[patch net-next RFC 0/7] Add support for resource abstraction

[patch net-next RFC 0/7] Add support for resource abstraction

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Arkadi says:

Many of the ASIC's internal resources are limited and are shared between
several hardware procedures. For example, unified hash-based memory can
be used for many lookup purposes, like FDB and LPM. In many cases the user
can provide a partitioning scheme for such a resource in order to perform
fine tuning for his application. In many cases after setting the partitioning
of the resource driver reload is needed. This patchset add support for hot
reset of the driver.

The proposed interface will provide the user the ability to understand the
limitations of the hardware, and receive notification regarding its occupancy.
Furthermore, monitoring the resource occupancy can be done in real-time and
can be useful in many cases.

Arkadi Sharshevsky (7):
  devlink: Add support for resource abstraction
  devlink: Add per devlink instance lock
  devlink: Add support for reload
  mlxsw: pci: Add support for performing bus reset
  mlxsw: Register KVD resources with devlink
  mlxsw: pci: Add support for getting resource through devlink
  mlxsw: core: Add support for reload

 drivers/net/ethernet/mellanox/mlxsw/core.c     |  79 ++++-
 drivers/net/ethernet/mellanox/mlxsw/core.h     |  15 +-
 drivers/net/ethernet/mellanox/mlxsw/i2c.c      |   4 +-
 drivers/net/ethernet/mellanox/mlxsw/pci.c      | 100 ++++---
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c | 148 +++++++++-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h |  12 +
 include/net/devlink.h                          |  76 +++++
 include/uapi/linux/devlink.h                   |   9 +
 net/core/devlink.c                             | 385 ++++++++++++++++++++++++-
 9 files changed, 752 insertions(+), 76 deletions(-)

-- 
2.9.5

[patch net-next RFC 1/7] devlink: Add support for resource abstraction

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

Add support for hardware resource abstraction over devlink. Each resource
is identified via unique name, furthermore it contains information
regarding its size and its related sub resources. Each resource can also
provide its current occupancy.

In some cases the sizes of some resources can be changed, yet for those
changes to take place a hot driver reload may be needed. The reload
capability will be introduced in the next patch.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 include/net/devlink.h        |  74 ++++++++++
 include/uapi/linux/devlink.h |   8 ++
 net/core/devlink.c           | 317 +++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 399 insertions(+)

diff --git a/include/net/devlink.h b/include/net/devlink.h
index b9654e1..fe57534 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -26,6 +26,7 @@ struct devlink {
 	struct list_head port_list;
 	struct list_head sb_list;
 	struct list_head dpipe_table_list;
+	struct list_head resource_list;
 	struct devlink_dpipe_headers *dpipe_headers;
 	const struct devlink_ops *ops;
 	struct device *dev;
@@ -223,6 +224,43 @@ struct devlink_dpipe_headers {
 	unsigned int headers_count;
 };
 
+/**
+ * struct devlink_resource_ops - resource ops
+ * @occ_get - get the occupied size
+ * @size_validate - validate the size of the resource before update, reload
+ *                  is needed for changes to take place
+ */
+struct devlink_resource_ops {
+	u64 (*occ_get)(void *priv);
+	int (*size_validate)(void *priv, u64 size);
+};
+
+/**
+ * struct devlink_resource - devlink resource
+ * @priv - private
+ * @name - name of the resource
+ * @size - size of the resource
+ * @size_new - updated size of the resource, reload is needed
+ * @id - id, per devlink instance
+ * @parent - parent resource
+ * @list - parent list
+ * @sub_resource_list - list of sub-resources
+ * @resource_ops - resource ops
+ */
+struct devlink_resource {
+	void *priv;
+	const char *name;
+	u64 size;
+	u64 size_new;
+	u64 id;
+	struct devlink_resource *parent;
+	struct list_head list;
+	struct list_head sub_resource_list;
+	struct devlink_resource_ops *resource_ops;
+};
+
+#define DEVLINK_RESOURCE_ID_PARENT_TOP 0
+
 struct devlink_ops {
 	int (*port_type_set)(struct devlink_port *devlink_port,
 			     enum devlink_port_type port_type);
@@ -331,6 +369,18 @@ int devlink_dpipe_match_put(struct sk_buff *skb,
 extern struct devlink_dpipe_header devlink_dpipe_header_ethernet;
 extern struct devlink_dpipe_header devlink_dpipe_header_ipv4;
 extern struct devlink_dpipe_header devlink_dpipe_header_ipv6;
+int devlink_resource_register(struct devlink *devlink,
+			      const char *resource_name,
+			      bool top_hirarchy,
+			      u64 resource_size,
+			      u64 resource_id,
+			      u64 parent_resource_id,
+			      struct devlink_resource_ops *resource_ops,
+			      void *priv);
+int devlink_resource_size_get(struct devlink *devlink,
+			      u64 resource_id,
+			      u64 *p_resource_size);
+void devlink_resources_clear(struct devlink *devlink);
 
 #else
 
@@ -468,6 +518,30 @@ devlink_dpipe_match_put(struct sk_buff *skb,
 	return 0;
 }
 
+static inline int
+devlink_resource_register(struct devlink *devlink,
+			  const char *resource_name,
+			  bool top_hirarchy,
+			  u64 resource_size,
+			  u64 resource_id,
+			  u64 parent_resource_id,
+			  struct devlink_resource_ops *resource_ops,
+			  void *priv)
+{
+	return 0;
+}
+
+static inline int
+devlink_resource_size_get(struct devlink *devlink, u64 resource_id,
+			  u64 *p_resource_size)
+{
+	return -EINVAL;
+}
+
+static inline void devlink_resources_clear(struct devlink *devlink)
+{
+}
+
 #endif
 
 #endif /* _NET_DEVLINK_H_ */
diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
index 0cbca96..9db1d70 100644
--- a/include/uapi/linux/devlink.h
+++ b/include/uapi/linux/devlink.h
@@ -69,6 +69,8 @@ enum devlink_command {
 	DEVLINK_CMD_DPIPE_ENTRIES_GET,
 	DEVLINK_CMD_DPIPE_HEADERS_GET,
 	DEVLINK_CMD_DPIPE_TABLE_COUNTERS_SET,
+	DEVLINK_CMD_RESOURCE_SET,
+	DEVLINK_CMD_RESOURCE_DUMP,
 
 	/* add new commands above here */
 	__DEVLINK_CMD_MAX,
@@ -201,6 +203,12 @@ enum devlink_attr {
 	DEVLINK_ATTR_PAD,
 
 	DEVLINK_ATTR_ESWITCH_ENCAP_MODE,	/* u8 */
+	DEVLINK_ATTR_RESOURCES,			/* nested */
+	DEVLINK_ATTR_RESOURCE,			/* nested */
+	DEVLINK_ATTR_RESOURCE_NAME,		/* string */
+	DEVLINK_ATTR_RESOURCE_SIZE,		/* u64 */
+	DEVLINK_ATTR_RESOURCE_SIZE_NEW,		/* u64 */
+	DEVLINK_ATTR_RESOURCE_ID,		/* u64 */
 
 	/* add new attributes above here, update the policy in devlink.c */
 
diff --git a/net/core/devlink.c b/net/core/devlink.c
index 7d430c1..a48dcd6 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -2273,6 +2273,198 @@ static int devlink_nl_cmd_dpipe_table_counters_set(struct sk_buff *skb,
 						counters_enable);
 }
 
+struct devlink_resource *
+devlink_subresource_find(struct devlink_resource *resource, u64 resource_id)
+{
+	struct devlink_resource *sub_resource;
+
+	list_for_each_entry(sub_resource, &resource->sub_resource_list, list) {
+		struct devlink_resource *__sub_resource;
+
+		if (sub_resource->id == resource_id)
+			return sub_resource;
+
+		__sub_resource = devlink_subresource_find(sub_resource,
+							  resource_id);
+		if (__sub_resource)
+			return __sub_resource;
+	}
+	return NULL;
+}
+
+struct devlink_resource *
+devlink_resource_find(struct devlink *devlink, u64 resource_id)
+{
+	struct devlink_resource *resource;
+
+	list_for_each_entry(resource, &devlink->resource_list, list) {
+		struct devlink_resource *sub_resource;
+
+		if (resource->id == resource_id)
+			return resource;
+
+		sub_resource = devlink_subresource_find(resource, resource_id);
+		if (sub_resource)
+			return sub_resource;
+	}
+	return NULL;
+}
+
+static int devlink_nl_cmd_resource_set(struct sk_buff *skb,
+				       struct genl_info *info)
+{
+	struct devlink *devlink = info->user_ptr[0];
+	struct devlink_resource *resource;
+	u64 resource_id;
+	u64 size;
+	int err;
+
+	if (!info->attrs[DEVLINK_ATTR_RESOURCE_ID] ||
+	    !info->attrs[DEVLINK_ATTR_RESOURCE_SIZE])
+		return -EINVAL;
+	resource_id = nla_get_u64(info->attrs[DEVLINK_ATTR_RESOURCE_ID]);
+
+	resource = devlink_resource_find(devlink, resource_id);
+	if (!resource)
+		return -EINVAL;
+
+	if (!resource->resource_ops->size_validate)
+		return -EINVAL;
+
+	size = nla_get_u64(info->attrs[DEVLINK_ATTR_RESOURCE_SIZE]);
+	err = resource->resource_ops->size_validate(resource->priv, size);
+	if (err)
+		return err;
+
+	resource->size_new = size;
+	return 0;
+}
+
+static int devlink_resource_put(struct sk_buff *skb,
+				struct devlink_resource *resource)
+{
+	struct devlink_resource *sub_resource;
+	struct nlattr *sub_resource_attr;
+	struct nlattr *resource_attr;
+
+	resource_attr = nla_nest_start(skb, DEVLINK_ATTR_RESOURCE);
+	if (!resource_attr)
+		return -EMSGSIZE;
+
+	if (nla_put_string(skb, DEVLINK_ATTR_RESOURCE_NAME, resource->name) ||
+	    nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE, resource->size,
+			      DEVLINK_ATTR_PAD) ||
+	    nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_ID, resource->id,
+			      DEVLINK_ATTR_PAD))
+		goto nla_put_failure;
+	if (resource->size != resource->size_new)
+		nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE_NEW,
+				  resource->size_new, DEVLINK_ATTR_PAD);
+	if (list_empty(&resource->sub_resource_list))
+		goto out;
+
+	sub_resource_attr = nla_nest_start(skb, DEVLINK_ATTR_RESOURCES);
+	if (!sub_resource_attr)
+		goto nla_put_failure;
+
+	list_for_each_entry(sub_resource, &resource->sub_resource_list, list) {
+		if (devlink_resource_put(skb, sub_resource))
+			goto resource_put_failure;
+	}
+
+	nla_nest_end(skb, sub_resource_attr);
+out:
+	nla_nest_end(skb, resource_attr);
+	return 0;
+
+resource_put_failure:
+	nla_nest_cancel(skb, sub_resource_attr);
+nla_put_failure:
+	nla_nest_cancel(skb, resource_attr);
+	return -EMSGSIZE;
+}
+
+static int devlink_resource_fill(struct genl_info *info,
+				 enum devlink_command cmd, int flags)
+{
+	struct devlink *devlink = info->user_ptr[0];
+	struct devlink_resource *resource;
+	struct nlattr *resources_attr;
+	struct sk_buff *skb = NULL;
+	struct nlmsghdr *nlh;
+	bool incomplete;
+	void *hdr;
+	int i;
+	int err;
+
+	resource = list_first_entry(&devlink->resource_list,
+				    struct devlink_resource, list);
+start_again:
+	err = devlink_dpipe_send_and_alloc_skb(&skb, info);
+	if (err)
+		return err;
+
+	hdr = genlmsg_put(skb, info->snd_portid, info->snd_seq,
+			  &devlink_nl_family, NLM_F_MULTI, cmd);
+	if (!hdr) {
+		nlmsg_free(skb);
+		return -EMSGSIZE;
+	}
+
+	if (devlink_nl_put_handle(skb, devlink))
+		goto nla_put_failure;
+
+	resources_attr = nla_nest_start(skb, DEVLINK_ATTR_RESOURCES);
+	if (!resources_attr)
+		goto nla_put_failure;
+
+	incomplete = false;
+	i = 0;
+	list_for_each_entry_from(resource, &devlink->resource_list, list) {
+		err = devlink_resource_put(skb, resource);
+		if (err) {
+			if (!i)
+				goto err_resource_put;
+			incomplete = true;
+			break;
+		}
+		i++;
+	}
+	nla_nest_end(skb, resources_attr);
+	genlmsg_end(skb, hdr);
+	if (incomplete)
+		goto start_again;
+send_done:
+	nlh = nlmsg_put(skb, info->snd_portid, info->snd_seq,
+			NLMSG_DONE, 0, flags | NLM_F_MULTI);
+	if (!nlh) {
+		err = devlink_dpipe_send_and_alloc_skb(&skb, info);
+		if (err)
+			goto err_skb_send_alloc;
+		goto send_done;
+	}
+	return genlmsg_reply(skb, info);
+
+nla_put_failure:
+	err = -EMSGSIZE;
+err_resource_put:
+err_skb_send_alloc:
+	genlmsg_cancel(skb, hdr);
+	nlmsg_free(skb);
+	return err;
+}
+
+static int devlink_nl_cmd_resource_dump(struct sk_buff *skb,
+					struct genl_info *info)
+{
+	struct devlink *devlink = info->user_ptr[0];
+
+	if (list_empty(&devlink->resource_list))
+		return -EOPNOTSUPP;
+
+	return devlink_resource_fill(info, DEVLINK_CMD_RESOURCE_DUMP, 0);
+}
+
 static const struct nla_policy devlink_nl_policy[DEVLINK_ATTR_MAX + 1] = {
 	[DEVLINK_ATTR_BUS_NAME] = { .type = NLA_NUL_STRING },
 	[DEVLINK_ATTR_DEV_NAME] = { .type = NLA_NUL_STRING },
@@ -2451,6 +2643,20 @@ static const struct genl_ops devlink_nl_ops[] = {
 		.flags = GENL_ADMIN_PERM,
 		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK,
 	},
+	{
+		.cmd = DEVLINK_CMD_RESOURCE_SET,
+		.doit = devlink_nl_cmd_resource_set,
+		.policy = devlink_nl_policy,
+		.flags = GENL_ADMIN_PERM,
+		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK,
+	},
+	{
+		.cmd = DEVLINK_CMD_RESOURCE_DUMP,
+		.doit = devlink_nl_cmd_resource_dump,
+		.policy = devlink_nl_policy,
+		.flags = GENL_ADMIN_PERM,
+		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK,
+	},
 };
 
 static struct genl_family devlink_nl_family __ro_after_init = {
@@ -2488,6 +2694,7 @@ struct devlink *devlink_alloc(const struct devlink_ops *ops, size_t priv_size)
 	INIT_LIST_HEAD(&devlink->port_list);
 	INIT_LIST_HEAD(&devlink->sb_list);
 	INIT_LIST_HEAD_RCU(&devlink->dpipe_table_list);
+	INIT_LIST_HEAD(&devlink->resource_list);
 	return devlink;
 }
 EXPORT_SYMBOL_GPL(devlink_alloc);
@@ -2815,6 +3022,116 @@ void devlink_dpipe_table_unregister(struct devlink *devlink,
 }
 EXPORT_SYMBOL_GPL(devlink_dpipe_table_unregister);
 
+/**
+ *	devlink_resource_register - devlink resource register
+ *
+ *	@devlink: devlink
+ *	@resource_name: resource's name
+ *	@top_hirarchy: top hirarchy
+ *	@resource_size: resource's size
+ *	@resource_id: resource's id
+ *	@parent_reosurce_id: resource's parent id
+ *	@resource_ops: resource ops
+ *	@priv: priv
+ */
+int devlink_resource_register(struct devlink *devlink,
+			      const char *resource_name,
+			      bool top_hirarchy,
+			      u64 resource_size,
+			      u64 resource_id,
+			      u64 parent_resource_id,
+			      struct devlink_resource_ops *resource_ops,
+			      void *priv)
+{
+	struct devlink_resource *resource;
+	struct list_head *resource_list;
+
+	resource = devlink_resource_find(devlink, resource_id);
+	if (resource)
+		return -EINVAL;
+
+	if (top_hirarchy) {
+		resource_list = &devlink->resource_list;
+	} else {
+		struct devlink_resource *parent_resource;
+
+		parent_resource = devlink_resource_find(devlink,
+							parent_resource_id);
+		if (parent_resource)
+			resource_list = &parent_resource->sub_resource_list;
+		else
+			return -EINVAL;
+	}
+
+	resource = kzalloc(sizeof(*resource), GFP_KERNEL);
+	if (!resource)
+		return -ENOMEM;
+
+	resource->name = resource_name;
+	resource->size = resource_size;
+	resource->size_new = resource_size;
+	resource->id = resource_id;
+	resource->resource_ops = resource_ops;
+	resource->priv = priv;
+	INIT_LIST_HEAD(&resource->sub_resource_list);
+
+	list_add_tail(&resource->list, resource_list);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(devlink_resource_register);
+
+static void devlink_subresources_clear(struct devlink_resource *resource)
+{
+	struct devlink_resource *sub_resource, *tmp;
+
+	list_for_each_entry_safe(sub_resource, tmp,
+				 &resource->sub_resource_list, list) {
+		devlink_subresources_clear(sub_resource);
+		list_del(&sub_resource->list);
+		kfree(sub_resource);
+	}
+}
+
+/**
+ *	devlink_resource_clear - frees all resources
+ *
+ *	@devlink: devlink
+ */
+void devlink_resources_clear(struct devlink *devlink)
+{
+	struct devlink_resource *resource, *tmp;
+
+	list_for_each_entry_safe(resource, tmp,
+				 &devlink->resource_list, list) {
+		devlink_subresources_clear(resource);
+		list_del(&resource->list);
+		kfree(resource);
+	}
+}
+EXPORT_SYMBOL_GPL(devlink_resources_clear);
+
+/**
+ *	devlink_resource_size_get - get and update size
+ *
+ *	@devlink: devlink
+ *	@resource_id: the requested resource id
+ *	@p_resource_size: ptr to update
+ */
+int devlink_resource_size_get(struct devlink *devlink,
+			      u64 resource_id,
+			      u64 *p_resource_size)
+{
+	struct devlink_resource *resource;
+
+	resource = devlink_resource_find(devlink, resource_id);
+	if (!resource)
+		return -EINVAL;
+	*p_resource_size = resource->size_new;
+	resource->size = resource->size_new;
+	return 0;
+}
+EXPORT_SYMBOL_GPL(devlink_resource_size_get);
+
 static int __init devlink_module_init(void)
 {
 	return genl_register_family(&devlink_nl_family);
-- 
2.9.5

[patch net-next RFC 2/7] devlink: Add per devlink instance lock

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

This is a preparation before introducing hot reload support. Currently
there is one global lock protecting all devlink access. This patch adds
per devlink instance lock which protects the internal members which are
the sb/dpipe/resources.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 include/net/devlink.h |  1 +
 net/core/devlink.c    | 50 +++++++++++++++++++++++++++++++++++---------------
 2 files changed, 36 insertions(+), 15 deletions(-)

diff --git a/include/net/devlink.h b/include/net/devlink.h
index fe57534..789e606 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -31,6 +31,7 @@ struct devlink {
 	const struct devlink_ops *ops;
 	struct device *dev;
 	possible_net_t _net;
+	struct mutex lock;
 	char priv[0] __aligned(NETDEV_ALIGN);
 };
 
diff --git a/net/core/devlink.c b/net/core/devlink.c
index a48dcd6..86dfec9 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -336,6 +336,8 @@ devlink_sb_tc_index_get_from_info(struct devlink_sb *devlink_sb,
 #define DEVLINK_NL_FLAG_NEED_PORT	BIT(1)
 #define DEVLINK_NL_FLAG_NEED_SB		BIT(2)
 #define DEVLINK_NL_FLAG_LOCK_PORTS	BIT(3)
+#define DEVLINK_NL_FLAG_NO_LOCK		BIT(4)
+
 	/* port is not needed but we need to ensure they don't
 	 * change in the middle of command
 	 */
@@ -380,15 +382,24 @@ static int devlink_nl_pre_doit(const struct genl_ops *ops,
 		}
 		info->user_ptr[1] = devlink_sb;
 	}
+	if (!(ops->internal_flags & DEVLINK_NL_FLAG_NO_LOCK))
+		mutex_lock(&devlink->lock);
+
 	return 0;
 }
 
 static void devlink_nl_post_doit(const struct genl_ops *ops,
 				 struct sk_buff *skb, struct genl_info *info)
 {
+	struct devlink *devlink;
+
+	devlink = devlink_get_from_info(info);
+
 	if (ops->internal_flags & DEVLINK_NL_FLAG_NEED_PORT ||
 	    ops->internal_flags & DEVLINK_NL_FLAG_LOCK_PORTS)
 		mutex_unlock(&devlink_port_mutex);
+	if (!(ops->internal_flags & DEVLINK_NL_FLAG_NO_LOCK))
+		mutex_unlock(&devlink->lock);
 	mutex_unlock(&devlink_mutex);
 }
 
@@ -801,6 +812,7 @@ static int devlink_nl_cmd_sb_get_dumpit(struct sk_buff *msg,
 	list_for_each_entry(devlink, &devlink_list, list) {
 		if (!net_eq(devlink_net(devlink), sock_net(msg->sk)))
 			continue;
+		mutex_lock(&devlink->lock);
 		list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 			if (idx < start) {
 				idx++;
@@ -811,10 +823,13 @@ static int devlink_nl_cmd_sb_get_dumpit(struct sk_buff *msg,
 						 NETLINK_CB(cb->skb).portid,
 						 cb->nlh->nlmsg_seq,
 						 NLM_F_MULTI);
-			if (err)
+			if (err) {
+				mutex_unlock(&devlink->lock);
 				goto out;
+			}
 			idx++;
 		}
+		mutex_unlock(&devlink->lock);
 	}
 out:
 	mutex_unlock(&devlink_mutex);
@@ -935,14 +950,18 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg,
 		if (!net_eq(devlink_net(devlink), sock_net(msg->sk)) ||
 		    !devlink->ops || !devlink->ops->sb_pool_get)
 			continue;
+		mutex_lock(&devlink->lock);
 		list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 			err = __sb_pool_get_dumpit(msg, start, &idx, devlink,
 						   devlink_sb,
 						   NETLINK_CB(cb->skb).portid,
 						   cb->nlh->nlmsg_seq);
-			if (err && err != -EOPNOTSUPP)
+			if (err && err != -EOPNOTSUPP) {
+				mutex_unlock(&devlink->lock);
 				goto out;
+			}
 		}
+		mutex_unlock(&devlink->lock);
 	}
 out:
 	mutex_unlock(&devlink_mutex);
@@ -2695,6 +2714,7 @@ struct devlink *devlink_alloc(const struct devlink_ops *ops, size_t priv_size)
 	INIT_LIST_HEAD(&devlink->sb_list);
 	INIT_LIST_HEAD_RCU(&devlink->dpipe_table_list);
 	INIT_LIST_HEAD(&devlink->resource_list);
+	mutex_init(&devlink->lock);
 	return devlink;
 }
 EXPORT_SYMBOL_GPL(devlink_alloc);
@@ -2858,7 +2878,7 @@ int devlink_sb_register(struct devlink *devlink, unsigned int sb_index,
 	struct devlink_sb *devlink_sb;
 	int err = 0;
 
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	if (devlink_sb_index_exists(devlink, sb_index)) {
 		err = -EEXIST;
 		goto unlock;
@@ -2877,7 +2897,7 @@ int devlink_sb_register(struct devlink *devlink, unsigned int sb_index,
 	devlink_sb->egress_tc_count = egress_tc_count;
 	list_add_tail(&devlink_sb->list, &devlink->sb_list);
 unlock:
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 	return err;
 }
 EXPORT_SYMBOL_GPL(devlink_sb_register);
@@ -2886,11 +2906,11 @@ void devlink_sb_unregister(struct devlink *devlink, unsigned int sb_index)
 {
 	struct devlink_sb *devlink_sb;
 
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	devlink_sb = devlink_sb_get_by_index(devlink, sb_index);
 	WARN_ON(!devlink_sb);
 	list_del(&devlink_sb->list);
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 	kfree(devlink_sb);
 }
 EXPORT_SYMBOL_GPL(devlink_sb_unregister);
@@ -2906,9 +2926,9 @@ EXPORT_SYMBOL_GPL(devlink_sb_unregister);
 int devlink_dpipe_headers_register(struct devlink *devlink,
 				   struct devlink_dpipe_headers *dpipe_headers)
 {
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	devlink->dpipe_headers = dpipe_headers;
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 	return 0;
 }
 EXPORT_SYMBOL_GPL(devlink_dpipe_headers_register);
@@ -2922,9 +2942,9 @@ EXPORT_SYMBOL_GPL(devlink_dpipe_headers_register);
  */
 void devlink_dpipe_headers_unregister(struct devlink *devlink)
 {
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	devlink->dpipe_headers = NULL;
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 }
 EXPORT_SYMBOL_GPL(devlink_dpipe_headers_unregister);
 
@@ -2990,9 +3010,9 @@ int devlink_dpipe_table_register(struct devlink *devlink,
 	table->priv = priv;
 	table->counter_control_extern = counter_control_extern;
 
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	list_add_tail_rcu(&table->list, &devlink->dpipe_table_list);
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 	return 0;
 }
 EXPORT_SYMBOL_GPL(devlink_dpipe_table_register);
@@ -3008,17 +3028,17 @@ void devlink_dpipe_table_unregister(struct devlink *devlink,
 {
 	struct devlink_dpipe_table *table;
 
-	mutex_lock(&devlink_mutex);
+	mutex_lock(&devlink->lock);
 	table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 					 table_name);
 	if (!table)
 		goto unlock;
 	list_del_rcu(&table->list);
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 	kfree_rcu(table, rcu);
 	return;
 unlock:
-	mutex_unlock(&devlink_mutex);
+	mutex_unlock(&devlink->lock);
 }
 EXPORT_SYMBOL_GPL(devlink_dpipe_table_unregister);
 
-- 
2.9.5

[patch net-next RFC 3/7] devlink: Add support for reload

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

Add support for performing driver hot reload.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 include/net/devlink.h        |  1 +
 include/uapi/linux/devlink.h |  1 +
 net/core/devlink.c           | 18 ++++++++++++++++++
 3 files changed, 20 insertions(+)

diff --git a/include/net/devlink.h b/include/net/devlink.h
index 789e606..ff6e384 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -263,6 +263,7 @@ struct devlink_resource {
 #define DEVLINK_RESOURCE_ID_PARENT_TOP 0
 
 struct devlink_ops {
+	int (*reload)(struct devlink *devlink);
 	int (*port_type_set)(struct devlink_port *devlink_port,
 			     enum devlink_port_type port_type);
 	int (*port_split)(struct devlink *devlink, unsigned int port_index,
diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
index 9db1d70..86a329a 100644
--- a/include/uapi/linux/devlink.h
+++ b/include/uapi/linux/devlink.h
@@ -71,6 +71,7 @@ enum devlink_command {
 	DEVLINK_CMD_DPIPE_TABLE_COUNTERS_SET,
 	DEVLINK_CMD_RESOURCE_SET,
 	DEVLINK_CMD_RESOURCE_DUMP,
+	DEVLINK_CMD_RELOAD,
 
 	/* add new commands above here */
 	__DEVLINK_CMD_MAX,
diff --git a/net/core/devlink.c b/net/core/devlink.c
index 86dfec9..3a9fcda 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -2484,6 +2484,16 @@ static int devlink_nl_cmd_resource_dump(struct sk_buff *skb,
 	return devlink_resource_fill(info, DEVLINK_CMD_RESOURCE_DUMP, 0);
 }
 
+static int devlink_nl_cmd_reload(struct sk_buff *skb, struct genl_info *info)
+{
+	struct devlink *devlink = info->user_ptr[0];
+
+	if (!devlink->ops->reload)
+		return -EOPNOTSUPP;
+
+	return devlink->ops->reload(devlink);
+}
+
 static const struct nla_policy devlink_nl_policy[DEVLINK_ATTR_MAX + 1] = {
 	[DEVLINK_ATTR_BUS_NAME] = { .type = NLA_NUL_STRING },
 	[DEVLINK_ATTR_DEV_NAME] = { .type = NLA_NUL_STRING },
@@ -2676,6 +2686,14 @@ static const struct genl_ops devlink_nl_ops[] = {
 		.flags = GENL_ADMIN_PERM,
 		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK,
 	},
+	{
+		.cmd = DEVLINK_CMD_RELOAD,
+		.doit = devlink_nl_cmd_reload,
+		.policy = devlink_nl_policy,
+		.flags = GENL_ADMIN_PERM,
+		.internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK |
+				  DEVLINK_NL_FLAG_NO_LOCK,
+	},
 };
 
 static struct genl_family devlink_nl_family __ro_after_init = {
-- 
2.9.5

[patch net-next RFC 4/7] mlxsw: pci: Add support for performing bus reset

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

This is a preparation stage before introducing hot reload. During the
reload process the ASIC should be resetted by accessing the PCI BAR due
to unavailability of the mailbox/emad interfaces.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/core.h |  1 +
 drivers/net/ethernet/mellanox/mlxsw/pci.c  | 56 +++++++++++++++++++++++-------
 2 files changed, 44 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.h b/drivers/net/ethernet/mellanox/mlxsw/core.h
index 6e966af..34dda96 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.h
@@ -332,6 +332,7 @@ struct mlxsw_bus {
 		    const struct mlxsw_config_profile *profile,
 		    struct mlxsw_res *res);
 	void (*fini)(void *bus_priv);
+	void (*reset)(void *bus_priv);
 	bool (*skb_transmit_busy)(void *bus_priv,
 				  const struct mlxsw_tx_info *tx_info);
 	int (*skb_transmit)(void *bus_priv, struct sk_buff *skb,
diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c
index 23f7d82..00c9155 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c
@@ -154,6 +154,7 @@ struct mlxsw_pci {
 		} comp;
 	} cmd;
 	struct mlxsw_bus_info bus_info;
+	const struct pci_device_id *id;
 };
 
 static void mlxsw_pci_queue_tasklet_schedule(struct mlxsw_pci_queue *q)
@@ -1622,16 +1623,6 @@ static int mlxsw_pci_cmd_exec(void *bus_priv, u16 opcode, u8 opcode_mod,
 	return err;
 }
 
-static const struct mlxsw_bus mlxsw_pci_bus = {
-	.kind			= "pci",
-	.init			= mlxsw_pci_init,
-	.fini			= mlxsw_pci_fini,
-	.skb_transmit_busy	= mlxsw_pci_skb_transmit_busy,
-	.skb_transmit		= mlxsw_pci_skb_transmit,
-	.cmd_exec		= mlxsw_pci_cmd_exec,
-	.features		= MLXSW_BUS_F_TXRX,
-};
-
 static int mlxsw_pci_sw_reset(struct mlxsw_pci *mlxsw_pci,
 			      const struct pci_device_id *id)
 {
@@ -1655,6 +1646,44 @@ static int mlxsw_pci_sw_reset(struct mlxsw_pci *mlxsw_pci,
 	return 0;
 }
 
+static void mlxsw_pci_free_irq_vectors(struct mlxsw_pci *mlxsw_pci)
+{
+	pci_free_irq_vectors(mlxsw_pci->pdev);
+}
+
+static int mlxsw_pci_alloc_irq_vectors(struct mlxsw_pci *mlxsw_pci)
+{
+	int err;
+
+	err = pci_alloc_irq_vectors(mlxsw_pci->pdev, 1, 1, PCI_IRQ_MSIX);
+	if (err < 0) {
+		dev_err(&mlxsw_pci->pdev->dev, "MSI-X init failed\n");
+		return err;
+	}
+
+	return 0;
+}
+
+static void mlxsw_pci_reset(void *bus_priv)
+{
+	struct mlxsw_pci *mlxsw_pci = bus_priv;
+
+	mlxsw_pci_free_irq_vectors(mlxsw_pci);
+	mlxsw_pci_sw_reset(mlxsw_pci, mlxsw_pci->id);
+	mlxsw_pci_alloc_irq_vectors(mlxsw_pci);
+}
+
+static const struct mlxsw_bus mlxsw_pci_bus = {
+	.kind			= "pci",
+	.init			= mlxsw_pci_init,
+	.fini			= mlxsw_pci_fini,
+	.skb_transmit_busy	= mlxsw_pci_skb_transmit_busy,
+	.skb_transmit		= mlxsw_pci_skb_transmit,
+	.cmd_exec		= mlxsw_pci_cmd_exec,
+	.features		= MLXSW_BUS_F_TXRX,
+	.reset			= mlxsw_pci_reset,
+};
+
 static int mlxsw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	const char *driver_name = pdev->driver->name;
@@ -1716,7 +1745,7 @@ static int mlxsw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		goto err_sw_reset;
 	}
 
-	err = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_MSIX);
+	err = mlxsw_pci_alloc_irq_vectors(mlxsw_pci);
 	if (err < 0) {
 		dev_err(&pdev->dev, "MSI-X init failed\n");
 		goto err_msix_init;
@@ -1725,6 +1754,7 @@ static int mlxsw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	mlxsw_pci->bus_info.device_kind = driver_name;
 	mlxsw_pci->bus_info.device_name = pci_name(mlxsw_pci->pdev);
 	mlxsw_pci->bus_info.dev = &pdev->dev;
+	mlxsw_pci->id = id;
 
 	err = mlxsw_core_bus_device_register(&mlxsw_pci->bus_info,
 					     &mlxsw_pci_bus, mlxsw_pci);
@@ -1736,7 +1766,7 @@ static int mlxsw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	return 0;
 
 err_bus_device_register:
-	pci_free_irq_vectors(mlxsw_pci->pdev);
+	mlxsw_pci_free_irq_vectors(mlxsw_pci);
 err_msix_init:
 err_sw_reset:
 	iounmap(mlxsw_pci->hw_addr);
@@ -1756,7 +1786,7 @@ static void mlxsw_pci_remove(struct pci_dev *pdev)
 	struct mlxsw_pci *mlxsw_pci = pci_get_drvdata(pdev);
 
 	mlxsw_core_bus_device_unregister(mlxsw_pci->core);
-	pci_free_irq_vectors(mlxsw_pci->pdev);
+	mlxsw_pci_free_irq_vectors(mlxsw_pci);
 	iounmap(mlxsw_pci->hw_addr);
 	pci_release_regions(mlxsw_pci->pdev);
 	pci_disable_device(mlxsw_pci->pdev);
-- 
2.9.5

[patch net-next RFC 5/7] mlxsw: Register KVD resources with devlink

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

Register the KVD resources with devlink. The KVD is a memory resource
which is subdivided into three partitions which are the linear, hash
single and hash double.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/core.c     |  9 +++
 drivers/net/ethernet/mellanox/mlxsw/core.h     |  1 +
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c | 91 +++++++++++++++++++++++++-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h | 12 ++++
 4 files changed, 112 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.c b/drivers/net/ethernet/mellanox/mlxsw/core.c
index f3315bc..ee41036 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.c
@@ -1012,6 +1012,12 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 	if (err)
 		goto err_bus_init;
 
+	if (mlxsw_driver->resources_register) {
+		err = mlxsw_driver->resources_register(mlxsw_core);
+		if (err)
+			goto err_register_resources;
+	}
+
 	err = mlxsw_ports_init(mlxsw_core);
 	if (err)
 		goto err_ports_init;
@@ -1067,6 +1073,8 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 err_ports_init:
 	mlxsw_bus->fini(bus_priv);
 err_bus_init:
+	devlink_resources_clear(devlink);
+err_register_resources:
 	devlink_free(devlink);
 err_devlink_alloc:
 	mlxsw_core_driver_put(device_kind);
@@ -1086,6 +1094,7 @@ void mlxsw_core_bus_device_unregister(struct mlxsw_core *mlxsw_core)
 	mlxsw_emad_fini(mlxsw_core);
 	kfree(mlxsw_core->lag.mapping);
 	mlxsw_ports_fini(mlxsw_core);
+	devlink_resources_clear(devlink);
 	mlxsw_core->bus->fini(mlxsw_core->bus_priv);
 	devlink_free(devlink);
 	mlxsw_core_driver_put(device_kind);
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.h b/drivers/net/ethernet/mellanox/mlxsw/core.h
index 34dda96..e23f83b 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.h
@@ -308,6 +308,7 @@ struct mlxsw_driver {
 				       u32 *p_cur, u32 *p_max);
 	void (*txhdr_construct)(struct sk_buff *skb,
 				const struct mlxsw_tx_info *tx_info);
+	int (*resources_register)(struct mlxsw_core *mlxsw_core);
 	u8 txhdr_len;
 	const struct mlxsw_config_profile *profile;
 };
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
index 12b6ac4..565f94e 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
@@ -3864,7 +3864,6 @@ static void mlxsw_sp_fini(struct mlxsw_core *mlxsw_core)
 	mlxsw_sp_fids_fini(mlxsw_sp);
 	mlxsw_sp_kvdl_fini(mlxsw_sp);
 }
-
 static const struct mlxsw_config_profile mlxsw_sp_config_profile = {
 	.used_max_vepa_channels		= 1,
 	.max_vepa_channels		= 0,
@@ -3897,6 +3896,95 @@ static const struct mlxsw_config_profile mlxsw_sp_config_profile = {
 	.resource_query_enable		= 1,
 };
 
+int mlxsw_sp_resource_kvd_size_validate(void *priv, u64 size)
+{
+	/* KVD total size cannot be set*/
+	return -EINVAL;
+}
+
+int mlxsw_sp_resource_kvd_sub_size_validate(void *priv, u64 size)
+{
+	const struct mlxsw_config_profile *profile;
+
+	profile = &mlxsw_sp_config_profile;
+	if (size % profile->kvd_hash_granularity)
+		return -EINVAL;
+	return 0;
+}
+
+static struct devlink_resource_ops mlxsw_sp_resource_kvd_ops = {
+	.size_validate = mlxsw_sp_resource_kvd_size_validate,
+};
+
+static struct devlink_resource_ops mlxsw_sp_resource_kvd_linear_ops = {
+	.size_validate = mlxsw_sp_resource_kvd_sub_size_validate,
+};
+
+static struct devlink_resource_ops mlxsw_sp_resource_kvd_hash_single_ops = {
+	.size_validate = mlxsw_sp_resource_kvd_sub_size_validate,
+};
+
+static struct devlink_resource_ops mlxsw_sp_resource_kvd_hash_double_ops = {
+	.size_validate = mlxsw_sp_resource_kvd_sub_size_validate,
+};
+
+static int mlxsw_sp_resources_register(struct mlxsw_core *mlxsw_core)
+{
+	struct devlink *devlink = priv_to_devlink(mlxsw_core);
+	u64 kvd_size, single_size, double_size, linear_size;
+	const struct mlxsw_config_profile *profile;
+	int err;
+
+	profile = &mlxsw_sp_config_profile;
+	if (!MLXSW_CORE_RES_VALID(mlxsw_core, KVD_SIZE))
+		return -EIO;
+
+	kvd_size = MLXSW_CORE_RES_GET(mlxsw_core, KVD_SIZE);
+	err = devlink_resource_register(devlink, MLXSW_SP_RESOURCE_NAME_KVD,
+					true, kvd_size,
+					MLXSW_SP_RESOURCE_KVD,
+					DEVLINK_RESOURCE_ID_PARENT_TOP,
+					&mlxsw_sp_resource_kvd_ops, mlxsw_core);
+	if (err)
+		return err;
+
+	linear_size = profile->kvd_linear_size;
+	err = devlink_resource_register(devlink, MLXSW_SP_RESOURCE_NAME_KVD_LINEAR,
+					false, linear_size,
+					MLXSW_SP_RESOURCE_KVD_LINEAR,
+					MLXSW_SP_RESOURCE_KVD,
+					&mlxsw_sp_resource_kvd_linear_ops,
+					mlxsw_core);
+	if (err)
+		return err;
+
+	double_size = kvd_size - linear_size;
+	double_size *= profile->kvd_hash_double_parts;
+	double_size /= profile->kvd_hash_double_parts +
+		       profile->kvd_hash_single_parts;
+	double_size = rounddown(double_size, profile->kvd_hash_granularity);
+	err = devlink_resource_register(devlink, MLXSW_SP_RESOURCE_NAME_KVD_HASH_DOUBLE,
+					false, double_size,
+					MLXSW_SP_RESOURCE_KVD_HASH_DOUBLE,
+					MLXSW_SP_RESOURCE_KVD,
+					&mlxsw_sp_resource_kvd_hash_double_ops,
+					mlxsw_core);
+	if (err)
+		return err;
+
+	single_size = kvd_size - double_size - linear_size;
+	err = devlink_resource_register(devlink, MLXSW_SP_RESOURCE_NAME_KVD_HASH_SINGLE,
+					false, single_size,
+					MLXSW_SP_RESOURCE_KVD_HASH_SINGLE,
+					MLXSW_SP_RESOURCE_KVD,
+					&mlxsw_sp_resource_kvd_hash_single_ops,
+					mlxsw_core);
+	if (err)
+		return err;
+
+	return 0;
+}
+
 static struct mlxsw_driver mlxsw_sp_driver = {
 	.kind				= mlxsw_sp_driver_name,
 	.priv_size			= sizeof(struct mlxsw_sp),
@@ -3916,6 +4004,7 @@ static struct mlxsw_driver mlxsw_sp_driver = {
 	.sb_occ_port_pool_get		= mlxsw_sp_sb_occ_port_pool_get,
 	.sb_occ_tc_port_bind_get	= mlxsw_sp_sb_occ_tc_port_bind_get,
 	.txhdr_construct		= mlxsw_sp_txhdr_construct,
+	.resources_register		= mlxsw_sp_resources_register,
 	.txhdr_len			= MLXSW_TXHDR_LEN,
 	.profile			= &mlxsw_sp_config_profile,
 };
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
index dc1b739..9413481 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
@@ -65,6 +65,18 @@
 #define MLXSW_SP_KVD_LINEAR_SIZE 98304 /* entries */
 #define MLXSW_SP_KVD_GRANULARITY 128
 
+#define MLXSW_SP_RESOURCE_NAME_KVD "kvd"
+#define MLXSW_SP_RESOURCE_NAME_KVD_LINEAR "linear"
+#define MLXSW_SP_RESOURCE_NAME_KVD_HASH_SINGLE "hash_single"
+#define MLXSW_SP_RESOURCE_NAME_KVD_HASH_DOUBLE "hash_double"
+
+enum mlxsw_sp_resource_id {
+	MLXSW_SP_RESOURCE_KVD,
+	MLXSW_SP_RESOURCE_KVD_LINEAR,
+	MLXSW_SP_RESOURCE_KVD_HASH_SINGLE,
+	MLXSW_SP_RESOURCE_KVD_HASH_DOUBLE,
+};
+
 struct mlxsw_sp_port;
 struct mlxsw_sp_rif;
 
-- 
2.9.5

[patch net-next RFC 6/7] mlxsw: pci: Add support for getting resource through devlink

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

Up until now the KVD partition was static. This patch introduces the
ability to get the resource sizes via devlink. In case the resource is not
available the default configuration is used.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/core.c     | 16 ++++++++
 drivers/net/ethernet/mellanox/mlxsw/core.h     |  9 ++++
 drivers/net/ethernet/mellanox/mlxsw/pci.c      | 40 +++++-------------
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c | 57 ++++++++++++++++++++++++++
 4 files changed, 92 insertions(+), 30 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.c b/drivers/net/ethernet/mellanox/mlxsw/core.c
index ee41036..ab93437 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.c
@@ -1800,6 +1800,22 @@ void mlxsw_core_flush_owq(void)
 }
 EXPORT_SYMBOL(mlxsw_core_flush_owq);
 
+int mlxsw_core_kvd_sizes_get(struct mlxsw_core *mlxsw_core,
+			     const struct mlxsw_config_profile *profile,
+			     u64 *p_single_size, u64 *p_double_size,
+			     u64 *p_linear_size)
+{
+	struct mlxsw_driver *driver =  mlxsw_core->driver;
+
+	if (!driver->kvd_sizes_get)
+		return -EINVAL;
+
+	return driver->kvd_sizes_get(mlxsw_core, profile,
+				     p_single_size, p_double_size,
+				     p_linear_size);
+}
+EXPORT_SYMBOL(mlxsw_core_kvd_sizes_get);
+
 static int __init mlxsw_core_module_init(void)
 {
 	int err;
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.h b/drivers/net/ethernet/mellanox/mlxsw/core.h
index e23f83b..e44061d 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.h
@@ -309,10 +309,19 @@ struct mlxsw_driver {
 	void (*txhdr_construct)(struct sk_buff *skb,
 				const struct mlxsw_tx_info *tx_info);
 	int (*resources_register)(struct mlxsw_core *mlxsw_core);
+	int (*kvd_sizes_get)(struct mlxsw_core *mlxsw_core,
+			     const struct mlxsw_config_profile *profile,
+			     u64 *p_single_size, u64 *p_double_size,
+			     u64 *p_linear_size);
 	u8 txhdr_len;
 	const struct mlxsw_config_profile *profile;
 };
 
+int mlxsw_core_kvd_sizes_get(struct mlxsw_core *mlxsw_core,
+			     const struct mlxsw_config_profile *profile,
+			     u64 *p_single_size, u64 *p_double_size,
+			     u64 *p_linear_size);
+
 bool mlxsw_core_res_valid(struct mlxsw_core *mlxsw_core,
 			  enum mlxsw_res_id res_id);
 
diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c
index 00c9155..da08f08 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c
@@ -1053,38 +1053,18 @@ static int mlxsw_pci_resources_query(struct mlxsw_pci *mlxsw_pci, char *mbox,
 }
 
 static int
-mlxsw_pci_profile_get_kvd_sizes(const struct mlxsw_config_profile *profile,
+mlxsw_pci_profile_get_kvd_sizes(const struct mlxsw_pci *mlxsw_pci,
+				const struct mlxsw_config_profile *profile,
 				struct mlxsw_res *res)
 {
-	u32 single_size, double_size, linear_size;
-
-	if (!MLXSW_RES_VALID(res, KVD_SINGLE_MIN_SIZE) ||
-	    !MLXSW_RES_VALID(res, KVD_DOUBLE_MIN_SIZE) ||
-	    !profile->used_kvd_split_data)
-		return -EIO;
-
-	linear_size = profile->kvd_linear_size;
+	u64 single_size, double_size, linear_size;
+	int err;
 
-	/* The hash part is what left of the kvd without the
-	 * linear part. It is split to the single size and
-	 * double size by the parts ratio from the profile.
-	 * Both sizes must be a multiplications of the
-	 * granularity from the profile.
-	 */
-	double_size = MLXSW_RES_GET(res, KVD_SIZE) - linear_size;
-	double_size *= profile->kvd_hash_double_parts;
-	double_size /= profile->kvd_hash_double_parts +
-		       profile->kvd_hash_single_parts;
-	double_size /= profile->kvd_hash_granularity;
-	double_size *= profile->kvd_hash_granularity;
-	single_size = MLXSW_RES_GET(res, KVD_SIZE) - double_size -
-		      linear_size;
-
-	/* Check results are legal. */
-	if (single_size < MLXSW_RES_GET(res, KVD_SINGLE_MIN_SIZE) ||
-	    double_size < MLXSW_RES_GET(res, KVD_DOUBLE_MIN_SIZE) ||
-	    MLXSW_RES_GET(res, KVD_SIZE) < linear_size)
-		return -EIO;
+	err = mlxsw_core_kvd_sizes_get(mlxsw_pci->core, profile,
+				       &single_size, &double_size,
+				       &linear_size);
+	if (err)
+		return err;
 
 	MLXSW_RES_SET(res, KVD_SINGLE_SIZE, single_size);
 	MLXSW_RES_SET(res, KVD_DOUBLE_SIZE, double_size);
@@ -1185,7 +1165,7 @@ static int mlxsw_pci_config_profile(struct mlxsw_pci *mlxsw_pci, char *mbox,
 			mbox, profile->adaptive_routing_group_cap);
 	}
 	if (MLXSW_RES_VALID(res, KVD_SIZE)) {
-		err = mlxsw_pci_profile_get_kvd_sizes(profile, res);
+		err = mlxsw_pci_profile_get_kvd_sizes(mlxsw_pci, profile, res);
 		if (err)
 			return err;
 
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
index 565f94e..8f8db6c 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
@@ -3985,6 +3985,62 @@ static int mlxsw_sp_resources_register(struct mlxsw_core *mlxsw_core)
 	return 0;
 }
 
+static int mlxsw_sp_kvd_sizes_get(struct mlxsw_core *mlxsw_core,
+				  const struct mlxsw_config_profile *profile,
+				  u64 *p_single_size, u64 *p_double_size,
+				  u64 *p_linear_size)
+{
+	struct devlink *devlink = priv_to_devlink(mlxsw_core);
+	u64 double_size;
+	int err;
+
+	if (!MLXSW_CORE_RES_VALID(mlxsw_core, KVD_SINGLE_MIN_SIZE) ||
+	    !MLXSW_CORE_RES_VALID(mlxsw_core, KVD_DOUBLE_MIN_SIZE) ||
+	    !profile->used_kvd_split_data)
+		return -EIO;
+
+	/* The hash part is what left of the kvd without the
+	 * linear part. It is split to the single size and
+	 * double size by the parts ratio from the profile.
+	 * Both sizes must be a multiplications of the
+	 * granularity from the profile. In case the user
+	 * provided the sizes they are obtained via devlink.
+	 */
+	err = devlink_resource_size_get(devlink,
+					MLXSW_SP_RESOURCE_KVD_LINEAR,
+					p_linear_size);
+	if (err)
+		*p_linear_size = profile->kvd_linear_size;
+
+	err = devlink_resource_size_get(devlink,
+					MLXSW_SP_RESOURCE_KVD_HASH_DOUBLE,
+					p_double_size);
+	if (err) {
+		double_size = MLXSW_CORE_RES_GET(mlxsw_core, KVD_SIZE) -
+			      *p_linear_size;
+		double_size *= profile->kvd_hash_double_parts;
+		double_size /= profile->kvd_hash_double_parts +
+			       profile->kvd_hash_single_parts;
+		*p_double_size = rounddown(double_size,
+					   profile->kvd_hash_granularity);
+	}
+
+	err = devlink_resource_size_get(devlink,
+					MLXSW_SP_RESOURCE_KVD_HASH_SINGLE,
+					p_single_size);
+	if (err)
+		*p_single_size = MLXSW_CORE_RES_GET(mlxsw_core, KVD_SIZE) -
+				 *p_double_size - *p_linear_size;
+
+	/* Check results are legal. */
+	if (*p_single_size < MLXSW_CORE_RES_GET(mlxsw_core, KVD_SINGLE_MIN_SIZE) ||
+	    *p_double_size < MLXSW_CORE_RES_GET(mlxsw_core, KVD_DOUBLE_MIN_SIZE) ||
+	    MLXSW_CORE_RES_GET(mlxsw_core, KVD_SIZE) < *p_linear_size)
+		return -EIO;
+
+	return 0;
+}
+
 static struct mlxsw_driver mlxsw_sp_driver = {
 	.kind				= mlxsw_sp_driver_name,
 	.priv_size			= sizeof(struct mlxsw_sp),
@@ -4005,6 +4061,7 @@ static struct mlxsw_driver mlxsw_sp_driver = {
 	.sb_occ_tc_port_bind_get	= mlxsw_sp_sb_occ_tc_port_bind_get,
 	.txhdr_construct		= mlxsw_sp_txhdr_construct,
 	.resources_register		= mlxsw_sp_resources_register,
+	.kvd_sizes_get			= mlxsw_sp_kvd_sizes_get,
 	.txhdr_len			= MLXSW_TXHDR_LEN,
 	.profile			= &mlxsw_sp_config_profile,
 };
-- 
2.9.5

[patch net-next RFC 7/7] mlxsw: core: Add support for reload

[Jiri Pirko]

From: Arkadi Sharshevsky <arkadis@mellanox.com>

Add support for hot reload. First, all the driver/core resources are
released but the PCI and devlink instances, then reset is performed
through the PCI interface. Finally the driver performs initialization.

In case of reload failure the driver is left in a partially initialized
state. Special care is taken during the driver removal in order to
properly handle this state.

Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/core.c | 58 +++++++++++++++++++++++-------
 drivers/net/ethernet/mellanox/mlxsw/core.h |  4 +--
 drivers/net/ethernet/mellanox/mlxsw/i2c.c  |  4 +--
 drivers/net/ethernet/mellanox/mlxsw/pci.c  |  4 +--
 4 files changed, 51 insertions(+), 19 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.c b/drivers/net/ethernet/mellanox/mlxsw/core.c
index ab93437..2fcde3c 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.c
@@ -113,6 +113,7 @@ struct mlxsw_core {
 	struct mlxsw_thermal *thermal;
 	struct mlxsw_core_port *ports;
 	unsigned int max_ports;
+	bool reload_fail;
 	unsigned long driver_priv[0];
 	/* driver_priv has to be always the last item */
 };
@@ -962,7 +963,24 @@ mlxsw_devlink_sb_occ_tc_port_bind_get(struct devlink_port *devlink_port,
 						     pool_type, p_cur, p_max);
 }
 
+static int mlxsw_devlink_core_bus_device_reload(struct devlink *devlink)
+{
+	struct mlxsw_core *mlxsw_core = devlink_priv(devlink);
+	const struct mlxsw_bus *mlxsw_bus = mlxsw_core->bus;
+	int err;
+
+	mlxsw_core_bus_device_unregister(mlxsw_core, true);
+	mlxsw_bus->reset(mlxsw_core->bus_priv);
+	err = mlxsw_core_bus_device_register(mlxsw_core->bus_info,
+					     mlxsw_core->bus,
+					     mlxsw_core->bus_priv, true);
+	if (err)
+		mlxsw_core->reload_fail = true;
+	return err;
+}
+
 static const struct devlink_ops mlxsw_devlink_ops = {
+	.reload				= mlxsw_devlink_core_bus_device_reload,
 	.port_type_set			= mlxsw_devlink_port_type_set,
 	.port_split			= mlxsw_devlink_port_split,
 	.port_unsplit			= mlxsw_devlink_port_unsplit,
@@ -980,7 +998,7 @@ static const struct devlink_ops mlxsw_devlink_ops = {
 
 int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 				   const struct mlxsw_bus *mlxsw_bus,
-				   void *bus_priv)
+				   void *bus_priv, bool reload)
 {
 	const char *device_kind = mlxsw_bus_info->device_kind;
 	struct mlxsw_core *mlxsw_core;
@@ -992,11 +1010,14 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 	mlxsw_driver = mlxsw_core_driver_get(device_kind);
 	if (!mlxsw_driver)
 		return -EINVAL;
-	alloc_size = sizeof(*mlxsw_core) + mlxsw_driver->priv_size;
-	devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
-	if (!devlink) {
-		err = -ENOMEM;
-		goto err_devlink_alloc;
+
+	if (!reload) {
+		alloc_size = sizeof(*mlxsw_core) + mlxsw_driver->priv_size;
+		devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
+		if (!devlink) {
+			err = -ENOMEM;
+			goto err_devlink_alloc;
+		}
 	}
 
 	mlxsw_core = devlink_priv(devlink);
@@ -1012,7 +1033,7 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 	if (err)
 		goto err_bus_init;
 
-	if (mlxsw_driver->resources_register) {
+	if (mlxsw_driver->resources_register && !reload) {
 		err = mlxsw_driver->resources_register(mlxsw_core);
 		if (err)
 			goto err_register_resources;
@@ -1038,9 +1059,11 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 	if (err)
 		goto err_emad_init;
 
-	err = devlink_register(devlink, mlxsw_bus_info->dev);
-	if (err)
-		goto err_devlink_register;
+	if (!reload) {
+		err = devlink_register(devlink, mlxsw_bus_info->dev);
+		if (err)
+			goto err_devlink_register;
+	}
 
 	err = mlxsw_hwmon_init(mlxsw_core, mlxsw_bus_info, &mlxsw_core->hwmon);
 	if (err)
@@ -1082,20 +1105,29 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 }
 EXPORT_SYMBOL(mlxsw_core_bus_device_register);
 
-void mlxsw_core_bus_device_unregister(struct mlxsw_core *mlxsw_core)
+void mlxsw_core_bus_device_unregister(struct mlxsw_core *mlxsw_core,
+				      bool reload)
 {
 	const char *device_kind = mlxsw_core->bus_info->device_kind;
 	struct devlink *devlink = priv_to_devlink(mlxsw_core);
 
+	if (mlxsw_core->reload_fail)
+		goto out;
+
 	if (mlxsw_core->driver->fini)
 		mlxsw_core->driver->fini(mlxsw_core);
 	mlxsw_thermal_fini(mlxsw_core->thermal);
-	devlink_unregister(devlink);
+	if (!reload)
+		devlink_unregister(devlink);
 	mlxsw_emad_fini(mlxsw_core);
 	kfree(mlxsw_core->lag.mapping);
 	mlxsw_ports_fini(mlxsw_core);
-	devlink_resources_clear(devlink);
+	if (!reload)
+		devlink_resources_clear(devlink);
 	mlxsw_core->bus->fini(mlxsw_core->bus_priv);
+	if (reload)
+		return;
+out:
 	devlink_free(devlink);
 	mlxsw_core_driver_put(device_kind);
 }
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.h b/drivers/net/ethernet/mellanox/mlxsw/core.h
index e44061d..1a62da5 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.h
@@ -66,8 +66,8 @@ void mlxsw_core_driver_unregister(struct mlxsw_driver *mlxsw_driver);
 
 int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
 				   const struct mlxsw_bus *mlxsw_bus,
-				   void *bus_priv);
-void mlxsw_core_bus_device_unregister(struct mlxsw_core *mlxsw_core);
+				   void *bus_priv, bool reload);
+void mlxsw_core_bus_device_unregister(struct mlxsw_core *mlxsw_core, bool reload);
 
 struct mlxsw_tx_info {
 	u8 local_port;
diff --git a/drivers/net/ethernet/mellanox/mlxsw/i2c.c b/drivers/net/ethernet/mellanox/mlxsw/i2c.c
index 12c3a44..d7a5562 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/i2c.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/i2c.c
@@ -539,7 +539,7 @@ static int mlxsw_i2c_probe(struct i2c_client *client,
 	mlxsw_i2c->dev = &client->dev;
 
 	err = mlxsw_core_bus_device_register(&mlxsw_i2c->bus_info,
-					     &mlxsw_i2c_bus, mlxsw_i2c);
+					     &mlxsw_i2c_bus, mlxsw_i2c, false);
 	if (err) {
 		dev_err(&client->dev, "Fail to register core bus\n");
 		return err;
@@ -557,7 +557,7 @@ static int mlxsw_i2c_remove(struct i2c_client *client)
 {
 	struct mlxsw_i2c *mlxsw_i2c = i2c_get_clientdata(client);
 
-	mlxsw_core_bus_device_unregister(mlxsw_i2c->core);
+	mlxsw_core_bus_device_unregister(mlxsw_i2c->core, false);
 	mutex_destroy(&mlxsw_i2c->cmd.lock);
 
 	return 0;
diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c
index da08f08..b69508d 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c
@@ -1737,7 +1737,7 @@ static int mlxsw_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	mlxsw_pci->id = id;
 
 	err = mlxsw_core_bus_device_register(&mlxsw_pci->bus_info,
-					     &mlxsw_pci_bus, mlxsw_pci);
+					     &mlxsw_pci_bus, mlxsw_pci, false);
 	if (err) {
 		dev_err(&pdev->dev, "cannot register bus device\n");
 		goto err_bus_device_register;
@@ -1765,7 +1765,7 @@ static void mlxsw_pci_remove(struct pci_dev *pdev)
 {
 	struct mlxsw_pci *mlxsw_pci = pci_get_drvdata(pdev);
 
-	mlxsw_core_bus_device_unregister(mlxsw_pci->core);
+	mlxsw_core_bus_device_unregister(mlxsw_pci->core, false);
 	mlxsw_pci_free_irq_vectors(mlxsw_pci);
 	iounmap(mlxsw_pci->hw_addr);
 	pci_release_regions(mlxsw_pci->pdev);
-- 
2.9.5

Re: [patch net-next RFC 1/7] devlink: Add support for resource abstraction

[David Ahern]

On 10/24/17 3:22 AM, Jiri Pirko wrote:
> diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
> index 0cbca96..9db1d70 100644
> --- a/include/uapi/linux/devlink.h
> +++ b/include/uapi/linux/devlink.h
> @@ -69,6 +69,8 @@ enum devlink_command {
>  	DEVLINK_CMD_DPIPE_ENTRIES_GET,
>  	DEVLINK_CMD_DPIPE_HEADERS_GET,
>  	DEVLINK_CMD_DPIPE_TABLE_COUNTERS_SET,
> +	DEVLINK_CMD_RESOURCE_SET,
> +	DEVLINK_CMD_RESOURCE_DUMP,
>  
>  	/* add new commands above here */
>  	__DEVLINK_CMD_MAX,
> @@ -201,6 +203,12 @@ enum devlink_attr {
>  	DEVLINK_ATTR_PAD,
>  
>  	DEVLINK_ATTR_ESWITCH_ENCAP_MODE,	/* u8 */
> +	DEVLINK_ATTR_RESOURCES,			/* nested */
> +	DEVLINK_ATTR_RESOURCE,			/* nested */
> +	DEVLINK_ATTR_RESOURCE_NAME,		/* string */
> +	DEVLINK_ATTR_RESOURCE_SIZE,		/* u64 */
> +	DEVLINK_ATTR_RESOURCE_SIZE_NEW,		/* u64 */
> +	DEVLINK_ATTR_RESOURCE_ID,		/* u64 */
>  
>  	/* add new attributes above here, update the policy in devlink.c */
>  

Where can I find the userspace patch to try out the RFC?

Re: [patch net-next RFC 7/7] mlxsw: core: Add support for reload

[David Ahern]

On 10/24/17 3:22 AM, Jiri Pirko wrote:
> @@ -980,7 +998,7 @@ static const struct devlink_ops mlxsw_devlink_ops = {
>  
>  int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
>  				   const struct mlxsw_bus *mlxsw_bus,
> -				   void *bus_priv)
> +				   void *bus_priv, bool reload)
>  {
>  	const char *device_kind = mlxsw_bus_info->device_kind;
>  	struct mlxsw_core *mlxsw_core;
> @@ -992,11 +1010,14 @@ int mlxsw_core_bus_device_register(const struct mlxsw_bus_info *mlxsw_bus_info,
>  	mlxsw_driver = mlxsw_core_driver_get(device_kind);
>  	if (!mlxsw_driver)
>  		return -EINVAL;
> -	alloc_size = sizeof(*mlxsw_core) + mlxsw_driver->priv_size;
> -	devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
> -	if (!devlink) {
> -		err = -ENOMEM;
> -		goto err_devlink_alloc;
> +
> +	if (!reload) {
> +		alloc_size = sizeof(*mlxsw_core) + mlxsw_driver->priv_size;
> +		devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
> +		if (!devlink) {
> +			err = -ENOMEM;
> +			goto err_devlink_alloc;
> +		}
>  	}
>  
>  	mlxsw_core = devlink_priv(devlink);

devlink is not set if reload is true, yet is used in the line above.

Re: [patch net-next RFC 5/7] mlxsw: Register KVD resources with devlink

[David Ahern]

On 10/24/17 3:22 AM, Jiri Pirko wrote:
> diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
> index 12b6ac4..565f94e 100644
> --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
> +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
> @@ -3864,7 +3864,6 @@ static void mlxsw_sp_fini(struct mlxsw_core *mlxsw_core)
>  	mlxsw_sp_fids_fini(mlxsw_sp);
>  	mlxsw_sp_kvdl_fini(mlxsw_sp);
>  }
> -
>  static const struct mlxsw_config_profile mlxsw_sp_config_profile = {
>  	.used_max_vepa_channels		= 1,
>  	.max_vepa_channels		= 0,

Why remove the spacer line?

> @@ -3897,6 +3896,95 @@ static const struct mlxsw_config_profile mlxsw_sp_config_profile = {
>  	.resource_query_enable		= 1,
>  };
>  
> +int mlxsw_sp_resource_kvd_size_validate(void *priv, u64 size)
> +{
> +	/* KVD total size cannot be set*/
> +	return -EINVAL;

-EOPNOTSUPP? There is nothing wrong with the value; it's the action that
is not supported.

> +}
> +
> +int mlxsw_sp_resource_kvd_sub_size_validate(void *priv, u64 size)
> +{
> +	const struct mlxsw_config_profile *profile;
> +
> +	profile = &mlxsw_sp_config_profile;
> +	if (size % profile->kvd_hash_granularity)
> +		return -EINVAL;

devlink needs extack capability to return messages to the user.

Re: [patch net-next RFC 1/7] devlink: Add support for resource abstraction

[Arkadi Sharshevsky]

On 10/25/2017 06:26 PM, David Ahern wrote:
> On 10/24/17 3:22 AM, Jiri Pirko wrote:
>> diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
>> index 0cbca96..9db1d70 100644
>> --- a/include/uapi/linux/devlink.h
>> +++ b/include/uapi/linux/devlink.h
>> @@ -69,6 +69,8 @@ enum devlink_command {
>>  	DEVLINK_CMD_DPIPE_ENTRIES_GET,
>>  	DEVLINK_CMD_DPIPE_HEADERS_GET,
>>  	DEVLINK_CMD_DPIPE_TABLE_COUNTERS_SET,
>> +	DEVLINK_CMD_RESOURCE_SET,
>> +	DEVLINK_CMD_RESOURCE_DUMP,
>>  
>>  	/* add new commands above here */
>>  	__DEVLINK_CMD_MAX,
>> @@ -201,6 +203,12 @@ enum devlink_attr {
>>  	DEVLINK_ATTR_PAD,
>>  
>>  	DEVLINK_ATTR_ESWITCH_ENCAP_MODE,	/* u8 */
>> +	DEVLINK_ATTR_RESOURCES,			/* nested */
>> +	DEVLINK_ATTR_RESOURCE,			/* nested */
>> +	DEVLINK_ATTR_RESOURCE_NAME,		/* string */
>> +	DEVLINK_ATTR_RESOURCE_SIZE,		/* u64 */
>> +	DEVLINK_ATTR_RESOURCE_SIZE_NEW,		/* u64 */
>> +	DEVLINK_ATTR_RESOURCE_ID,		/* u64 */
>>  
>>  	/* add new attributes above here, update the policy in devlink.c */
>>  
> 
> Where can I find the userspace patch to try out the RFC?
> 

Hi David,

Thanks for the review. I will send the updated version this week
with the userspace part.