On Mon, Nov 28, 2016 at 06:53:35AM -0500, John Ferlan wrote:
>
>
> On 11/25/2016 10:28 AM, Erik Skultety wrote:
> > On Mon, Nov 21, 2016 at 06:35:48PM -0500, John Ferlan wrote:
> >> Rather than using negative logic and setting the maxparams to a lesser
> >> value based on which capabilities exist, alter the logic to modify the
> >> maxparams based on a base value plus the found capabilities. Reduces the
> >> chance that some backported feature produces an incorrect value.
> >>
> >> Signed-off-by: John Ferlan <jferlan@xxxxxxxxxx>
> >> ---
> >> src/qemu/qemu_driver.c | 23 ++++++++++++++---------
> >> 1 file changed, 14 insertions(+), 9 deletions(-)
> >>
> >> diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
> >> index d039255..87d219f 100644
> >> --- a/src/qemu/qemu_driver.c
> >> +++ b/src/qemu/qemu_driver.c
> >> @@ -112,9 +112,12 @@ VIR_LOG_INIT("qemu.qemu_driver");
> >>
> >> #define QEMU_NB_MEM_PARAM 3
> >>
> >> -#define QEMU_NB_BLOCK_IO_TUNE_PARAM 6
> >> -#define QEMU_NB_BLOCK_IO_TUNE_PARAM_MAX 13
> >> -#define QEMU_NB_BLOCK_IO_TUNE_PARAM_MAX_LENGTH 19
> >> +#define QEMU_NB_BLOCK_IO_TUNE_BASE_PARAMS 6
> >> +#define QEMU_NB_BLOCK_IO_TUNE_MAX_PARAMS 7
> >> +#define QEMU_NB_BLOCK_IO_TUNE_LENGTH_PARAMS 6
> >> +#define QEMU_NB_BLOCK_IO_TUNE_ALL_PARAMS (QEMU_NB_BLOCK_IO_TUNE_BASE_PARAMS + \
> >> + QEMU_NB_BLOCK_IO_TUNE_MAX_PARAMS + \
> >> + QEMU_NB_BLOCK_IO_TUNE_LENGTH_PARAMS)
> >>
> >> #define QEMU_NB_NUMA_PARAM 2
> >>
> >> @@ -17719,7 +17722,7 @@ qemuDomainGetBlockIoTune(virDomainPtr dom,
> >> virDomainBlockIoTuneInfo reply;
> >> char *device = NULL;
> >> int ret = -1;
> >> - int maxparams = QEMU_NB_BLOCK_IO_TUNE_PARAM_MAX_LENGTH;
>
> NB: "current" persistent def max - moved just to make it clearer what
> the max was vis-a-vis persistent vs. live.
>
> >> + int maxparams;
> >>
> >> virCheckFlags(VIR_DOMAIN_AFFECT_LIVE |
> >> VIR_DOMAIN_AFFECT_CONFIG |
> >> @@ -17753,11 +17756,13 @@ qemuDomainGetBlockIoTune(virDomainPtr dom,
> >> goto endjob;
> >> }
> >>
> >> - if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_IOTUNE_MAX))
> >> - maxparams = QEMU_NB_BLOCK_IO_TUNE_PARAM;
> >> - else if (!virQEMUCapsGet(priv->qemuCaps,
> >> - QEMU_CAPS_DRIVE_IOTUNE_MAX_LENGTH))
> >> - maxparams = QEMU_NB_BLOCK_IO_TUNE_PARAM_MAX;
> >> + maxparams = QEMU_NB_BLOCK_IO_TUNE_BASE_PARAMS;
> >> + if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_IOTUNE_MAX))
> >> + maxparams += QEMU_NB_BLOCK_IO_TUNE_MAX_PARAMS;
> >> + if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_IOTUNE_MAX_LENGTH))
> >> + maxparams += QEMU_NB_BLOCK_IO_TUNE_LENGTH_PARAMS;
> >> + } else {
> >> + maxparams = QEMU_NB_BLOCK_IO_TUNE_ALL_PARAMS;
> >> }
> >>
> >
> > I'm curious about the else branch. Can you elaborate more on why do you assume
> > qemu supports all the features when you retrieve a persistentdef vs livedef?
>
> History mainly. How does one choose otherwise? When it's persistent,
> then you can get/return all 'valid' values vs. when it's live you can
> only get/return those that the domain can retrieve. Typically the order
> of fetching is done weighted towards which version the value was added.
> That's why I stuck group_name between the "max" and "max_length" - so it
> has a chance of being fetched instead of one max_length value as I know
> it was added to qemu in between the two.
>
> >
> > From what I understand, this is one of the APIs that you have to call twice,
> > since the RPC layer does not allocate a structure of an appropriate size
> > automatically. So with the first run, you say, please allocate a structure of
> > size maxparams equal _ALL_PARAMS (without knowing that qemu actually supports
> > all these features) and rerun.
>
> True - it's one of those pass 0 or pass 0 and/or some other value in
> order to indicate that you want to get a count so that you can allocate
> an appropriately sized buffer.
>
> > Then during the second run, you run BLOCK_IOTUNE_ASSIGN on all of them, what am
> > I missing?
>
> Not quite sure what you're asking - if dig into the macro you'll note a
> comparison "if (*nparams < maxparams && " which essentially ensures that
> we'll only ever fill in as many as we should. On the "second" pass
> through the code we'll hit the else of the "if (*nparams == 0) {" which
> can further 'refine' maxparams to be whatever was passed in *nparams
> (e.g. whatever was allocated by the caller). Similar to any other chunk
> of code there certainly is assumption that the allocation was done
> correctly.
Nah, I missed this line: reply = disk->blkdeviotune; - it all makes more sense
now and all the 'invalid' - more like 'unsupported' - attributes will just be
zeroed out automatically during the initial allocation.
Thanks anyway,
Erik
>
> FWIW: Let's say on your first pass the domain isn't running and you
> return ALL, but on the second pass the domain is running and maxparams
> gets set to MAX. That means we'll fetch up to 'MAX' from the running
> domain with the other entries being left at whatever they were
> initialized to by the caller. We'll also return MAX (a lesser value) in
> nparams from the call indicating to the caller that less values were
> fetched than the size of the buffer. It's up to the caller to recognize
> that if they so desire. They could also print out whatever size they
> passed in.
>
>
> John
>
> >
> > Erik
> >
> >> if (*nparams == 0) {
> >> --
> >> 2.7.4
> >>
> >> --
> >> libvir-list mailing list
> >> libvir-list@xxxxxxxxxx
> >> https://www.redhat.com/mailman/listinfo/libvir-list
Attachment:
signature.asc
Description: PGP signature
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
