Re: [PATCH v2 3/5] conf, qemu: enable NVDIMM support for ppc64
On 3/20/20 4:53 AM, Shivaprasad bhat wrote:
On Thu, Mar 12, 2020 at 3:00 AM Daniel Henrique Barboza
<danielhb413(a)gmail.com> wrote:
>
> Using the 'uuid' element for ppc64 NVDIMM memory added in the
> previous patch, use it in qemuBuildMemoryDeviceStr() to pass
> it over to QEMU.
>
> Another ppc64 restriction is the necessity of a mem->labelsize,
> given than ppc64 only support label-area backed NVDIMMs.
>
> Finally, we don't want ppc64 NVDIMMs to align up due to the
> high risk of going beyond the end of file with a 256MiB
> increment that the user didn't predict. Align it down
> instead.
>
> Signed-off-by: Daniel Henrique Barboza <danielhb413(a)gmail.com>
> ---
...
> +static void
> +qemuDomainNVDimmAlignSizePseries(virDomainDefPtr def,
> + virDomainMemoryDefPtr mem)
> +{
> + /* For NVDIMMs in ppc64 in we want to align down the guest
> + * visible space, instead of align up, to avoid writing
> + * beyond the end of file by adding a potential 256MiB
> + * to the user specified size.
> + *
> + * The label-size is mandatory for ppc64 as well, meaning that
> + * the guest visible space will be target_size-label_size.
> + *
> + * Finally, target_size must include label_size.
> + *
> + * The above can be summed up as follows:
> + *
> + * target_size = AlignDown(target_size - label_size) + label_size
> + */
> + unsigned long long ppc64AlignSize = qemuDomainGetMemorySizeAlignment(def);
> + unsigned long long guestArea = mem->size - mem->labelsize;
> +
> + /* Align down guest_area. 256MiB is the minimum size. */
> + guestArea = (guestArea/ppc64AlignSize) * ppc64AlignSize;
> + guestArea = MAX(guestArea, ppc64AlignSize);
The math is correct, but we need additional checks when a backing file of
size less than 256MB is attempted to be hot-plugged. The qemu errors
out like below if
the backing file is 240MB and MAX of (240, 256) is chosen.
-object memory-backend-file,id=memnvdimm1,prealloc=yes,mem-path
=/tmp/nvdimm,share=yes,size=268566528: backing store size 0xf000000 does
not match 'size' option 0x10020000
Got it. I'll error out in this case.
> +
> + mem->size = guestArea + mem->labelsize;
> +}
> +
> +....
> + </label>
> </target>
> <address type='dimm' slot='0'/>
> </memory>
> diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c
> index 35d413d40b..077f7e7650 100644
> --- a/tests/qemuxml2argvtest.c
> +++ b/tests/qemuxml2argvtest.c
> @@ -2790,6 +2790,10 @@ mymain(void)
> DO_TEST_CAPS_LATEST("memory-hotplug-nvdimm-align");
> DO_TEST_CAPS_LATEST("memory-hotplug-nvdimm-pmem");
> DO_TEST_CAPS_LATEST("memory-hotplug-nvdimm-readonly");
> + DO_TEST("memory-hotplug-nvdimm-ppc64", QEMU_CAPS_KVM,
QEMU_CAPS_DEVICE_NVDIMM,
I see the qemucapabilitiesdata/caps-5*.ppc64* not updated with the
nvdimm support. Can you update it as well?
Then this can be changed to just DO_TEST_CAPS_LATEST_PPC64, without
this elaborate list of caps.
> + QEMU_CAPS_NUMA, QEMU_CAPS_DEVICE_SPAPR_PCI_HOST_BRIDGE,
> + QEMU_CAPS_OBJECT_MEMORY_RAM,
...
Roger that!
DHB
Thanks,
Shiva