[PATCH v2 0/3] spapr, spapr_numa: fix max-associativity-domains

[PATCH v2 0/3] spapr, spapr_numa: fix max-associativity-domains

[Daniel Henrique Barboza]

changes from v1:
- Patch 01: added Greg's R-b
- Patch 02:
    * removed trivial comment in spapr.c
    * removed the 'NUMA' caps in spapr_numa_initial_nvgpu_numa_id
    * added Greg's R-b
- Patch 03:
    * open-coded the logic inside spapr_numa_write_rtas_dt() instead of
      creating a helper
- added Cédric's 'Tested-by' in all patches
v1 link: https://lists.gnu.org/archive/html/qemu-devel/2021-01/msg07324.html

Daniel Henrique Barboza (3):
  spapr: move spapr_machine_using_legacy_numa() to spapr_numa.c
  spapr_numa.c: create spapr_numa_initial_nvgpu_numa_id() helper
  spapr_numa.c: fix ibm,max-associativity-domains calculation

 hw/ppc/spapr.c              | 20 +-------------------
 hw/ppc/spapr_numa.c         | 27 ++++++++++++++++++++++++++-
 include/hw/ppc/spapr.h      |  1 -
 include/hw/ppc/spapr_numa.h |  1 +
 4 files changed, 28 insertions(+), 21 deletions(-)

-- 
2.26.2

[PATCH v2 1/3] spapr: move spapr_machine_using_legacy_numa() to spapr_numa.c

[Daniel Henrique Barboza]

This function is used only in spapr_numa.c.

Tested-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
---
 hw/ppc/spapr.c         | 9 ---------
 hw/ppc/spapr_numa.c    | 9 +++++++++
 include/hw/ppc/spapr.h | 1 -
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 6c47466fc2..2d60c6f594 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -295,15 +295,6 @@ static hwaddr spapr_node0_size(MachineState *machine)
     return machine->ram_size;
 }
 
-bool spapr_machine_using_legacy_numa(SpaprMachineState *spapr)
-{
-    MachineState *machine = MACHINE(spapr);
-    SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
-
-    return smc->pre_5_2_numa_associativity ||
-           machine->numa_state->num_nodes <= 1;
-}
-
 static void add_str(GString *s, const gchar *s1)
 {
     g_string_append_len(s, s1, strlen(s1) + 1);
diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
index b50796bbe3..261810525b 100644
--- a/hw/ppc/spapr_numa.c
+++ b/hw/ppc/spapr_numa.c
@@ -19,6 +19,15 @@
 /* Moved from hw/ppc/spapr_pci_nvlink2.c */
 #define SPAPR_GPU_NUMA_ID           (cpu_to_be32(1))
 
+static bool spapr_machine_using_legacy_numa(SpaprMachineState *spapr)
+{
+    MachineState *machine = MACHINE(spapr);
+    SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
+
+    return smc->pre_5_2_numa_associativity ||
+           machine->numa_state->num_nodes <= 1;
+}
+
 static bool spapr_numa_is_symmetrical(MachineState *ms)
 {
     int src, dst;
diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
index c27c7ce515..ccbeeca1de 100644
--- a/include/hw/ppc/spapr.h
+++ b/include/hw/ppc/spapr.h
@@ -851,7 +851,6 @@ int spapr_max_server_number(SpaprMachineState *spapr);
 void spapr_store_hpte(PowerPCCPU *cpu, hwaddr ptex,
                       uint64_t pte0, uint64_t pte1);
 void spapr_mce_req_event(PowerPCCPU *cpu, bool recovered);
-bool spapr_machine_using_legacy_numa(SpaprMachineState *spapr);
 
 /* DRC callbacks. */
 void spapr_core_release(DeviceState *dev);
-- 
2.26.2

[PATCH v2 2/3] spapr_numa.c: create spapr_numa_initial_nvgpu_numa_id() helper

[Daniel Henrique Barboza]

We'll need to check the initial value given to spapr->gpu_numa_id when
building the rtas DT, so put it in a helper for easier access and to
avoid repetition.

Tested-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
---
 hw/ppc/spapr.c              | 11 +----------
 hw/ppc/spapr_numa.c         | 14 ++++++++++++++
 include/hw/ppc/spapr_numa.h |  1 +
 3 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
index 2d60c6f594..bc3fa276ff 100644
--- a/hw/ppc/spapr.c
+++ b/hw/ppc/spapr.c
@@ -2765,16 +2765,7 @@ static void spapr_machine_init(MachineState *machine)
 
     }
 
-    /*
-     * NVLink2-connected GPU RAM needs to be placed on a separate NUMA node.
-     * We assign a new numa ID per GPU in spapr_pci_collect_nvgpu() which is
-     * called from vPHB reset handler so we initialize the counter here.
-     * If no NUMA is configured from the QEMU side, we start from 1 as GPU RAM
-     * must be equally distant from any other node.
-     * The final value of spapr->gpu_numa_id is going to be written to
-     * max-associativity-domains in spapr_build_fdt().
-     */
-    spapr->gpu_numa_id = MAX(1, machine->numa_state->num_nodes);
+    spapr->gpu_numa_id = spapr_numa_initial_nvgpu_numa_id(machine);
 
     /* Init numa_assoc_array */
     spapr_numa_associativity_init(spapr, machine);
diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
index 261810525b..a757dd88b8 100644
--- a/hw/ppc/spapr_numa.c
+++ b/hw/ppc/spapr_numa.c
@@ -46,6 +46,20 @@ static bool spapr_numa_is_symmetrical(MachineState *ms)
     return true;
 }
 
+/*
+ * NVLink2-connected GPU RAM needs to be placed on a separate NUMA node.
+ * We assign a new numa ID per GPU in spapr_pci_collect_nvgpu() which is
+ * called from vPHB reset handler so we initialize the counter here.
+ * If no NUMA is configured from the QEMU side, we start from 1 as GPU RAM
+ * must be equally distant from any other node.
+ * The final value of spapr->gpu_numa_id is going to be written to
+ * max-associativity-domains in spapr_build_fdt().
+ */
+unsigned int spapr_numa_initial_nvgpu_numa_id(MachineState *machine)
+{
+    return MAX(1, machine->numa_state->num_nodes);
+}
+
 /*
  * This function will translate the user distances into
  * what the kernel understand as possible values: 10
diff --git a/include/hw/ppc/spapr_numa.h b/include/hw/ppc/spapr_numa.h
index b3fd950634..6f9f02d3de 100644
--- a/include/hw/ppc/spapr_numa.h
+++ b/include/hw/ppc/spapr_numa.h
@@ -31,5 +31,6 @@ int spapr_numa_fixup_cpu_dt(SpaprMachineState *spapr, void *fdt,
                             int offset, PowerPCCPU *cpu);
 int spapr_numa_write_assoc_lookup_arrays(SpaprMachineState *spapr, void *fdt,
                                          int offset);
+unsigned int spapr_numa_initial_nvgpu_numa_id(MachineState *machine);
 
 #endif /* HW_SPAPR_NUMA_H */
-- 
2.26.2

Re: [PATCH v2 2/3] spapr_numa.c: create spapr_numa_initial_nvgpu_numa_id() helper

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 3466 bytes --]

On Thu, Jan 28, 2021 at 02:42:12PM -0300, Daniel Henrique Barboza wrote:
> We'll need to check the initial value given to spapr->gpu_numa_id when
> building the rtas DT, so put it in a helper for easier access and to
> avoid repetition.
> 
> Tested-by: Cédric Le Goater <clg@kaod.org>
> Reviewed-by: Greg Kurz <groug@kaod.org>
> Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

Applied to ppc-for-6.0, thanks.

> ---
>  hw/ppc/spapr.c              | 11 +----------
>  hw/ppc/spapr_numa.c         | 14 ++++++++++++++
>  include/hw/ppc/spapr_numa.h |  1 +
>  3 files changed, 16 insertions(+), 10 deletions(-)
> 
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index 2d60c6f594..bc3fa276ff 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -2765,16 +2765,7 @@ static void spapr_machine_init(MachineState *machine)
>  
>      }
>  
> -    /*
> -     * NVLink2-connected GPU RAM needs to be placed on a separate NUMA node.
> -     * We assign a new numa ID per GPU in spapr_pci_collect_nvgpu() which is
> -     * called from vPHB reset handler so we initialize the counter here.
> -     * If no NUMA is configured from the QEMU side, we start from 1 as GPU RAM
> -     * must be equally distant from any other node.
> -     * The final value of spapr->gpu_numa_id is going to be written to
> -     * max-associativity-domains in spapr_build_fdt().
> -     */
> -    spapr->gpu_numa_id = MAX(1, machine->numa_state->num_nodes);
> +    spapr->gpu_numa_id = spapr_numa_initial_nvgpu_numa_id(machine);
>  
>      /* Init numa_assoc_array */
>      spapr_numa_associativity_init(spapr, machine);
> diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
> index 261810525b..a757dd88b8 100644
> --- a/hw/ppc/spapr_numa.c
> +++ b/hw/ppc/spapr_numa.c
> @@ -46,6 +46,20 @@ static bool spapr_numa_is_symmetrical(MachineState *ms)
>      return true;
>  }
>  
> +/*
> + * NVLink2-connected GPU RAM needs to be placed on a separate NUMA node.
> + * We assign a new numa ID per GPU in spapr_pci_collect_nvgpu() which is
> + * called from vPHB reset handler so we initialize the counter here.
> + * If no NUMA is configured from the QEMU side, we start from 1 as GPU RAM
> + * must be equally distant from any other node.
> + * The final value of spapr->gpu_numa_id is going to be written to
> + * max-associativity-domains in spapr_build_fdt().
> + */
> +unsigned int spapr_numa_initial_nvgpu_numa_id(MachineState *machine)
> +{
> +    return MAX(1, machine->numa_state->num_nodes);
> +}
> +
>  /*
>   * This function will translate the user distances into
>   * what the kernel understand as possible values: 10
> diff --git a/include/hw/ppc/spapr_numa.h b/include/hw/ppc/spapr_numa.h
> index b3fd950634..6f9f02d3de 100644
> --- a/include/hw/ppc/spapr_numa.h
> +++ b/include/hw/ppc/spapr_numa.h
> @@ -31,5 +31,6 @@ int spapr_numa_fixup_cpu_dt(SpaprMachineState *spapr, void *fdt,
>                              int offset, PowerPCCPU *cpu);
>  int spapr_numa_write_assoc_lookup_arrays(SpaprMachineState *spapr, void *fdt,
>                                           int offset);
> +unsigned int spapr_numa_initial_nvgpu_numa_id(MachineState *machine);
>  
>  #endif /* HW_SPAPR_NUMA_H */

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[PATCH v2 3/3] spapr_numa.c: fix ibm, max-associativity-domains calculation

[Daniel Henrique Barboza]

The current logic for calculating 'maxdomain' making it a sum of
numa_state->num_nodes with spapr->gpu_numa_id. spapr->gpu_numa_id is
used as a index to determine the next available NUMA id that a
given NVGPU can use.

The problem is that the initial value of gpu_numa_id, for any topology
that has more than one NUMA node, is equal to numa_state->num_nodes.
This means that our maxdomain will always be, at least, twice the
amount of existing NUMA nodes. This means that a guest with 4 NUMA
nodes will end up with the following max-associativity-domains:

rtas/ibm,max-associativity-domains
                 00000004 00000008 00000008 00000008 00000008

This overtuning of maxdomains doesn't go unnoticed in the guest, being
detected in SLUB during boot:

 dmesg | grep SLUB
[    0.000000] SLUB: HWalign=128, Order=0-3, MinObjects=0, CPUs=4, Nodes=8

SLUB is detecting 8 total nodes, with 4 nodes being online.

This patch fixes ibm,max-associativity-domains by considering the amount
of NVGPUs NUMA nodes presented in the guest, instead of just
spapr->gpu_numa_id.

Reported-by: Cédric Le Goater <clg@kaod.org>
Tested-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
---
 hw/ppc/spapr_numa.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
index a757dd88b8..779f18b994 100644
--- a/hw/ppc/spapr_numa.c
+++ b/hw/ppc/spapr_numa.c
@@ -311,6 +311,8 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
 {
     MachineState *ms = MACHINE(spapr);
     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
+    uint32_t number_nvgpus_nodes = spapr->gpu_numa_id -
+                                   spapr_numa_initial_nvgpu_numa_id(ms);
     uint32_t refpoints[] = {
         cpu_to_be32(0x4),
         cpu_to_be32(0x3),
@@ -318,7 +320,7 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
         cpu_to_be32(0x1),
     };
     uint32_t nr_refpoints = ARRAY_SIZE(refpoints);
-    uint32_t maxdomain = ms->numa_state->num_nodes + spapr->gpu_numa_id;
+    uint32_t maxdomain = ms->numa_state->num_nodes + number_nvgpus_nodes;
     uint32_t maxdomains[] = {
         cpu_to_be32(4),
         cpu_to_be32(maxdomain),
-- 
2.26.2

Re: [PATCH v2 3/3] spapr_numa.c: fix ibm,max-associativity-domains calculation

[Greg Kurz]

On Thu, 28 Jan 2021 14:42:13 -0300
Daniel Henrique Barboza <danielhb413@gmail.com> wrote:

> The current logic for calculating 'maxdomain' making it a sum of
> numa_state->num_nodes with spapr->gpu_numa_id. spapr->gpu_numa_id is
> used as a index to determine the next available NUMA id that a
> given NVGPU can use.
> 
> The problem is that the initial value of gpu_numa_id, for any topology
> that has more than one NUMA node, is equal to numa_state->num_nodes.
> This means that our maxdomain will always be, at least, twice the
> amount of existing NUMA nodes. This means that a guest with 4 NUMA
> nodes will end up with the following max-associativity-domains:
> 
> rtas/ibm,max-associativity-domains
>                  00000004 00000008 00000008 00000008 00000008
> 
> This overtuning of maxdomains doesn't go unnoticed in the guest, being
> detected in SLUB during boot:
> 
>  dmesg | grep SLUB
> [    0.000000] SLUB: HWalign=128, Order=0-3, MinObjects=0, CPUs=4, Nodes=8
> 
> SLUB is detecting 8 total nodes, with 4 nodes being online.
> 
> This patch fixes ibm,max-associativity-domains by considering the amount
> of NVGPUs NUMA nodes presented in the guest, instead of just
> spapr->gpu_numa_id.
> 
> Reported-by: Cédric Le Goater <clg@kaod.org>
> Tested-by: Cédric Le Goater <clg@kaod.org>
> Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
> ---

Reviewed-by: Greg Kurz <groug@kaod.org>

>  hw/ppc/spapr_numa.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
> index a757dd88b8..779f18b994 100644
> --- a/hw/ppc/spapr_numa.c
> +++ b/hw/ppc/spapr_numa.c
> @@ -311,6 +311,8 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
>  {
>      MachineState *ms = MACHINE(spapr);
>      SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
> +    uint32_t number_nvgpus_nodes = spapr->gpu_numa_id -
> +                                   spapr_numa_initial_nvgpu_numa_id(ms);
>      uint32_t refpoints[] = {
>          cpu_to_be32(0x4),
>          cpu_to_be32(0x3),
> @@ -318,7 +320,7 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
>          cpu_to_be32(0x1),
>      };
>      uint32_t nr_refpoints = ARRAY_SIZE(refpoints);
> -    uint32_t maxdomain = ms->numa_state->num_nodes + spapr->gpu_numa_id;
> +    uint32_t maxdomain = ms->numa_state->num_nodes + number_nvgpus_nodes;
>      uint32_t maxdomains[] = {
>          cpu_to_be32(4),
>          cpu_to_be32(maxdomain),

Re: [PATCH v2 3/3] spapr_numa.c: fix ibm,max-associativity-domains calculation

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 2743 bytes --]


On Thu, Jan 28, 2021 at 02:42:13PM -0300, Daniel Henrique Barboza wrote:
> The current logic for calculating 'maxdomain' making it a sum of
> numa_state->num_nodes with spapr->gpu_numa_id. spapr->gpu_numa_id is
> used as a index to determine the next available NUMA id that a
> given NVGPU can use.
> 
> The problem is that the initial value of gpu_numa_id, for any topology
> that has more than one NUMA node, is equal to numa_state->num_nodes.
> This means that our maxdomain will always be, at least, twice the
> amount of existing NUMA nodes. This means that a guest with 4 NUMA
> nodes will end up with the following max-associativity-domains:
> 
> rtas/ibm,max-associativity-domains
>                  00000004 00000008 00000008 00000008 00000008
> 
> This overtuning of maxdomains doesn't go unnoticed in the guest, being
> detected in SLUB during boot:
> 
>  dmesg | grep SLUB
> [    0.000000] SLUB: HWalign=128, Order=0-3, MinObjects=0, CPUs=4, Nodes=8
> 
> SLUB is detecting 8 total nodes, with 4 nodes being online.
> 
> This patch fixes ibm,max-associativity-domains by considering the amount
> of NVGPUs NUMA nodes presented in the guest, instead of just
> spapr->gpu_numa_id.
> 
> Reported-by: Cédric Le Goater <clg@kaod.org>
> Tested-by: Cédric Le Goater <clg@kaod.org>
> Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

Applied, thanks.

> ---
>  hw/ppc/spapr_numa.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
> index a757dd88b8..779f18b994 100644
> --- a/hw/ppc/spapr_numa.c
> +++ b/hw/ppc/spapr_numa.c
> @@ -311,6 +311,8 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
>  {
>      MachineState *ms = MACHINE(spapr);
>      SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
> +    uint32_t number_nvgpus_nodes = spapr->gpu_numa_id -
> +                                   spapr_numa_initial_nvgpu_numa_id(ms);
>      uint32_t refpoints[] = {
>          cpu_to_be32(0x4),
>          cpu_to_be32(0x3),
> @@ -318,7 +320,7 @@ void spapr_numa_write_rtas_dt(SpaprMachineState *spapr, void *fdt, int rtas)
>          cpu_to_be32(0x1),
>      };
>      uint32_t nr_refpoints = ARRAY_SIZE(refpoints);
> -    uint32_t maxdomain = ms->numa_state->num_nodes + spapr->gpu_numa_id;
> +    uint32_t maxdomain = ms->numa_state->num_nodes + number_nvgpus_nodes;
>      uint32_t maxdomains[] = {
>          cpu_to_be32(4),
>          cpu_to_be32(maxdomain),

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]