From: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
To: linux-mm@kvack.org, akpm@linux-foundation.org
Cc: Wei Xu <weixugc@google.com>, Huang Ying <ying.huang@intel.com>,
Yang Shi <shy828301@gmail.com>,
Davidlohr Bueso <dave@stgolabs.net>,
Tim C Chen <tim.c.chen@intel.com>,
Michal Hocko <mhocko@kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Hesham Almatary <hesham.almatary@huawei.com>,
Dave Hansen <dave.hansen@intel.com>,
Jonathan Cameron <Jonathan.Cameron@huawei.com>,
Alistair Popple <apopple@nvidia.com>,
Dan Williams <dan.j.williams@intel.com>,
Johannes Weiner <hannes@cmpxchg.org>,
jvgediya.oss@gmail.com, Bharata B Rao <bharata@amd.com>,
"Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Subject: [PATCH mm-unstable] mm/demotion: Assign correct memory type for multiple dax devices with the same node affinity
Date: Fri, 26 Aug 2022 15:32:24 +0530 [thread overview]
Message-ID: <20220826100224.542312-1-aneesh.kumar@linux.ibm.com> (raw)
With multiple dax devices having the same node affinity, the kernel wrongly assigned
default_dram memory type to some devices after the memory hotplug operation. Fix this by
not clearing node_memory_types on the dax device remove.
The current kernel cleared node_memory_type on successful removal of a dax device.
But then we can have multiple dax devices with the same node affinity. Clearing the
node_memory_type results in assigning other dax devices to the default dram type when
we bring them online.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
mm/memory-tiers.c | 37 +++++++++++++++++++++++++++++--------
1 file changed, 29 insertions(+), 8 deletions(-)
diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
index ba844fe9cc8c..c4bd6d052a33 100644
--- a/mm/memory-tiers.c
+++ b/mm/memory-tiers.c
@@ -27,9 +27,14 @@ struct demotion_nodes {
nodemask_t preferred;
};
+struct node_memory_type_map {
+ struct memory_dev_type *memtype;
+ int map_count;
+};
+
static DEFINE_MUTEX(memory_tier_lock);
static LIST_HEAD(memory_tiers);
-static struct memory_dev_type *node_memory_types[MAX_NUMNODES];
+static struct node_memory_type_map node_memory_types[MAX_NUMNODES];
static struct memory_dev_type *default_dram_type;
#ifdef CONFIG_MIGRATION
static int top_tier_adistance;
@@ -386,9 +391,19 @@ static inline void establish_demotion_targets(void) {}
static inline void __init_node_memory_type(int node, struct memory_dev_type *memtype)
{
- if (!node_memory_types[node]) {
- node_memory_types[node] = memtype;
- kref_get(&memtype->kref);
+ if (!node_memory_types[node].memtype)
+ node_memory_types[node].memtype = memtype;
+ /*
+ * for each device getting added in the same NUMA node
+ * with this specific memtype, bump the map count. We
+ * Only take memtype device reference once, so that
+ * changing a node memtype can be done by droping the
+ * only reference count taken here.
+ */
+
+ if (node_memory_types[node].memtype == memtype) {
+ if (!node_memory_types[node].map_count++)
+ kref_get(&memtype->kref);
}
}
@@ -406,7 +421,7 @@ static struct memory_tier *set_node_memory_tier(int node)
__init_node_memory_type(node, default_dram_type);
- memtype = node_memory_types[node];
+ memtype = node_memory_types[node].memtype;
node_set(node, memtype->nodes);
memtier = find_create_memory_tier(memtype);
if (!IS_ERR(memtier))
@@ -448,7 +463,7 @@ static bool clear_node_memory_tier(int node)
rcu_assign_pointer(pgdat->memtier, NULL);
synchronize_rcu();
- memtype = node_memory_types[node];
+ memtype = node_memory_types[node].memtype;
node_clear(node, memtype->nodes);
if (nodes_empty(memtype->nodes)) {
list_del_init(&memtype->tier_sibiling);
@@ -502,8 +517,14 @@ EXPORT_SYMBOL_GPL(init_node_memory_type);
void clear_node_memory_type(int node, struct memory_dev_type *memtype)
{
mutex_lock(&memory_tier_lock);
- if (node_memory_types[node] == memtype) {
- node_memory_types[node] = NULL;
+ if (node_memory_types[node].memtype == memtype)
+ node_memory_types[node].map_count--;
+ /*
+ * If we umapped all the attached devices to this node,
+ * clear the node memory type.
+ */
+ if (!node_memory_types[node].map_count) {
+ node_memory_types[node].memtype = NULL;
kref_put(&memtype->kref, release_memtype);
}
mutex_unlock(&memory_tier_lock);
--
2.37.2
next reply other threads:[~2022-08-26 10:03 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-26 10:02 Aneesh Kumar K.V [this message]
2022-08-27 3:00 ` [PATCH mm-unstable] mm/demotion: Assign correct memory type for multiple dax devices with the same node affinity Andrew Morton
2022-09-01 6:15 ` Huang, Ying
2022-09-01 6:24 ` Aneesh Kumar K V
2022-09-01 6:45 ` Huang, Ying
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20220826100224.542312-1-aneesh.kumar@linux.ibm.com \
--to=aneesh.kumar@linux.ibm.com \
--cc=Jonathan.Cameron@huawei.com \
--cc=akpm@linux-foundation.org \
--cc=apopple@nvidia.com \
--cc=bharata@amd.com \
--cc=dan.j.williams@intel.com \
--cc=dave.hansen@intel.com \
--cc=dave@stgolabs.net \
--cc=hannes@cmpxchg.org \
--cc=hesham.almatary@huawei.com \
--cc=jvgediya.oss@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@kernel.org \
--cc=shy828301@gmail.com \
--cc=tim.c.chen@intel.com \
--cc=weixugc@google.com \
--cc=ying.huang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).