[libvirt] [PATCH v3 07/17] qemu: Implement NVDIMM
John Ferlan
jferlan at redhat.com
Tue Mar 14 13:36:15 UTC 2017
On 03/09/2017 11:06 AM, Michal Privoznik wrote:
> So, majority of the code is just ready as-is. Well, with one
> slight change: differentiate between dimm and nvdimm in places
> like device alias generation, generating the command line and so
> on.
>
> Speaking of the command line, we also need to append 'nvdimm=on'
> to the '-machine' argument so that the nvdimm feature is
> advertised in the ACPI tables properly.
>
> Signed-off-by: Michal Privoznik <mprivozn at redhat.com>
> ---
> src/qemu/qemu_alias.c | 10 ++-
> src/qemu/qemu_command.c | 76 +++++++++++++++-------
> src/qemu/qemu_domain.c | 42 ++++++++----
> .../qemuxml2argv-memory-hotplug-nvdimm.args | 26 ++++++++
> tests/qemuxml2argvtest.c | 4 +-
> 5 files changed, 121 insertions(+), 37 deletions(-)
> create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-memory-hotplug-nvdimm.args
>
[...]
> diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
> index 07178f839..a66ce6645 100644
> --- a/src/qemu/qemu_command.c
> +++ b/src/qemu/qemu_command.c
> @@ -3118,7 +3118,8 @@ qemuBuildMemoryBackendStr(virJSONValuePtr *backendProps,
> const long system_page_size = virGetSystemPageSizeKB();
> virDomainMemoryAccess memAccess = VIR_DOMAIN_MEMORY_ACCESS_DEFAULT;
> size_t i;
> - char *mem_path = NULL;
> + char *memPath = NULL;
> + bool prealloc = false;
> virBitmapPtr nodemask = NULL;
> int ret = -1;
> virJSONValuePtr props = NULL;
> @@ -3199,26 +3200,31 @@ qemuBuildMemoryBackendStr(virJSONValuePtr *backendProps,
> if (!(props = virJSONValueNewObject()))
> return -1;
>
> - if (pagesize || def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE) {
> + if (pagesize || mem->nvdimmPath ||
> + def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE) {
> *backendType = "memory-backend-file";
>
> - if (def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE) {
> - /* we can have both pagesize and mem source, then check mem source first */
> - if (virJSONValueObjectAdd(props,
> - "s:mem-path", cfg->memoryBackingDir,
> - NULL) < 0)
> + if (mem->nvdimmPath) {
> + if (VIR_STRDUP(memPath, mem->nvdimmPath) < 0)
> + goto cleanup;
> + prealloc = true;
> + } else if (def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE) {
> + /* We can have both pagesize and mem source,
> + * then check mem source first. */
> + if (VIR_STRDUP(memPath, cfg->memoryBackingDir) < 0)
> goto cleanup;
> } else {
> - if (qemuGetDomainHupageMemPath(def, cfg, pagesize, &mem_path) < 0)
> - goto cleanup;
> -
> - if (virJSONValueObjectAdd(props,
> - "b:prealloc", true,
> - "s:mem-path", mem_path,
> - NULL) < 0)
> + if (qemuGetDomainHupageMemPath(def, cfg, pagesize, &memPath) < 0)
> goto cleanup;
> + prealloc = true;
> }
>
FWIW: In my v2 review I alluded to a double comma thing (e.g. code that
needs virQEMUBuildBufferEscapeComma)... This is what I was thinking
about, but IIRC it's only something for certain command line options and
not JSON object building...
> + if (virJSONValueObjectAdd(props,
> + "B:prealloc", prealloc,
> + "s:mem-path", memPath,
> + NULL) < 0)
> + goto cleanup;
> +
> switch (memAccess) {
> case VIR_DOMAIN_MEMORY_ACCESS_SHARED:
> if (virJSONValueObjectAdd(props, "b:share", true, NULL) < 0)
> @@ -3274,6 +3280,7 @@ qemuBuildMemoryBackendStr(virJSONValuePtr *backendProps,
>
> /* If none of the following is requested... */
> if (!needHugepage && !mem->sourceNodes && !nodeSpecified &&
> + !mem->nvdimmPath &&
> memAccess == VIR_DOMAIN_MEMORY_ACCESS_DEFAULT &&
> def->mem.source != VIR_DOMAIN_MEMORY_SOURCE_FILE && !force) {
> /* report back that using the new backend is not necessary
> @@ -3303,8 +3310,7 @@ qemuBuildMemoryBackendStr(virJSONValuePtr *backendProps,
>
> cleanup:
> virJSONValueFree(props);
> - VIR_FREE(mem_path);
> -
> + VIR_FREE(memPath);
> return ret;
> }
[...]
> diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
> index 66c0e5911..495242981 100644
> --- a/src/qemu/qemu_domain.c
> +++ b/src/qemu/qemu_domain.c
> @@ -5919,6 +5919,7 @@ qemuDomainDefValidateMemoryHotplugDevice(const virDomainMemoryDef *mem,
> {
> switch ((virDomainMemoryModel) mem->model) {
> case VIR_DOMAIN_MEMORY_MODEL_DIMM:
> + case VIR_DOMAIN_MEMORY_MODEL_NVDIMM:
> if (mem->info.type != VIR_DOMAIN_DEVICE_ADDRESS_TYPE_DIMM &&
> mem->info.type != VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE) {
> virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> @@ -5951,11 +5952,6 @@ qemuDomainDefValidateMemoryHotplugDevice(const virDomainMemoryDef *mem,
> }
> break;
>
> - case VIR_DOMAIN_MEMORY_MODEL_NVDIMM:
> - virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> - _("nvdimm hotplug not supported yet"));
> - return -1;
> -
> case VIR_DOMAIN_MEMORY_MODEL_NONE:
> case VIR_DOMAIN_MEMORY_MODEL_LAST:
> return -1;
> @@ -5986,6 +5982,8 @@ qemuDomainDefValidateMemoryHotplug(const virDomainDef *def,
> unsigned int nmems = def->nmems;
> unsigned long long hotplugSpace;
> unsigned long long hotplugMemory = 0;
> + bool needPCDimmCap = false;
> + bool needNvdimmCap = false;
needNVDimm could be used.... although I see no reason for these bool's
the way the rest is written.
> size_t i;
>
> hotplugSpace = def->mem.max_memory - virDomainDefGetMemoryInitial(def);
> @@ -6009,12 +6007,6 @@ qemuDomainDefValidateMemoryHotplug(const virDomainDef *def,
> return 0;
> }
>
> - if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_PC_DIMM)) {
> - virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> - _("memory hotplug isn't supported by this QEMU binary"));
> - return -1;
> - }
> -
> if (!ARCH_IS_PPC64(def->os.arch)) {
> /* due to guest support, qemu would silently enable NUMA with one node
> * once the memory hotplug backend is enabled. To avoid possible
> @@ -6038,6 +6030,34 @@ qemuDomainDefValidateMemoryHotplug(const virDomainDef *def,
> for (i = 0; i < def->nmems; i++) {
> hotplugMemory += def->mems[i]->size;
>
> + switch ((virDomainMemoryModel) def->mems[i]->model) {
> + case VIR_DOMAIN_MEMORY_MODEL_DIMM:
> + needPCDimmCap = true;
> + break;
> +
> + case VIR_DOMAIN_MEMORY_MODEL_NVDIMM:
> + needNvdimmCap = true;
> + break;
> +
> + case VIR_DOMAIN_MEMORY_MODEL_NONE:
> + case VIR_DOMAIN_MEMORY_MODEL_LAST:
> + break;
> + }
> +
> + if (needPCDimmCap &&
> + !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_PC_DIMM)) {
> + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> + _("memory hotplug isn't supported by this QEMU binary"));
> + return -1;
> + }
> +
> + if (needNvdimmCap &&
> + !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_NVDIMM)) {
> + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> + _("nvdimm isn't supported by this QEMU binary"));
> + return -1;
> + }
> +
Still inefficient as virQEMUCapsGet will get called each time through
the for loop as soon as need*Cap = true... Perhaps even moreso since
one or the other will be called both times for a single pass if there
both types of *Dimm defs are in the XML (once one or the other is seen).
I think the bool's should be turned into 'checkedPCDimmCap' and
'checkedNVDimmCap' though and that would be less inefficient.
e.g. within the case:
case VIR_DOMAIN_MEMORY_MODEL_DIMM:
if (!checkedPCDimmCap) {
if (!virQEMUCapsGet())
failure
checkedPCDimmCap = true;
}
> /* already existing devices don't need to be checked on hotplug */
> if (!mem &&
> qemuDomainDefValidateMemoryHotplugDevice(def->mems[i], def) < 0)
ACK - I think it would be good to not make a CapsGet on every pass
through though
John
More information about the libvir-list
mailing list