search for: cpumap

Hello, Third version of the NUMA placement series Xen 4.2. All the comments received during v2''s review have been addressed (more details in single changelogs). The most notable changes are the following: - the libxl_cpumap --> libxl_bitmap renaming has been rebased on top of the recent patches that allows us to allocate bitmaps of different sizes; - the heuristics for deciding which NUMA placement is the best one has been redesigned, so that it now provides total ordering. Here it is what this posting cont...

...h was not an hard requirement, AFAIUI), because I don't think it is worthwhile in this case. Let me know what you think about it. Thanks and Regards, Dario --- Dario Faggioli (5): xl: update the manpage about "cpus=" and NUMA node-affinity libxl: introduce libxl_node_to_cpumap xl: allow for node-wise specification of vcpu pinning xl: implement and enable dryrun mode for `xl vcpu-pin' xl: test script for the cpumap parser (for vCPU pinning) docs/man/xl.cfg.pod.5 | 42 +++++-- tools/libxl/check-xl-vcpupin-parse | 229 +++++++++++++++...

...t; >> This series tries to re-enable XDP_REDIRECT for mergeable buffer which >> was removed since commit 7324f5399b06 ("virtio_net: disable >> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: >> >> - not enough tailroom was reserved which breaks cpumap > To address this at a more fundamental level, I would suggest that we/you > instead extend XDP to know it's buffers "frame" size/end. (The > assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but > ixgbe+virtio_net broke that assumption). > > It should...

...t; >> This series tries to re-enable XDP_REDIRECT for mergeable buffer which >> was removed since commit 7324f5399b06 ("virtio_net: disable >> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: >> >> - not enough tailroom was reserved which breaks cpumap > To address this at a more fundamental level, I would suggest that we/you > instead extend XDP to know it's buffers "frame" size/end. (The > assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but > ixgbe+virtio_net broke that assumption). > > It should...

...re-enable XDP_REDIRECT for mergeable buffer which >>>> was removed since commit 7324f5399b06 ("virtio_net: disable >>>> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: >>>> >>>> - not enough tailroom was reserved which breaks cpumap >>> To address this at a more fundamental level, I would suggest that we/you >>> instead extend XDP to know it's buffers "frame" size/end. (The >>> assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but >>> ixgbe+virtio_net broke that...

...re-enable XDP_REDIRECT for mergeable buffer which >>>> was removed since commit 7324f5399b06 ("virtio_net: disable >>>> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: >>>> >>>> - not enough tailroom was reserved which breaks cpumap >>> To address this at a more fundamental level, I would suggest that we/you >>> instead extend XDP to know it's buffers "frame" size/end. (The >>> assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but >>> ixgbe+virtio_net broke that...

Hi all, I am trying to add vcpupin in the guest xml file. I am working with openstack and the code I have is python bound. I investigated through the code and found that the elemets in xml file are set in get_guest_config function. Now the thing is I am not able to set vcpupin element. I mean I tried guest.cputune_vcpupin but it's not working. Help me out, please! Thanks.   ~Peeyush Gupta

...o VCPUOP_initialize. The separation allows for build-time allocation of vcpus which becomes more important when trying to allocate memory in a NUMA-aware manner (i.e. knowing which physical cpus are being used in a domain before we allocate the memory). The second patch adds a new domain allocmap cpumap_t which is a bitmap of physical cpus from which vcpus are to be allocated. As vcpus are created, the selection of which physical cpu is balanced across the set of physical cpus within the map. The patch lets us control vcpu allocation at a high level (start all VCPUS on hyperthreads, NODE1, CPU2)...

...ies tries to re-enable XDP_REDIRECT for mergeable buffer which > >> was removed since commit 7324f5399b06 ("virtio_net: disable > >> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: > >> > >> - not enough tailroom was reserved which breaks cpumap > > > To address this at a more fundamental level, I would suggest that we/you > > instead extend XDP to know it's buffers "frame" size/end. (The > > assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but > > ixgbe+virtio_net broke that assum...

...config function. Now the thing is I am not able to set > vcpupin element. I mean I tried guest.cputune_vcpupin but it's not > working. Hi Peeyush In libvirt python -binding . You can find the following functions for pining vcpus. class virDomain: ..... def pinVcpu(self, vcpu, cpumap): """Dynamically change the real CPUs which can be allocated to a virtual CPU. This function requires privileged access to the hypervisor. """ ret = libvirtmod.virDomainPinVcpu(self._o, vcpu, cpumap) if ret == -1: raise libvirtError ('vi...

...e_ticks)) / total_ticks; + return load; +} +#else + +#include <xen/interface/platform.h> +static int dbs_calc_load(struct cpu_dbs_info_s *this_dbs_info) +{ + int load = 0; + struct xen_platform_op op; + uint64_t idletime[NR_CPUS]; + struct cpufreq_policy *policy; + unsigned int j; + cpumask_t cpumap; + + op.cmd = XENPF_getidletime; + set_xen_guest_handle(op.u.getidletime.cpumap_bitmap, (uint8_t *) cpus_addr(cpumap)); + op.u.getidletime.cpumap_nr_cpus = NR_CPUS;// num_online_cpus(); + set_xen_guest_handle(op.u.getidletime.idletime, idletime); + if (HYPERVISOR_platfor...

...ECT for mergeable buffer which > >>>> was removed since commit 7324f5399b06 ("virtio_net: disable > >>>> XDP_REDIRECT in receive_mergeable() case"). Main concerns are: > >>>> > >>>> - not enough tailroom was reserved which breaks cpumap > >>> To address this at a more fundamental level, I would suggest that we/you > >>> instead extend XDP to know it's buffers "frame" size/end. (The > >>> assumption use to be, xdp_buff->data_hard_start + PAGE_SIZE, but > >>> ixgbe+...

...t; helping for the performance. So this patch remove it and switch to use > generic XDP routine to handle this case. I don't like where this is going. I don't like intermixing the native XDP and generic XDP in this way, for several reasons: 1. XDP generic is not feature complete, e.g. cpumap will drop these packets. It might not be possible to implement some features, think of (AF_XDP) zero-copy. 2. This can easily cause out-of-order packets. 3. It makes it harder to troubleshoot, when diagnosing issues around #1, we have a hard time determining what path an XDP packet to...

...t; helping for the performance. So this patch remove it and switch to use > generic XDP routine to handle this case. I don't like where this is going. I don't like intermixing the native XDP and generic XDP in this way, for several reasons: 1. XDP generic is not feature complete, e.g. cpumap will drop these packets. It might not be possible to implement some features, think of (AF_XDP) zero-copy. 2. This can easily cause out-of-order packets. 3. It makes it harder to troubleshoot, when diagnosing issues around #1, we have a hard time determining what path an XDP packet to...

...formance. So this patch remove it and switch to use >> generic XDP routine to handle this case. > I don't like where this is going. I don't like intermixing the native > XDP and generic XDP in this way, for several reasons: > > 1. XDP generic is not feature complete, e.g. cpumap will drop these > packets. Well, then we'd better fix it, otherwise it would be even worse than not having it. For cpumap, it can be done through queuing skb in the pointer ring with some encoding/decoding. > It might not be possible to implement some features, think > of...

...se` +(it is `true` by default) with something like the below, to prevent +any placement from happening: + + libxl_defbool_set(&domain_build_info->numa_placement, false); + +Also, if `numa_placement` is set to `true`, the domain must not +have any cpu affinity (i.e., `domain_build_info->cpumap` must +have all its bits set, as it is by default), or domain creation +will fail returning `ERROR_INVAL`. + +Besides than that, looking and/or tweaking the placement algorithm +search "Automatic NUMA placement" in libxl\_internal.h. + +Note this may change in future versions of Xen/libxl...

...ns(+), 9 deletions(-) -- diff -r baff9c7cc4b3 tools/libxc/xc_domain.c --- a/tools/libxc/xc_domain.c Fri May 04 11:23:25 2007 -0400 +++ b/tools/libxc/xc_domain.c Fri May 04 17:35:33 2007 -0400 @@ -108,7 +108,7 @@ int xc_vcpu_setaffinity(int xc_handle, set_xen_guest_handle(domctl.u.vcpuaffinity.cpumap.bitmap, local); domctl.u.vcpuaffinity.cpumap.nr_cpus = sizeof(cpumap) * 8; - + if ( lock_pages(local, sizeof(local)) != 0 ) { PERROR("Could not lock memory for Xen hypercall"); @@ -139,7 +139,7 @@ int xc_vcpu_getaffinity(int xc_handle, set_xen_guest_ha...

Modify the cpufreq ondemand governor so that it can get idle and total ticks from the Xen hypervisor. Linux and Xen have different ideas of what an idle tick is, so the Xen values for both have to be returned in the same platform hypercall. Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com> diff -r b4278beaf354 xen/arch/x86/platform_hypercall.c --- a/xen/arch/x86/platform_hypercall.c

Hi: This series tries to re-enable XDP_REDIRECT for mergeable buffer which was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). Main concerns are: - not enough tailroom was reserved which breaks cpumap - complex logic like EWMA and linearizing during XDP processing Fix those by: - reserve enough tailroom during refill - disable EWMA and use fixed size of rx buffer - drop linearizing logic and offload it to generic XDP routine, this could happen only when the buffer were refilled before XDP se...

Hi: This series tries to re-enable XDP_REDIRECT for mergeable buffer which was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). Main concerns are: - not enough tailroom was reserved which breaks cpumap - complex logic like EWMA and linearizing during XDP processing Fix those by: - reserve enough tailroom during refill - disable EWMA and use fixed size of rx buffer - drop linearizing logic and offload it to generic XDP routine, this could happen only when the buffer were refilled before XDP se...