On 11/15/18 4:55 AM, Michal Privoznik wrote: > On 11/05/2018 01:58 PM, John Ferlan wrote: >> Create a new API that will allow an adjustment of IOThread >> polling parameters for the specified IOThread. These parameters >> will not be saved in the guest XML. Currently the only parameters >> supported will allow the hypervisor to adjust the parameters used >> to limit and alter the scope of the polling interval. The polling >> interval allows the IOThread to spend more or less time processing >> in the guest. >> >> Based on code originally posted by Pavel Hrdina <phrdina@xxxxxxxxxx> >> to add virDomainAddIOThreadParams and virDomainModIOThreadParams. >> Modification of those changes to use virDomainSetIOThreadParams >> instead and remove concepts related to saving the data in guest >> XML as well as the way to specifically enable the polling parameters. >> >> Signed-off-by: John Ferlan <jferlan@xxxxxxxxxx> >> ACKed-by: Michal Privoznik <mprivozn@xxxxxxxxxx> >> --- >> include/libvirt/libvirt-domain.h | 44 ++++++++++++++++++++ >> src/driver-hypervisor.h | 8 ++++ >> src/libvirt-domain.c | 70 ++++++++++++++++++++++++++++++++ >> src/libvirt_public.syms | 5 +++ >> src/remote/remote_driver.c | 1 + >> src/remote/remote_protocol.x | 21 +++++++++- >> src/remote_protocol-structs | 10 +++++ >> 7 files changed, 158 insertions(+), 1 deletion(-) >> >> diff --git a/include/libvirt/libvirt-domain.h b/include/libvirt/libvirt-domain.h >> index 58fd4bc10c..bf89d0149f 100644 >> --- a/include/libvirt/libvirt-domain.h >> +++ b/include/libvirt/libvirt-domain.h >> @@ -1911,6 +1911,50 @@ int virDomainDelIOThread(virDomainPtr domain, >> unsigned int iothread_id, >> unsigned int flags); >> >> +/* IOThread set parameters */ >> + >> +/** >> + * VIR_DOMAIN_IOTHREAD_POLL_MAX_NS: >> + * >> + * The maximum polling time that can be used by polling algorithm in ns. >> + * The polling time starts at 0 (zero) and is the time spent by the guest >> + * to process IOThread data before returning the CPU to the host. The >> + * polling time will be dynamically modified over time based on the >> + * poll_grow and poll_shrink parameters provided. A value set too large >> + * will cause more CPU time to be allocated the guest. A value set too >> + * small will not provide enough cycles for the guest to process data. >> + * The polling interval is not available for statistical purposes. >> + */ >> +# define VIR_DOMAIN_IOTHREAD_POLL_MAX_NS "poll_max_ns" >> + >> +/** >> + * VIR_DOMAIN_IOTHREAD_POLL_GROW: >> + * >> + * This provides a value for the dynamic polling adjustment algorithm to >> + * use to grow its polling interval up to the poll_max_ns value. A value >> + * of 0 (zero) allows the hypervisor to choose its own value. The algorithm >> + * to use for adjustment is hypervisor specific. >> + */ >> +# define VIR_DOMAIN_IOTHREAD_POLL_GROW "poll_grow" >> + >> +/** >> + * VIR_DOMAIN_IOTHREAD_POLL_SHRINK: >> + * >> + * This provides a value for the dynamic polling adjustment algorithm to >> + * use to shrink its polling interval when the polling interval exceeds >> + * the poll_max_ns value. A value of 0 (zero) allows the hypervisor to >> + * choose its own value. The algorithm to use for adjustment is hypervisor >> + * specific. >> + */ >> +# define VIR_DOMAIN_IOTHREAD_POLL_SHRINK "poll_shrink" >> + >> +int virDomainSetIOThreadParams(virDomainPtr domain, > > On a completely unrelated note, this is stupid. I mean the amount of > spaces after 'int'. I wonder if a patch that reformats all the header > files would be accepted. I agree with the spacing... I assumed it had something to do with the docs pages generation; however, a quick test shows that by removing the extraneous spaces doesn't change the format that I can see in the docs. Furthermore there are other API's which don't do any spacing. So perhaps a nice task for a first time contributor to remove the spaces *and* make sure that it doesn't affect the webpage. Tks - John Since it'll conflict - I'll wait for you to push the 'memfd' stuff first w/ capabilities changes before pushing this... but don't wait too long ;-) > > The ACK still holds. > >> + unsigned int iothread_id, >> + virTypedParameterPtr params, >> + int nparams, >> + unsigned int flags); > > Michal > -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
