[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[PATCH v1 1/4] qemu: revert latest pSeries NVDIMM design changes



In [1], changes were made to remove the existing auto-alignment
for pSeries NVDIMM devices. That design promotes strange situations
where the NVDIMM size reported in the domain XML is different
from what QEMU is actually using. We removed the auto-alignment
and relied on standard size validation.

However, this goes against Libvirt design philosophy of not
tampering with existing guest behavior, as pointed out by Daniel
in [2]. Since we can't know for sure whether there are guests that
are relying on the auto-alignment feature to work, the changes
made in [1] are a direct violation of this rule.

This patch reverts [1] entirely, re-enabling auto-alignment for
pSeries NVDIMM as it was before. Changes will be made to ease
the limitations of this design without hurting existing
guests.

This reverts the following commits:

- commit 0ee56369c8b4f2f898b6aa1ff1f51ab033be1c02,
"qemu_domain.c: change qemuDomainMemoryDeviceAlignSize() return type".

- commit 07de813924caf37e535855541c0c1183d9d382e2,
"qemu_domain.c: do not auto-align ppc64 NVDIMMs"

- commit 0ccceaa57c50e5ee528f7073fa8723afd62b88b7,
"qemu_validate.c: add pSeries NVDIMM size alignment validation"

- commit 4fa2202d884414ad34d9952e72fb39b1d93c7e14,
"qemu_domain.c: make qemuDomainGetMemorySizeAlignment() public"

- commit 2d93cbdea9d1b8dbf36bc0ffee6cb73d83d208c7,
"Revert "formatdomain.html.in: mention pSeries NVDIMM 'align down' mechanic""

[1] https://www.redhat.com/archives/libvir-list/2020-July/msg02010.html
[2] https://www.redhat.com/archives/libvir-list/2020-September/msg00572.html
---
 docs/formatdomain.rst                         |  6 +-
 src/qemu/qemu_domain.c                        | 57 +++++++++++++++++--
 src/qemu/qemu_domain.h                        |  6 +-
 src/qemu/qemu_hotplug.c                       |  6 +-
 src/qemu/qemu_validate.c                      | 42 ++------------
 ...ory-hotplug-nvdimm-ppc64.ppc64-latest.args |  2 +-
 .../memory-hotplug-nvdimm-ppc64.xml           |  2 +-
 .../memory-hotplug-nvdimm-ppc64.xml           |  2 +-
 8 files changed, 70 insertions(+), 53 deletions(-)

diff --git a/docs/formatdomain.rst b/docs/formatdomain.rst
index 18e237c157..d5930a4ac1 100644
--- a/docs/formatdomain.rst
+++ b/docs/formatdomain.rst
@@ -7216,8 +7216,10 @@ Example: usage of the memory devices
       following restrictions apply:
 
       #. the minimum label size is 128KiB,
-      #. the remaining size (total-size - label-size) will be aligned
-         to 4KiB as default.
+      #. the remaining size (total-size - label-size), also called guest area,
+         will be aligned to 4KiB as default. For pSeries guests, the guest area
+         will be aligned down to 256MiB, and the minimum size of the guest area
+         must be at least 256MiB.
 
    ``readonly``
       The ``readonly`` element is used to mark the vNVDIMM as read-only. Only
diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index 03917cf000..4f796bef4a 100644
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -8037,7 +8037,7 @@ qemuDomainAgentAvailable(virDomainObjPtr vm,
 }
 
 
-unsigned long long
+static unsigned long long
 qemuDomainGetMemorySizeAlignment(const virDomainDef *def)
 {
     /* PPC requires the memory sizes to be rounded to 256MiB increments, so
@@ -8067,6 +8067,44 @@ qemuDomainGetMemoryModuleSizeAlignment(const virDomainDef *def,
 }
 
 
+static int
+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. Error
+     * out if target_size is smaller than 256MiB + label_size,
+     * since aligning it up will cause QEMU errors. */
+    if (mem->size < (ppc64AlignSize + mem->labelsize)) {
+        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                       _("minimum target size for the NVDIMM "
+                         "must be 256MB plus the label size"));
+        return -1;
+    }
+
+    guestArea = (guestArea/ppc64AlignSize) * ppc64AlignSize;
+    mem->size = guestArea + mem->labelsize;
+
+    return 0;
+}
+
+
 int
 qemuDomainAlignMemorySizes(virDomainDefPtr def)
 {
@@ -8113,8 +8151,11 @@ qemuDomainAlignMemorySizes(virDomainDefPtr def)
 
     /* Align memory module sizes */
     for (i = 0; i < def->nmems; i++) {
-        if (!(def->mems[i]->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM &&
-              ARCH_IS_PPC64(def->os.arch))) {
+        if (def->mems[i]->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM &&
+            ARCH_IS_PPC64(def->os.arch)) {
+            if (qemuDomainNVDimmAlignSizePseries(def, def->mems[i]) < 0)
+                return -1;
+        } else {
             align = qemuDomainGetMemoryModuleSizeAlignment(def, def->mems[i]);
             def->mems[i]->size = VIR_ROUND_UP(def->mems[i]->size, align);
         }
@@ -8143,15 +8184,19 @@ qemuDomainAlignMemorySizes(virDomainDefPtr def)
  * inplace. Default rounding is now to 1 MiB (qemu requires rounding to page,
  * size so this should be safe).
  */
-void
+int
 qemuDomainMemoryDeviceAlignSize(virDomainDefPtr def,
                                 virDomainMemoryDefPtr mem)
 {
-    if (!(mem->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM &&
-          ARCH_IS_PPC64(def->os.arch))) {
+    if (mem->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM &&
+        ARCH_IS_PPC64(def->os.arch)) {
+        return qemuDomainNVDimmAlignSizePseries(def, mem);
+    } else {
         mem->size = VIR_ROUND_UP(mem->size,
                                  qemuDomainGetMemorySizeAlignment(def));
     }
+
+    return 0;
 }
 
 
diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h
index adba79aded..e6d4acb8d4 100644
--- a/src/qemu/qemu_domain.h
+++ b/src/qemu/qemu_domain.h
@@ -751,11 +751,9 @@ bool qemuDomainDiskBlockJobIsActive(virDomainDiskDefPtr disk);
 bool qemuDomainHasBlockjob(virDomainObjPtr vm, bool copy_only)
     ATTRIBUTE_NONNULL(1);
 
-unsigned long long qemuDomainGetMemorySizeAlignment(const virDomainDef *def);
-
 int qemuDomainAlignMemorySizes(virDomainDefPtr def);
-void qemuDomainMemoryDeviceAlignSize(virDomainDefPtr def,
-                                     virDomainMemoryDefPtr mem);
+int qemuDomainMemoryDeviceAlignSize(virDomainDefPtr def,
+                                    virDomainMemoryDefPtr mem);
 
 virDomainChrDefPtr qemuFindAgentConfig(virDomainDefPtr def);
 
diff --git a/src/qemu/qemu_hotplug.c b/src/qemu/qemu_hotplug.c
index e2c6e14c2e..3a780d00af 100644
--- a/src/qemu/qemu_hotplug.c
+++ b/src/qemu/qemu_hotplug.c
@@ -2368,7 +2368,8 @@ qemuDomainAttachMemory(virQEMUDriverPtr driver,
     int id;
     int ret = -1;
 
-    qemuDomainMemoryDeviceAlignSize(vm->def, mem);
+    if (qemuDomainMemoryDeviceAlignSize(vm->def, mem) < 0)
+        goto cleanup;
 
     if (qemuDomainDefValidateMemoryHotplug(vm->def, priv->qemuCaps, mem) < 0)
         goto cleanup;
@@ -5612,7 +5613,8 @@ qemuDomainDetachPrepMemory(virDomainObjPtr vm,
     virDomainMemoryDefPtr mem;
     int idx;
 
-    qemuDomainMemoryDeviceAlignSize(vm->def, match);
+    if (qemuDomainMemoryDeviceAlignSize(vm->def, match) < 0)
+        return -1;
 
     if ((idx = virDomainMemoryFindByDef(vm->def, match)) < 0) {
         virReportError(VIR_ERR_DEVICE_MISSING,
diff --git a/src/qemu/qemu_validate.c b/src/qemu/qemu_validate.c
index 070f1c962b..c56ad70623 100644
--- a/src/qemu/qemu_validate.c
+++ b/src/qemu/qemu_validate.c
@@ -4034,45 +4034,15 @@ qemuValidateDomainDeviceDefHub(virDomainHubDefPtr hub,
 }
 
 
-static unsigned long long
-qemuValidateGetNVDIMMAlignedSizePseries(virDomainMemoryDefPtr mem,
-                                        const virDomainDef *def)
-{
-    unsigned long long ppc64AlignSize = qemuDomainGetMemorySizeAlignment(def);
-    unsigned long long guestArea = mem->size - mem->labelsize;
-
-    /* NVDIMM is already aligned */
-    if (guestArea % ppc64AlignSize == 0)
-        return mem->size;
-
-    /* Suggested aligned size is rounded up */
-    guestArea = (guestArea/ppc64AlignSize + 1) * ppc64AlignSize;
-    return guestArea + mem->labelsize;
-}
-
 static int
 qemuValidateDomainDeviceDefMemory(virDomainMemoryDefPtr mem,
-                                  const virDomainDef *def,
                                   virQEMUCapsPtr qemuCaps)
 {
-    if (mem->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM) {
-        if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_NVDIMM)) {
-            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
-                           _("nvdimm isn't supported by this QEMU binary"));
-            return -1;
-        }
-
-        if (qemuDomainIsPSeries(def)) {
-            unsigned long long alignedNVDIMMSize =
-                qemuValidateGetNVDIMMAlignedSizePseries(mem, def);
-
-            if (mem->size != alignedNVDIMMSize) {
-                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
-                               _("nvdimm size is not aligned. Suggested aligned "
-                                 "size: %llu KiB"), alignedNVDIMMSize);
-                return -1;
-            }
-        }
+    if (mem->model == VIR_DOMAIN_MEMORY_MODEL_NVDIMM &&
+        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_NVDIMM)) {
+        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                       _("nvdimm isn't supported by this QEMU binary"));
+        return -1;
     }
 
     return 0;
@@ -4190,7 +4160,7 @@ qemuValidateDomainDeviceDef(const virDomainDeviceDef *dev,
         break;
 
     case VIR_DOMAIN_DEVICE_MEMORY:
-        ret = qemuValidateDomainDeviceDefMemory(dev->data.memory, def, qemuCaps);
+        ret = qemuValidateDomainDeviceDefMemory(dev->data.memory, qemuCaps);
         break;
 
     case VIR_DOMAIN_DEVICE_LEASE:
diff --git a/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.ppc64-latest.args b/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.ppc64-latest.args
index 58e3f9e161..eff80dcf80 100644
--- a/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.ppc64-latest.args
+++ b/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.ppc64-latest.args
@@ -19,7 +19,7 @@ file=/tmp/lib/domain--1-QEMUGuest1/master-key.aes \
 -smp 2,sockets=2,dies=1,cores=1,threads=1 \
 -numa node,nodeid=0,cpus=0-1,mem=1024 \
 -object memory-backend-file,id=memnvdimm0,prealloc=yes,mem-path=/tmp/nvdimm,\
-size=805437440 \
+size=537001984 \
 -device nvdimm,node=0,label-size=131072,\
 uuid=49545eb3-75e1-2d0a-acdd-f0294406c99e,memdev=memnvdimm0,id=nvdimm0,slot=0 \
 -uuid c7a5fdbd-edaf-9455-926a-d65c16db1809 \
diff --git a/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.xml b/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.xml
index 10c146e8cf..ae5a17d3c8 100644
--- a/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.xml
+++ b/tests/qemuxml2argvdata/memory-hotplug-nvdimm-ppc64.xml
@@ -38,7 +38,7 @@
         <path>/tmp/nvdimm</path>
       </source>
       <target>
-        <size unit='KiB'>786560</size>
+        <size unit='KiB'>550000</size>
         <node>0</node>
         <label>
           <size unit='KiB'>128</size>
diff --git a/tests/qemuxml2xmloutdata/memory-hotplug-nvdimm-ppc64.xml b/tests/qemuxml2xmloutdata/memory-hotplug-nvdimm-ppc64.xml
index 10c146e8cf..ae5a17d3c8 100644
--- a/tests/qemuxml2xmloutdata/memory-hotplug-nvdimm-ppc64.xml
+++ b/tests/qemuxml2xmloutdata/memory-hotplug-nvdimm-ppc64.xml
@@ -38,7 +38,7 @@
         <path>/tmp/nvdimm</path>
       </source>
       <target>
-        <size unit='KiB'>786560</size>
+        <size unit='KiB'>550000</size>
         <node>0</node>
         <label>
           <size unit='KiB'>128</size>
-- 
2.26.2


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]