[PATCH 0/2] null_blk: fix wrong use of nr_online_nodes

[PATCH 0/2] null_blk: fix wrong use of nr_online_nodes

[Peng Liu]

Helper node_available is introduced to judge whether a node can be
used, and it is used to fix the wrong use of nr_online_nodes when
numa node is sparse.

Peng Liu (2):
  include/linux/nodemask.h: create node_available() helper
  null_blk: fix wrong use of nr_online_nodes

 arch/ia64/hp/common/sba_iommu.c |  2 +-
 arch/x86/pci/acpi.c             |  2 +-
 drivers/acpi/arm64/iort.c       |  2 +-
 drivers/block/null_blk/main.c   | 45 ++++++++++++++++++++-------------
 drivers/pci/pci-sysfs.c         |  2 +-
 include/linux/nodemask.h        |  3 +++
 mm/mempolicy.c                  |  2 +-
 7 files changed, 36 insertions(+), 22 deletions(-)

-- 
2.25.1

[PATCH 1/2] include/linux/nodemask.h: create node_available() helper

[Peng Liu]

Lots of code dose
	node != NUMA_NO_NODE && !node_online(node)
or
	node == NUMA_NO_NODE || node_online(node)
so create node_available to do this to simplify code.

Signed-off-by: Peng Liu <liupeng256@huawei.com>
---
 arch/ia64/hp/common/sba_iommu.c | 2 +-
 arch/x86/pci/acpi.c             | 2 +-
 drivers/acpi/arm64/iort.c       | 2 +-
 drivers/pci/pci-sysfs.c         | 2 +-
 include/linux/nodemask.h        | 3 +++
 mm/mempolicy.c                  | 2 +-
 6 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/arch/ia64/hp/common/sba_iommu.c b/arch/ia64/hp/common/sba_iommu.c
index 8ad6946521d8..da3a010bb5fb 100644
--- a/arch/ia64/hp/common/sba_iommu.c
+++ b/arch/ia64/hp/common/sba_iommu.c
@@ -1969,7 +1969,7 @@ sba_map_ioc_to_node(struct ioc *ioc, acpi_handle handle)
 	unsigned int node;
 
 	node = acpi_get_node(handle);
-	if (node != NUMA_NO_NODE && !node_online(node))
+	if (!node_available(node))
 		node = NUMA_NO_NODE;
 
 	ioc->node = node;
diff --git a/arch/x86/pci/acpi.c b/arch/x86/pci/acpi.c
index 052f1d78a562..d4909daa44d9 100644
--- a/arch/x86/pci/acpi.c
+++ b/arch/x86/pci/acpi.c
@@ -254,7 +254,7 @@ static int pci_acpi_root_get_node(struct acpi_pci_root *root)
 			dev_info(&device->dev, FW_BUG "no _PXM; falling back to node %d from hardware (may be inconsistent with ACPI node numbers)\n",
 				node);
 	}
-	if (node != NUMA_NO_NODE && !node_online(node))
+	if (!node_available(node))
 		node = NUMA_NO_NODE;
 
 	return node;
diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
index f2f8f05662de..bdf690010b97 100644
--- a/drivers/acpi/arm64/iort.c
+++ b/drivers/acpi/arm64/iort.c
@@ -1251,7 +1251,7 @@ static int  __init arm_smmu_v3_set_proximity(struct device *dev,
 	if (smmu->flags & ACPI_IORT_SMMU_V3_PXM_VALID) {
 		int dev_node = pxm_to_node(smmu->pxm);
 
-		if (dev_node != NUMA_NO_NODE && !node_online(dev_node))
+		if (!node_available(dev_node))
 			return -EINVAL;
 
 		set_dev_node(dev, dev_node);
diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
index c263ffc5884a..502490d26c1d 100644
--- a/drivers/pci/pci-sysfs.c
+++ b/drivers/pci/pci-sysfs.c
@@ -344,7 +344,7 @@ static ssize_t numa_node_store(struct device *dev,
 	if ((node < 0 && node != NUMA_NO_NODE) || node >= MAX_NUMNODES)
 		return -EINVAL;
 
-	if (node != NUMA_NO_NODE && !node_online(node))
+	if (!node_available(node))
 		return -EINVAL;
 
 	add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);
diff --git a/include/linux/nodemask.h b/include/linux/nodemask.h
index 567c3ddba2c4..591201a1e646 100644
--- a/include/linux/nodemask.h
+++ b/include/linux/nodemask.h
@@ -70,6 +70,7 @@
  *
  * int node_online(node)		Is some node online?
  * int node_possible(node)		Is some node possible?
+ * int node_available(node)		Is some node available(online or NUMA_NO_NODE)?
  *
  * node_set_online(node)		set bit 'node' in node_online_map
  * node_set_offline(node)		clear bit 'node' in node_online_map
@@ -510,6 +511,8 @@ static inline int node_random(const nodemask_t *mask)
 #define num_possible_nodes()	num_node_state(N_POSSIBLE)
 #define node_online(node)	node_state((node), N_ONLINE)
 #define node_possible(node)	node_state((node), N_POSSIBLE)
+#define node_available(node) \
+		((node) == NUMA_NO_NODE || node_state((node), N_ONLINE))
 
 #define for_each_node(node)	   for_each_node_state(node, N_POSSIBLE)
 #define for_each_online_node(node) for_each_node_state(node, N_ONLINE)
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 8c74107a2b15..7a31b006c5e8 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -141,7 +141,7 @@ int numa_map_to_online_node(int node)
 {
 	int min_dist = INT_MAX, dist, n, min_node;
 
-	if (node == NUMA_NO_NODE || node_online(node))
+	if (node_available(node))
 		return node;
 
 	min_node = node;
-- 
2.25.1

Re: [PATCH 1/2] include/linux/nodemask.h: create node_available() helper

[Bjorn Helgaas]

Subject line convention looks like "numa: ..."

On Fri, May 06, 2022 at 01:58:00AM +0000, Peng Liu wrote:
> Lots of code dose
               does

> 	node != NUMA_NO_NODE && !node_online(node)
> or
> 	node == NUMA_NO_NODE || node_online(node)
> so create node_available to do this to simplify code.
            node_available()

I'm not really sure what meaning "node_available" conveys, though.
Probably just because I don't understand NUMA.

Should the test for NUMA_NO_NODE be folded into node_state() or
node_online() directly instead of adding a new node_available()
interface?

NUMA_NO_NODE is -1.  It's not clear to me that node_state()/
node_isset()/test_bit() would do the right thing given -1.  I doubt
all node_online() callers ensure they don't pass NUMA_NO_NODE.

> --- a/include/linux/nodemask.h
> +++ b/include/linux/nodemask.h
> @@ -70,6 +70,7 @@
>   *
>   * int node_online(node)		Is some node online?
>   * int node_possible(node)		Is some node possible?
> + * int node_available(node)		Is some node available(online or NUMA_NO_NODE)?

Existing file generally fits in 80 columns; follow that lead unless
you have a really good reason.  E.g., maybe this?

  + * int node_available(node)		Node online or NUMA_NO_NODE

[PATCH 2/2] null_blk: fix wrong use of nr_online_nodes

[Peng Liu]

Certain systems are designed to have sparse/discontiguous nodes,
a valid node may be greater than nr_online_nodes. So, the use of
"nid >= nr_online_nodes" to judge if a node is online is wrong.

Node id is a basic parameter of the system, a user-configured node
must be checked as early as possible. Otherwise, it may cause panic
when calling some vulnerable functions such as node_online which
will cause panic if a very big node is received.

Check g_home_node once users config it, and use node_available to
make node-checking compatible with sparse/discontiguous nodes.

Fixes: 7ff684a683d7 ("null_blk: prevent crash from bad home_node value")
Signed-off-by: Peng Liu <liupeng256@huawei.com>
Suggested-by: Davidlohr Bueso <dave@stgolabs.net>
---
 drivers/block/null_blk/main.c | 45 ++++++++++++++++++++++-------------
 1 file changed, 28 insertions(+), 17 deletions(-)

diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c
index 05b1120e6623..995348d6e7e7 100644
--- a/drivers/block/null_blk/main.c
+++ b/drivers/block/null_blk/main.c
@@ -97,7 +97,33 @@ module_param_named(poll_queues, g_poll_queues, int, 0444);
 MODULE_PARM_DESC(poll_queues, "Number of IOPOLL submission queues");
 
 static int g_home_node = NUMA_NO_NODE;
-module_param_named(home_node, g_home_node, int, 0444);
+
+static int null_param_store_val(const char *str, int *val, int min, int max)
+{
+	int ret, new_val;
+
+	ret = kstrtoint(str, 10, &new_val);
+	if (ret)
+		return -EINVAL;
+
+	if (new_val < min || new_val > max)
+		return -EINVAL;
+
+	*val = new_val;
+	return 0;
+}
+
+static int null_set_home_node(const char *str, const struct kernel_param *kp)
+{
+	return null_param_store_val(str, &g_home_node, 0, MAX_NUMNODES - 1);
+}
+
+static const struct kernel_param_ops null_home_node_param_ops = {
+	.set	= null_set_home_node,
+	.get	= param_get_int,
+};
+
+device_param_cb(home_node, &null_home_node_param_ops, &g_home_node, 0444);
 MODULE_PARM_DESC(home_node, "Home node for the device");
 
 #ifdef CONFIG_BLK_DEV_NULL_BLK_FAULT_INJECTION
@@ -120,21 +146,6 @@ MODULE_PARM_DESC(init_hctx, "Fault injection to fail hctx init. init_hctx=<inter
 
 static int g_queue_mode = NULL_Q_MQ;
 
-static int null_param_store_val(const char *str, int *val, int min, int max)
-{
-	int ret, new_val;
-
-	ret = kstrtoint(str, 10, &new_val);
-	if (ret)
-		return -EINVAL;
-
-	if (new_val < min || new_val > max)
-		return -EINVAL;
-
-	*val = new_val;
-	return 0;
-}
-
 static int null_set_queue_mode(const char *str, const struct kernel_param *kp)
 {
 	return null_param_store_val(str, &g_queue_mode, NULL_Q_BIO, NULL_Q_MQ);
@@ -2107,7 +2118,7 @@ static int __init null_init(void)
 		g_max_sectors = BLK_DEF_MAX_SECTORS;
 	}
 
-	if (g_home_node != NUMA_NO_NODE && g_home_node >= nr_online_nodes) {
+	if (!node_available(g_home_node)) {
 		pr_err("invalid home_node value\n");
 		g_home_node = NUMA_NO_NODE;
 	}
-- 
2.25.1