird to design a solution where libvirt would ask QEMU to provide the
> >>>> mapping only so that it can be configured back. Not only because of the
> >>>> extra step, but also because QEMU can't then remove the mapping anyway.
> >>>> I might be misunderstanding the issue though.  
> >>> if '-numa node,cpus' is removed, we no longer will be using cpu_index as
> >>> configuration interface with user, that would allow QEMU start pruning
> >>> it from HMP/QMP interfaces and then probably remove it internally.
> >>> (I haven't explored yet if we could get rid of it completely but
> >>> I'd expect migration stream would be the only reason to keep it intrenally).
> >>>
> >>> I'm quite reluctant to add cpu_index to modern query-hotpluggable-cpus output,
> >>> since the whole goal is to get rid of the index, which don't actually work
> >>> with SPAPR where CPU entity is a core and threads are internal impl. detail
> >>> (while cpu_index has 1:1 mapping with threads).
> >>>
> >>> However if it will let QEMU to drop '-numa node,cpus=', we can discuss
> >>> adding optional 'x-cpu-index' to query-hotpluggable-cpus, that will be available
> >>> for old machine types for the sole purpose to help libvirt map old CLI to new one.
> >>> New machines shouldn't care about index though, since they should be using
> >>> '-numa cpu'.  
> >>
> >> The problem here is that so far, all that libvirt users see are vCPU
> >> IDs. They use them to assign vCPUs to NUMA nodes. And in order to make
> >> libvirt switch to the new command line it needs a way to map IDs to
> >> socket=,core=,thread=. I will play more with the preconfig and let you know.  
> > 
> > If libvirt's vCPU IDs are mirroring cpu_index, I'd say it shouldn't be doing
> > so, see Daniel's response
> > https://lists.gnu.org/archive/html/qemu-devel/2017-10/msg04369.html  
> 
> No. Libvirt is not mirroring anything. Daniel's right, and libvirt 
> doesn't need to know actual qemu IDs. But it needs to ensure the 
> topology. This is exactly what you objected to in the very first reply 
> to the 4/5 patch. And I agree.
> 
> > 
> > FYI:
> > I didn't read through all the history of -preconfig patches but QEMU options
> > for topology aware (sane) numa configuration on the table were:
> >    1. -numa node,cpus[cpu_index]
> >        libvirt needs to duplicate internal QEMU algorithms that map cpu_index
> >        values to topology info and use it to map vCPUs to numa nodes
> >        (and keep in sync with QEMU as it's machine versioned moving target)  
> 
> I'm not sure I follow. So libvirt would continue to use -numa 
> node,cpus=. How does topology step into that?
keyword was 'were',
it was one of the options that were considered back then (that wasn't accepted)

 
> >    2. -numa cpu CLI option,
> >        libvirt needs to duplicate internal QEMU algorithms that calculate
> >        target depended values for socket/core/thread ids. (basically it's
> >        the same as #1), the only difference is that CLI user interface is
> >        expressed in topology properties.  
> 
> Sure, this is what I tried to do. But you suggested using preconfig + 
> query-hotpluggable-cpus.
it wasn't acceptable because #2 was fragile. It involves duplicating complex
architecture specific topology from QEMU (including machine version compact
hacks) and it has to be kept in sync (as it's sometimes extended to support
new layout)

> >    3. when we discussed it in the past #2 wasn't going to fly as it still
> >       had tha same burden as #1 (duplicating code and keeping it in sync).
> >       so we ended up with runtime configuration (-preconfig) to avoid QEMU
> >       restart just for querying, where libvirt could get list of possible CPUs
> >       from QEMU instance and complete numa configuration on the fly (at least
> >       for the first time, results could be cached and re-used with -numa cpu).
> >   
> 
> Yeah, this is the burden I am talking about. I feel like we are not 
> talking about the same thing. Maybe I'm misunderstanding something.
Perhaps we should 'meet' and discuss it (it might be easier to resove
misunderstanding).

> 
> Michal





More information about the libvir-list mailing list