[v6] NUMA: Enable adding NUMA node implicitly

Message ID 1510626841-9928-1-git-send-email-douly.fnst@cn.fujitsu.com
State New
Headers show
Series
  • [v6] NUMA: Enable adding NUMA node implicitly
Related show

Commit Message

Dou Liyang Nov. 14, 2017, 2:34 a.m.
Linux and Windows need ACPI SRAT table to make memory hotplug work properly,
however currently QEMU doesn't create SRAT table if numa options aren't present
on CLI.

Which breaks both linux and windows guests in certain conditions:
 * Windows: won't enable memory hotplug without SRAT table at all
 * Linux: if QEMU is started with initial memory all below 4Gb and no SRAT table
   present, guest kernel will use nommu DMA ops, which breaks 32bit hw drivers
   when memory is hotplugged and guest tries to use it with that drivers.

Fix above issues by automatically creating a numa node when QEMU is started with
memory hotplug enabled but without '-numa' options on CLI.
(PS: auto-create numa node only for new machine types so not to break migration).

Which would provide SRAT table to guests without explicit -numa options on CLI
and would allow:
 * Windows: to enable memory hotplug
 * Linux: switch to SWIOTLB DMA ops, to bounce DMA transfers to 32bit allocated
   buffers that legacy drivers/hw can handle.

[Rewritten by Igor]

Reported-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Suggested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Dou Liyang <douly.fnst@cn.fujitsu.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Cc: Alistair Francis <alistair23@gmail.com>
Cc: Takao Indoh <indou.takao@jp.fujitsu.com>
Cc: Izumi Taku <izumi.taku@jp.fujitsu.com>
---
changelog V5 --> V6:
  - rebase it to avoid building failure
  - test again
---
 hw/i386/pc.c        |  1 +
 hw/i386/pc_piix.c   |  1 +
 hw/i386/pc_q35.c    |  1 +
 include/hw/boards.h |  1 +
 numa.c              | 21 ++++++++++++++++++++-
 vl.c                |  3 +--
 6 files changed, 25 insertions(+), 3 deletions(-)

Comments

Igor Mammedov Nov. 14, 2017, 3:30 p.m. | #1
On Tue, 14 Nov 2017 10:34:01 +0800
Dou Liyang <douly.fnst@cn.fujitsu.com> wrote:

> Linux and Windows need ACPI SRAT table to make memory hotplug work properly,
> however currently QEMU doesn't create SRAT table if numa options aren't present
> on CLI.
> 
> Which breaks both linux and windows guests in certain conditions:
>  * Windows: won't enable memory hotplug without SRAT table at all
>  * Linux: if QEMU is started with initial memory all below 4Gb and no SRAT table
>    present, guest kernel will use nommu DMA ops, which breaks 32bit hw drivers
>    when memory is hotplugged and guest tries to use it with that drivers.
> 
> Fix above issues by automatically creating a numa node when QEMU is started with
> memory hotplug enabled but without '-numa' options on CLI.
> (PS: auto-create numa node only for new machine types so not to break migration).
> 
> Which would provide SRAT table to guests without explicit -numa options on CLI
> and would allow:
>  * Windows: to enable memory hotplug
>  * Linux: switch to SWIOTLB DMA ops, to bounce DMA transfers to 32bit allocated
>    buffers that legacy drivers/hw can handle.
> 
> [Rewritten by Igor]
> 
> Reported-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
> Suggested-by: Igor Mammedov <imammedo@redhat.com>
> Signed-off-by: Dou Liyang <douly.fnst@cn.fujitsu.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Richard Henderson <rth@twiddle.net>
> Cc: Eduardo Habkost <ehabkost@redhat.com>
> Cc: "Michael S. Tsirkin" <mst@redhat.com>
> Cc: Marcel Apfelbaum <marcel@redhat.com>
> Cc: Igor Mammedov <imammedo@redhat.com>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Thomas Huth <thuth@redhat.com>
> Cc: Alistair Francis <alistair23@gmail.com>
> Cc: Takao Indoh <indou.takao@jp.fujitsu.com>
> Cc: Izumi Taku <izumi.taku@jp.fujitsu.com>
> ---
> changelog V5 --> V6:
>   - rebase it to avoid building failure
>   - test again
> ---
>  hw/i386/pc.c        |  1 +
>  hw/i386/pc_piix.c   |  1 +
>  hw/i386/pc_q35.c    |  1 +
>  include/hw/boards.h |  1 +
>  numa.c              | 21 ++++++++++++++++++++-
>  vl.c                |  3 +--
>  6 files changed, 25 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> index e11a65b..156501c 100644
> --- a/hw/i386/pc.c
> +++ b/hw/i386/pc.c
> @@ -2325,6 +2325,7 @@ static void pc_machine_class_init(ObjectClass *oc, void *data)
>      mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
>      mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
>      mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
> +    mc->auto_enable_numa_with_memhp = true;
>      mc->has_hotpluggable_cpus = true;
>      mc->default_boot_order = "cad";
>      mc->hot_add_cpu = pc_hot_add_cpu;
> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
> index f79d5cb..5e47528 100644
> --- a/hw/i386/pc_piix.c
> +++ b/hw/i386/pc_piix.c
> @@ -446,6 +446,7 @@ static void pc_i440fx_2_10_machine_options(MachineClass *m)
>      m->is_default = 0;
>      m->alias = NULL;
>      SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
> +    m->auto_enable_numa_with_memhp = false;
>  }
>  
>  DEFINE_I440FX_MACHINE(v2_10, "pc-i440fx-2.10", NULL,
> diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
> index da3ea60..d606004 100644
> --- a/hw/i386/pc_q35.c
> +++ b/hw/i386/pc_q35.c
> @@ -318,6 +318,7 @@ static void pc_q35_2_10_machine_options(MachineClass *m)
>      m->alias = NULL;
>      SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
>      m->numa_auto_assign_ram = numa_legacy_auto_assign_ram;
> +    m->auto_enable_numa_with_memhp = false;
>  }
>  
>  DEFINE_Q35_MACHINE(v2_10, "pc-q35-2.10", NULL,
> diff --git a/include/hw/boards.h b/include/hw/boards.h
> index 191a5b3..f1077f1 100644
> --- a/include/hw/boards.h
> +++ b/include/hw/boards.h
> @@ -192,6 +192,7 @@ struct MachineClass {
>      bool ignore_memory_transaction_failures;
>      int numa_mem_align_shift;
>      const char **valid_cpu_types;
> +    bool auto_enable_numa_with_memhp;
>      void (*numa_auto_assign_ram)(MachineClass *mc, NodeInfo *nodes,
>                                   int nb_nodes, ram_addr_t size);
>  
> diff --git a/numa.c b/numa.c
> index 8d78d95..7151b24 100644
> --- a/numa.c
> +++ b/numa.c
> @@ -216,6 +216,7 @@ static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
>      }
>      numa_info[nodenr].present = true;
>      max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
> +    nb_numa_nodes++;
>  }
>  
>  static void parse_numa_distance(NumaDistOptions *dist, Error **errp)
> @@ -282,7 +283,6 @@ static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
>          if (err) {
>              goto end;
>          }
> -        nb_numa_nodes++;
>          break;
>      case NUMA_OPTIONS_TYPE_DIST:
>          parse_numa_distance(&object->u.dist, &err);
> @@ -433,6 +433,25 @@ void parse_numa_opts(MachineState *ms)
>          exit(1);
>      }
>  
> +    /*
> +     * If memory hotplug is enabled (slots > 0) but without '-numa'
> +     * options explicitly on CLI, guestes will break.
s/guestes/guests/

> +     *
> +     *   Windows: won't enable memory hotplug without SRAT table at all
> +     *
> +     *   Linux: if QEMU is started with initial memory all below 4Gb
> +     *   and no SRAT table present, guest kernel will use nommu DMA ops,
> +     *   which breaks 32bit hw drivers when memory is hotplugged and
> +     *   guest tries to use it with that drivers.
> +     *
> +     * Enable NUMA implicitly by adding a new NUMA node automatically.
> +     */
> +    if (ms->ram_slots > 0 && nb_numa_nodes == 0 &&
> +        mc->auto_enable_numa_with_memhp) {
> +            NumaNodeOptions node = { };
> +            parse_numa_node(ms, &node, NULL);
I get build break here:

numa.c:451:13: error: too few arguments to function ‘parse_numa_node’
             parse_numa_node(ms, &node, NULL);

> +    }
> +
>      assert(max_numa_nodeid <= MAX_NODES);
>  
>      /* No support for sparse NUMA node IDs yet: */
> diff --git a/vl.c b/vl.c
> index ec29909..be332d1 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -4675,8 +4675,6 @@ int main(int argc, char **argv, char **envp)
>      default_drive(default_floppy, snapshot, IF_FLOPPY, 0, FD_OPTS);
>      default_drive(default_sdcard, snapshot, IF_SD, 0, SD_OPTS);
>  
> -    parse_numa_opts(current_machine);
> -
>      if (qemu_opts_foreach(qemu_find_opts("mon"),
>                            mon_init_func, NULL, NULL)) {
>          exit(1);
> @@ -4726,6 +4724,7 @@ int main(int argc, char **argv, char **envp)
>      current_machine->boot_order = boot_order;
>      current_machine->cpu_model = cpu_model;
>  
> +    parse_numa_opts(current_machine);
>  
>      /* parse features once if machine provides default cpu_type */
>      if (machine_class->default_cpu_type) {
Dou Liyang Nov. 15, 2017, 1:29 a.m. | #2
Hi Igor,

[...]
>> +            parse_numa_node(ms, &node, NULL);
> I get build break here:
>
> numa.c:451:13: error: too few arguments to function ‘parse_numa_node’
>              parse_numa_node(ms, &node, NULL);
>

In upstream tree, your commit

   cc001888b780 ("numa: fixup parsed NumaNodeOptions earlier")

removed a argument from parse_numa_node() recently. this definition
of function becomes

static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
                             Error **errp)

this patch is based on the upstream tree, parse_numa_node() should have
three arguments.

I am not sure why you got this building failure log, can you tell me
which branch did you test?

Thanks,
	dou
Igor Mammedov Nov. 15, 2017, 10:32 a.m. | #3
On Wed, 15 Nov 2017 09:29:22 +0800
Dou Liyang <douly.fnst@cn.fujitsu.com> wrote:

> Hi Igor,
> 
> [...]
> >> +            parse_numa_node(ms, &node, NULL);  
> > I get build break here:
> >
> > numa.c:451:13: error: too few arguments to function ‘parse_numa_node’
> >              parse_numa_node(ms, &node, NULL);
> >  
> 
> In upstream tree, your commit
> 
>    cc001888b780 ("numa: fixup parsed NumaNodeOptions earlier")
> 
> removed a argument from parse_numa_node() recently. this definition
> of function becomes
> 
> static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
>                              Error **errp)
> 
> this patch is based on the upstream tree, parse_numa_node() should have
> three arguments.
> 
> I am not sure why you got this building failure log, can you tell me
> which branch did you test?
it looks like I've failed to update master branch and applied it to old master,
so after refetching and applying it again build and tests passes.

(Sorry for  noise :/)

> 
> Thanks,
> 	dou
> 
> 
>
Igor Mammedov Nov. 15, 2017, 10:35 a.m. | #4
On Tue, 14 Nov 2017 10:34:01 +0800
Dou Liyang <douly.fnst@cn.fujitsu.com> wrote:

> Linux and Windows need ACPI SRAT table to make memory hotplug work properly,
> however currently QEMU doesn't create SRAT table if numa options aren't present
> on CLI.
> 
> Which breaks both linux and windows guests in certain conditions:
>  * Windows: won't enable memory hotplug without SRAT table at all
>  * Linux: if QEMU is started with initial memory all below 4Gb and no SRAT table
>    present, guest kernel will use nommu DMA ops, which breaks 32bit hw drivers
>    when memory is hotplugged and guest tries to use it with that drivers.
> 
> Fix above issues by automatically creating a numa node when QEMU is started with
> memory hotplug enabled but without '-numa' options on CLI.
> (PS: auto-create numa node only for new machine types so not to break migration).
> 
> Which would provide SRAT table to guests without explicit -numa options on CLI
> and would allow:
>  * Windows: to enable memory hotplug
>  * Linux: switch to SWIOTLB DMA ops, to bounce DMA transfers to 32bit allocated
>    buffers that legacy drivers/hw can handle.
> 
> [Rewritten by Igor]
> 
> Reported-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
> Suggested-by: Igor Mammedov <imammedo@redhat.com>
> Signed-off-by: Dou Liyang <douly.fnst@cn.fujitsu.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Richard Henderson <rth@twiddle.net>
> Cc: Eduardo Habkost <ehabkost@redhat.com>
> Cc: "Michael S. Tsirkin" <mst@redhat.com>
> Cc: Marcel Apfelbaum <marcel@redhat.com>
> Cc: Igor Mammedov <imammedo@redhat.com>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Thomas Huth <thuth@redhat.com>
> Cc: Alistair Francis <alistair23@gmail.com>
> Cc: Takao Indoh <indou.takao@jp.fujitsu.com>
> Cc: Izumi Taku <izumi.taku@jp.fujitsu.com>
> ---
> changelog V5 --> V6:
>   - rebase it to avoid building failure
>   - test again
Reviewed-by: Igor Mammedov <imammedo@redhat.com>


> ---
>  hw/i386/pc.c        |  1 +
>  hw/i386/pc_piix.c   |  1 +
>  hw/i386/pc_q35.c    |  1 +
>  include/hw/boards.h |  1 +
>  numa.c              | 21 ++++++++++++++++++++-
>  vl.c                |  3 +--
>  6 files changed, 25 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> index e11a65b..156501c 100644
> --- a/hw/i386/pc.c
> +++ b/hw/i386/pc.c
> @@ -2325,6 +2325,7 @@ static void pc_machine_class_init(ObjectClass *oc, void *data)
>      mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
>      mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
>      mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
> +    mc->auto_enable_numa_with_memhp = true;
>      mc->has_hotpluggable_cpus = true;
>      mc->default_boot_order = "cad";
>      mc->hot_add_cpu = pc_hot_add_cpu;
> diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
> index f79d5cb..5e47528 100644
> --- a/hw/i386/pc_piix.c
> +++ b/hw/i386/pc_piix.c
> @@ -446,6 +446,7 @@ static void pc_i440fx_2_10_machine_options(MachineClass *m)
>      m->is_default = 0;
>      m->alias = NULL;
>      SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
> +    m->auto_enable_numa_with_memhp = false;
>  }
>  
>  DEFINE_I440FX_MACHINE(v2_10, "pc-i440fx-2.10", NULL,
> diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
> index da3ea60..d606004 100644
> --- a/hw/i386/pc_q35.c
> +++ b/hw/i386/pc_q35.c
> @@ -318,6 +318,7 @@ static void pc_q35_2_10_machine_options(MachineClass *m)
>      m->alias = NULL;
>      SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
>      m->numa_auto_assign_ram = numa_legacy_auto_assign_ram;
> +    m->auto_enable_numa_with_memhp = false;
>  }
>  
>  DEFINE_Q35_MACHINE(v2_10, "pc-q35-2.10", NULL,
> diff --git a/include/hw/boards.h b/include/hw/boards.h
> index 191a5b3..f1077f1 100644
> --- a/include/hw/boards.h
> +++ b/include/hw/boards.h
> @@ -192,6 +192,7 @@ struct MachineClass {
>      bool ignore_memory_transaction_failures;
>      int numa_mem_align_shift;
>      const char **valid_cpu_types;
> +    bool auto_enable_numa_with_memhp;
>      void (*numa_auto_assign_ram)(MachineClass *mc, NodeInfo *nodes,
>                                   int nb_nodes, ram_addr_t size);
>  
> diff --git a/numa.c b/numa.c
> index 8d78d95..7151b24 100644
> --- a/numa.c
> +++ b/numa.c
> @@ -216,6 +216,7 @@ static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
>      }
>      numa_info[nodenr].present = true;
>      max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
> +    nb_numa_nodes++;
>  }
>  
>  static void parse_numa_distance(NumaDistOptions *dist, Error **errp)
> @@ -282,7 +283,6 @@ static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
>          if (err) {
>              goto end;
>          }
> -        nb_numa_nodes++;
>          break;
>      case NUMA_OPTIONS_TYPE_DIST:
>          parse_numa_distance(&object->u.dist, &err);
> @@ -433,6 +433,25 @@ void parse_numa_opts(MachineState *ms)
>          exit(1);
>      }
>  
> +    /*
> +     * If memory hotplug is enabled (slots > 0) but without '-numa'
> +     * options explicitly on CLI, guestes will break.
> +     *
> +     *   Windows: won't enable memory hotplug without SRAT table at all
> +     *
> +     *   Linux: if QEMU is started with initial memory all below 4Gb
> +     *   and no SRAT table present, guest kernel will use nommu DMA ops,
> +     *   which breaks 32bit hw drivers when memory is hotplugged and
> +     *   guest tries to use it with that drivers.
> +     *
> +     * Enable NUMA implicitly by adding a new NUMA node automatically.
> +     */
> +    if (ms->ram_slots > 0 && nb_numa_nodes == 0 &&
> +        mc->auto_enable_numa_with_memhp) {
> +            NumaNodeOptions node = { };
> +            parse_numa_node(ms, &node, NULL);
> +    }
> +
>      assert(max_numa_nodeid <= MAX_NODES);
>  
>      /* No support for sparse NUMA node IDs yet: */
> diff --git a/vl.c b/vl.c
> index ec29909..be332d1 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -4675,8 +4675,6 @@ int main(int argc, char **argv, char **envp)
>      default_drive(default_floppy, snapshot, IF_FLOPPY, 0, FD_OPTS);
>      default_drive(default_sdcard, snapshot, IF_SD, 0, SD_OPTS);
>  
> -    parse_numa_opts(current_machine);
> -
>      if (qemu_opts_foreach(qemu_find_opts("mon"),
>                            mon_init_func, NULL, NULL)) {
>          exit(1);
> @@ -4726,6 +4724,7 @@ int main(int argc, char **argv, char **envp)
>      current_machine->boot_order = boot_order;
>      current_machine->cpu_model = cpu_model;
>  
> +    parse_numa_opts(current_machine);
>  
>      /* parse features once if machine provides default cpu_type */
>      if (machine_class->default_cpu_type) {

Patch

diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index e11a65b..156501c 100644
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -2325,6 +2325,7 @@  static void pc_machine_class_init(ObjectClass *oc, void *data)
     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
+    mc->auto_enable_numa_with_memhp = true;
     mc->has_hotpluggable_cpus = true;
     mc->default_boot_order = "cad";
     mc->hot_add_cpu = pc_hot_add_cpu;
diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
index f79d5cb..5e47528 100644
--- a/hw/i386/pc_piix.c
+++ b/hw/i386/pc_piix.c
@@ -446,6 +446,7 @@  static void pc_i440fx_2_10_machine_options(MachineClass *m)
     m->is_default = 0;
     m->alias = NULL;
     SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
+    m->auto_enable_numa_with_memhp = false;
 }
 
 DEFINE_I440FX_MACHINE(v2_10, "pc-i440fx-2.10", NULL,
diff --git a/hw/i386/pc_q35.c b/hw/i386/pc_q35.c
index da3ea60..d606004 100644
--- a/hw/i386/pc_q35.c
+++ b/hw/i386/pc_q35.c
@@ -318,6 +318,7 @@  static void pc_q35_2_10_machine_options(MachineClass *m)
     m->alias = NULL;
     SET_MACHINE_COMPAT(m, PC_COMPAT_2_10);
     m->numa_auto_assign_ram = numa_legacy_auto_assign_ram;
+    m->auto_enable_numa_with_memhp = false;
 }
 
 DEFINE_Q35_MACHINE(v2_10, "pc-q35-2.10", NULL,
diff --git a/include/hw/boards.h b/include/hw/boards.h
index 191a5b3..f1077f1 100644
--- a/include/hw/boards.h
+++ b/include/hw/boards.h
@@ -192,6 +192,7 @@  struct MachineClass {
     bool ignore_memory_transaction_failures;
     int numa_mem_align_shift;
     const char **valid_cpu_types;
+    bool auto_enable_numa_with_memhp;
     void (*numa_auto_assign_ram)(MachineClass *mc, NodeInfo *nodes,
                                  int nb_nodes, ram_addr_t size);
 
diff --git a/numa.c b/numa.c
index 8d78d95..7151b24 100644
--- a/numa.c
+++ b/numa.c
@@ -216,6 +216,7 @@  static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
     }
     numa_info[nodenr].present = true;
     max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
+    nb_numa_nodes++;
 }
 
 static void parse_numa_distance(NumaDistOptions *dist, Error **errp)
@@ -282,7 +283,6 @@  static int parse_numa(void *opaque, QemuOpts *opts, Error **errp)
         if (err) {
             goto end;
         }
-        nb_numa_nodes++;
         break;
     case NUMA_OPTIONS_TYPE_DIST:
         parse_numa_distance(&object->u.dist, &err);
@@ -433,6 +433,25 @@  void parse_numa_opts(MachineState *ms)
         exit(1);
     }
 
+    /*
+     * If memory hotplug is enabled (slots > 0) but without '-numa'
+     * options explicitly on CLI, guestes will break.
+     *
+     *   Windows: won't enable memory hotplug without SRAT table at all
+     *
+     *   Linux: if QEMU is started with initial memory all below 4Gb
+     *   and no SRAT table present, guest kernel will use nommu DMA ops,
+     *   which breaks 32bit hw drivers when memory is hotplugged and
+     *   guest tries to use it with that drivers.
+     *
+     * Enable NUMA implicitly by adding a new NUMA node automatically.
+     */
+    if (ms->ram_slots > 0 && nb_numa_nodes == 0 &&
+        mc->auto_enable_numa_with_memhp) {
+            NumaNodeOptions node = { };
+            parse_numa_node(ms, &node, NULL);
+    }
+
     assert(max_numa_nodeid <= MAX_NODES);
 
     /* No support for sparse NUMA node IDs yet: */
diff --git a/vl.c b/vl.c
index ec29909..be332d1 100644
--- a/vl.c
+++ b/vl.c
@@ -4675,8 +4675,6 @@  int main(int argc, char **argv, char **envp)
     default_drive(default_floppy, snapshot, IF_FLOPPY, 0, FD_OPTS);
     default_drive(default_sdcard, snapshot, IF_SD, 0, SD_OPTS);
 
-    parse_numa_opts(current_machine);
-
     if (qemu_opts_foreach(qemu_find_opts("mon"),
                           mon_init_func, NULL, NULL)) {
         exit(1);
@@ -4726,6 +4724,7 @@  int main(int argc, char **argv, char **envp)
     current_machine->boot_order = boot_order;
     current_machine->cpu_model = cpu_model;
 
+    parse_numa_opts(current_machine);
 
     /* parse features once if machine provides default cpu_type */
     if (machine_class->default_cpu_type) {