After a LPM, the device tree node ibm,dynamic-reconfiguration-memory may be
updated by the hypervisor in the case the NUMA topology of the LPAR's
memory is updated.
This is caught by the kernel, but the memory's node is updated because
there is no way to move a memory block between nodes.
If later a memory block is added or removed, drmem_update_dt() is called
and it is overwriting the DT node to match the added or removed LMB. But
the LMB's associativity node has not been updated after the DT node update
and thus the node is overwritten by the Linux's topology instead of the
hypervisor one.
Introduce a hook called when the ibm,dynamic-reconfiguration-memory node is
updated to force an update of the LMB's associativity.
Signed-off-by: Laurent Dufour <[email protected]>
---
arch/powerpc/include/asm/drmem.h | 1 +
arch/powerpc/mm/drmem.c | 48 +++++++++++++++++++++++
arch/powerpc/platforms/pseries/mobility.c | 9 +++++
3 files changed, 58 insertions(+)
diff --git a/arch/powerpc/include/asm/drmem.h b/arch/powerpc/include/asm/drmem.h
index bf2402fed3e0..55c2c25085b0 100644
--- a/arch/powerpc/include/asm/drmem.h
+++ b/arch/powerpc/include/asm/drmem.h
@@ -111,6 +111,7 @@ int drmem_update_dt(void);
int __init
walk_drmem_lmbs_early(unsigned long node, void *data,
int (*func)(struct drmem_lmb *, const __be32 **, void *));
+void drmem_update_lmbs(void);
#endif
static inline void invalidate_lmb_associativity_index(struct drmem_lmb *lmb)
diff --git a/arch/powerpc/mm/drmem.c b/arch/powerpc/mm/drmem.c
index 9af3832c9d8d..46074bdfdb3c 100644
--- a/arch/powerpc/mm/drmem.c
+++ b/arch/powerpc/mm/drmem.c
@@ -307,6 +307,54 @@ int __init walk_drmem_lmbs_early(unsigned long node, void *data,
return ret;
}
+/*
+ * Update the LMB associativity index.
+ */
+static int update_lmb(struct drmem_lmb *updated_lmb,
+ __maybe_unused const __be32 **usm,
+ __maybe_unused void *data)
+{
+ struct drmem_lmb *lmb;
+
+ /*
+ * Brut force there may be better way to fetch the LMB
+ */
+ for_each_drmem_lmb(lmb) {
+ if (lmb->drc_index != updated_lmb->drc_index)
+ continue;
+
+ lmb->aa_index = updated_lmb->aa_index;
+ break;
+ }
+ return 0;
+}
+
+/*
+ * Update the LMB associativity index.
+ *
+ * This needs to be called when the hypervisor is updating the
+ * dynamic-reconfiguration-memory node property.
+ */
+void drmem_update_lmbs(void)
+{
+ struct device_node *node;
+ const __be32 *prop;
+
+ node = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
+ if (!node)
+ return;
+
+ prop = of_get_property(node, "ibm,dynamic-memory", NULL);
+ if (prop) {
+ __walk_drmem_v1_lmbs(prop, NULL, NULL, update_lmb);
+ } else {
+ prop = of_get_property(node, "ibm,dynamic-memory-v2", NULL);
+ if (prop)
+ __walk_drmem_v2_lmbs(prop, NULL, NULL, update_lmb);
+ }
+
+ of_node_put(node);
+}
#endif
static int init_drmem_lmb_size(struct device_node *dn)
diff --git a/arch/powerpc/platforms/pseries/mobility.c b/arch/powerpc/platforms/pseries/mobility.c
index ea4d6a660e0d..c68eccc6e8df 100644
--- a/arch/powerpc/platforms/pseries/mobility.c
+++ b/arch/powerpc/platforms/pseries/mobility.c
@@ -25,6 +25,7 @@
#include <asm/machdep.h>
#include <asm/rtas.h>
+#include <asm/drmem.h>
#include "pseries.h"
#include "../../kernel/cacheinfo.h"
@@ -237,6 +238,7 @@ int pseries_devicetree_update(s32 scope)
__be32 *data;
int update_nodes_token;
int rc;
+ bool drmem_updated = false;
update_nodes_token = rtas_token("ibm,update-nodes");
if (update_nodes_token == RTAS_UNKNOWN_SERVICE)
@@ -271,6 +273,10 @@ int pseries_devicetree_update(s32 scope)
continue;
}
+ if (!strcmp(np->full_name,
+ "ibm,dynamic-reconfiguration-memory"))
+ drmem_updated = true;
+
switch (action) {
case DELETE_DT_NODE:
delete_dt_node(np);
@@ -293,6 +299,9 @@ int pseries_devicetree_update(s32 scope)
} while (rc == 1);
kfree(rtas_buf);
+
+ if (drmem_updated)
+ drmem_update_lmbs();
return rc;
}
--
2.31.1
On 4/27/21 8:01 AM, Laurent Dufour wrote:
> After a LPM, the device tree node ibm,dynamic-reconfiguration-memory may be
> updated by the hypervisor in the case the NUMA topology of the LPAR's
> memory is updated.
>
> This is caught by the kernel, but the memory's node is updated because
> there is no way to move a memory block between nodes.
>
> If later a memory block is added or removed, drmem_update_dt() is called
> and it is overwriting the DT node to match the added or removed LMB. But
> the LMB's associativity node has not been updated after the DT node update
> and thus the node is overwritten by the Linux's topology instead of the
> hypervisor one.
>
> Introduce a hook called when the ibm,dynamic-reconfiguration-memory node is
> updated to force an update of the LMB's associativity.
>
> Signed-off-by: Laurent Dufour <[email protected]>
> ---
> arch/powerpc/include/asm/drmem.h | 1 +
> arch/powerpc/mm/drmem.c | 48 +++++++++++++++++++++++
> arch/powerpc/platforms/pseries/mobility.c | 9 +++++
> 3 files changed, 58 insertions(+)
>
> diff --git a/arch/powerpc/include/asm/drmem.h b/arch/powerpc/include/asm/drmem.h
> index bf2402fed3e0..55c2c25085b0 100644
> --- a/arch/powerpc/include/asm/drmem.h
> +++ b/arch/powerpc/include/asm/drmem.h
> @@ -111,6 +111,7 @@ int drmem_update_dt(void);
> int __init
> walk_drmem_lmbs_early(unsigned long node, void *data,
> int (*func)(struct drmem_lmb *, const __be32 **, void *));
> +void drmem_update_lmbs(void);
> #endif
>
> static inline void invalidate_lmb_associativity_index(struct drmem_lmb *lmb)
> diff --git a/arch/powerpc/mm/drmem.c b/arch/powerpc/mm/drmem.c
> index 9af3832c9d8d..46074bdfdb3c 100644
> --- a/arch/powerpc/mm/drmem.c
> +++ b/arch/powerpc/mm/drmem.c
> @@ -307,6 +307,54 @@ int __init walk_drmem_lmbs_early(unsigned long node, void *data,
> return ret;
> }
>
> +/*
> + * Update the LMB associativity index.
> + */
> +static int update_lmb(struct drmem_lmb *updated_lmb,
> + __maybe_unused const __be32 **usm,
> + __maybe_unused void *data)
> +{
> + struct drmem_lmb *lmb;
> +
> + /*
> + * Brut force there may be better way to fetch the LMB
> + */
> + for_each_drmem_lmb(lmb) {
> + if (lmb->drc_index != updated_lmb->drc_index)
> + continue;
> +
> + lmb->aa_index = updated_lmb->aa_index;
> + break;
> + }
> + return 0;
> +}
> +
> +/*
> + * Update the LMB associativity index.
> + *
> + * This needs to be called when the hypervisor is updating the
> + * dynamic-reconfiguration-memory node property.
> + */
> +void drmem_update_lmbs(void)
> +{
> + struct device_node *node;
> + const __be32 *prop;
> +
> + node = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
> + if (!node)
> + return;
> +
> + prop = of_get_property(node, "ibm,dynamic-memory", NULL);
> + if (prop) {
> + __walk_drmem_v1_lmbs(prop, NULL, NULL, update_lmb);
> + } else {
> + prop = of_get_property(node, "ibm,dynamic-memory-v2", NULL);
> + if (prop)
> + __walk_drmem_v2_lmbs(prop, NULL, NULL, update_lmb);
> + }
> +
> + of_node_put(node);
> +}
> #endif
>
> static int init_drmem_lmb_size(struct device_node *dn)
> diff --git a/arch/powerpc/platforms/pseries/mobility.c b/arch/powerpc/platforms/pseries/mobility.c
> index ea4d6a660e0d..c68eccc6e8df 100644
> --- a/arch/powerpc/platforms/pseries/mobility.c
> +++ b/arch/powerpc/platforms/pseries/mobility.c
> @@ -25,6 +25,7 @@
>
> #include <asm/machdep.h>
> #include <asm/rtas.h>
> +#include <asm/drmem.h>
> #include "pseries.h"
> #include "../../kernel/cacheinfo.h"
>
> @@ -237,6 +238,7 @@ int pseries_devicetree_update(s32 scope)
> __be32 *data;
> int update_nodes_token;
> int rc;
> + bool drmem_updated = false;
>
> update_nodes_token = rtas_token("ibm,update-nodes");
> if (update_nodes_token == RTAS_UNKNOWN_SERVICE)
> @@ -271,6 +273,10 @@ int pseries_devicetree_update(s32 scope)
> continue;
> }
>
> + if (!strcmp(np->full_name,
> + "ibm,dynamic-reconfiguration-memory"))
> + drmem_updated = true;
Is there a reason that we can't use the existing pseries_memory_notifier()
callback in pseries/hotplug-memory.c to trigger the drmem_update_lmbs() when
either the ibm,dynamic-memory or ibm,dynamic-memory-v2 properties are updated?
Something like:
static int pseries_memory_notifier(struct notifier_block *nb,
unsigned long action, void *data)
{
struct of_reconfig_data *rd = data;
int err = 0;
switch (action) {
case OF_RECONFIG_ATTACH_NODE:
err = pseries_add_mem_node(rd->dn);
break;
case OF_RECONFIG_DETACH_NODE:
err = pseries_remove_mem_node(rd->dn);
break;
case OF_RECONFIG_UPDATE_PROPERTY:
if (!strcmp(rd->dn->full_name, "ibm,dynamic-reconfiguration-memory"));
drmem_update_lmbs(rd->prop);
break;
}
return notifier_from_errno(err);
}
Your drmem_update_lmbs() would need to be updated to take a property and to
check the property name against ibm,dyanmic-memory[-v2].
-Tyrel
> +
> switch (action) {
> case DELETE_DT_NODE:
> delete_dt_node(np);
> @@ -293,6 +299,9 @@ int pseries_devicetree_update(s32 scope)
> } while (rc == 1);
>
> kfree(rtas_buf);
> +
> + if (drmem_updated)
> + drmem_update_lmbs();
> return rc;
> }
>
Le 27/04/2021 à 19:01, Tyrel Datwyler a écrit :
> On 4/27/21 8:01 AM, Laurent Dufour wrote:
>> After a LPM, the device tree node ibm,dynamic-reconfiguration-memory may be
>> updated by the hypervisor in the case the NUMA topology of the LPAR's
>> memory is updated.
>>
>> This is caught by the kernel, but the memory's node is updated because
>> there is no way to move a memory block between nodes.
>>
>> If later a memory block is added or removed, drmem_update_dt() is called
>> and it is overwriting the DT node to match the added or removed LMB. But
>> the LMB's associativity node has not been updated after the DT node update
>> and thus the node is overwritten by the Linux's topology instead of the
>> hypervisor one.
>>
>> Introduce a hook called when the ibm,dynamic-reconfiguration-memory node is
>> updated to force an update of the LMB's associativity.
>>
>> Signed-off-by: Laurent Dufour <[email protected]>
>> ---
>> arch/powerpc/include/asm/drmem.h | 1 +
>> arch/powerpc/mm/drmem.c | 48 +++++++++++++++++++++++
>> arch/powerpc/platforms/pseries/mobility.c | 9 +++++
>> 3 files changed, 58 insertions(+)
>>
>> diff --git a/arch/powerpc/include/asm/drmem.h b/arch/powerpc/include/asm/drmem.h
>> index bf2402fed3e0..55c2c25085b0 100644
>> --- a/arch/powerpc/include/asm/drmem.h
>> +++ b/arch/powerpc/include/asm/drmem.h
>> @@ -111,6 +111,7 @@ int drmem_update_dt(void);
>> int __init
>> walk_drmem_lmbs_early(unsigned long node, void *data,
>> int (*func)(struct drmem_lmb *, const __be32 **, void *));
>> +void drmem_update_lmbs(void);
>> #endif
>>
>> static inline void invalidate_lmb_associativity_index(struct drmem_lmb *lmb)
>> diff --git a/arch/powerpc/mm/drmem.c b/arch/powerpc/mm/drmem.c
>> index 9af3832c9d8d..46074bdfdb3c 100644
>> --- a/arch/powerpc/mm/drmem.c
>> +++ b/arch/powerpc/mm/drmem.c
>> @@ -307,6 +307,54 @@ int __init walk_drmem_lmbs_early(unsigned long node, void *data,
>> return ret;
>> }
>>
>> +/*
>> + * Update the LMB associativity index.
>> + */
>> +static int update_lmb(struct drmem_lmb *updated_lmb,
>> + __maybe_unused const __be32 **usm,
>> + __maybe_unused void *data)
>> +{
>> + struct drmem_lmb *lmb;
>> +
>> + /*
>> + * Brut force there may be better way to fetch the LMB
>> + */
>> + for_each_drmem_lmb(lmb) {
>> + if (lmb->drc_index != updated_lmb->drc_index)
>> + continue;
>> +
>> + lmb->aa_index = updated_lmb->aa_index;
>> + break;
>> + }
>> + return 0;
>> +}
>> +
>> +/*
>> + * Update the LMB associativity index.
>> + *
>> + * This needs to be called when the hypervisor is updating the
>> + * dynamic-reconfiguration-memory node property.
>> + */
>> +void drmem_update_lmbs(void)
>> +{
>> + struct device_node *node;
>> + const __be32 *prop;
>> +
>> + node = of_find_node_by_path("/ibm,dynamic-reconfiguration-memory");
>> + if (!node)
>> + return;
>> +
>> + prop = of_get_property(node, "ibm,dynamic-memory", NULL);
>> + if (prop) {
>> + __walk_drmem_v1_lmbs(prop, NULL, NULL, update_lmb);
>> + } else {
>> + prop = of_get_property(node, "ibm,dynamic-memory-v2", NULL);
>> + if (prop)
>> + __walk_drmem_v2_lmbs(prop, NULL, NULL, update_lmb);
>> + }
>> +
>> + of_node_put(node);
>> +}
>> #endif
>>
>> static int init_drmem_lmb_size(struct device_node *dn)
>> diff --git a/arch/powerpc/platforms/pseries/mobility.c b/arch/powerpc/platforms/pseries/mobility.c
>> index ea4d6a660e0d..c68eccc6e8df 100644
>> --- a/arch/powerpc/platforms/pseries/mobility.c
>> +++ b/arch/powerpc/platforms/pseries/mobility.c
>> @@ -25,6 +25,7 @@
>>
>> #include <asm/machdep.h>
>> #include <asm/rtas.h>
>> +#include <asm/drmem.h>
>> #include "pseries.h"
>> #include "../../kernel/cacheinfo.h"
>>
>> @@ -237,6 +238,7 @@ int pseries_devicetree_update(s32 scope)
>> __be32 *data;
>> int update_nodes_token;
>> int rc;
>> + bool drmem_updated = false;
>>
>> update_nodes_token = rtas_token("ibm,update-nodes");
>> if (update_nodes_token == RTAS_UNKNOWN_SERVICE)
>> @@ -271,6 +273,10 @@ int pseries_devicetree_update(s32 scope)
>> continue;
>> }
>>
>> + if (!strcmp(np->full_name,
>> + "ibm,dynamic-reconfiguration-memory"))
>> + drmem_updated = true;
>
> Is there a reason that we can't use the existing pseries_memory_notifier()
> callback in pseries/hotplug-memory.c to trigger the drmem_update_lmbs() when
> either the ibm,dynamic-memory or ibm,dynamic-memory-v2 properties are updated?
Thanks a lot Tyrel!
That's far more elegant, I'll send a v2 soon.
Laurent.
> Something like:
>
> static int pseries_memory_notifier(struct notifier_block *nb,
> unsigned long action, void *data)
> {
> struct of_reconfig_data *rd = data;
> int err = 0;
>
> switch (action) {
> case OF_RECONFIG_ATTACH_NODE:
> err = pseries_add_mem_node(rd->dn);
> break;
> case OF_RECONFIG_DETACH_NODE:
> err = pseries_remove_mem_node(rd->dn);
> break;
> case OF_RECONFIG_UPDATE_PROPERTY:
> if (!strcmp(rd->dn->full_name, "ibm,dynamic-reconfiguration-memory"));
> drmem_update_lmbs(rd->prop);
> break;
> }
> return notifier_from_errno(err);
> }
>
> Your drmem_update_lmbs() would need to be updated to take a property and to
> check the property name against ibm,dyanmic-memory[-v2].
>
> -Tyrel
>
>> +
>> switch (action) {
>> case DELETE_DT_NODE:
>> delete_dt_node(np);
>> @@ -293,6 +299,9 @@ int pseries_devicetree_update(s32 scope)
>> } while (rc == 1);
>>
>> kfree(rtas_buf);
>> +
>> + if (drmem_updated)
>> + drmem_update_lmbs();
>> return rc;
>> }
>>
>