Skip to contentSkip to conversation

ethtool netlink interface, part 1

PATCH Superseded Michal Kubecek 6 participants 109 messages Mar 25, 2019 – Mar 29, 2019

Subsystems (5)

v5 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v5 cover ethtool netlink interface, part 1

This is first part of netlink based alternative userspace interface for ethtool. The aim is to address some long known issues with the ioctl interface, mainly lack of extensibility, raciness, limited error reporting and absence of notifications. The goal is to allow userspace ethtool utility to provide all features it currently does but without using the ioctl interface.

The interface uses generic netlink family "ethtool"; it provides multicast group "monitor" which is used for notifications. Documentation for the interface is in Documentation/networking/ethtool-netlink.txt file. The netlink interface is optional, it is built when CONFIG_ETHTOOL_NETLINK (bool) option is enabled.

There are three types of requests distinguished by prefix "GET" (query for information), "SET" (modify parameters) and "ACT" (perform an action). The GET_* and SET_* messages are paired even if the data they deal with is read only because SET_* message is also used for reply to corresponding GET_* message. Notifications on parameter changes are also sent as SET_* messages in the same format as reply to corresponding GET request.

Basic concepts:

- make extensions easier not only by allowing new attributes but also by imposing as few artificial limits as possible, e.g. by using arbitrary size bit sets for most bitmap attributes or by not using fixed size strings
- use extack for error reporting and warnings
- send netlink notifications on changes (even if they were done using the ioctl interfaces)
- avoid the racy read/modify/write cycle between kernel and userspace by sending only attributes which userspace wants to change; there is still a read/modify/write cycle between generic kernel code and ethtool_ops handler in NIC driver but it is only in kernel and under a lock
- reduce the number of value lists that need to be kept in sync between kernel and userspace (e.g. recognized link modes)
- where feasible, allow dump requests to query specific information for all network devices
- as the lack of extensibility of the ioctl interface led to having too many commands, group some of these together to one netlink message but allow querying only part(s) of the information (using "info mask" bitmap) and modifying only some of the parameters (by providing only some attributes)
- as parsing and generating netlink messages is more complicated than simply copying data structures between userspace API and ethtool_ops handlers (which most ioctl commands do), split the code into multiple files in net/ethtool directory; move net/core/ethtool.c also to this directory and rename it to ioctl.c

The full (work in progress) series, together with the (userspace) ethtool counterpart can be found at https://github.com/mkubecek/ethnl

Changes between v4 and v5:

- do not panic on failed initialization, only WARN()

Main changes between RFC v3 and v4:

- use more kerneldoc style comments
- strict attribute policy checking
- use macros for tables of link mode names and parameters
- provide permanent hardware address in rtnetlink
- coding style cleanup
- split too long patches, reorder
- wrap more ETHA_SETTINGS_* attributes in nests
- add also some SET_* implementation into submitted part

Main changes between RFC v2 and RFC v3:

- do not allow building as a module (no netdev notifiers needed)
- drop some obsolete fields
- add permanent hw address, timestamping and private flags support
- rework bitset handling to get rid of variable length arrays
- notify monitor on device renames
- restructure GET_SETTINGS/SET_SETTINGS messages
- split too long patches and submit only first part of the series

Main changes between RFC v1 and RFC v2:

- support dumps for all "get" requests
- provide notifications for changes related to supported request types
- support getting string sets (both global and per device)
- support getting/setting device features
- get rid of family specific header, everything passed as attributes
- split netlink code into multiple files in net/ethtool/ directory

ToDo / open questions:

- some features provided by ethtool would rather belong to devlink (and some are already superseded by devlink); however, only few drivers provide devlink interface at the moment and as recent discussion on flashing revealed, we cannot rely on devlink's presence

- while the netlink interface allows easy future extensions, ethtool_ops interface does not; some settings could be implemented using tunables and accessed via relevant netlink messages (as well as tunables) from userspace but in the long term, something better will be needed

- currently, all communication with drivers via ethtool_ops is done under RTNL as this is what ioctl interface does and likely many ethtool_ops handlers rely on that; if we are going to rework ethtool_ops in the future ("phase two"), it would be nice to get rid of it

- ethtool_ops should pass extack pointer to allow drivers more meaningful error reporting; it's not clear, however, how to pass information about offending attribute

- notifications are sent whenever a change is done via netlink API or ioctl API and for netdev features also whenever they are updated using netdev_change_features(); it would be desirable to notify also about link state and negotiation result (speed/duplex and partner link modes) but it would be more tricky

Michal Kubecek (22):
rtnetlink: provide permanent hardware address in RTM_NEWLINK
netlink: introduce nla_put_bitfield32()
netlink: add strict version of nla_parse_nested()
ethtool: move to its own directory
ethtool: introduce ethtool netlink interface
ethtool: helper functions for netlink interface
ethtool: netlink bitset handling
ethtool: support for netlink notifications
ethtool: implement EVENT notifications
ethtool: generic handlers for GET requests
ethtool: move string arrays into common file
ethtool: provide string sets with GET_STRSET request
ethtool: provide driver/device information in GET_INFO request
ethtool: provide timestamping information in GET_INFO request
ethtool: provide link mode names as a string set
ethtool: provide link settings and link modes in GET_SETTINGS request
ethtool: set link settings and link modes with SET_SETTINGS request
ethtool: provide link state in GET_SETTINGS request
ethtool: provide WoL information in GET_SETTINGS request
ethtool: set WoL settings with SET_SETTINGS request
ethtool: provide message level in GET_SETTINGS request
ethtool: set message level with SET_SETTINGS request

 Documentation/networking/ethtool-netlink.txt | 458 +++++++++++
 include/linux/ethtool.h                      |   4 +
 include/linux/ethtool_netlink.h              |  17 +
 include/linux/netdevice.h                    |  14 +
 include/net/netlink.h                        |  24 +
 include/uapi/linux/ethtool.h                 |  10 +
 include/uapi/linux/ethtool_netlink.h         | 274 ++++++
 include/uapi/linux/if_link.h                 |   1 +
 include/uapi/linux/net_tstamp.h              |  13 +
 net/Kconfig                                  |   8 +
 net/Makefile                                 |   2 +-
 net/core/Makefile                            |   2 +-
 net/core/rtnetlink.c                         |   4 +
 net/ethtool/Makefile                         |   7 +
 net/ethtool/bitset.c                         | 597 ++++++++++++++
 net/ethtool/bitset.h                         |  40 +
 net/ethtool/common.c                         | 225 +++++
 net/ethtool/common.h                         |  27 +
 net/ethtool/info.c                           | 296 +++++++
 net/{core/ethtool.c => ethtool/ioctl.c}      | 237 +-----
 net/ethtool/netlink.c                        | 708 ++++++++++++++++
 net/ethtool/netlink.h                        | 296 +++++++
 net/ethtool/settings.c                       | 823 +++++++++++++++++++
 net/ethtool/strset.c                         | 471 +++++++++++
 24 files changed, 4355 insertions(+), 203 deletions(-)
 create mode 100644 Documentation/networking/ethtool-netlink.txt
 create mode 100644 include/linux/ethtool_netlink.h
 create mode 100644 include/uapi/linux/ethtool_netlink.h
 create mode 100644 net/ethtool/Makefile
 create mode 100644 net/ethtool/bitset.c
 create mode 100644 net/ethtool/bitset.h
 create mode 100644 net/ethtool/common.c
 create mode 100644 net/ethtool/common.h
 create mode 100644 net/ethtool/info.c
 rename net/{core/ethtool.c => ethtool/ioctl.c} (91%)
 create mode 100644 net/ethtool/netlink.c
 create mode 100644 net/ethtool/netlink.h
 create mode 100644 net/ethtool/settings.c
 create mode 100644 net/ethtool/strset.c
Signature · 1 line
2.21.0
#1 lore raw

PATCH v5 1/22 rtnetlink: provide permanent hardware address in RTM_NEWLINK

Permanent hardware address of a network device was traditionally provided via ethtool ioctl interface but as Jiri Pirko pointed out in a review of ethtool netlink interface, rtnetlink is much more suitable for it so let's add it to the RTM_NEWLINK message.

As permanent address is not modifiable, reject userspace requests containing IFLA_PERM_ADDRESS attribute.

Note: we already provide permanent hardware address for bond slaves; unfortunately we cannot drop that attribute for backward compatibility reasons.

---
 include/uapi/linux/if_link.h | 1 +
 net/core/rtnetlink.c         | 4 ++++
 2 files changed, 5 insertions(+)
include/uapi/linux/if_link.h +1 -0
index 5b225ff63b48..351ef746b8b0 100644
@@ -167,6 +167,7 @@ enum {
 	IFLA_NEW_IFINDEX,
 	IFLA_MIN_MTU,
 	IFLA_MAX_MTU,
+	IFLA_PERM_ADDRESS,
 	__IFLA_MAX
 };
 
net/core/rtnetlink.c +4 -0
index a51cab95ba64..a72e8f4d777b 100644
@@ -1026,6 +1026,7 @@ static noinline size_t if_nlmsg_size(const struct net_device *dev,
 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
 	       + nla_total_size(4)  /* IFLA_MIN_MTU */
 	       + nla_total_size(4)  /* IFLA_MAX_MTU */
+	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
 	       + 0;
 }
 
@@ -1683,6 +1684,8 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
 		goto nla_put_failure;
 
+	if (nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
+		goto nla_put_failure;
 
 	rcu_read_lock();
 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
@@ -1742,6 +1745,7 @@ static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
 	[IFLA_MIN_MTU]		= { .type = NLA_U32 },
 	[IFLA_MAX_MTU]		= { .type = NLA_U32 },
+	[IFLA_PERM_ADDRESS]	= { .type = NLA_REJECT },
 };
 
 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
Signature · 1 line
2.21.0
#2 lore raw

Re: [PATCH net-next v5 01/22] rtnetlink: provide permanent hardware address in RTM_NEWLINK

Quoted 43 lines · Mon, Mar 25, 2019 at 06:07:57PM CET, [email protected] wrote:

Permanent hardware address of a network device was traditionally provided via ethtool ioctl interface but as Jiri Pirko pointed out in a review of ethtool netlink interface, rtnetlink is much more suitable for it so let's add it to the RTM_NEWLINK message.

As permanent address is not modifiable, reject userspace requests containing IFLA_PERM_ADDRESS attribute.

Note: we already provide permanent hardware address for bond slaves; unfortunately we cannot drop that attribute for backward compatibility reasons.

---
include/uapi/linux/if_link.h | 1 +
net/core/rtnetlink.c | 4 ++++

2 files changed, 5 insertions(+)
include/uapi/linux/if_link.h +1 -0
index 5b225ff63b48..351ef746b8b0 100644
@@ -167,6 +167,7 @@ enum {
 	IFLA_NEW_IFINDEX,
 	IFLA_MIN_MTU,
 	IFLA_MAX_MTU,
+	IFLA_PERM_ADDRESS,
 	__IFLA_MAX
 };
 
net/core/rtnetlink.c +1 -0
index a51cab95ba64..a72e8f4d777b 100644
@@ -1026,6 +1026,7 @@ static noinline size_t if_nlmsg_size(const struct net_device *dev,
 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
 	       + nla_total_size(4)  /* IFLA_MIN_MTU */
 	       + nla_total_size(4)  /* IFLA_MAX_MTU */
+	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
 	       + 0;
 }
 

net/core/rtnetlink.c static int rtnl_fill_ifinfo(struct sk_buff *skb, · line 1688

Quoted 5 lines
net/core/rtnetlink.c +2 -0
@@ -1683,6 +1684,8 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
 		goto nla_put_failure;
 
+	if (nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
+		goto nla_put_failure;

I don't think we should put permaddr if driver did not set it. 2 solutions:
1) provide a helper that driver will use to set the perm_addr. This helper sets a "valid bit". Then you only put IFLA_PERM_ADDRESS in case the "valid bit" is set.
2) Assuming that no driver would set permaddr to all zeroes, don't put IFLA_PERM_ADDRESS in case permadd is all zeroes.

Quoted 14 lines
net/core/rtnetlink.c +1 -0
 	rcu_read_lock();
 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
@@ -1742,6 +1745,7 @@ static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
 	[IFLA_MIN_MTU]		= { .type = NLA_U32 },
 	[IFLA_MAX_MTU]		= { .type = NLA_U32 },
+	[IFLA_PERM_ADDRESS]	= { .type = NLA_REJECT },
 };
 
 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
Signature · 1 line
2.21.0
#3 lore raw

Re: [PATCH net-next v5 01/22] rtnetlink: provide permanent hardware address in RTM_NEWLINK

Quoted 8 lines · Jiri Pirko

I don't think we should put permaddr if driver did not set it. 2 solutions:
1) provide a helper that driver will use to set the perm_addr. This helper sets a "valid bit". Then you only put IFLA_PERM_ADDRESS in case the "valid bit" is set.
2) Assuming that no driver would set permaddr to all zeroes, don't put IFLA_PERM_ADDRESS in case permadd is all zeroes.

I already replied to similar suggestion in v4 discussion:

http://patchwork.ozlabs.org/patch/1060164/#2117512

But I don't have really strong opinion about this. The problem with not being able to distinguish between "no/unknown permanent address" and "old kernel not providing the information" is going to become less important over time.

Michal

#4 lore raw

Re: [PATCH net-next v5 01/22] rtnetlink: provide permanent hardware address in RTM_NEWLINK

Quoted 10 lines · Tue, Mar 26, 2019 at 11:31:15AM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko

I don't think we should put permaddr if driver did not set it. 2 solutions:
1) provide a helper that driver will use to set the perm_addr. This helper sets a "valid bit". Then you only put IFLA_PERM_ADDRESS in case the "valid bit" is set.
2) Assuming that no driver would set permaddr to all zeroes, don't put IFLA_PERM_ADDRESS in case permadd is all zeroes.

Quoted 8 lines

I already replied to similar suggestion in v4 discussion:

http://patchwork.ozlabs.org/patch/1060164/#2117512

But I don't have really strong opinion about this. The problem with not being able to distinguish between "no/unknown permanent address" and "old kernel not providing the information" is going to become less important over time.

If the attribute is sent to userspace, it should mean the permaddr is there and valid.

Quoted 2 lines

Michal

maintainer
#5 lore raw

Re: [PATCH net-next v5 01/22] rtnetlink: provide permanent hardware address in RTM_NEWLINK

In reply to Michal Kubecek

From: Michal Kubecek <[email protected]>
Date: Mon, 25 Mar 2019 18:07:57 +0100 (CET)

net/core/rtnetlink.c static int rtnl_fill_ifinfo(struct sk_buff *skb, · line 1691

Quoted 9 lines
net/core/rtnetlink.c +2 -0
@@ -1683,6 +1684,8 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
 		goto nla_put_failure;
 
+	if (nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
+		goto nla_put_failure;
 
 	rcu_read_lock();
 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))

I guess, as Jiri says, we really do have to check if the driver actually set this before pushing the attribute out to userspace.

But seriously, you chould just check for dev_addr_len zeros.

All zeros is not a valid link address on any link type.

#6 lore raw

PATCH v5 2/22 netlink: introduce nla_put_bitfield32()

In reply to Michal Kubecek

Similar to other data types, this helper puts NLA_BITFIELD32 attribute into a netlink message. It takes separate value and selector arguments, if there is a struct nla_bitfield32 already, one can use nla_put().

---
 include/net/netlink.h | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
include/net/netlink.h +15 -0
index 23f27b0b3cef..bc0497076bec 100644
@@ -1211,6 +1211,21 @@ static inline int nla_put_in6_addr(struct sk_buff *skb, int attrtype,
 	return nla_put(skb, attrtype, sizeof(*addr), addr);
 }
 
+/**
+ * nla_put_bitfield32 - Add a bitfield32 value/selector attribute to
+ * a socket buffer
+ * @skb: socket buffer to add attribute to
+ * @value: 32-bit value bitmap
+ * @selector: 32-bit selector bitmap
+ */
+static inline int nla_put_bitfield32(struct sk_buff *skb, int attrtype,
+				     u32 value, u32 selector)
+{
+	struct nla_bitfield32 tmp = { .value = value, .selector = selector };
+
+	return nla_put(skb, attrtype, sizeof(tmp), &tmp);
+}
+
 /**
  * nla_get_u32 - return payload of u32 attribute
  * @nla: u32 netlink attribute
Signature · 1 line
2.21.0
Acked-by on v5 2/22
#7 lore raw

Re: [PATCH net-next v5 02/22] netlink: introduce nla_put_bitfield32()

Quoted 5 lines · Mon, Mar 25, 2019 at 06:08:00PM CET, [email protected] wrote:

Similar to other data types, this helper puts NLA_BITFIELD32 attribute into a netlink message. It takes separate value and selector arguments, if there is a struct nla_bitfield32 already, one can use nla_put().

Reviewed-by on v5 2/22
#8 lore raw

Re: [PATCH net-next v5 02/22] netlink: introduce nla_put_bitfield32()

In reply to Michal Kubecek
Quoted 5 lines · Michal Kubecek

Similar to other data types, this helper puts NLA_BITFIELD32 attribute into a netlink message. It takes separate value and selector arguments, if there is a struct nla_bitfield32 already, one can use nla_put().

Signature · 1 line
Florian
#9 lore raw

PATCH v5 3/22 netlink: add strict version of nla_parse_nested()

In reply to Michal Kubecek

Similar to nla_parse_strict() and nlmsg_parse_strict(), add also nla_parse_nested_strict() as a version of nla_parse_nested() with strict policy checking.

---
 include/net/netlink.h | 9 +++++++++
 1 file changed, 9 insertions(+)
include/net/netlink.h +9 -0
index bc0497076bec..723139637ba4 100644
@@ -913,6 +913,15 @@ static inline int nla_parse_nested(struct nlattr *tb[], int maxtype,
 			 extack);
 }
 
+static inline int nla_parse_nested_strict(struct nlattr *tb[], int maxtype,
+					  const struct nlattr *nla,
+					  const struct nla_policy *policy,
+					  struct netlink_ext_ack *extack)
+{
+	return nla_parse_strict(tb, maxtype, nla_data(nla), nla_len(nla),
+				policy, extack);
+}
+
 /**
  * nla_put_u8 - Add a u8 netlink attribute to a socket buffer
  * @skb: socket buffer to add attribute to
Signature · 1 line
2.21.0
#12 lore raw

PATCH v5 4/22 ethtool: move to its own directory

In reply to Michal Kubecek

The ethtool netlink interface is going to be split into multiple files so that it will be more convenient to put all of them in a separate directory net/ethtool. Start by moving current ethtool.c with ioctl interface into this directory and renaming it to ioctl.c.

---
 net/Makefile                            | 2 +-
 net/core/Makefile                       | 2 +-
 net/ethtool/Makefile                    | 3 +++
 net/{core/ethtool.c => ethtool/ioctl.c} | 0
 4 files changed, 5 insertions(+), 2 deletions(-)
 create mode 100644 net/ethtool/Makefile
 rename net/{core/ethtool.c => ethtool/ioctl.c} (100%)
net/Makefile +1 -1
index 449fc0b221f8..848303d98d3d 100644
@@ -13,7 +13,7 @@ obj-$(CONFIG_NET)		+= $(tmp-y)
 
 # LLC has to be linked before the files in net/802/
 obj-$(CONFIG_LLC)		+= llc/
-obj-$(CONFIG_NET)		+= ethernet/ 802/ sched/ netlink/ bpf/
+obj-$(CONFIG_NET)		+= ethernet/ 802/ sched/ netlink/ bpf/ ethtool/
 obj-$(CONFIG_NETFILTER)		+= netfilter/
 obj-$(CONFIG_INET)		+= ipv4/
 obj-$(CONFIG_TLS)		+= tls/
net/core/Makefile +1 -1
index f97d6254e564..9ddc2f3ecd97 100644
@@ -8,7 +8,7 @@ obj-y := sock.o request_sock.o skbuff.o datagram.o stream.o scm.o \
 
 obj-$(CONFIG_SYSCTL) += sysctl_net_core.o
 
-obj-y		     += dev.o ethtool.o dev_addr_lists.o dst.o netevent.o \
+obj-y		     += dev.o dev_addr_lists.o dst.o netevent.o \
 			neighbour.o rtnetlink.o utils.o link_watch.o filter.o \
 			sock_diag.o dev_ioctl.o tso.o sock_reuseport.o \
 			fib_notifier.o xdp.o flow_offload.o
net/ethtool/Makefile new file +3 -0
new file mode 100644
index 000000000000..3ebfab2bca66
@@ -0,0 +1,3 @@
+# SPDX-License-Identifier: GPL-2.0
+
+obj-y		+= ioctl.o
net/ethtool/ioctl.c renamed from net/core/ethtool.c +0 -0
similarity index 100%
rename from net/core/ethtool.c
rename to net/ethtool/ioctl.c
Signature · 1 line
2.21.0
Acked-by on v5 4/22
#13 lore raw

Re: [PATCH net-next v5 04/22] ethtool: move to its own directory

Quoted 6 lines · Mon, Mar 25, 2019 at 06:08:06PM CET, [email protected] wrote:

The ethtool netlink interface is going to be split into multiple files so that it will be more convenient to put all of them in a separate directory net/ethtool. Start by moving current ethtool.c with ioctl interface into this directory and renaming it to ioctl.c.

Reviewed-by on v5 4/22
#14 lore raw

Re: [PATCH net-next v5 04/22] ethtool: move to its own directory

In reply to Michal Kubecek
Quoted 6 lines · Michal Kubecek

The ethtool netlink interface is going to be split into multiple files so that it will be more convenient to put all of them in a separate directory net/ethtool. Start by moving current ethtool.c with ioctl interface into this directory and renaming it to ioctl.c.

Signature · 1 line
Florian
#15 lore raw

PATCH v5 5/22 ethtool: introduce ethtool netlink interface

In reply to Michal Kubecek

Basic genetlink and init infrastructure for the netlink interface, register genetlink family "ethtool". Introduce CONFIG_ETHTOOL_NETLINK Kconfig option. Add interface description into Documentation/networking.

---
 Documentation/networking/ethtool-netlink.txt | 168 +++++++++++++++++++
 include/linux/ethtool_netlink.h              |   9 +
 include/uapi/linux/ethtool_netlink.h         |  25 +++
 net/Kconfig                                  |   8 +
 net/ethtool/Makefile                         |   6 +-
 net/ethtool/netlink.c                        |  34 ++++
 net/ethtool/netlink.h                        |  12 ++
 7 files changed, 261 insertions(+), 1 deletion(-)
 create mode 100644 Documentation/networking/ethtool-netlink.txt
 create mode 100644 include/linux/ethtool_netlink.h
 create mode 100644 include/uapi/linux/ethtool_netlink.h
 create mode 100644 net/ethtool/netlink.c
 create mode 100644 net/ethtool/netlink.h
Documentation/networking/ethtool-netlink.txt new file +168 -0
new file mode 100644
index 000000000000..377d64c9b7fa
@@ -0,0 +1,168 @@
+                        Netlink interface for ethtool
+                        =============================
+
+
+Basic information
+-----------------
+
+Netlink interface for ethtool uses generic netlink family "ethtool" (userspace
+application should use macros ETHTOOL_GENL_NAME and ETHTOOL_GENL_VERSION
+defined in <linux/ethtool_netlink.h> uapi header). This family does not use
+a specific header, all information in requests and replies is passed using
+netlink attributes.
+
+In requests, device can be identified by ifindex or by name; if both are used,
+they must match. In replies, kernel fills both. The meaning of flags,
+info_mask and index fields depends on request type.
+
+The ethtool netlink interface uses extended ACK for error and warning
+reporting, userspace application developers are encouraged to make these
+messages available to user in a suitable way.
+
+Requests can be divided into three categories: "get" (retrieving information),
+"set" (setting parameters) and "action" (invoking an action).
+
+All "set" and "action" type requests require admin privileges (CAP_NET_ADMIN
+in the namespace). Most "get" type request are allowed for anyone but there
+are exceptions (where the response contains sensitive information). In some
+cases, the request as such is allowed for anyone but unprivileged users have
+attributes with sensitive information (e.g. wake-on-lan password) omitted.
+
+
+Conventions
+-----------
+
+Attributes which represent a boolean value usually use u8 type so that we can
+distinguish three states: "on", "off" and "not present" (meaning the
+information is not available in "get" requests or value is not to be changed
+in "set" requests). For these attributes, the "true" value should be passed as
+number 1 but any non-zero value should be understood as "true" by recipient.
+
+Some request types allow passing an attribute named ETHA_*_INFOMASK with
+a bitmask telling kernel that we are only interested in some parts of the
+information. If info mask is omitted, all available information is returned.
+Meaning of info mask bits depends on request type and is listed below.
+
+
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain
+
+    ETHA_DEV_INDEX	(u32)		device ifindex
+    ETHA_DEV_NAME	(string)	device name
+
+In device related requests, one of these is sufficient; if both are used, they
+must match (i.e. identify the same device). In device related replies both are
+provided by kernel. In dump requests, device is not specified and kernel
+replies with one message per network device (only those for which the request
+is supported).
+
+
+List of message types
+---------------------
+
+All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
+to indicate the type.
+
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;
+the same message can be also used to set (some of) the parameters, except for
+messages marked as "response only" in the table above. "Get" messages with
+NLM_F_DUMP flags and no device identification dump the information for all
+devices supporting the request.
+
+Later sections describe the format and semantics of these request messages.
+
+
+Request translation
+-------------------
+
+The following table maps iosctl commands to netlink commands providing their
+functionality. Entries with "n/a" in right column are commands which do not
+have their netlink replacement yet.
+
+ioctl command			netlink command
+---------------------------------------------------------------------
+ETHTOOL_GSET			n/a
+ETHTOOL_SSET			n/a
+ETHTOOL_GDRVINFO		n/a
+ETHTOOL_GREGS			n/a
+ETHTOOL_GWOL			n/a
+ETHTOOL_SWOL			n/a
+ETHTOOL_GMSGLVL			n/a
+ETHTOOL_SMSGLVL			n/a
+ETHTOOL_NWAY_RST		n/a
+ETHTOOL_GLINK			n/a
+ETHTOOL_GEEPROM			n/a
+ETHTOOL_SEEPROM			n/a
+ETHTOOL_GCOALESCE		n/a
+ETHTOOL_SCOALESCE		n/a
+ETHTOOL_GRINGPARAM		n/a
+ETHTOOL_SRINGPARAM		n/a
+ETHTOOL_GPAUSEPARAM		n/a
+ETHTOOL_SPAUSEPARAM		n/a
+ETHTOOL_GRXCSUM			n/a
+ETHTOOL_SRXCSUM			n/a
+ETHTOOL_GTXCSUM			n/a
+ETHTOOL_STXCSUM			n/a
+ETHTOOL_GSG			n/a
+ETHTOOL_SSG			n/a
+ETHTOOL_TEST			n/a
+ETHTOOL_GSTRINGS		n/a
+ETHTOOL_PHYS_ID			n/a
+ETHTOOL_GSTATS			n/a
+ETHTOOL_GTSO			n/a
+ETHTOOL_STSO			n/a
+ETHTOOL_GPERMADDR		rtnetlink RTM_GETLINK
+ETHTOOL_GUFO			n/a
+ETHTOOL_SUFO			n/a
+ETHTOOL_GGSO			n/a
+ETHTOOL_SGSO			n/a
+ETHTOOL_GFLAGS			n/a
+ETHTOOL_SFLAGS			n/a
+ETHTOOL_GPFLAGS			n/a
+ETHTOOL_SPFLAGS			n/a
+ETHTOOL_GRXFH			n/a
+ETHTOOL_SRXFH			n/a
+ETHTOOL_GGRO			n/a
+ETHTOOL_SGRO			n/a
+ETHTOOL_GRXRINGS		n/a
+ETHTOOL_GRXCLSRLCNT		n/a
+ETHTOOL_GRXCLSRULE		n/a
+ETHTOOL_GRXCLSRLALL		n/a
+ETHTOOL_SRXCLSRLDEL		n/a
+ETHTOOL_SRXCLSRLINS		n/a
+ETHTOOL_FLASHDEV		n/a
+ETHTOOL_RESET			n/a
+ETHTOOL_SRXNTUPLE		n/a
+ETHTOOL_GRXNTUPLE		n/a
+ETHTOOL_GSSET_INFO		n/a
+ETHTOOL_GRXFHINDIR		n/a
+ETHTOOL_SRXFHINDIR		n/a
+ETHTOOL_GFEATURES		n/a
+ETHTOOL_SFEATURES		n/a
+ETHTOOL_GCHANNELS		n/a
+ETHTOOL_SCHANNELS		n/a
+ETHTOOL_SET_DUMP		n/a
+ETHTOOL_GET_DUMP_FLAG		n/a
+ETHTOOL_GET_DUMP_DATA		n/a
+ETHTOOL_GET_TS_INFO		n/a
+ETHTOOL_GMODULEINFO		n/a
+ETHTOOL_GMODULEEEPROM		n/a
+ETHTOOL_GEEE			n/a
+ETHTOOL_SEEE			n/a
+ETHTOOL_GRSSH			n/a
+ETHTOOL_SRSSH			n/a
+ETHTOOL_GTUNABLE		n/a
+ETHTOOL_STUNABLE		n/a
+ETHTOOL_GPHYSTATS		n/a
+ETHTOOL_PERQUEUE		n/a
+ETHTOOL_GLINKSETTINGS		n/a
+ETHTOOL_SLINKSETTINGS		n/a
+ETHTOOL_PHY_GTUNABLE		n/a
+ETHTOOL_PHY_STUNABLE		n/a
+ETHTOOL_GFECPARAM		n/a
+ETHTOOL_SFECPARAM		n/a
include/linux/ethtool_netlink.h new file +9 -0
new file mode 100644
index 000000000000..0412adb4f42f
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _LINUX_ETHTOOL_NETLINK_H_
+#define _LINUX_ETHTOOL_NETLINK_H_
+
+#include <uapi/linux/ethtool_netlink.h>
+#include <linux/ethtool.h>
+
+#endif /* _LINUX_ETHTOOL_NETLINK_H_ */
include/uapi/linux/ethtool_netlink.h new file +25 -0
new file mode 100644
index 000000000000..6aa267451542
@@ -0,0 +1,25 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/*
+ * include/uapi/linux/ethtool_netlink.h - netlink interface for ethtool
+ *
+ * See Documentation/networking/ethtool-netlink.txt in kernel source tree for
+ * doucumentation of the interface.
+ */
+
+#ifndef _UAPI_LINUX_ETHTOOL_NETLINK_H_
+#define _UAPI_LINUX_ETHTOOL_NETLINK_H_
+
+#include <linux/ethtool.h>
+
+enum {
+	ETHNL_CMD_NOOP,
+
+	__ETHNL_CMD_CNT,
+	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
+};
+
+/* generic netlink info */
+#define ETHTOOL_GENL_NAME "ethtool"
+#define ETHTOOL_GENL_VERSION 1
+
+#endif /* _UAPI_LINUX_ETHTOOL_NETLINK_H_ */
net/Kconfig +8 -0
index 3e8fdd688329..75c600b45775 100644
@@ -448,6 +448,14 @@ config FAILOVER
 	  migration of VMs with direct attached VFs by failing over to the
 	  paravirtual datapath when the VF is unplugged.
 
+config ETHTOOL_NETLINK
+	bool "Netlink interface for ethtool"
+	default y
+	help
+	  An alternative userspace interface for ethtool based on generic
+	  netlink. It provides better extensibility and some new features,
+	  e.g. notification messages.
+
 endif   # if NET
 
 # Used by archs to tell that they support BPF JIT compiler plus which flavour.
net/ethtool/Makefile +5 -1
index 3ebfab2bca66..f30e0da88be5 100644
@@ -1,3 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 
-obj-y		+= ioctl.o
+obj-y				+= ioctl.o
+
+obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
+
+ethtool_nl-y	:= netlink.o
net/ethtool/netlink.c new file +34 -0
new file mode 100644
index 000000000000..85dd6dac71a2
@@ -0,0 +1,34 @@
+// SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
+
+#include <linux/ethtool_netlink.h>
+#include "netlink.h"
+
+/* genetlink setup */
+
+static const struct genl_ops ethtool_genl_ops[] = {
+};
+
+struct genl_family ethtool_genl_family = {
+	.hdrsize	= 0,
+	.name		= ETHTOOL_GENL_NAME,
+	.version	= ETHTOOL_GENL_VERSION,
+	.netnsok	= true,
+	.parallel_ops	= true,
+	.ops		= ethtool_genl_ops,
+	.n_ops		= ARRAY_SIZE(ethtool_genl_ops),
+};
+
+/* module setup */
+
+static int __init ethnl_init(void)
+{
+	int ret;
+
+	ret = genl_register_family(&ethtool_genl_family);
+	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))
+		return ret;
+
+	return 0;
+}
+
+subsys_initcall(ethnl_init);
net/ethtool/netlink.h new file +12 -0
new file mode 100644
index 000000000000..63063b582ca2
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _NET_ETHTOOL_NETLINK_H
+#define _NET_ETHTOOL_NETLINK_H
+
+#include <linux/ethtool_netlink.h>
+#include <linux/netdevice.h>
+#include <net/genetlink.h>
+
+extern struct genl_family ethtool_genl_family;
+
+#endif /* _NET_ETHTOOL_NETLINK_H */
Signature · 1 line
2.21.0
#16 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 42 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:

Basic genetlink and init infrastructure for the netlink interface, register genetlink family "ethtool". Introduce CONFIG_ETHTOOL_NETLINK Kconfig option. Add interface description into Documentation/networking.

---

Documentation/networking/ethtool-netlink.txt | 168 +++++++++++++++++++
include/linux/ethtool_netlink.h              |   9 +
include/uapi/linux/ethtool_netlink.h         |  25 +++
net/Kconfig                                  |   8 +
net/ethtool/Makefile                         |   6 +-
net/ethtool/netlink.c                        |  34 ++++
net/ethtool/netlink.h                        |  12 ++
7 files changed, 261 insertions(+), 1 deletion(-)

create mode 100644 Documentation/networking/ethtool-netlink.txt create mode 100644 include/linux/ethtool_netlink.h create mode 100644 include/uapi/linux/ethtool_netlink.h create mode 100644 net/ethtool/netlink.c
create mode 100644 net/ethtool/netlink.h

Documentation/networking/ethtool-netlink.txt new file +16 -0
new file mode 100644
index 000000000000..377d64c9b7fa
@@ -0,0 +1,168 @@
+                        Netlink interface for ethtool
+                        =============================
+
+
+Basic information
+-----------------
+
+Netlink interface for ethtool uses generic netlink family "ethtool" (userspace
+application should use macros ETHTOOL_GENL_NAME and ETHTOOL_GENL_VERSION
+defined in <linux/ethtool_netlink.h> uapi header). This family does not use
+a specific header, all information in requests and replies is passed using
+netlink attributes.
+
+In requests, device can be identified by ifindex or by name; if both are used,
+they must match. In replies, kernel fills both. The meaning of flags,
+info_mask and index fields depends on request type.

Documentation/networking/ethtool-netlink.txt line 26

Quoted 10 lines
Documentation/networking/ethtool-netlink.txt new file +10 -0
+
+The ethtool netlink interface uses extended ACK for error and warning
+reporting, userspace application developers are encouraged to make these
+messages available to user in a suitable way.
+
+Requests can be divided into three categories: "get" (retrieving information),
+"set" (setting parameters) and "action" (invoking an action).
+
+All "set" and "action" type requests require admin privileges (CAP_NET_ADMIN
+in the namespace). Most "get" type request are allowed for anyone but there

s/request/requests/

Quoted 15 lines
Documentation/networking/ethtool-netlink.txt new file +15 -0
+are exceptions (where the response contains sensitive information). In some
+cases, the request as such is allowed for anyone but unprivileged users have
+attributes with sensitive information (e.g. wake-on-lan password) omitted.
+
+
+Conventions
+-----------
+
+Attributes which represent a boolean value usually use u8 type so that we can
+distinguish three states: "on", "off" and "not present" (meaning the
+information is not available in "get" requests or value is not to be changed
+in "set" requests). For these attributes, the "true" value should be passed as
+number 1 but any non-zero value should be understood as "true" by recipient.
+
+Some request types allow passing an attribute named ETHA_*_INFOMASK with

Documentation/networking/ethtool-netlink.txt line 51

Quoted 10 lines
Documentation/networking/ethtool-netlink.txt new file +10 -0
+a bitmask telling kernel that we are only interested in some parts of the
+information. If info mask is omitted, all available information is returned.
+Meaning of info mask bits depends on request type and is listed below.
+
+
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain

Isn't it ETHA_DEV_*? I must admit I'm a bit confused.

Documentation/networking/ethtool-netlink.txt line 57

Quoted 6 lines
Documentation/networking/ethtool-netlink.txt new file +6 -0
+
+    ETHA_DEV_INDEX	(u32)		device ifindex
+    ETHA_DEV_NAME	(string)	device name
+
+In device related requests, one of these is sufficient; if both are used, they
+must match (i.e. identify the same device). In device related replies both are

You say this now for the second time. First time this was said in second para.

Documentation/networking/ethtool-netlink.txt line 66

Quoted 9 lines
Documentation/networking/ethtool-netlink.txt new file +9 -0
+provided by kernel. In dump requests, device is not specified and kernel
+replies with one message per network device (only those for which the request
+is supported).
+
+
+List of message types
+---------------------
+
+All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"

Why "usually"? Why not "always"?

Documentation/networking/ethtool-netlink.txt line 71

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt new file +5 -0
+to indicate the type.
+
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;

Okay. Do we want reply to "*_cmd_something_get" command to be "*_cmd_something_set". That sounds odd. Why reply has to be "cmd"? Why not something like "reply" or "response"?
This should work for both "doit/dumpit" and notifications.

Quoted 124 lines
Documentation/networking/ethtool-netlink.txt new file +97 -0
+the same message can be also used to set (some of) the parameters, except for
+messages marked as "response only" in the table above. "Get" messages with
+NLM_F_DUMP flags and no device identification dump the information for all
+devices supporting the request.
+
+Later sections describe the format and semantics of these request messages.
+
+
+Request translation
+-------------------
+
+The following table maps iosctl commands to netlink commands providing their
+functionality. Entries with "n/a" in right column are commands which do not
+have their netlink replacement yet.
+
+ioctl command			netlink command
+---------------------------------------------------------------------
+ETHTOOL_GSET			n/a
+ETHTOOL_SSET			n/a
+ETHTOOL_GDRVINFO		n/a
+ETHTOOL_GREGS			n/a
+ETHTOOL_GWOL			n/a
+ETHTOOL_SWOL			n/a
+ETHTOOL_GMSGLVL			n/a
+ETHTOOL_SMSGLVL			n/a
+ETHTOOL_NWAY_RST		n/a
+ETHTOOL_GLINK			n/a
+ETHTOOL_GEEPROM			n/a
+ETHTOOL_SEEPROM			n/a
+ETHTOOL_GCOALESCE		n/a
+ETHTOOL_SCOALESCE		n/a
+ETHTOOL_GRINGPARAM		n/a
+ETHTOOL_SRINGPARAM		n/a
+ETHTOOL_GPAUSEPARAM		n/a
+ETHTOOL_SPAUSEPARAM		n/a
+ETHTOOL_GRXCSUM			n/a
+ETHTOOL_SRXCSUM			n/a
+ETHTOOL_GTXCSUM			n/a
+ETHTOOL_STXCSUM			n/a
+ETHTOOL_GSG			n/a
+ETHTOOL_SSG			n/a
+ETHTOOL_TEST			n/a
+ETHTOOL_GSTRINGS		n/a
+ETHTOOL_PHYS_ID			n/a
+ETHTOOL_GSTATS			n/a
+ETHTOOL_GTSO			n/a
+ETHTOOL_STSO			n/a
+ETHTOOL_GPERMADDR		rtnetlink RTM_GETLINK
+ETHTOOL_GUFO			n/a
+ETHTOOL_SUFO			n/a
+ETHTOOL_GGSO			n/a
+ETHTOOL_SGSO			n/a
+ETHTOOL_GFLAGS			n/a
+ETHTOOL_SFLAGS			n/a
+ETHTOOL_GPFLAGS			n/a
+ETHTOOL_SPFLAGS			n/a
+ETHTOOL_GRXFH			n/a
+ETHTOOL_SRXFH			n/a
+ETHTOOL_GGRO			n/a
+ETHTOOL_SGRO			n/a
+ETHTOOL_GRXRINGS		n/a
+ETHTOOL_GRXCLSRLCNT		n/a
+ETHTOOL_GRXCLSRULE		n/a
+ETHTOOL_GRXCLSRLALL		n/a
+ETHTOOL_SRXCLSRLDEL		n/a
+ETHTOOL_SRXCLSRLINS		n/a
+ETHTOOL_FLASHDEV		n/a
+ETHTOOL_RESET			n/a
+ETHTOOL_SRXNTUPLE		n/a
+ETHTOOL_GRXNTUPLE		n/a
+ETHTOOL_GSSET_INFO		n/a
+ETHTOOL_GRXFHINDIR		n/a
+ETHTOOL_SRXFHINDIR		n/a
+ETHTOOL_GFEATURES		n/a
+ETHTOOL_SFEATURES		n/a
+ETHTOOL_GCHANNELS		n/a
+ETHTOOL_SCHANNELS		n/a
+ETHTOOL_SET_DUMP		n/a
+ETHTOOL_GET_DUMP_FLAG		n/a
+ETHTOOL_GET_DUMP_DATA		n/a
+ETHTOOL_GET_TS_INFO		n/a
+ETHTOOL_GMODULEINFO		n/a
+ETHTOOL_GMODULEEEPROM		n/a
+ETHTOOL_GEEE			n/a
+ETHTOOL_SEEE			n/a
+ETHTOOL_GRSSH			n/a
+ETHTOOL_SRSSH			n/a
+ETHTOOL_GTUNABLE		n/a
+ETHTOOL_STUNABLE		n/a
+ETHTOOL_GPHYSTATS		n/a
+ETHTOOL_PERQUEUE		n/a
+ETHTOOL_GLINKSETTINGS		n/a
+ETHTOOL_SLINKSETTINGS		n/a
+ETHTOOL_PHY_GTUNABLE		n/a
+ETHTOOL_PHY_STUNABLE		n/a
+ETHTOOL_GFECPARAM		n/a
+ETHTOOL_SFECPARAM		n/a
include/linux/ethtool_netlink.h new file +9 -0
new file mode 100644
index 000000000000..0412adb4f42f
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _LINUX_ETHTOOL_NETLINK_H_
+#define _LINUX_ETHTOOL_NETLINK_H_
+
+#include <uapi/linux/ethtool_netlink.h>
+#include <linux/ethtool.h>
+
+#endif /* _LINUX_ETHTOOL_NETLINK_H_ */
include/uapi/linux/ethtool_netlink.h new file +6 -0
new file mode 100644
index 000000000000..6aa267451542
@@ -0,0 +1,25 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/*
+ * include/uapi/linux/ethtool_netlink.h - netlink interface for ethtool
+ *
+ * See Documentation/networking/ethtool-netlink.txt in kernel source tree for
+ * doucumentation of the interface.

include/uapi/linux/ethtool_netlink.h line 16

Quoted 10 lines
include/uapi/linux/ethtool_netlink.h new file +10 -0
+ */
+
+#ifndef _UAPI_LINUX_ETHTOOL_NETLINK_H_
+#define _UAPI_LINUX_ETHTOOL_NETLINK_H_
+
+#include <linux/ethtool.h>
+
+enum {
+	ETHNL_CMD_NOOP,
+

Usually headers have something like:
/* add new commands above here */
here.

Quoted 33 lines
include/uapi/linux/ethtool_netlink.h new file +9 -0
+	__ETHNL_CMD_CNT,
+	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
+};
+
+/* generic netlink info */
+#define ETHTOOL_GENL_NAME "ethtool"
+#define ETHTOOL_GENL_VERSION 1
+
+#endif /* _UAPI_LINUX_ETHTOOL_NETLINK_H_ */
net/Kconfig +8 -0
index 3e8fdd688329..75c600b45775 100644
@@ -448,6 +448,14 @@ config FAILOVER
 	  migration of VMs with direct attached VFs by failing over to the
 	  paravirtual datapath when the VF is unplugged.
 
+config ETHTOOL_NETLINK
+	bool "Netlink interface for ethtool"
+	default y
+	help
+	  An alternative userspace interface for ethtool based on generic
+	  netlink. It provides better extensibility and some new features,
+	  e.g. notification messages.
+

endif # if NET

# Used by archs to tell that they support BPF JIT compiler plus which flavour.

net/ethtool/Makefile +0 -0
index 3ebfab2bca66..f30e0da88be5 100644

net/ethtool/Makefile line 5

Quoted 6 lines
net/ethtool/Makefile +3 -1
@@ -1,3 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 
-obj-y		+= ioctl.o
+obj-y				+= ioctl.o
+
+obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o

Why? I believe this should be always build-in same as ioctl.

Quoted 10 lines
net/ethtool/Makefile +2 -0
+
+ethtool_nl-y	:= netlink.o
net/ethtool/netlink.c new file +2 -0
new file mode 100644
index 000000000000..85dd6dac71a2
@@ -0,0 +1,34 @@
+// SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
+

net/ethtool/netlink.c line 12

Quoted 10 lines
net/ethtool/netlink.c new file +10 -0
+#include <linux/ethtool_netlink.h>
+#include "netlink.h"
+
+/* genetlink setup */
+
+static const struct genl_ops ethtool_genl_ops[] = {
+};
+
+struct genl_family ethtool_genl_family = {
+	.hdrsize	= 0,

No need to set 0.

Quoted 28 lines
net/ethtool/netlink.c new file +22 -0
+	.name		= ETHTOOL_GENL_NAME,
+	.version	= ETHTOOL_GENL_VERSION,
+	.netnsok	= true,
+	.parallel_ops	= true,
+	.ops		= ethtool_genl_ops,
+	.n_ops		= ARRAY_SIZE(ethtool_genl_ops),
+};
+
+/* module setup */
+
+static int __init ethnl_init(void)
+{
+	int ret;
+
+	ret = genl_register_family(&ethtool_genl_family);
+	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))
+		return ret;
+
+	return 0;
+}
+
+subsys_initcall(ethnl_init);
net/ethtool/netlink.h new file +0 -0
new file mode 100644
index 000000000000..63063b582ca2

net/ethtool/netlink.h line 10

Quoted 10 lines
net/ethtool/netlink.h new file +10 -0
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _NET_ETHTOOL_NETLINK_H
+#define _NET_ETHTOOL_NETLINK_H
+
+#include <linux/ethtool_netlink.h>
+#include <linux/netdevice.h>
+#include <net/genetlink.h>
+
+extern struct genl_family ethtool_genl_family;

Why? You need this just within "netlink.c", don't you?

Quoted 5 lines
net/ethtool/netlink.h new file +2 -0
+
+#endif /* _NET_ETHTOOL_NETLINK_H */
Signature · 1 line
2.21.0
#17 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 5 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt line 26

Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt new file +2 -0
+All "set" and "action" type requests require admin privileges (CAP_NET_ADMIN
+in the namespace). Most "get" type request are allowed for anyone but there

s/request/requests/

Will fix.

Quoted 7 lines

Documentation/networking/ethtool-netlink.txt line 51

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt new file +5 -0
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain

Isn't it ETHA_DEV_*? I must admit I'm a bit confused.

ETHA_*_DEV is the nesting attribute (e.g. ETHA_SETTINGS_DEV), ETHA_DEV_* (ETHA_DEV_INDEX and ETHA_DEV_NAME) are in the nest.

Quoted 11 lines

Documentation/networking/ethtool-netlink.txt line 57

Quoted 8 lines
Documentation/networking/ethtool-netlink.txt new file +6 -0
+
+    ETHA_DEV_INDEX	(u32)		device ifindex
+    ETHA_DEV_NAME	(string)	device name
+
+In device related requests, one of these is sufficient; if both are used, they
+must match (i.e. identify the same device). In device related replies both are

You say this now for the second time. First time this was said in second para.

I'll drop one of them.

Quoted 6 lines

Documentation/networking/ethtool-netlink.txt line 66

Quoted 4 lines
Documentation/networking/ethtool-netlink.txt new file +4 -0
+List of message types
+---------------------
+
+All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"

Why "usually"? Why not "always"?

Right, it's always. And if it changes one day, the sentence will have to be rewritten anyway.

Quoted 8 lines

Documentation/networking/ethtool-netlink.txt line 71

Quoted 3 lines
Documentation/networking/ethtool-netlink.txt new file +3 -0
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;

Okay. Do we want reply to "*_cmd_something_get" command to be "*_cmd_something_set". That sounds odd. Why reply has to be "cmd"? Why not something like "reply" or "response"?
This should work for both "doit/dumpit" and notifications.

As stated right below, the aim is to use the same format for replies to GET requests as userspace uses for related SET requests. We could use different id (genlmsghdr::cmd) but that seemed like a waste for no actual gain.

Quoted 5 lines
Quoted 4 lines
Documentation/networking/ethtool-netlink.txt new file +4 -0
+the same message can be also used to set (some of) the parameters, except for
+messages marked as "response only" in the table above. "Get" messages with
+NLM_F_DUMP flags and no device identification dump the information for all
+devices supporting the request.
Quoted 8 lines

include/uapi/linux/ethtool_netlink.h line 16

Quoted 4 lines
include/uapi/linux/ethtool_netlink.h new file +4 -0
+
+enum {
+	ETHNL_CMD_NOOP,
+

Usually headers have something like:
/* add new commands above here */
here.

OK

Quoted 7 lines
Quoted 7 lines
net/ethtool/Makefile +0 -0
index 3ebfab2bca66..f30e0da88be5 100644
@@ -1,3 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 
Quoted 6 lines

net/ethtool/Makefile line 5

Quoted 4 lines
net/ethtool/Makefile +3 -1
-obj-y		+= ioctl.o
+obj-y				+= ioctl.o
+
+obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o

Why? I believe this should be always build-in same as ioctl.

I would like to make the ioctl interface optional as well, eventually. As someone noted in one of the earlier discussions, there may be some special minimalistic setups where ethtool interface may be of no use.

Quoted 4 lines

net/ethtool/Makefile line 7

Quoted 2 lines
net/ethtool/Makefile +2 -0
+struct genl_family ethtool_genl_family = {
+	.hdrsize	= 0,

No need to set 0.

OK

Quoted 4 lines

net/ethtool/Makefile

Quoted 2 lines
net/ethtool/Makefile +2 -0
+
+extern struct genl_family ethtool_genl_family;

Why? You need this just within "netlink.c", don't you?

In the submitted part, yes. But one of the later patches adds specific notify handler (different from ethnl_std_notify()) which is not in netlink.c and needs to use pointer to ethtool_genl_family for a call to genlmsg_put() and genlmsg_multicast().

But I can make it static for now and change to extern when it's needed.

Michal

#18 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 17 lines · Tue, Mar 26, 2019 at 02:24:27PM CET, [email protected] wrote:
Quoted 5 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt line 26

Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt new file +2 -0
+All "set" and "action" type requests require admin privileges (CAP_NET_ADMIN
+in the namespace). Most "get" type request are allowed for anyone but there

s/request/requests/

Will fix.

Quoted 7 lines

Documentation/networking/ethtool-netlink.txt line 51

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt new file +5 -0
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain

Isn't it ETHA_DEV_*? I must admit I'm a bit confused.

Quoted 2 lines

ETHA_*_DEV is the nesting attribute (e.g. ETHA_SETTINGS_DEV), ETHA_DEV_* (ETHA_DEV_INDEX and ETHA_DEV_NAME) are in the nest.

Yeah. I wonder why you need to duplicate this. Can this be in top-lever attr enum that is shared among all commands? It is there anyway and looks a bit silly to have "DEV" attr separate for every command. Something like this:

ATTR_IFINDEX
ATTR_IFNAME
ATTR_SOMEOTHER (flags perhaps)
ATTR_CMD_SPECIFIC_NEST_START
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
ATTR_CMD_SPECIFIC_NEST_END

Quoted 22 lines
Quoted 12 lines

Documentation/networking/ethtool-netlink.txt line 57

Quoted 9 lines
Documentation/networking/ethtool-netlink.txt new file +6 -0
+
+    ETHA_DEV_INDEX	(u32)		device ifindex
+    ETHA_DEV_NAME	(string)	device name
+
+In device related requests, one of these is sufficient; if both are used, they
+must match (i.e. identify the same device). In device related replies both are

You say this now for the second time. First time this was said in second para.

I'll drop one of them.

Quoted 6 lines

Documentation/networking/ethtool-netlink.txt line 66

Quoted 4 lines
Documentation/networking/ethtool-netlink.txt new file +4 -0
+List of message types
+---------------------
+
+All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"

Why "usually"? Why not "always"?

Quoted 2 lines

Right, it's always. And if it changes one day, the sentence will have to be rewritten anyway.

Okay.

Quoted 10 lines
Quoted 9 lines

Documentation/networking/ethtool-netlink.txt line 71

Quoted 4 lines
Documentation/networking/ethtool-netlink.txt new file +3 -0
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;

Okay. Do we want reply to "*_cmd_something_get" command to be "*_cmd_something_set". That sounds odd. Why reply has to be "cmd"? Why not something like "reply" or "response"?
This should work for both "doit/dumpit" and notifications.

Quoted 4 lines

As stated right below, the aim is to use the same format for replies to GET requests as userspace uses for related SET requests. We could use different id (genlmsghdr::cmd) but that seemed like a waste for no actual gain.

I understand. I just wonder if the replies/notifications could use the same name, not having "set" in it. I know we have it like this in many netlink ifaces, it is however confusing to users. So once we are doing this from scratch, we can do it differently.

Quoted 31 lines
Quoted 6 lines
Quoted 5 lines
Documentation/networking/ethtool-netlink.txt new file +4 -0
+the same message can be also used to set (some of) the parameters, except for
+messages marked as "response only" in the table above. "Get" messages with
+NLM_F_DUMP flags and no device identification dump the information for all
+devices supporting the request.
Quoted 8 lines

include/uapi/linux/ethtool_netlink.h line 16

Quoted 4 lines
include/uapi/linux/ethtool_netlink.h new file +4 -0
+
+enum {
+	ETHNL_CMD_NOOP,
+

Usually headers have something like:
/* add new commands above here */
here.

OK

Quoted 7 lines
Quoted 7 lines
net/ethtool/Makefile +0 -0
index 3ebfab2bca66..f30e0da88be5 100644
@@ -1,3 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 
Quoted 6 lines

net/ethtool/Makefile line 5

Quoted 4 lines
net/ethtool/Makefile +3 -1
-obj-y		+= ioctl.o
+obj-y				+= ioctl.o
+
+obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o

Why? I believe this should be always build-in same as ioctl.

Quoted 3 lines

I would like to make the ioctl interface optional as well, eventually. As someone noted in one of the earlier discussions, there may be some special minimalistic setups where ethtool interface may be of no use.

Okay, fair enough.

Quoted 13 lines
Quoted 5 lines

net/ethtool/Makefile line 7

Quoted 3 lines
net/ethtool/Makefile +2 -0
+struct genl_family ethtool_genl_family = {
+	.hdrsize	= 0,

No need to set 0.

OK

Quoted 4 lines

net/ethtool/Makefile

Quoted 2 lines
net/ethtool/Makefile +2 -0
+
+extern struct genl_family ethtool_genl_family;

Why? You need this just within "netlink.c", don't you?

Quoted 6 lines

In the submitted part, yes. But one of the later patches adds specific notify handler (different from ethnl_std_notify()) which is not in netlink.c and needs to use pointer to ethtool_genl_family for a call to genlmsg_put() and genlmsg_multicast().

But I can make it static for now and change to extern when it's needed.

Please do.

Quoted 3 lines

Michal

#19 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 19 lines · Jiri Pirko
Quoted 10 lines · Tue, Mar 26, 2019 at 02:24:27PM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt line 51

Quoted 5 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt new file +5 -0
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain

Isn't it ETHA_DEV_*? I must admit I'm a bit confused.

Quoted 2 lines

ETHA_*_DEV is the nesting attribute (e.g. ETHA_SETTINGS_DEV), ETHA_DEV_* (ETHA_DEV_INDEX and ETHA_DEV_NAME) are in the nest.

Yeah. I wonder why you need to duplicate this. Can this be in top-lever attr enum that is shared among all commands? It is there anyway and looks a bit silly to have "DEV" attr separate for every command. Something like this:

Quoted 8 lines

ATTR_IFINDEX
ATTR_IFNAME
ATTR_SOMEOTHER (flags perhaps)
ATTR_CMD_SPECIFIC_NEST_START
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
ATTR_CMD_SPECIFIC_NEST_END

I would rather prefer the opposite:

ATTR_HEADER
ATTR_IFINDEX
ATTR_IFNAME
ATTR_INFO_MASK
ATTR_PER_QUEUE
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
...

This way the "header" with universal attributes (not specfic to a particular message type) would be kept together at the beginning even after we need to add some more later and command specific attributes would still have fixed numbers (starting from 2). I'll think about it some more and check what would be pros and cons of the two variants when parsing and generating the messages.

Quoted 14 lines
Quoted 9 lines
Quoted 8 lines

Documentation/networking/ethtool-netlink.txt line 71

Quoted 3 lines
Documentation/networking/ethtool-netlink.txt new file +3 -0
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;

Okay. Do we want reply to "*_cmd_something_get" command to be "*_cmd_something_set". That sounds odd. Why reply has to be "cmd"? Why not something like "reply" or "response"?
This should work for both "doit/dumpit" and notifications.

Quoted 4 lines

As stated right below, the aim is to use the same format for replies to GET requests as userspace uses for related SET requests. We could use different id (genlmsghdr::cmd) but that seemed like a waste for no actual gain.

Quoted 4 lines

I understand. I just wonder if the replies/notifications could use the same name, not having "set" in it. I know we have it like this in many netlink ifaces, it is however confusing to users. So once we are doing this from scratch, we can do it differently.

How about

  ETHTOOL_MSG_GET_FOO  for get requests
  ETHTOOL_MSG_FOO      for get replies, notifications and set requests
  ETHTOOL_MSG_ACT_FOO  for actions (renegotiation, reset, blinking, ...)

?

Michal

#20 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 41 lines · Wed, Mar 27, 2019 at 10:26:04AM CET, [email protected] wrote:
Quoted 19 lines · Jiri Pirko
Quoted 10 lines · Tue, Mar 26, 2019 at 02:24:27PM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt line 51

Quoted 5 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt new file +5 -0
+Device identification
+---------------------
+
+When appropriate, network device is identified by a nested attribute named
+ETHA_*_DEV. This attribute can contain

Isn't it ETHA_DEV_*? I must admit I'm a bit confused.

Quoted 2 lines

ETHA_*_DEV is the nesting attribute (e.g. ETHA_SETTINGS_DEV), ETHA_DEV_* (ETHA_DEV_INDEX and ETHA_DEV_NAME) are in the nest.

Yeah. I wonder why you need to duplicate this. Can this be in top-lever attr enum that is shared among all commands? It is there anyway and looks a bit silly to have "DEV" attr separate for every command. Something like this:

Quoted 8 lines

ATTR_IFINDEX
ATTR_IFNAME
ATTR_SOMEOTHER (flags perhaps)
ATTR_CMD_SPECIFIC_NEST_START
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
ATTR_CMD_SPECIFIC_NEST_END

I would rather prefer the opposite:

ATTR_HEADER
ATTR_IFINDEX
ATTR_IFNAME
ATTR_INFO_MASK
ATTR_PER_QUEUE
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
...

Quoted 6 lines

This way the "header" with universal attributes (not specfic to a particular message type) would be kept together at the beginning even after we need to add some more later and command specific attributes would still have fixed numbers (starting from 2). I'll think about it some more and check what would be pros and cons of the two variants when parsing and generating the messages.

Okay, so what you suggest is per-cmd top level attr enum. That leads to duplications of common attributes:
You would have to have HEADER attr defined in every cmd enum:

enum cmdx {
ATTR_CMDX_HEADER
ATTR_CMDX_SOMETHING
ATTR_CMDX_SOMETHING2
ATTR_CMDX_SOMETHING3
};

enum cmdy {
ATTR_CMDY_HEADER
ATTR_CMDY_SOMETHING
ATTR_CMDY_SOMETHING2
ATTR_CMDY_SOMETHING3
};

That is odd. TC has it and I hate it there :)

I think that the rtnetlink example is better. The generic things are in generic top level enum. Then you have IFLA_LINKINFO with per-type enums.

Quoted 20 lines
Quoted 15 lines
Quoted 10 lines
Quoted 9 lines

Documentation/networking/ethtool-netlink.txt line 71

Quoted 4 lines
Documentation/networking/ethtool-netlink.txt new file +3 -0
+Messages of type "get" are used by userspace to request information and
+usually do not contain any attributes (that may be added later for dump
+filtering). Kernel response is in the form of corresponding "set" message;

Okay. Do we want reply to "*_cmd_something_get" command to be "*_cmd_something_set". That sounds odd. Why reply has to be "cmd"? Why not something like "reply" or "response"?
This should work for both "doit/dumpit" and notifications.

Quoted 4 lines

As stated right below, the aim is to use the same format for replies to GET requests as userspace uses for related SET requests. We could use different id (genlmsghdr::cmd) but that seemed like a waste for no actual gain.

Quoted 4 lines

I understand. I just wonder if the replies/notifications could use the same name, not having "set" in it. I know we have it like this in many netlink ifaces, it is however confusing to users. So once we are doing this from scratch, we can do it differently.

Quoted 7 lines

How about

 ETHTOOL_MSG_GET_FOO  for get requests
 ETHTOOL_MSG_FOO      for get replies, notifications and set requests
 ETHTOOL_MSG_ACT_FOO  for actions (renegotiation, reset, blinking, ...)

?

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

What do you think?

Quoted 2 lines

Michal

#21 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 10 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Signature · 1 line
Florian
#22 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 12 lines · Thu, Mar 28, 2019 at 03:05:14AM CET, [email protected] wrote:
Quoted 9 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

Quoted 4 lines
Quoted 1 line

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Sound good. Something like:

ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_GET_RPLY /* kernel->userspace replies to get */ ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT
ETHTOOL_MSG_FOO_NTF /* kernel->userspace async messages - notifications */

#23 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 18 lines · Jiri Pirko
Quoted 12 lines · Thu, Mar 28, 2019 at 03:05:14AM CET, [email protected] wrote:
Quoted 9 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

Quoted 4 lines
Quoted 1 line

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Quoted 7 lines

Sound good. Something like:

ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_GET_RPLY /* kernel->userspace replies to get */ ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT
ETHTOOL_MSG_FOO_NTF /* kernel->userspace async messages - notifications */

The names sound fine to me and having different message ids would still allow processing messages by the same handler easily.

But there is one potential issue I would like to point out: this way we spend 4 message ids for a get/set pair rather than 2. These message ids (genlmsghdr::cmd) are u8, i.e. the resource is not as infinite as one would wish. There are 80 ioctl commands (43 "get" and 29 "set") at the moment.

Netlink API should be less greedy in general. I already combined some ioctl commands into one netlink request type and with an easy way to add new attributes to existing commands, we won't need to add new commands as often (certainly not in a way which left us with 9 "get" and 9 "set" ioctl commands for netdev features). So even with 4 ids per get/set pair, we might be safe for reasonably long time. But it's still something to keep in mind.

Michal

#24 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 36 lines · Thu, Mar 28, 2019 at 10:37:46AM CET, [email protected] wrote:
Quoted 18 lines · Jiri Pirko
Quoted 12 lines · Thu, Mar 28, 2019 at 03:05:14AM CET, [email protected] wrote:
Quoted 9 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

Quoted 4 lines
Quoted 1 line

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Quoted 7 lines

Sound good. Something like:

ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_GET_RPLY /* kernel->userspace replies to get */ ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT
ETHTOOL_MSG_FOO_NTF /* kernel->userspace async messages - notifications */

The names sound fine to me and having different message ids would still allow processing messages by the same handler easily.

But there is one potential issue I would like to point out: this way we spend 4 message ids for a get/set pair rather than 2. These message ids (genlmsghdr::cmd) are u8, i.e. the resource is not as infinite as one would wish. There are 80 ioctl commands (43 "get" and 29 "set") at the moment.

Quoted 7 lines

Netlink API should be less greedy in general. I already combined some ioctl commands into one netlink request type and with an easy way to add new attributes to existing commands, we won't need to add new commands as often (certainly not in a way which left us with 9 "get" and 9 "set" ioctl commands for netdev features). So even with 4 ids per get/set pair, we might be safe for reasonably long time. But it's still something to keep in mind.

There are still 16 bits reserve in genl msg header:

struct genlmsghdr {
        __u8    cmd;
        __u8    version;
        __u16   reserved;
};
#25 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 45 lines · Jiri Pirko
Quoted 36 lines · Thu, Mar 28, 2019 at 10:37:46AM CET, [email protected] wrote:
Quoted 18 lines · Jiri Pirko
Quoted 12 lines · Thu, Mar 28, 2019 at 03:05:14AM CET, [email protected] wrote:
Quoted 9 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

Quoted 4 lines
Quoted 1 line

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Quoted 7 lines

Sound good. Something like:

ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_GET_RPLY /* kernel->userspace replies to get */ ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT
ETHTOOL_MSG_FOO_NTF /* kernel->userspace async messages - notifications */

The names sound fine to me and having different message ids would still allow processing messages by the same handler easily.

But there is one potential issue I would like to point out: this way we spend 4 message ids for a get/set pair rather than 2. These message ids (genlmsghdr::cmd) are u8, i.e. the resource is not as infinite as one would wish. There are 80 ioctl commands (43 "get" and 29 "set") at the moment.

Quoted 7 lines

Netlink API should be less greedy in general. I already combined some ioctl commands into one netlink request type and with an easy way to add new attributes to existing commands, we won't need to add new commands as often (certainly not in a way which left us with 9 "get" and 9 "set" ioctl commands for netdev features). So even with 4 ids per get/set pair, we might be safe for reasonably long time. But it's still something to keep in mind.

Quoted 7 lines

There are still 16 bits reserve in genl msg header:

struct genlmsghdr {
        __u8    cmd;
        __u8    version;
        __u16   reserved;
};

And you know not all message IDs will be making sense depending on the direction, so aliasing specific message IDs to an existing value should be fine?

Signature · 1 line
Florian
#26 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 54 lines · Thu, Mar 28, 2019 at 06:00:20PM CET, [email protected] wrote:
Quoted 45 lines · Jiri Pirko
Quoted 36 lines · Thu, Mar 28, 2019 at 10:37:46AM CET, [email protected] wrote:
Quoted 18 lines · Jiri Pirko
Quoted 12 lines · Thu, Mar 28, 2019 at 03:05:14AM CET, [email protected] wrote:
Quoted 9 lines · Jiri Pirko

Why don't you have ETHTOOL_MSG_SET_FOO for set? I think that for kerne->userspace the ETHTOOL_MSG_FOO if fine. I would change the ordering of words thought, but it is cosmetics:
ETHTOOL_MSG_FOO /* kernel->userspace messages - replies, notifications */ ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT

Quoted 4 lines
Quoted 1 line

What do you think?

We could even name the notification explicitly with: ETHTOOL_MSG_NOTIF or ETHTOOL_MSG_NTF just so we spell out exactly what those messages are.

Quoted 7 lines

Sound good. Something like:

ETHTOOL_MSG_FOO_GET
ETHTOOL_MSG_FOO_GET_RPLY /* kernel->userspace replies to get */ ETHTOOL_MSG_FOO_SET
ETHTOOL_MSG_FOO_ACT
ETHTOOL_MSG_FOO_NTF /* kernel->userspace async messages - notifications */

The names sound fine to me and having different message ids would still allow processing messages by the same handler easily.

But there is one potential issue I would like to point out: this way we spend 4 message ids for a get/set pair rather than 2. These message ids (genlmsghdr::cmd) are u8, i.e. the resource is not as infinite as one would wish. There are 80 ioctl commands (43 "get" and 29 "set") at the moment.

Quoted 7 lines

Netlink API should be less greedy in general. I already combined some ioctl commands into one netlink request type and with an easy way to add new attributes to existing commands, we won't need to add new commands as often (certainly not in a way which left us with 9 "get" and 9 "set" ioctl commands for netdev features). So even with 4 ids per get/set pair, we might be safe for reasonably long time. But it's still something to keep in mind.

Quoted 7 lines

There are still 16 bits reserve in genl msg header:

struct genlmsghdr {
        __u8    cmd;
        __u8    version;
        __u16   reserved;
};
Quoted 3 lines

And you know not all message IDs will be making sense depending on the direction, so aliasing specific message IDs to an existing value should be fine?

You are right, good idea. There can be 2 enums one for in, one for out.

#27 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

In reply to Michal Kubecek
Quoted 6 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:

Basic genetlink and init infrastructure for the netlink interface, register genetlink family "ethtool". Introduce CONFIG_ETHTOOL_NETLINK Kconfig option. Add interface description into Documentation/networking.

---

[...]

Quoted 6 lines
net/ethtool/netlink.c new file +0 -0
new file mode 100644
index 000000000000..85dd6dac71a2

net/ethtool/netlink.c line 8

Quoted 8 lines
net/ethtool/netlink.c new file +8 -0
@@ -0,0 +1,34 @@
+// SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
+
+#include <linux/ethtool_netlink.h>
+#include "netlink.h"
+
+/* genetlink setup */
+
+static const struct genl_ops ethtool_genl_ops[] = {

Please be consistent with prefixes. Either use "ethtool_" or "ethnl_" for all functions and variables in this code.

Quoted 10 lines
net/ethtool/netlink.c new file +10 -0
+};
+
+struct genl_family ethtool_genl_family = {
+	.hdrsize	= 0,
+	.name		= ETHTOOL_GENL_NAME,
+	.version	= ETHTOOL_GENL_VERSION,
+	.netnsok	= true,
+	.parallel_ops	= true,
+	.ops		= ethtool_genl_ops,
+	.n_ops		= ARRAY_SIZE(ethtool_genl_ops),

net/ethtool/netlink.c line 28

Quoted 10 lines
net/ethtool/netlink.c new file +10 -0
+};
+
+/* module setup */
+
+static int __init ethnl_init(void)
+{
+	int ret;
+
+	ret = genl_register_family(&ethtool_genl_family);
+	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))

Why do you need this warning? Please avoid it.

Quoted 27 lines
net/ethtool/netlink.c new file +6 -0
+		return ret;
+
+	return 0;
+}
+
+subsys_initcall(ethnl_init);
net/ethtool/netlink.h new file +12 -0
new file mode 100644
index 000000000000..63063b582ca2
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _NET_ETHTOOL_NETLINK_H
+#define _NET_ETHTOOL_NETLINK_H
+
+#include <linux/ethtool_netlink.h>
+#include <linux/netdevice.h>
+#include <net/genetlink.h>
+
+extern struct genl_family ethtool_genl_family;
+
+#endif /* _NET_ETHTOOL_NETLINK_H */
Signature · 1 line
2.21.0
#28 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c line 8

Quoted 3 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
net/ethtool/netlink.c new file +3 -0
+/* genetlink setup */
+
+static const struct genl_ops ethtool_genl_ops[] = {

Please be consistent with prefixes. Either use "ethtool_" or "ethnl_" for all functions and variables in this code.

OK

Quoted 10 lines

net/ethtool/netlink.c line 28

Quoted 8 lines
net/ethtool/netlink.c new file +8 -0
+/* module setup */
+
+static int __init ethnl_init(void)
+{
+	int ret;
+
+	ret = genl_register_family(&ethtool_genl_family);
+	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))

Why do you need this warning? Please avoid it.

I'm confused now... few days ago you replied "+1" to the idea:

http://lkml.kernel.org/r/20190321162105.GU2087@nanopsycho

I agreed that panic() (which is what e.g. rtnetlink does) would be an overkill but I would be definitely opposed to not having anything in the log at all and just silently going on without the interface (which may result in misconfigured network). I believe that if this fails, it is a sign of something going very wrong inside the kernel so that the "W" taint flag would be appropriate.

Michal

#29 lore raw

Re: [PATCH net-next v5 05/22] ethtool: introduce ethtool netlink interface

Quoted 24 lines · Tue, Mar 26, 2019 at 06:32:15PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c line 8

Quoted 3 lines · Mon, Mar 25, 2019 at 06:08:09PM CET, [email protected] wrote:
net/ethtool/netlink.c new file +3 -0
+/* genetlink setup */
+
+static const struct genl_ops ethtool_genl_ops[] = {

Please be consistent with prefixes. Either use "ethtool_" or "ethnl_" for all functions and variables in this code.

OK

Quoted 10 lines

net/ethtool/netlink.c line 28

Quoted 8 lines
net/ethtool/netlink.c new file +8 -0
+/* module setup */
+
+static int __init ethnl_init(void)
+{
+	int ret;
+
+	ret = genl_register_family(&ethtool_genl_family);
+	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))

Why do you need this warning? Please avoid it.

I'm confused now... few days ago you replied "+1" to the idea:

Quoted 8 lines

http://lkml.kernel.org/r/20190321162105.GU2087@nanopsycho

I agreed that panic() (which is what e.g. rtnetlink does) would be an overkill but I would be definitely opposed to not having anything in the log at all and just silently going on without the interface (which may result in misconfigured network). I believe that if this fails, it is a sign of something going very wrong inside the kernel so that the "W" taint flag would be appropriate.

Okay. Fair. It's just that I looked over the code and did not find any other genl_register_family() call with WARN.

Quoted 2 lines

Michal

#30 lore raw

PATCH v5 6/22 ethtool: helper functions for netlink interface

In reply to Michal Kubecek

Various helpers used by ethtool netlink code.

---
 include/uapi/linux/ethtool_netlink.h |  11 ++
 net/ethtool/netlink.c                | 144 +++++++++++++++++++++++++++
 net/ethtool/netlink.h                | 144 +++++++++++++++++++++++++++
 3 files changed, 299 insertions(+)
include/uapi/linux/ethtool_netlink.h +11 -0
index 6aa267451542..59240a2cda56 100644
@@ -18,6 +18,17 @@ enum {
 	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
 };
 
+/* device specification */
+
+enum {
+	ETHA_DEV_UNSPEC,
+	ETHA_DEV_INDEX,				/* u32 */
+	ETHA_DEV_NAME,				/* string */
+
+	__ETHA_DEV_CNT,
+	ETHA_DEV_MAX = (__ETHA_DEV_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/netlink.c +144 -0

152 lines not on this page · raw message

net/ethtool/netlink.h +144 -0

151 lines not on this page · raw message

Signature · 1 line
2.21.0
#31 lore raw

Re: [PATCH net-next v5 06/22] ethtool: helper functions for netlink interface

Quoted 14 lines · Mon, Mar 25, 2019 at 06:08:12PM CET, [email protected] wrote:

Various helpers used by ethtool netlink code.

---

include/uapi/linux/ethtool_netlink.h |  11 ++
net/ethtool/netlink.c                | 144 +++++++++++++++++++++++++++
net/ethtool/netlink.h                | 144 +++++++++++++++++++++++++++
3 files changed, 299 insertions(+)
include/uapi/linux/ethtool_netlink.h +0 -0
index 6aa267451542..59240a2cda56 100644

include/uapi/linux/ethtool_netlink.h enum { · line 24

Quoted 7 lines
include/uapi/linux/ethtool_netlink.h +4 -0
@@ -18,6 +18,17 @@ enum {
 	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
 };
 
+/* device specification */
+
+enum {
+	ETHA_DEV_UNSPEC,

ETHNL/ETHA looks odd.
I would prefer prefix "ETHTOOL_* for everything in uapi:
ETHTOOL_CMD_* for commands.
ETHTOOL_ATTR_* for attributes.

Looks much nicer.

Quoted 15 lines
include/uapi/linux/ethtool_netlink.h +7 -0
+	ETHA_DEV_INDEX,				/* u32 */
+	ETHA_DEV_NAME,				/* string */
+
+	__ETHA_DEV_CNT,
+	ETHA_DEV_MAX = (__ETHA_DEV_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/netlink.c +0 -0
index 85dd6dac71a2..cc6829ba4331 100644

net/ethtool/netlink.c line 7

Quoted 7 lines
net/ethtool/netlink.c +2 -0
@@ -1,8 +1,152 @@
 // SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
 
+#include <net/sock.h>
 #include <linux/ethtool_netlink.h>
 #include "netlink.h"
 
+static const struct nla_policy dev_policy[ETHA_DEV_MAX + 1] = {

ethnl_dev_policy. Please maintain the namespace if possible. "dev_policy" looks way too generic if anyone sees it in the code.

Quoted 9 lines
net/ethtool/netlink.c +9 -0
+	[ETHA_DEV_UNSPEC]	= { .type = NLA_REJECT },
+	[ETHA_DEV_INDEX]	= { .type = NLA_U32 },
+	[ETHA_DEV_NAME]		= { .type = NLA_NUL_STRING,
+				    .len = IFNAMSIZ - 1 },
+};
+
+/**
+ * ethnl_dev_get() - get device identified by nested attribute
+ * @info: genetlink info (also used for extack error reporting)

net/ethtool/netlink.c line 26

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+ * @nest: nest attribute with device identification
+ *
+ * Finds the network device identified by ETHA_DEV_INDEX (ifindex) or
+ * ETHA_DEV_NAME (name) attributes in a nested attribute @nest. If both
+ * are supplied, they must identify the same device. If successful, takes
+ * a reference to the device which is to be released by caller.
+ *
+ * Return: pointer to the device if successful, ERR_PTR(err) on error
+ */
+struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest)

I wonder, why you have this in nested attr? Aren't ifindex/ifname always used as a handle for commands/notifications? Would be just in toplevel (see devlink/nl80211)

net/ethtool/netlink.c line 35

Quoted 9 lines
net/ethtool/netlink.c +9 -0
+{
+	struct net *net = genl_info_net(info);
+	struct nlattr *tb[ETHA_DEV_MAX + 1];
+	struct net_device *dev;
+	int ret;
+
+	if (!nest) {
+		ETHNL_SET_ERRMSG(info,
+				 "mandatory device identification missing");

No need to wrap.

net/ethtool/netlink.c line 40

Quoted 5 lines
net/ethtool/netlink.c +5 -0
+		return ERR_PTR(-EINVAL);
+	}
+	ret = nla_parse_nested_strict(tb, ETHA_DEV_MAX, nest, dev_policy,
+				      info->extack);
+	if (ret < 0)

"if (ret)" is enough. "Returns 0 on success or a negative error code."

Quoted 4 lines
net/ethtool/netlink.c +4 -0
+		return ERR_PTR(ret);
+
+	if (tb[ETHA_DEV_INDEX]) {
+		dev = dev_get_by_index(net, nla_get_u32(tb[ETHA_DEV_INDEX]));

net/ethtool/netlink.c line 54

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+		if (!dev)
+			return ERR_PTR(-ENODEV);
+		/* if both ifindex and ifname are passed, they must match */
+		if (tb[ETHA_DEV_NAME]) {
+			const char *nl_name = nla_data(tb[ETHA_DEV_NAME]);
+
+			if (strncmp(dev->name, nl_name, IFNAMSIZ)) {
+				dev_put(dev);
+				ETHNL_SET_ERRMSG(info,
+						 "ifindex and ifname do not match");

No need to wrap.

Quoted 39 lines
net/ethtool/netlink.c +39 -0
+				return ERR_PTR(-ENODEV);
+			}
+		}
+		return dev;
+	} else if (tb[ETHA_DEV_NAME]) {
+		dev = dev_get_by_name(net, nla_data(tb[ETHA_DEV_NAME]));
+		if (!dev)
+			return ERR_PTR(-ENODEV);
+	} else {
+		ETHNL_SET_ERRMSG(info, "either ifindex or ifname required");
+		return ERR_PTR(-EINVAL);
+	}
+
+	if (!netif_device_present(dev)) {
+		dev_put(dev);
+		ETHNL_SET_ERRMSG(info, "device not present");
+		return ERR_PTR(-ENODEV);
+	}
+	return dev;
+}
+
+/**
+ * ethnl_fill_dev() - Put device identification nest into a message
+ * @msg:      skb with the message
+ * @dev:      network device to describe
+ * @attrtype: attribute type to use for the nest
+ *
+ * Create a nested attribute with attributes describing given network device.
+ * Clean up on error.
+ *
+ * Return: 0 on success, error value (-EMSGSIZE only) on error
+ */
+int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype)
+{
+	struct nlattr *nest;
+	int ret = -EMSGSIZE;
+
+	nest = ethnl_nest_start(msg, attrtype);
+	if (!nest)

net/ethtool/netlink.c line 103

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+		return -EMSGSIZE;
+
+	if (nla_put_u32(msg, ETHA_DEV_INDEX, (u32)dev->ifindex))
+		goto err;
+	if (nla_put_string(msg, ETHA_DEV_NAME, dev->name))
+		goto err;
+
+	nla_nest_end(msg, nest);
+	return 0;
+err:

Usually this label is called "nla_put_failure". Same for the rest. "err" could be confused with error return variable.

have "ret = -EMSGSIZE" here. Easier to follow the error path.

Quoted 9 lines
net/ethtool/netlink.c +9 -0
+	nla_nest_cancel(msg, nest);
+	return ret;
+}
+
+/**
+ * ethnl_reply_init() - Create skb for a reply and fill device identification
+ * @payload: payload length (without netlink and genetlink header)
+ * @dev:     device the reply is about (may be null)
+ * @cmd:     ETHNL_CMD_* command for reply

net/ethtool/netlink.c line 122

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+ * @info:    genetlink info of the received packet we respond to
+ * @ehdrp:   place to store payload pointer returned by genlmsg_new()
+ *
+ * Return: pointer to allocated skb on success, NULL on error
+ */
+struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
+				 u16 dev_attrtype, struct genl_info *info,
+				 void **ehdrp)
+{
+	struct sk_buff *rskb;

Could be just "skb", or "msg". You have "msg" in ethnl_fill_dev(). Please have it consistent.

net/ethtool/netlink.c line 128

Quoted 6 lines
net/ethtool/netlink.c +6 -0
+	void *ehdr;
+
+	rskb = genlmsg_new(payload, GFP_KERNEL);
+	if (!rskb) {
+		ETHNL_SET_ERRMSG(info,
+				 "failed to allocate reply message");

No need to wrap.

net/ethtool/netlink.c line 140

Quoted 12 lines
net/ethtool/netlink.c +12 -0
+		return NULL;
+	}
+
+	ehdr = genlmsg_put_reply(rskb, info, &ethtool_genl_family, 0, cmd);
+	if (!ehdr)
+		goto err;
+	if (ehdrp)
+		*ehdrp = ehdr;
+	if (dev) {
+		int ret = ethnl_fill_dev(rskb, dev, dev_attrtype);
+
+		if (ret < 0)

"if (ret)" is enough.

Quoted 17 lines
net/ethtool/netlink.c +9 -0
+			goto err;
+	}
+
+	return rskb;
+err:
+	nlmsg_free(rskb);
+	return NULL;
+}
+
 /* genetlink setup */
 
 static const struct genl_ops ethtool_genl_ops[] = {
net/ethtool/netlink.h +0 -0
index 63063b582ca2..db90d95410b1 100644

net/ethtool/netlink.h line 12

Quoted 7 lines
net/ethtool/netlink.h +4 -0
@@ -6,7 +6,151 @@
 #include <linux/ethtool_netlink.h>
 #include <linux/netdevice.h>
 #include <net/genetlink.h>
+#include <net/sock.h>
+
+#define ETHNL_SET_ERRMSG(info, msg) \
+	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)

Why do you need this macro? Can info be null?
In general, macros like this should be avoided.

net/ethtool/netlink.h line 25

Quoted 13 lines
net/ethtool/netlink.h +10 -0
 extern struct genl_family ethtool_genl_family;
 
+struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest);
+int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype);
+
+struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
+				 u16 dev_attrtype, struct genl_info *info,
+				 void **ehdrp);
+
+static inline int ethnl_str_size(const char *s)
+{
+	return nla_total_size(strlen(s) + 1);

This looks like more generic helper, not tight to ethtool.

net/ethtool/netlink.h line 28

Quoted 3 lines
net/ethtool/netlink.h +3 -0
+}
+
+static inline int ethnl_str_ifne_size(const char *s)

Eh? "ifne"? What is this good for?

net/ethtool/netlink.h line 38

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+{
+	return s[0] ? ethnl_str_size(s) : 0;
+}
+
+static inline int ethnl_put_str_ifne(struct sk_buff *skb, int attrtype,
+				     const char *s)
+{
+	if (!s[0])
+		return 0;
+	return nla_put_string(skb, attrtype, s);

I don't like helpers like this. Do the check in the caller and put or not put the string there. It's a single if.

net/ethtool/netlink.h line 44

Quoted 6 lines
net/ethtool/netlink.h +6 -0
+}
+
+static inline struct nlattr *ethnl_nest_start(struct sk_buff *skb,
+					      int attrtype)
+{
+	return nla_nest_start(skb, attrtype | NLA_F_NESTED);

Please use nla_nest_start directly and avoid helpers like this.

net/ethtool/netlink.h line 53

Quoted 9 lines
net/ethtool/netlink.h +9 -0
+}
+
+static inline int ethnlmsg_parse(const struct nlmsghdr *nlh,
+				 struct nlattr *tb[], int maxtype,
+				 const struct nla_policy *policy,
+				 struct genl_info *info)
+{
+	return nlmsg_parse_strict(nlh, GENL_HDRLEN, tb, maxtype, policy,
+				  info ? info->extack : NULL);

Same thing, please use nlmsg_parse_strict directly.

Quoted 65 lines
net/ethtool/netlink.h +65 -0
+}
+
+/* ethnl_update_* return true if the value is changed */
+static inline bool ethnl_update_u32(u32 *dst, struct nlattr *attr)
+{
+	u32 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u32(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_u8(u8 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u8(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+/* update u32 value used as bool from NLA_U8 */
+static inline bool ethnl_update_bool32(u32 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = !!nla_get_u8(attr);
+	if (!!*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_binary(u8 *dst, unsigned int len,
+				       struct nlattr *attr)
+{
+	if (!attr)
+		return false;
+	if (nla_len(attr) < len)
+		len = nla_len(attr);
+	if (!memcmp(dst, nla_data(attr), len))
+		return false;
+
+	memcpy(dst, nla_data(attr), len);
+	return true;
+}
+
+static inline bool ethnl_update_bitfield32(u32 *dst, struct nlattr *attr)
+{
+	struct nla_bitfield32 change;
+	u32 newval;
+

net/ethtool/netlink.h line 128

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+	if (!attr)
+		return false;
+	change = nla_get_bitfield32(attr);
+	newval = (*dst & ~change.selector) | (change.value & change.selector);
+	if (*dst == newval)
+		return false;
+
+	*dst = newval;
+	return true;
+}

I don't understand puspose of these "update" helper functions. Try to avoid them. In general, please try to avoid wrappers around netlink api.

Quoted 5 lines
net/ethtool/netlink.h +5 -0
+
+static inline void warn_partial_info(struct genl_info *info)
+{
+	ETHNL_SET_ERRMSG(info, "not all requested data could be retrieved");
+}

net/ethtool/netlink.h line 143

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+
+/* Check user privileges explicitly to allow finer access control based on
+ * context of the request or hiding part of the information from unprivileged
+ * users
+ */
+static inline bool ethnl_is_privileged(struct sk_buff *skb)
+{
+	struct net *net = sock_net(skb->sk);
+
+	return netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN);

Same here.

net/ethtool/netlink.h line 150

Quoted 7 lines
net/ethtool/netlink.h +7 -0
+}
+
+/* total size of ETHA_*_DEV nested attribute; this is an upper estimate so that
+ * we do not need to hold RTNL longer than necessary to prevent rename between
+ * estimating the size and composing the message
+ */
+static inline unsigned int dev_ident_size(void)

You are missing "ethnl_" prefix.

Quoted 9 lines
net/ethtool/netlink.h +5 -0
+{
+	return nla_total_size(nla_total_size(sizeof(u32)) +
+			      nla_total_size(IFNAMSIZ));
+}
+
 #endif /* _NET_ETHTOOL_NETLINK_H */
Signature · 1 line
2.21.0
#32 lore raw

Re: [PATCH net-next v5 06/22] ethtool: helper functions for netlink interface

Quoted 11 lines · Jiri Pirko

include/uapi/linux/ethtool_netlink.h enum { · line 24

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:12PM CET, [email protected] wrote:
include/uapi/linux/ethtool_netlink.h +4 -0
+/* device specification */
+
+enum {
+	ETHA_DEV_UNSPEC,

ETHNL/ETHA looks odd.
I would prefer prefix "ETHTOOL_* for everything in uapi:
ETHTOOL_CMD_* for commands.
ETHTOOL_ATTR_* for attributes.

Quoted 1 line

Looks much nicer.

Even with ETHA_ prefix, there are places where having to fit into 80 columns makes the code hard to read. ETHTOOL_ATTR_ would make it even harder.

Quoted 4 lines
Quoted 1 line

+static const struct nla_policy dev_policy[ETHA_DEV_MAX + 1] = {

ethnl_dev_policy. Please maintain the namespace if possible. "dev_policy" looks way too generic if anyone sees it in the code.

OK

Quoted 15 lines
Quoted 15 lines
net/ethtool/netlink.c +15 -0
+	[ETHA_DEV_UNSPEC]	= { .type = NLA_REJECT },
+	[ETHA_DEV_INDEX]	= { .type = NLA_U32 },
+	[ETHA_DEV_NAME]		= { .type = NLA_NUL_STRING,
+				    .len = IFNAMSIZ - 1 },
+};
+
+/**
+ * ethnl_dev_get() - get device identified by nested attribute
+ * @info: genetlink info (also used for extack error reporting)
+ * @nest: nest attribute with device identification
+ *
+ * Finds the network device identified by ETHA_DEV_INDEX (ifindex) or
+ * ETHA_DEV_NAME (name) attributes in a nested attribute @nest. If both
+ * are supplied, they must identify the same device. If successful, takes
+ * a reference to the device which is to be released by caller.
Quoted 8 lines
Quoted 4 lines

+ *
+ * Return: pointer to the device if successful, ERR_PTR(err) on error
+ */
+struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest)

I wonder, why you have this in nested attr? Aren't ifindex/ifname always used as a handle for commands/notifications? Would be just in toplevel (see devlink/nl80211)

It makes things only slightly easier now but I would like to make the API as future proof as possible. If something needs to be added later (e.g. netns id) or something else changes (e.g. 64-bit ifindex), only one enum and these two helpers would need to be adjusted.

Quoted 5 lines

net/ethtool/netlink.c line 35

Quoted 3 lines
net/ethtool/netlink.c +3 -0
+	if (!nest) {
+		ETHNL_SET_ERRMSG(info,
+				 "mandatory device identification missing");

No need to wrap.

OK

Quoted 5 lines

net/ethtool/netlink.c line 40

Quoted 3 lines
net/ethtool/netlink.c +3 -0
+	ret = nla_parse_nested_strict(tb, ETHA_DEV_MAX, nest, dev_policy,
+				      info->extack);
+	if (ret < 0)

"if (ret)" is enough. "Returns 0 on success or a negative error code."

OK

Quoted 6 lines

net/ethtool/netlink.c line 54

Quoted 4 lines
net/ethtool/netlink.c +4 -0
+			if (strncmp(dev->name, nl_name, IFNAMSIZ)) {
+				dev_put(dev);
+				ETHNL_SET_ERRMSG(info,
+						 "ifindex and ifname do not match");

No need to wrap.

OK

Quoted 31 lines
Quoted 18 lines
net/ethtool/netlink.c +18 -0
+/**
+ * ethnl_fill_dev() - Put device identification nest into a message
+ * @msg:      skb with the message
+ * @dev:      network device to describe
+ * @attrtype: attribute type to use for the nest
+ *
+ * Create a nested attribute with attributes describing given network device.
+ * Clean up on error.
+ *
+ * Return: 0 on success, error value (-EMSGSIZE only) on error
+ */
+int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype)
+{
+	struct nlattr *nest;
+	int ret = -EMSGSIZE;
+
+	nest = ethnl_nest_start(msg, attrtype);
+	if (!nest)

net/ethtool/netlink.c line 103

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+		return -EMSGSIZE;
+
+	if (nla_put_u32(msg, ETHA_DEV_INDEX, (u32)dev->ifindex))
+		goto err;
+	if (nla_put_string(msg, ETHA_DEV_NAME, dev->name))
+		goto err;
+
+	nla_nest_end(msg, nest);
+	return 0;
+err:

Usually this label is called "nla_put_failure". Same for the rest. "err" could be confused with error return variable.

Quoted 1 line

have "ret = -EMSGSIZE" here. Easier to follow the error path.

OK

Quoted 14 lines
Quoted 14 lines
net/ethtool/netlink.c +14 -0
+	nla_nest_cancel(msg, nest);
+	return ret;
+}
+
+/**
+ * ethnl_reply_init() - Create skb for a reply and fill device identification
+ * @payload: payload length (without netlink and genetlink header)
+ * @dev:     device the reply is about (may be null)
+ * @cmd:     ETHNL_CMD_* command for reply
+ * @info:    genetlink info of the received packet we respond to
+ * @ehdrp:   place to store payload pointer returned by genlmsg_new()
+ *
+ * Return: pointer to allocated skb on success, NULL on error
+ */
Quoted 8 lines

net/ethtool/netlink.c line 122

Quoted 5 lines
net/ethtool/netlink.c +5 -0
+struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
+				 u16 dev_attrtype, struct genl_info *info,
+				 void **ehdrp)
+{
+	struct sk_buff *rskb;

Could be just "skb", or "msg". You have "msg" in ethnl_fill_dev(). Please have it consistent.

OK

Quoted 6 lines

net/ethtool/netlink.c line 128

Quoted 4 lines
net/ethtool/netlink.c +4 -0
+	rskb = genlmsg_new(payload, GFP_KERNEL);
+	if (!rskb) {
+		ETHNL_SET_ERRMSG(info,
+				 "failed to allocate reply message");

No need to wrap.

OK

Quoted 5 lines

net/ethtool/netlink.c line 140

Quoted 3 lines
net/ethtool/netlink.c +3 -0
+		int ret = ethnl_fill_dev(rskb, dev, dev_attrtype);
+
+		if (ret < 0)

"if (ret)" is enough.

OK

Quoted 5 lines

net/ethtool/netlink.h line 12

Quoted 3 lines
net/ethtool/netlink.h +3 -0
+
+#define ETHNL_SET_ERRMSG(info, msg) \
+	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)

Why do you need this macro? Can info be null?

The same functions are used for generating replies to GET requests (info is not null) and notifications (info is null, no extack).

Quoted 1 line

In general, macros like this should be avoided.

At the expense of repeating the same patterns. In this case it would be acceptable but it would mean one more indentation level for lines with the error/warning messages.

Quoted 5 lines

net/ethtool/netlink.h line 25

Quoted 3 lines
net/ethtool/netlink.h +3 -0
+static inline int ethnl_str_size(const char *s)
+{
+	return nla_total_size(strlen(s) + 1);

This looks like more generic helper, not tight to ethtool.

OK

Quoted 3 lines
Quoted 1 line

+static inline int ethnl_str_ifne_size(const char *s)

Eh? "ifne"? What is this good for?

It's for "if not empty". Structures like ethtool_drvinfo use fixed size char arrays for strings where empty string means the information is not available. So if the string is empty, netlink attribute is omitted.

Quoted 5 lines
Quoted 5 lines
net/ethtool/netlink.h +5 -0
+{
+	return s[0] ? ethnl_str_size(s) : 0;
+}
+
+static inline int ethnl_put_str_ifne(struct sk_buff *skb, int attrtype,
Quoted 8 lines

net/ethtool/netlink.h line 38

Quoted 5 lines
net/ethtool/netlink.h +5 -0
+				     const char *s)
+{
+	if (!s[0])
+		return 0;
+	return nla_put_string(skb, attrtype, s);

I don't like helpers like this. Do the check in the caller and put or not put the string there. It's a single if.

OK

Quoted 6 lines

net/ethtool/netlink.h line 44

Quoted 4 lines
net/ethtool/netlink.h +4 -0
+static inline struct nlattr *ethnl_nest_start(struct sk_buff *skb,
+					      int attrtype)
+{
+	return nla_nest_start(skb, attrtype | NLA_F_NESTED);

Please use nla_nest_start directly and avoid helpers like this.

OK. It's one more thing to check (and forget) in many places, though.

Quoted 9 lines

net/ethtool/netlink.h line 53

Quoted 7 lines
net/ethtool/netlink.h +7 -0
+static inline int ethnlmsg_parse(const struct nlmsghdr *nlh,
+				 struct nlattr *tb[], int maxtype,
+				 const struct nla_policy *policy,
+				 struct genl_info *info)
+{
+	return nlmsg_parse_strict(nlh, GENL_HDRLEN, tb, maxtype, policy,
+				  info ? info->extack : NULL);

Same thing, please use nlmsg_parse_strict directly.

OK

Quoted 68 lines
Quoted 68 lines
net/ethtool/netlink.h +68 -0
+/* ethnl_update_* return true if the value is changed */
+static inline bool ethnl_update_u32(u32 *dst, struct nlattr *attr)
+{
+	u32 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u32(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_u8(u8 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u8(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+/* update u32 value used as bool from NLA_U8 */
+static inline bool ethnl_update_bool32(u32 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = !!nla_get_u8(attr);
+	if (!!*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_binary(u8 *dst, unsigned int len,
+				       struct nlattr *attr)
+{
+	if (!attr)
+		return false;
+	if (nla_len(attr) < len)
+		len = nla_len(attr);
+	if (!memcmp(dst, nla_data(attr), len))
+		return false;
+
+	memcpy(dst, nla_data(attr), len);
+	return true;
+}
+
+static inline bool ethnl_update_bitfield32(u32 *dst, struct nlattr *attr)
+{
+	struct nla_bitfield32 change;
+	u32 newval;
+
+	if (!attr)
+		return false;
+	change = nla_get_bitfield32(attr);
+	newval = (*dst & ~change.selector) | (change.value & change.selector);
+	if (*dst == newval)
Quoted 8 lines

net/ethtool/netlink.h line 128

Quoted 5 lines
net/ethtool/netlink.h +5 -0
+		return false;
+
+	*dst = newval;
+	return true;
+}

I don't understand puspose of these "update" helper functions. Try to avoid them. In general, please try to avoid wrappers around netlink api.

The purpose is to update a value according to a request attribute (if present) and tell caller if the value was actually modified. They are used like e.g.

	mod = false;
	if (ethnl_update_u32(&data.rx_pending, tb[ETHA_RING_RX_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.rx_mini_pending,
			     tb[ETHA_RING_RX_MINI_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.rx_jumbo_pending,
			     tb[ETHA_RING_RX_JUMBO_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.tx_pending, tb[ETHA_RING_TX_PENDING]))
		mod = true;

	if (!mod)
		return 0;
	/* call ethtool_ops handler and send notification */

so that we can omit calling the handler and sending a notification if the request was no-op. Expanding the helpers here would make the code needlessly complicated and error prone.

Quoted 8 lines
Quoted 8 lines
net/ethtool/netlink.h +8 -0
+static inline void warn_partial_info(struct genl_info *info)
+{
+	ETHNL_SET_ERRMSG(info, "not all requested data could be retrieved");
+}
+
+/* Check user privileges explicitly to allow finer access control based on
+ * context of the request or hiding part of the information from unprivileged
+ * users
Quoted 8 lines

net/ethtool/netlink.h line 143

Quoted 6 lines

+ */

net/ethtool/netlink.h +5 -0
+static inline bool ethnl_is_privileged(struct sk_buff *skb)
+{
+	struct net *net = sock_net(skb->sk);
+
+	return netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN);

Same here.

It's still the same, helpers allow replacing repeating code patterns with just saying what you are doing. But this one is called only from one place so I can live without it.

Quoted 9 lines

net/ethtool/netlink.h line 150

Quoted 7 lines
net/ethtool/netlink.h +7 -0
+}
+
+/* total size of ETHA_*_DEV nested attribute; this is an upper estimate so that
+ * we do not need to hold RTNL longer than necessary to prevent rename between
+ * estimating the size and composing the message
+ */
+static inline unsigned int dev_ident_size(void)

You are missing "ethnl_" prefix.

OK

Michal

#33 lore raw

Re: [PATCH net-next v5 06/22] ethtool: helper functions for netlink interface

Quoted 12 lines · Tue, Mar 26, 2019 at 03:19:55PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

include/uapi/linux/ethtool_netlink.h enum { · line 24

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:12PM CET, [email protected] wrote:
include/uapi/linux/ethtool_netlink.h +4 -0
+/* device specification */
+
+enum {
+	ETHA_DEV_UNSPEC,

ETHNL/ETHA looks odd.
I would prefer prefix "ETHTOOL_* for everything in uapi:
ETHTOOL_CMD_* for commands.
ETHTOOL_ATTR_* for attributes.

Quoted 5 lines
Quoted 1 line

Looks much nicer.

Even with ETHA_ prefix, there are places where having to fit into 80 columns makes the code hard to read. ETHTOOL_ATTR_ would make it even harder.

So cut the line. UAPI consistency is more important than to fit on a 80cols line.

ETH/ETHA is wrong prefix. This is "ETHTOOL". I can live with ETHTOOL_A_* for attributes.

Quoted 31 lines
Quoted 5 lines
Quoted 2 lines

+static const struct nla_policy dev_policy[ETHA_DEV_MAX + 1] = {

ethnl_dev_policy. Please maintain the namespace if possible. "dev_policy" looks way too generic if anyone sees it in the code.

OK

Quoted 15 lines
Quoted 15 lines
net/ethtool/netlink.c +15 -0
+	[ETHA_DEV_UNSPEC]	= { .type = NLA_REJECT },
+	[ETHA_DEV_INDEX]	= { .type = NLA_U32 },
+	[ETHA_DEV_NAME]		= { .type = NLA_NUL_STRING,
+				    .len = IFNAMSIZ - 1 },
+};
+
+/**
+ * ethnl_dev_get() - get device identified by nested attribute
+ * @info: genetlink info (also used for extack error reporting)
+ * @nest: nest attribute with device identification
+ *
+ * Finds the network device identified by ETHA_DEV_INDEX (ifindex) or
+ * ETHA_DEV_NAME (name) attributes in a nested attribute @nest. If both
+ * are supplied, they must identify the same device. If successful, takes
+ * a reference to the device which is to be released by caller.
Quoted 8 lines
Quoted 4 lines

+ *
+ * Return: pointer to the device if successful, ERR_PTR(err) on error
+ */
+struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest)

I wonder, why you have this in nested attr? Aren't ifindex/ifname always used as a handle for commands/notifications? Would be just in toplevel (see devlink/nl80211)

Quoted 4 lines

It makes things only slightly easier now but I would like to make the API as future proof as possible. If something needs to be added later (e.g. netns id) or something else changes (e.g. 64-bit ifindex), only one enum and these two helpers would need to be adjusted.

You can still have the helper without the nest. So if you don't nest this, everything is the same, both helper and extra attr in enum.

Quoted 103 lines
Quoted 6 lines

net/ethtool/netlink.c line 35

Quoted 4 lines
net/ethtool/netlink.c +3 -0
+	if (!nest) {
+		ETHNL_SET_ERRMSG(info,
+				 "mandatory device identification missing");

No need to wrap.

OK

Quoted 5 lines

net/ethtool/netlink.c line 40

Quoted 3 lines
net/ethtool/netlink.c +3 -0
+	ret = nla_parse_nested_strict(tb, ETHA_DEV_MAX, nest, dev_policy,
+				      info->extack);
+	if (ret < 0)

"if (ret)" is enough. "Returns 0 on success or a negative error code."

OK

Quoted 6 lines

net/ethtool/netlink.c line 54

Quoted 4 lines
net/ethtool/netlink.c +4 -0
+			if (strncmp(dev->name, nl_name, IFNAMSIZ)) {
+				dev_put(dev);
+				ETHNL_SET_ERRMSG(info,
+						 "ifindex and ifname do not match");

No need to wrap.

OK

Quoted 31 lines
Quoted 18 lines
net/ethtool/netlink.c +18 -0
+/**
+ * ethnl_fill_dev() - Put device identification nest into a message
+ * @msg:      skb with the message
+ * @dev:      network device to describe
+ * @attrtype: attribute type to use for the nest
+ *
+ * Create a nested attribute with attributes describing given network device.
+ * Clean up on error.
+ *
+ * Return: 0 on success, error value (-EMSGSIZE only) on error
+ */
+int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype)
+{
+	struct nlattr *nest;
+	int ret = -EMSGSIZE;
+
+	nest = ethnl_nest_start(msg, attrtype);
+	if (!nest)

net/ethtool/netlink.c line 103

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+		return -EMSGSIZE;
+
+	if (nla_put_u32(msg, ETHA_DEV_INDEX, (u32)dev->ifindex))
+		goto err;
+	if (nla_put_string(msg, ETHA_DEV_NAME, dev->name))
+		goto err;
+
+	nla_nest_end(msg, nest);
+	return 0;
+err:

Usually this label is called "nla_put_failure". Same for the rest. "err" could be confused with error return variable.

Quoted 1 line

have "ret = -EMSGSIZE" here. Easier to follow the error path.

OK

Quoted 14 lines
Quoted 14 lines
net/ethtool/netlink.c +14 -0
+	nla_nest_cancel(msg, nest);
+	return ret;
+}
+
+/**
+ * ethnl_reply_init() - Create skb for a reply and fill device identification
+ * @payload: payload length (without netlink and genetlink header)
+ * @dev:     device the reply is about (may be null)
+ * @cmd:     ETHNL_CMD_* command for reply
+ * @info:    genetlink info of the received packet we respond to
+ * @ehdrp:   place to store payload pointer returned by genlmsg_new()
+ *
+ * Return: pointer to allocated skb on success, NULL on error
+ */
Quoted 8 lines

net/ethtool/netlink.c line 122

Quoted 5 lines
net/ethtool/netlink.c +5 -0
+struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
+				 u16 dev_attrtype, struct genl_info *info,
+				 void **ehdrp)
+{
+	struct sk_buff *rskb;

Could be just "skb", or "msg". You have "msg" in ethnl_fill_dev(). Please have it consistent.

OK

Quoted 6 lines

net/ethtool/netlink.c line 128

Quoted 4 lines
net/ethtool/netlink.c +4 -0
+	rskb = genlmsg_new(payload, GFP_KERNEL);
+	if (!rskb) {
+		ETHNL_SET_ERRMSG(info,
+				 "failed to allocate reply message");

No need to wrap.

OK

Quoted 5 lines

net/ethtool/netlink.c line 140

Quoted 3 lines
net/ethtool/netlink.c +3 -0
+		int ret = ethnl_fill_dev(rskb, dev, dev_attrtype);
+
+		if (ret < 0)

"if (ret)" is enough.

OK

Quoted 8 lines
Quoted 5 lines

net/ethtool/netlink.h line 12

Quoted 3 lines
net/ethtool/netlink.h +3 -0
+
+#define ETHNL_SET_ERRMSG(info, msg) \
+	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)

Why do you need this macro? Can info be null?

The same functions are used for generating replies to GET requests (info is not null) and notifications (info is null, no extack).

So again, could this be generic? GENL_SET_ERR_MSG() could accept null and ignore.

Quoted 6 lines
Quoted 2 lines

In general, macros like this should be avoided.

At the expense of repeating the same patterns. In this case it would be acceptable but it would mean one more indentation level for lines with the error/warning messages.

No if you move the check to lower layer.

Quoted 152 lines
Quoted 6 lines

net/ethtool/netlink.h line 25

Quoted 4 lines
net/ethtool/netlink.h +3 -0
+static inline int ethnl_str_size(const char *s)
+{
+	return nla_total_size(strlen(s) + 1);

This looks like more generic helper, not tight to ethtool.

OK

Quoted 3 lines
Quoted 1 line

+static inline int ethnl_str_ifne_size(const char *s)

Eh? "ifne"? What is this good for?

It's for "if not empty". Structures like ethtool_drvinfo use fixed size char arrays for strings where empty string means the information is not available. So if the string is empty, netlink attribute is omitted.

Quoted 5 lines
Quoted 5 lines
net/ethtool/netlink.h +5 -0
+{
+	return s[0] ? ethnl_str_size(s) : 0;
+}
+
+static inline int ethnl_put_str_ifne(struct sk_buff *skb, int attrtype,
Quoted 8 lines

net/ethtool/netlink.h line 38

Quoted 5 lines
net/ethtool/netlink.h +5 -0
+				     const char *s)
+{
+	if (!s[0])
+		return 0;
+	return nla_put_string(skb, attrtype, s);

I don't like helpers like this. Do the check in the caller and put or not put the string there. It's a single if.

OK

Quoted 6 lines

net/ethtool/netlink.h line 44

Quoted 4 lines
net/ethtool/netlink.h +4 -0
+static inline struct nlattr *ethnl_nest_start(struct sk_buff *skb,
+					      int attrtype)
+{
+	return nla_nest_start(skb, attrtype | NLA_F_NESTED);

Please use nla_nest_start directly and avoid helpers like this.

OK. It's one more thing to check (and forget) in many places, though.

Quoted 9 lines

net/ethtool/netlink.h line 53

Quoted 7 lines
net/ethtool/netlink.h +7 -0
+static inline int ethnlmsg_parse(const struct nlmsghdr *nlh,
+				 struct nlattr *tb[], int maxtype,
+				 const struct nla_policy *policy,
+				 struct genl_info *info)
+{
+	return nlmsg_parse_strict(nlh, GENL_HDRLEN, tb, maxtype, policy,
+				  info ? info->extack : NULL);

Same thing, please use nlmsg_parse_strict directly.

OK

Quoted 68 lines
Quoted 68 lines
net/ethtool/netlink.h +68 -0
+/* ethnl_update_* return true if the value is changed */
+static inline bool ethnl_update_u32(u32 *dst, struct nlattr *attr)
+{
+	u32 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u32(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_u8(u8 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = nla_get_u8(attr);
+	if (*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+/* update u32 value used as bool from NLA_U8 */
+static inline bool ethnl_update_bool32(u32 *dst, struct nlattr *attr)
+{
+	u8 val;
+
+	if (!attr)
+		return false;
+	val = !!nla_get_u8(attr);
+	if (!!*dst == val)
+		return false;
+
+	*dst = val;
+	return true;
+}
+
+static inline bool ethnl_update_binary(u8 *dst, unsigned int len,
+				       struct nlattr *attr)
+{
+	if (!attr)
+		return false;
+	if (nla_len(attr) < len)
+		len = nla_len(attr);
+	if (!memcmp(dst, nla_data(attr), len))
+		return false;
+
+	memcpy(dst, nla_data(attr), len);
+	return true;
+}
+
+static inline bool ethnl_update_bitfield32(u32 *dst, struct nlattr *attr)
+{
+	struct nla_bitfield32 change;
+	u32 newval;
+
+	if (!attr)
+		return false;
+	change = nla_get_bitfield32(attr);
+	newval = (*dst & ~change.selector) | (change.value & change.selector);
+	if (*dst == newval)
Quoted 8 lines

net/ethtool/netlink.h line 128

Quoted 5 lines
net/ethtool/netlink.h +5 -0
+		return false;
+
+	*dst = newval;
+	return true;
+}

I don't understand puspose of these "update" helper functions. Try to avoid them. In general, please try to avoid wrappers around netlink api.

The purpose is to update a value according to a request attribute (if present) and tell caller if the value was actually modified. They are used like e.g.

	mod = false;
	if (ethnl_update_u32(&data.rx_pending, tb[ETHA_RING_RX_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.rx_mini_pending,
			     tb[ETHA_RING_RX_MINI_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.rx_jumbo_pending,
			     tb[ETHA_RING_RX_JUMBO_PENDING]))
		mod = true;
	if (ethnl_update_u32(&data.tx_pending, tb[ETHA_RING_TX_PENDING]))
		mod = true;

	if (!mod)
		return 0;
	/* call ethtool_ops handler and send notification */
Quoted 3 lines

so that we can omit calling the handler and sending a notification if the request was no-op. Expanding the helpers here would make the code needlessly complicated and error prone.

Got it. Makes sense.

Quoted 18 lines
Quoted 17 lines
Quoted 5 lines
net/ethtool/netlink.h +4 -0
+static inline void warn_partial_info(struct genl_info *info)
+{
+	ETHNL_SET_ERRMSG(info, "not all requested data could be retrieved");
+}

net/ethtool/netlink.h line 143

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+
+/* Check user privileges explicitly to allow finer access control based on
+ * context of the request or hiding part of the information from unprivileged
+ * users
+ */
+static inline bool ethnl_is_privileged(struct sk_buff *skb)
+{
+	struct net *net = sock_net(skb->sk);
+
+	return netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN);

Same here.

Quoted 3 lines

It's still the same, helpers allow replacing repeating code patterns with just saying what you are doing. But this one is called only from one place so I can live without it.

It is always a tradeoff. We have to be careful not to abuse it.

Quoted 14 lines
Quoted 10 lines

net/ethtool/netlink.h line 150

Quoted 8 lines
net/ethtool/netlink.h +7 -0
+}
+
+/* total size of ETHA_*_DEV nested attribute; this is an upper estimate so that
+ * we do not need to hold RTNL longer than necessary to prevent rename between
+ * estimating the size and composing the message
+ */
+static inline unsigned int dev_ident_size(void)

You are missing "ethnl_" prefix.

OK

Michal

#34 lore raw

PATCH v5 7/22 ethtool: netlink bitset handling

In reply to Michal Kubecek

Declare attribute type constants and add helper functions to generate and parse arbitrary length bit sets.

---
 Documentation/networking/ethtool-netlink.txt |  63 ++
 include/uapi/linux/ethtool_netlink.h         |  32 +
 net/ethtool/Makefile                         |   2 +-
 net/ethtool/bitset.c                         | 597 +++++++++++++++++++
 net/ethtool/bitset.h                         |  40 ++
 net/ethtool/netlink.h                        |   9 +
 6 files changed, 742 insertions(+), 1 deletion(-)
 create mode 100644 net/ethtool/bitset.c
 create mode 100644 net/ethtool/bitset.h
Documentation/networking/ethtool-netlink.txt +63 -0
index 377d64c9b7fa..e97218d820c0 100644
@@ -60,6 +60,69 @@ replies with one message per network device (only those for which the request
 is supported).
 
 
+Bit sets
+--------
+
+For short bitmaps of (reasonably) fixed length, standard NLA_BITFIELD32 type
+is used. For arbitrary length bitmaps, ethtool netlink uses a nested attribute
+with contents of one of two forms: compact (two binary bitmaps representing
+bit values and mask of affected bits) and bit-by-bit (list of bits identified
+by either index or name).
+
+Compact form: nested (bitset) atrribute contents:
+
+    ETHA_BITSET_LIST	(flag)		no mask, only a list
+    ETHA_BITSET_SIZE	(u32)		number of significant bits
+    ETHA_BITSET_VALUE	(binary)	bitmap of bit values
+    ETHA_BITSET_MASK	(binary)	bitmap of valid bits
+
+Value and mask must have length at least ETHA_BITSET_SIZE bits rounded up to
+a multiple of 32 bits. They consist of 32-bit words in host byte order, words
+ordered from least significant to most significant (i.e. the same way as
+bitmaps are passed with ioctl interface).
+
+For compact form, ETHA_BITSET_SIZE and ETHA_BITSET_VALUE are mandatory.
+Similar to BITFIELD32, a compact form bit set requests to set bits in the mask
+to 1 (if the bit is set in value) or 0 (if not) and preserve the rest. If
+ETHA_BITSET_LIST is present, there is no mask and bitset represents a simple
+list of bits.
+
+Kernel bit set length may differ from userspace length if older application is
+used on newer kernel or vice versa. If userspace bitmap is longer, an error is
+issued only if the request actually tries to set values of some bits not
+recognized by kernel.
+
+Bit-by-bit form: nested (bitset) attribute contents:
+
+    ETHA_BITSET_LIST	(flag)		no mask, only a list
+    ETHA_BITSET_SIZE	(u32)		number of significant bits (optional)
+    ETHA_BITSET_BITS	(nested)	array of bits
+	ETHA_BITSET_BIT
+	    ETHA_BIT_INDEX	(u32)		bit index (0 for LSB)
+            ETHA_BIT_NAME	(string)	bit name
+	    ETHA_BIT_VALUE	(flag)		present if bit is set
+        ETHA_BITSET_BIT
+	...
+
+Bit size is optional for bit-by-bit form. ETHA_BITSET_BITS nest can only
+contain ETHA_BITS_BIT attributes but there can be an arbitrary number of them.
+A bit may be identified by its index or by its name. When used in requests,
+listed bits are set to 0 or 1 according to ETHA_BIT_VALUE, the rest is
+preserved. A request fails if index exceeds kernel bit length or if name is
+not recognized.
+
+When ETHA_BITSET_LIST flag is present, bitset is interpreted as a simple bit
+list. ETHA_BIT_VALUE attributes are not used in such case. Bit list represents
+a bitmap with listed bits set and the rest zero.
+
+In requests, application can use either form. Form used by kernel in reply is
+determined by a flag in flags field of request header. Semantics of value and
+mask depends on the attribute. General idea is that flags control request
+processing, info_mask control which parts of the information are returned in
+"get" request and index identifies a particular subcommand or an object to
+which the request applies.
+
+
 List of message types
 ---------------------
 
include/uapi/linux/ethtool_netlink.h +32 -0
index 59240a2cda56..de18e076ed69 100644
@@ -29,6 +29,38 @@ enum {
 	ETHA_DEV_MAX = (__ETHA_DEV_CNT - 1)
 };
 
+/* bit sets */
+
+enum {
+	ETHA_BIT_UNSPEC,
+	ETHA_BIT_INDEX,				/* u32 */
+	ETHA_BIT_NAME,				/* string */
+	ETHA_BIT_VALUE,				/* flag */
+
+	__ETHA_BIT_CNT,
+	ETHA_BIT_MAX = (__ETHA_BIT_CNT - 1)
+};
+
+enum {
+	ETHA_BITS_UNSPEC,
+	ETHA_BITS_BIT,
+
+	__ETHA_BITS_CNT,
+	ETHA_BITS_MAX = (__ETHA_BITS_CNT - 1)
+};
+
+enum {
+	ETHA_BITSET_UNSPEC,
+	ETHA_BITSET_LIST,			/* flag */
+	ETHA_BITSET_SIZE,			/* u32 */
+	ETHA_BITSET_BITS,			/* nest - ETHA_BITS_* */
+	ETHA_BITSET_VALUE,			/* binary */
+	ETHA_BITSET_MASK,			/* binary */
+
+	__ETHA_BITSET_CNT,
+	ETHA_BITSET_MAX = (__ETHA_BITSET_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/Makefile +1 -1
index f30e0da88be5..482fdb9380fa 100644
@@ -4,4 +4,4 @@ obj-y				+= ioctl.o
 
 obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
 
-ethtool_nl-y	:= netlink.o
+ethtool_nl-y	:= netlink.o bitset.o
net/ethtool/bitset.c new file +597 -0

597 lines not on this page · raw message

net/ethtool/bitset.h new file +40 -0
new file mode 100644
index 000000000000..761d0c47fe23
@@ -0,0 +1,40 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _NET_ETHTOOL_BITSET_H
+#define _NET_ETHTOOL_BITSET_H
+
+/* when set, value and mask bitmaps are arrays of u32, when not, arrays of
+ * unsigned long
+ */
+#define ETHNL_BITSET_U32		BIT(0)
+/* generate a compact format bitset */
+#define ETHNL_BITSET_COMPACT		BIT(1)
+/* generate a bit list */
+#define ETHNL_BITSET_LIST		BIT(2)
+/* when set, names are interpreted as legacy string set (an array of
+ * char[ETH_GSTRING_LEN]), when not, as a simple array of char *
+ */
+#define ETHNL_BITSET_LEGACY_NAMES	BIT(3)
+
+int ethnl_bitset_is_compact(const struct nlattr *bitset, bool *compact);
+int ethnl_bitset_size(unsigned int size, const unsigned long *val,
+		      const unsigned long *mask, const void *names,
+		      unsigned int flags);
+int ethnl_bitset32_size(unsigned int size, const u32 *val, const u32 *mask,
+			const void *names, unsigned int flags);
+int ethnl_put_bitset(struct sk_buff *skb, int attrtype, unsigned int size,
+		     const unsigned long *val, const unsigned long *mask,
+		     const void *names, unsigned int flags);
+int ethnl_put_bitset32(struct sk_buff *skb, int attrtype, unsigned int size,
+		       const u32 *val, const u32 *mask, const void *names,
+		       unsigned int flags);
+bool ethnl_update_bitset(unsigned long *bitmap, unsigned long *bitmask,
+			 unsigned int nbits, const struct nlattr *attr,
+			 int *err, const void *names, bool legacy,
+			 struct genl_info *info);
+bool ethnl_update_bitset32(u32 *bitmap, u32 *bitmask, unsigned int nbits,
+			   const struct nlattr *attr, int *err,
+			   const void *names, bool legacy,
+			   struct genl_info *info);
+
+#endif /* _NET_ETHTOOL_BITSET_H */
net/ethtool/netlink.h +9 -0
index db90d95410b1..b8a6cd3dc3e3 100644
@@ -20,6 +20,15 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 				 u16 dev_attrtype, struct genl_info *info,
 				 void **ehdrp);
 
+#if BITS_PER_LONG == 64 && defined(__BIG_ENDIAN)
+void ethnl_bitmap_to_u32(unsigned long *bitmap, unsigned int nwords);
+#else
+static inline void ethnl_bitmap_to_u32(unsigned long *bitmap,
+				       unsigned int nwords)
+{
+}
+#endif
+
 static inline int ethnl_str_size(const char *s)
 {
 	return nla_total_size(strlen(s) + 1);
Signature · 1 line
2.21.0
#35 lore raw

Re: [PATCH net-next v5 07/22] ethtool: netlink bitset handling

Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:15PM CET, [email protected] wrote:

Declare attribute type constants and add helper functions to generate and parse arbitrary length bit sets.

Hmm, this looks like a lot of work. Two things:
1) This is generic. Not really related to ethtool in any way. Could this be done in netlink common code?
2) Did you think about leveraging NLA_BITFIELD32? What I mean is this:
NEST_START
NLA_BITFIELD32 index 0 (bit 0-31)
NLA_BITFIELD32 index 1 (bit 32-63)
NLA_BITFIELD32 index 2 (bit 64-95)
NLA_BITFIELD32 index 3 (bit 96-127)
....
NEST_END
It is basically an array of NLA_BITFIELD32.

#36 lore raw

Re: [PATCH net-next v5 07/22] ethtool: netlink bitset handling

Quoted 7 lines · Jiri Pirko
Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:15PM CET, [email protected] wrote:

Declare attribute type constants and add helper functions to generate and parse arbitrary length bit sets.

Hmm, this looks like a lot of work. Two things:
1) This is generic. Not really related to ethtool in any way. Could this be done in netlink common code?

I suppose it could if other netlink based APIs would be interested in using it. The only ethtool specific part is the support for "legacy style names" (fixed size strings) but that is something I'm not really happy about. Perhaps it's time to return to the original idea of supporting only arrays of (char *) and creating them around existing fixed size ones.

Quoted 9 lines

2) Did you think about leveraging NLA_BITFIELD32? What I mean is this:
NEST_START
NLA_BITFIELD32 index 0 (bit 0-31)
NLA_BITFIELD32 index 1 (bit 32-63)
NLA_BITFIELD32 index 2 (bit 64-95)
NLA_BITFIELD32 index 3 (bit 96-127)
....
NEST_END
It is basically an array of NLA_BITFIELD32.

That would be similar to compact form but it would introduce the interleaving and extra struct nlattr header for each block. I don't think it would make things easier.

The verbose form is meant to allow requests like

ethtool -K eth0 tx-vlan-stag-hw-insert on
ethtool -s eth0 advertise 1000baseT/Full off

without either keeping the table of available flags in sync between kernel and userspace (all userspace users of the API) or having to ask for the list first (in case of one shot requests as above; long running tools like "ethtool --monitor" or config management daemons would keep the tables and use compact form).

Michal

#37 lore raw

Re: [PATCH net-next v5 07/22] ethtool: netlink bitset handling

Quoted 9 lines · Tue, Mar 26, 2019 at 06:59:32PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko
Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:15PM CET, [email protected] wrote:

Declare attribute type constants and add helper functions to generate and parse arbitrary length bit sets.

Hmm, this looks like a lot of work. Two things:
1) This is generic. Not really related to ethtool in any way. Could this be done in netlink common code?

Quoted 6 lines

I suppose it could if other netlink based APIs would be interested in using it. The only ethtool specific part is the support for "legacy style names" (fixed size strings) but that is something I'm not really happy about. Perhaps it's time to return to the original idea of supporting only arrays of (char *) and creating them around existing fixed size ones.

Wait, could you please describe this more?

Quoted 11 lines
Quoted 10 lines

2) Did you think about leveraging NLA_BITFIELD32? What I mean is this:
NEST_START
NLA_BITFIELD32 index 0 (bit 0-31)
NLA_BITFIELD32 index 1 (bit 32-63)
NLA_BITFIELD32 index 2 (bit 64-95)
NLA_BITFIELD32 index 3 (bit 96-127)
....
NEST_END
It is basically an array of NLA_BITFIELD32.

Quoted 3 lines

That would be similar to compact form but it would introduce the interleaving and extra struct nlattr header for each block. I don't think it would make things easier.

I'm just trying to figure out if it would be possible to use existing building block instead of introducing new ones....

Quoted 11 lines

The verbose form is meant to allow requests like

ethtool -K eth0 tx-vlan-stag-hw-insert on
ethtool -s eth0 advertise 1000baseT/Full off

without either keeping the table of available flags in sync between kernel and userspace (all userspace users of the API) or having to ask for the list first (in case of one shot requests as above; long running tools like "ethtool --monitor" or config management daemons would keep the tables and use compact form).

Okay, I have to go over those patches but this smells.

Quoted 2 lines

Michal

#38 lore raw

Re: [PATCH net-next v5 07/22] ethtool: netlink bitset handling

Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Tue, Mar 26, 2019 at 06:59:32PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko
Quoted 2 lines · Mon, Mar 25, 2019 at 06:08:15PM CET, [email protected] wrote:

Declare attribute type constants and add helper functions to generate and parse arbitrary length bit sets.

Hmm, this looks like a lot of work. Two things:
1) This is generic. Not really related to ethtool in any way. Could this be done in netlink common code?

Quoted 8 lines
Quoted 6 lines

I suppose it could if other netlink based APIs would be interested in using it. The only ethtool specific part is the support for "legacy style names" (fixed size strings) but that is something I'm not really happy about. Perhaps it's time to return to the original idea of supporting only arrays of (char *) and creating them around existing fixed size ones.

Wait, could you please describe this more?

At the moment, there are two formats of bit name tables. Legacy format which is used by ioctl string sets, e.g.

const char netdev_features_strings[NETDEV_FEATURE_COUNT][ETH_GSTRING_LEN] = {
	...
};

(the same is provided by NIC drivers with ethtool_ops::get_strings()) and simple format used for newly added string sets, e.g.

const char *const link_mode_names[] = {
	...
};

If the bitset implementation is to be used by other APIs, legacy format support would make little sense so I would suggest to drop legacy format support from bitset code and build a (const char **) "index" for ethtool legacy string arrays before passing it to bitset related functions.

Michal

#39 lore raw

PATCH v5 8/22 ethtool: support for netlink notifications

In reply to Michal Kubecek

Add infrastructure for ethtool netlink notifications. There is only one multicast group "monitor" which is used to notify userspace about changes. Notifications are supposed to be broadcasted on every configuration change, whether it is done using the netlink interface or legacy ioctl one.

To trigger an ethtool notification, both ethtool netlink and external code use ethtool_notify() helper. This helper requires RTNL to be held and may sleep.

---
 include/linux/ethtool_netlink.h      |  5 ++++
 include/linux/netdevice.h            | 12 +++++++++
 include/uapi/linux/ethtool_netlink.h |  2 ++
 net/ethtool/netlink.c                | 37 ++++++++++++++++++++++++++++
 net/ethtool/netlink.h                |  2 ++
 5 files changed, 58 insertions(+)
include/linux/ethtool_netlink.h +5 -0
index 0412adb4f42f..2a15e64a16f3 100644
@@ -5,5 +5,10 @@
 
 #include <uapi/linux/ethtool_netlink.h>
 #include <linux/ethtool.h>
+#include <linux/netdevice.h>
+
+enum ethtool_multicast_groups {
+	ETHNL_MCGRP_MONITOR,
+};
 
 #endif /* _LINUX_ETHTOOL_NETLINK_H_ */
include/linux/netdevice.h +12 -0
index 166fdc0a78b4..bc761511edb4 100644
@@ -4353,6 +4353,18 @@ struct netdev_notifier_bonding_info {
 void netdev_bonding_info_change(struct net_device *dev,
 				struct netdev_bonding_info *bonding_info);
 
+#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data);
+#else
+static inline void ethtool_notify(struct net_device *dev,
+				  struct netlink_ext_ack *extack,
+				  unsigned int cmd, u32 req_mask,
+				  const void *data)
+{
+}
+#endif
+
 static inline
 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
 {
include/uapi/linux/ethtool_netlink.h +2 -0
index de18e076ed69..91e4d117957b 100644
@@ -65,4 +65,6 @@ enum {
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
 
+#define ETHTOOL_MCGRP_MONITOR_NAME "monitor"
+
 #endif /* _UAPI_LINUX_ETHTOOL_NETLINK_H_ */
net/ethtool/netlink.c +37 -0
index cc6829ba4331..4a31765165ea 100644
@@ -4,6 +4,10 @@
 #include <linux/ethtool_netlink.h>
 #include "netlink.h"
 
+u32 ethnl_bcast_seq;
+
+static bool ethnl_ok __read_mostly;
+
 static const struct nla_policy dev_policy[ETHA_DEV_MAX + 1] = {
 	[ETHA_DEV_UNSPEC]	= { .type = NLA_REJECT },
 	[ETHA_DEV_INDEX]	= { .type = NLA_U32 },
@@ -147,11 +151,41 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 	return NULL;
 }
 
+/* notifications */
+
+typedef void (*ethnl_notify_handler_t)(struct net_device *dev,
+				       struct netlink_ext_ack *extack,
+				       unsigned int cmd, u32 req_mask,
+				       const void *data);
+
+ethnl_notify_handler_t ethnl_notify_handlers[] = {
+};
+
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))
+		return;
+	ASSERT_RTNL();
+
+	if (likely(cmd < ARRAY_SIZE(ethnl_notify_handlers) &&
+		   ethnl_notify_handlers[cmd]))
+		ethnl_notify_handlers[cmd](dev, extack, cmd, req_mask, data);
+	else
+		WARN_ONCE(1, "notification %u not implemented (dev=%s, req_mask=0x%x)\n",
+			  cmd, netdev_name(dev), req_mask);
+}
+EXPORT_SYMBOL(ethtool_notify);
+
 /* genetlink setup */
 
 static const struct genl_ops ethtool_genl_ops[] = {
 };
 
+static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
+	[ETHNL_MCGRP_MONITOR] = { .name = ETHTOOL_MCGRP_MONITOR_NAME },
+};
+
 struct genl_family ethtool_genl_family = {
 	.hdrsize	= 0,
 	.name		= ETHTOOL_GENL_NAME,
@@ -160,6 +194,8 @@ struct genl_family ethtool_genl_family = {
 	.parallel_ops	= true,
 	.ops		= ethtool_genl_ops,
 	.n_ops		= ARRAY_SIZE(ethtool_genl_ops),
+	.mcgrps		= ethtool_nl_mcgrps,
+	.n_mcgrps	= ARRAY_SIZE(ethtool_nl_mcgrps),
 };
 
 /* module setup */
@@ -171,6 +207,7 @@ static int __init ethnl_init(void)
 	ret = genl_register_family(&ethtool_genl_family);
 	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))
 		return ret;
+	ethnl_ok = true;
 
 	return 0;
 }
net/ethtool/netlink.h +2 -0
index b8a6cd3dc3e3..5f2299548915 100644
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
+extern u32 ethnl_bcast_seq;
+
 extern struct genl_family ethtool_genl_family;
 
 struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest);
Signature · 1 line
2.21.0
#40 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 90 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:

Add infrastructure for ethtool netlink notifications. There is only one multicast group "monitor" which is used to notify userspace about changes. Notifications are supposed to be broadcasted on every configuration change, whether it is done using the netlink interface or legacy ioctl one.

To trigger an ethtool notification, both ethtool netlink and external code use ethtool_notify() helper. This helper requires RTNL to be held and may sleep.

---

include/linux/ethtool_netlink.h      |  5 ++++
include/linux/netdevice.h            | 12 +++++++++
include/uapi/linux/ethtool_netlink.h |  2 ++
net/ethtool/netlink.c                | 37 ++++++++++++++++++++++++++++
net/ethtool/netlink.h                |  2 ++
5 files changed, 58 insertions(+)
include/linux/ethtool_netlink.h +5 -0
index 0412adb4f42f..2a15e64a16f3 100644
@@ -5,5 +5,10 @@
 
 #include <uapi/linux/ethtool_netlink.h>
 #include <linux/ethtool.h>
+#include <linux/netdevice.h>
+
+enum ethtool_multicast_groups {
+	ETHNL_MCGRP_MONITOR,
+};
 
 #endif /* _LINUX_ETHTOOL_NETLINK_H_ */
include/linux/netdevice.h +12 -0
index 166fdc0a78b4..bc761511edb4 100644
@@ -4353,6 +4353,18 @@ struct netdev_notifier_bonding_info {
 void netdev_bonding_info_change(struct net_device *dev,
 				struct netdev_bonding_info *bonding_info);
 
+#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data);
+#else
+static inline void ethtool_notify(struct net_device *dev,
+				  struct netlink_ext_ack *extack,
+				  unsigned int cmd, u32 req_mask,
+				  const void *data)
+{
+}
+#endif
+

static inline
struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) {

include/uapi/linux/ethtool_netlink.h +2 -0
index de18e076ed69..91e4d117957b 100644
@@ -65,4 +65,6 @@ enum {
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
 
+#define ETHTOOL_MCGRP_MONITOR_NAME "monitor"
+
 #endif /* _UAPI_LINUX_ETHTOOL_NETLINK_H_ */
net/ethtool/netlink.c +8 -0
index cc6829ba4331..4a31765165ea 100644
@@ -4,6 +4,10 @@
 #include <linux/ethtool_netlink.h>
 #include "netlink.h"
 
+u32 ethnl_bcast_seq;
+
+static bool ethnl_ok __read_mostly;
+
 static const struct nla_policy dev_policy[ETHA_DEV_MAX + 1] = {
 	[ETHA_DEV_UNSPEC]	= { .type = NLA_REJECT },
 	[ETHA_DEV_INDEX]	= { .type = NLA_U32 },
@@ -147,11 +151,41 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 	return NULL;
 }
 
+/* notifications */
+
+typedef void (*ethnl_notify_handler_t)(struct net_device *dev,
+				       struct netlink_ext_ack *extack,

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 10 lines
net/ethtool/netlink.c +10 -0
+				       unsigned int cmd, u32 req_mask,
+				       const void *data);
+
+ethnl_notify_handler_t ethnl_notify_handlers[] = {
+};
+
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 178

Quoted 11 lines
net/ethtool/netlink.c +11 -0
+		return;
+	ASSERT_RTNL();
+
+	if (likely(cmd < ARRAY_SIZE(ethnl_notify_handlers) &&
+		   ethnl_notify_handlers[cmd]))
+		ethnl_notify_handlers[cmd](dev, extack, cmd, req_mask, data);
+	else
+		WARN_ONCE(1, "notification %u not implemented (dev=%s, req_mask=0x%x)\n",
+			  cmd, netdev_name(dev), req_mask);
+}
+EXPORT_SYMBOL(ethtool_notify);

Please be consistent with prefixes. Rest of the code (most of it) has prefix "ethnl".

Quoted 35 lines
net/ethtool/netlink.c +8 -0
+
 /* genetlink setup */
 
 static const struct genl_ops ethtool_genl_ops[] = {
 };
 
+static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
+	[ETHNL_MCGRP_MONITOR] = { .name = ETHTOOL_MCGRP_MONITOR_NAME },
+};
+
 struct genl_family ethtool_genl_family = {
 	.hdrsize	= 0,
 	.name		= ETHTOOL_GENL_NAME,
@@ -160,6 +194,8 @@ struct genl_family ethtool_genl_family = {
 	.parallel_ops	= true,
 	.ops		= ethtool_genl_ops,
 	.n_ops		= ARRAY_SIZE(ethtool_genl_ops),
+	.mcgrps		= ethtool_nl_mcgrps,
+	.n_mcgrps	= ARRAY_SIZE(ethtool_nl_mcgrps),
 };
 
 /* module setup */
@@ -171,6 +207,7 @@ static int __init ethnl_init(void)
 	ret = genl_register_family(&ethtool_genl_family);
 	if (WARN(ret < 0, "ethtool: genetlink family registration failed"))
 		return ret;
+	ethnl_ok = true;
 
 	return 0;
 }
net/ethtool/netlink.h +0 -0
index b8a6cd3dc3e3..5f2299548915 100644

net/ethtool/netlink.h line 14

Quoted 4 lines
net/ethtool/netlink.h +1 -0
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
+extern u32 ethnl_bcast_seq;

Why do you need to have this in header? Second, it is not used by anything. Please don't introduce variables that are not used. Introduce them only in patch where you use it.

Quoted 7 lines
net/ethtool/netlink.h +1 -0
+
 extern struct genl_family ethtool_genl_family;
 
 struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest);
Signature · 1 line
2.21.0
#41 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Quoted 6 lines
Quoted 6 lines
net/ethtool/netlink.c +6 -0
+		return;
+	ASSERT_RTNL();
+
+	if (likely(cmd < ARRAY_SIZE(ethnl_notify_handlers) &&
+		   ethnl_notify_handlers[cmd]))
+		ethnl_notify_handlers[cmd](dev, extack, cmd, req_mask, data);
Quoted 8 lines

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 178

Quoted 5 lines
net/ethtool/netlink.c +5 -0
+	else
+		WARN_ONCE(1, "notification %u not implemented (dev=%s, req_mask=0x%x)\n",
+			  cmd, netdev_name(dev), req_mask);
+}
+EXPORT_SYMBOL(ethtool_notify);

Please be consistent with prefixes. Rest of the code (most of it) has prefix "ethnl".

The reason why I used ethtool_ prefix here was that this function is an entry point which can be called from anywhere and other code doesn't care if we are using netlink or something else. But as any caller would need to include <linux/ethtool_netlink.h> anyway to get the constants for cmd and req_mask, there is no point hiding the fact. I'll rename it to ethnl_notify().

Quoted 8 lines
Quoted 8 lines
net/ethtool/netlink.h +0 -0
index b8a6cd3dc3e3..5f2299548915 100644
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
Quoted 5 lines

net/ethtool/netlink.h line 14

Quoted 1 line
net/ethtool/netlink.h +1 -0
+extern u32 ethnl_bcast_seq;

Why do you need to have this in header? Second, it is not used by anything. Please don't introduce variables that are not used. Introduce them only in patch where you use it.

It's the same as with ethtool_genl_family. I'll make it static as well until it's used in some other file.

Michal

#42 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 9 lines · Tue, Mar 26, 2019 at 07:17:20PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

Quoted 4 lines

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Again, haven't seen this in any other gen netlink implementation. Why do they not need it?

Quoted 16 lines
Quoted 15 lines

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 178

Quoted 12 lines
net/ethtool/netlink.c +11 -0
+		return;
+	ASSERT_RTNL();
+
+	if (likely(cmd < ARRAY_SIZE(ethnl_notify_handlers) &&
+		   ethnl_notify_handlers[cmd]))
+		ethnl_notify_handlers[cmd](dev, extack, cmd, req_mask, data);
+	else
+		WARN_ONCE(1, "notification %u not implemented (dev=%s, req_mask=0x%x)\n",
+			  cmd, netdev_name(dev), req_mask);
+}
+EXPORT_SYMBOL(ethtool_notify);

Please be consistent with prefixes. Rest of the code (most of it) has prefix "ethnl".

Quoted 6 lines

The reason why I used ethtool_ prefix here was that this function is an entry point which can be called from anywhere and other code doesn't care if we are using netlink or something else. But as any caller would need to include <linux/ethtool_netlink.h> anyway to get the constants for cmd and req_mask, there is no point hiding the fact. I'll rename it to ethnl_notify().

Okay.

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
net/ethtool/netlink.h +0 -0
index b8a6cd3dc3e3..5f2299548915 100644
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
Quoted 8 lines
Quoted 5 lines

net/ethtool/netlink.h line 14

Quoted 1 line
net/ethtool/netlink.h +1 -0
+extern u32 ethnl_bcast_seq;

Why do you need to have this in header? Second, it is not used by anything. Please don't introduce variables that are not used. Introduce them only in patch where you use it.

It's the same as with ethtool_genl_family. I'll make it static as well until it's used in some other file.

Not only static. You should remove it as you are not using it in this patch at all.

Quoted 2 lines

Michal

maintainer
#43 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Tue, Mar 26, 2019 at 07:17:20PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

Quoted 7 lines
Quoted 4 lines

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Again, haven't seen this in any other gen netlink implementation. Why do they not need it?

Hi Jiri

All other gen netlink kill the machine dead if they cannot register the socket.

Andrew

#44 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 16 lines · Wed, Mar 27, 2019 at 10:51:48AM CET, [email protected] wrote:
Quoted 15 lines · Jiri Pirko
Quoted 13 lines · Tue, Mar 26, 2019 at 07:17:20PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Quoted 6 lines
Quoted 2 lines

Again, haven't seen this in any other gen netlink implementation. Why do they not need it?

Hi Jiri

All other gen netlink kill the machine dead if they cannot register

Wait, what do you mean be "kill the machine dead"?

Quoted 1 line

the socket.

Example:
static int __init devlink_init(void)

{
        return genl_register_family(&devlink_nl_family);
}
maintainer
#45 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 1 line

Wait, what do you mean be "kill the machine dead"?

panic()

Andrew

#46 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 3 lines · Wed, Mar 27, 2019 at 11:16:28AM CET, [email protected] wrote:
Quoted 1 line

Wait, what do you mean be "kill the machine dead"?

panic()

I see panic only in net/ipv4/tcp_metrics.c.

Quoted 2 lines

Andrew

#47 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

In reply to Jiri Pirko
Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Tue, Mar 26, 2019 at 07:17:20PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

Quoted 7 lines
Quoted 4 lines

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Again, haven't seen this in any other gen netlink implementation. Why do they not need it?

Do they have notifications triggered from other code by directly calling a function (i.e. not through e.g. a netdev notifier)?

An alternative to a flag would be using a RCU pointer for this function (initialized to null and set once the family is registered) which is what e.g. netfilter is doing with its hooks but that would mean that each external notification would use an indirect call.

Quoted 14 lines
Quoted 14 lines
Quoted 13 lines

net/ethtool/netlink.h line 14

Quoted 9 lines
net/ethtool/netlink.h +1 -0
index b8a6cd3dc3e3..5f2299548915 100644
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
+extern u32 ethnl_bcast_seq;

Why do you need to have this in header? Second, it is not used by anything. Please don't introduce variables that are not used. Introduce them only in patch where you use it.

Quoted 5 lines
Quoted 2 lines

It's the same as with ethtool_genl_family. I'll make it static as well until it's used in some other file.

Not only static. You should remove it as you are not using it in this patch at all.

OK

Michal

#48 lore raw

Re: [PATCH net-next v5 08/22] ethtool: support for netlink notifications

Quoted 16 lines · Wed, Mar 27, 2019 at 10:59:50AM CET, [email protected] wrote:
Quoted 15 lines · Jiri Pirko
Quoted 13 lines · Tue, Mar 26, 2019 at 07:17:20PM CET, [email protected] wrote:
Quoted 7 lines · Jiri Pirko

net/ethtool/netlink.c struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd, · line 167

Quoted 4 lines · Mon, Mar 25, 2019 at 06:08:18PM CET, [email protected] wrote:
net/ethtool/netlink.c +4 -0
+void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
+		    unsigned int cmd, u32 req_mask, const void *data)
+{
+	if (unlikely(!ethnl_ok))

Why do you need this?

If genetlink family registration fails, ethtool_notify() can be still called from other code (e.g. the ethtool ioctl interface). In such case, better bail out right away than fail somewhere later (probably after preparing the message which can't be sent anyway).

Quoted 5 lines
Quoted 2 lines

Again, haven't seen this in any other gen netlink implementation. Why do they not need it?

Do they have notifications triggered from other code by directly calling a function (i.e. not through e.g. a netdev notifier)?

Okay, got it.

Quoted 5 lines

An alternative to a flag would be using a RCU pointer for this function (initialized to null and set once the family is registered) which is what e.g. netfilter is doing with its hooks but that would mean that each external notification would use an indirect call.

It's slow path, I don't think this kind of overhead is of any issue. Sounds good.

Quoted 24 lines
Quoted 15 lines
Quoted 15 lines
Quoted 14 lines

net/ethtool/netlink.h line 14

Quoted 10 lines
net/ethtool/netlink.h +1 -0
index b8a6cd3dc3e3..5f2299548915 100644
@@ -11,6 +11,8 @@
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
+extern u32 ethnl_bcast_seq;

Why do you need to have this in header? Second, it is not used by anything. Please don't introduce variables that are not used. Introduce them only in patch where you use it.

Quoted 5 lines
Quoted 2 lines

It's the same as with ethtool_genl_family. I'll make it static as well until it's used in some other file.

Not only static. You should remove it as you are not using it in this patch at all.

OK

Michal

#49 lore raw

PATCH v5 9/22 ethtool: implement EVENT notifications

In reply to Michal Kubecek

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

---
 Documentation/networking/ethtool-netlink.txt | 27 ++++++++
 include/uapi/linux/ethtool_netlink.h         | 37 +++++++++++
 net/ethtool/netlink.c                        | 65 +++++++++++++++++++-
 3 files changed, 128 insertions(+), 1 deletion(-)
Documentation/networking/ethtool-netlink.txt +27 -0
index e97218d820c0..5e5d785fe215 100644
@@ -126,6 +126,8 @@ which the request applies.
 List of message types
 ---------------------
 
+    ETHNL_CMD_EVENT			notification only
+
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
 
@@ -137,9 +139,34 @@ messages marked as "response only" in the table above. "Get" messages with
 NLM_F_DUMP flags and no device identification dump the information for all
 devices supporting the request.
 
+Type ETHNL_CMD_EVENT is special, these messages are never used in userspace
+requests or kernel replies. They are only sent by kernel to sockets listening
+to "monitor" multicast group to inform userspace about certain events.
+
 Later sections describe the format and semantics of these request messages.
 
 
+EVENT
+-----
+
+EVENT messages are only used in kernel multicast notifications. Atributes
+correspond to specific event types, the same type can appear multiple times.
+
+    ETHA_EVENT_NEWDEV		(nested)	new device was registered
+       ETHA_NEWDEV_DEV			(nested)	new device
+    ETHA_EVENT_DELDEV		(nested)	device was unregistered
+       ETHA_DELDEV_DEV			(nested)	removed device
+    ETHA_EVENT_RENAMEDEV	(nested)	device was renamed
+       ETHA_RENAMEDEV_DEV		(nested)	renamed device
+
+For ETHA_EVENT_RENAMEDEV, the name ETHA_RENAME_DEV/ETHA_DEV_NAME is the new
+name after the rename.
+
+Userspace application must expect multiple events to be present in one message
+and also multiple events of the same type (e.g. two or more newly registered
+devices).
+
+
 Request translation
 -------------------
 
include/uapi/linux/ethtool_netlink.h +37 -0
index 91e4d117957b..988519bc6e37 100644
@@ -13,6 +13,7 @@
 
 enum {
 	ETHNL_CMD_NOOP,
+	ETHNL_CMD_EVENT,		/* only for notifications */
 
 	__ETHNL_CMD_CNT,
 	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
@@ -61,6 +62,42 @@ enum {
 	ETHA_BITSET_MAX = (__ETHA_BITSET_CNT - 1)
 };
 
+/* events */
+
+enum {
+	ETHA_NEWDEV_UNSPEC,
+	ETHA_NEWDEV_DEV,			/* nest - ETHA_DEV_* */
+
+	__ETHA_NEWDEV_CNT,
+	ETHA_NEWDEV_MAX = (__ETHA_NEWDEV_CNT - 1)
+};
+
+enum {
+	ETHA_DELDEV_UNSPEC,
+	ETHA_DELDEV_DEV,			/* nest - ETHA_DEV_* */
+
+	__ETHA_DELDEV_CNT,
+	ETHA_DELDEV_MAX = (__ETHA_DELDEV_CNT - 1)
+};
+
+enum {
+	ETHA_RENAMEDEV_UNSPEC,
+	ETHA_RENAMEDEV_DEV,			/* nest - ETHA_DEV_* */
+
+	__ETHA_RENAMEDEV_CNT,
+	ETHA_RENAMEDEV_MAX = (__ETHA_RENAMEDEV_CNT - 1)
+};
+
+enum {
+	ETHA_EVENT_UNSPEC,
+	ETHA_EVENT_NEWDEV,			/* nest - ETHA_NEWDEV_* */
+	ETHA_EVENT_DELDEV,			/* nest - ETHA_DELDEV_* */
+	ETHA_EVENT_RENAMEDEV,			/* nest - ETHA_RENAMEDEV_* */
+
+	__ETHA_EVENT_CNT,
+	ETHA_EVENT_MAX = (__ETHA_EVENT_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/netlink.c +64 -1

77 lines not on this page · raw message

Signature · 1 line
2.21.0
#50 lore raw

Re: [PATCH net-next v5 09/22] ethtool: implement EVENT notifications

Quoted 6 lines · Mon, Mar 25, 2019 at 06:08:21PM CET, [email protected] wrote:

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

Quoted 6 lines

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

Wait. You duplicate events that are already going out through RTNETLINK. App should open RTNETLINK in order to get those. Other apps are doing that too. I don't think that duplications like this are desirable :/

#51 lore raw

Re: [PATCH net-next v5 09/22] ethtool: implement EVENT notifications

Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:21PM CET, [email protected] wrote:

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

Quoted 7 lines
Quoted 3 lines

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

Wait. You duplicate events that are already going out through RTNETLINK. App should open RTNETLINK in order to get those. Other apps are doing that too. I don't think that duplications like this are desirable :/

Is there a way to filter or at least recognize these events when using rtnetlink? I couldn't find any. The only way seems to be getting every RTM_NEWLINK message (there can be quite a lot of those), always perform the lookup in my device list and recognize what happened - only to almost always find that nothing interesting. It is possible, sure, but I would really like to avoid it.

Michal

#52 lore raw

Re: [PATCH net-next v5 09/22] ethtool: implement EVENT notifications

Quoted 19 lines · Michal Kubecek
Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:21PM CET, [email protected] wrote:

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

Quoted 7 lines
Quoted 3 lines

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

Wait. You duplicate events that are already going out through RTNETLINK. App should open RTNETLINK in order to get those. Other apps are doing that too. I don't think that duplications like this are desirable :/

Quoted 6 lines

Is there a way to filter or at least recognize these events when using rtnetlink? I couldn't find any. The only way seems to be getting every RTM_NEWLINK message (there can be quite a lot of those), always perform the lookup in my device list and recognize what happened - only to almost always find that nothing interesting. It is possible, sure, but I would really like to avoid it.

I am afraid you are right about this, would adding a filtering capability specifically for this in rtnetlink be a better route?

Signature · 1 line
Florian
#53 lore raw

Re: [PATCH net-next v5 09/22] ethtool: implement EVENT notifications

Quoted 29 lines · Florian Fainelli
Quoted 19 lines · Michal Kubecek
Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:21PM CET, [email protected] wrote:

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

Quoted 7 lines
Quoted 3 lines

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

Wait. You duplicate events that are already going out through RTNETLINK. App should open RTNETLINK in order to get those. Other apps are doing that too. I don't think that duplications like this are desirable :/

Quoted 6 lines

Is there a way to filter or at least recognize these events when using rtnetlink? I couldn't find any. The only way seems to be getting every RTM_NEWLINK message (there can be quite a lot of those), always perform the lookup in my device list and recognize what happened - only to almost always find that nothing interesting. It is possible, sure, but I would really like to avoid it.

Quoted 2 lines

I am afraid you are right about this, would adding a filtering capability specifically for this in rtnetlink be a better route?

Maybe we could add new IFLA_EVENT_* values and use IFLA_EVENT to mark RTM_NEWLINK messages announcing "new device" and "device rename". That way, monitoring application would still need to parse all RTM_NEWLINK messages but it would be able to recognize which announce a change in device list without a lookup in its structures.

Michal

#54 lore raw

Re: [PATCH net-next v5 09/22] ethtool: implement EVENT notifications

Quoted 30 lines · Thu, Mar 28, 2019 at 07:41:43AM CET, [email protected] wrote:
Quoted 29 lines · Florian Fainelli
Quoted 25 lines · Michal Kubecek
Quoted 10 lines · Jiri Pirko
Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:21PM CET, [email protected] wrote:

Three types of netlink notifications are introduced:

- ETHA_EVENT_NEWDEV to notify about newly registered network devices
- ETHA_EVENT_DELDEV to notify about unregistered network devices
- ETHA_EVENT_RENAMEDEV to notify about renamed network device

The notifications are triggered by NETDEV_REGISTER, NETDEV_UNREGISTER and NETDEV_CHANGENAME notifiers.

Quoted 7 lines
Quoted 3 lines

These notifications are intended for applications and daemons monitoring ethtool events to allow updating the list of existing devices without having to open another socket for rtnetlink.

Wait. You duplicate events that are already going out through RTNETLINK. App should open RTNETLINK in order to get those. Other apps are doing that too. I don't think that duplications like this are desirable :/

Is there a way to filter or at least recognize these events when using rtnetlink? I couldn't find any. The only way seems to be getting every RTM_NEWLINK message (there can be quite a lot of those), always perform the lookup in my device list and recognize what happened - only to almost always find that nothing interesting. It is possible, sure, but I would really like to avoid it.

Quoted 8 lines
Quoted 2 lines

I am afraid you are right about this, would adding a filtering capability specifically for this in rtnetlink be a better route?

Maybe we could add new IFLA_EVENT_* values and use IFLA_EVENT to mark RTM_NEWLINK messages announcing "new device" and "device rename". That way, monitoring application would still need to parse all RTM_NEWLINK messages but it would be able to recognize which announce a change in device list without a lookup in its structures.

Hmm, that sounds good.

Quoted 2 lines

Michal

#55 lore raw

PATCH v5 10/22 ethtool: generic handlers for GET requests

In reply to Michal Kubecek

Some parts of processing GET type requests and related notifications are independent of a command. Provide universal functions so that only four callbacks need to be defined for each command type:

parse_request() - parse incoming message

  prepare_data()  - retrieve data from driver or NIC
  reply_size()    - estimate reply message size
  fill_reply()    - compose reply message

These callback are defined in an instance of struct get_request_ops.

---
 net/ethtool/netlink.c | 319 ++++++++++++++++++++++++++++++++++++++++++
 net/ethtool/netlink.h | 111 +++++++++++++++
 2 files changed, 430 insertions(+)
net/ethtool/netlink.c +319 -0
index d84b9780dfa7..7371cdf98ce9 100644
@@ -151,6 +151,325 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 	return NULL;
 }
 
+/* GET request helpers */
+
+const struct get_request_ops *get_requests[__ETHNL_CMD_CNT] = {
+};
+
+/**
+ * ethnl_alloc_get_data() - Allocate and initialize data for a GET request
+ * @ops: instance of struct get_request_ops describing size and layout
+ *
+ * This initializes only the first part (req_info), second part (reply_data)
+ * is initialized before filling the reply data into it (which is done for
+ * each iteration in dump requests).
+ *
+ * Return: pointer to allocated and initialized data, NULL on error
+ */
+static struct common_req_info *
+ethnl_alloc_get_data(const struct get_request_ops *ops)
+{
+	struct common_req_info *req_info;
+
+	req_info = kmalloc(ops->data_size, GFP_KERNEL);
+	if (!req_info)
+		return NULL;
+
+	memset(req_info, '\0', ops->repdata_offset);
+	req_info->reply_data =
+		(struct common_reply_data *)((char *)req_info +
+					     ops->repdata_offset);
+
+	return req_info;
+}
+
+/**
+ * ethnl_free_get_data() - free GET request data
+ * @ops: instance of struct get_request_ops describing the layout
+ * @req_info: pointer to embedded struct common_req_info (at offset 0)
+ *
+ * Calls ->cleanup() handler if defined and frees the data block.
+ */
+static void ethnl_free_get_data(const struct get_request_ops *ops,
+				struct common_req_info *req_info)
+{
+	if (ops->cleanup)
+		ops->cleanup(req_info);
+	kfree(req_info);
+}
+
+/**
+ * ethnl_init_reply_data() - Initialize reply data for GET request
+ * @req_info: pointer to embedded struct common_req_info
+ * @ops:      instance of struct get_request_ops describing the layout
+ * @dev:      network device to initialize the reply for
+ *
+ * Fills the reply data part with zeros and sets the dev member. Must be called
+ * before calling the ->fill_reply() callback (for each iteration when handling
+ * dump requests).
+ */
+static void ethnl_init_reply_data(const struct common_req_info *req_info,
+				  const struct get_request_ops *ops,
+				  struct net_device *dev)
+{
+	memset(req_info->reply_data, '\0',
+	       ops->data_size - ops->repdata_offset);
+	req_info->reply_data->dev = dev;
+}
+
+/* generic ->doit() handler for GET type requests */
+int ethnl_get_doit(struct sk_buff *skb, struct genl_info *info)
+{
+	const u8 cmd = info->genlhdr->cmd;
+	struct common_req_info *req_info;
+	const struct get_request_ops *ops;
+	struct sk_buff *rskb;
+	void *reply_payload;
+	int reply_len;
+	int ret;
+
+	ops = get_requests[cmd];
+	if (WARN_ONCE(!ops, "cmd %u has no get_request_ops\n", cmd))
+		return -EOPNOTSUPP;
+	req_info = ethnl_alloc_get_data(ops);
+	if (!req_info)
+		return -ENOMEM;
+	ret = ops->parse_request(req_info, skb, info, info->nlhdr);
+	if (!ops->allow_nodev_do && !req_info->dev) {
+		ETHNL_SET_ERRMSG(info, "device not specified in do request");
+		ret = -EINVAL;
+	}
+	if (ret < 0)
+		goto err_dev;
+	ethnl_init_reply_data(req_info, ops, req_info->dev);
+
+	rtnl_lock();
+	ret = ops->prepare_data(req_info, info);
+	if (ret < 0)
+		goto err_rtnl;
+	reply_len = ops->reply_size(req_info);
+	if (ret < 0)
+		goto err_rtnl;
+	ret = -ENOMEM;
+	rskb = ethnl_reply_init(reply_len, req_info->dev, ops->reply_cmd,
+				ops->dev_attrtype, info, &reply_payload);
+	if (!rskb)
+		goto err_rtnl;
+	ret = ops->fill_reply(rskb, req_info);
+	if (ret < 0)
+		goto err;
+	rtnl_unlock();
+
+	genlmsg_end(rskb, reply_payload);
+	if (req_info->dev)
+		dev_put(req_info->dev);
+	ethnl_free_get_data(ops, req_info);
+	return genlmsg_reply(rskb, info);
+
+err:
+	WARN_ONCE(ret == -EMSGSIZE,
+		  "calculated message payload length (%d) not sufficient\n",
+		  reply_len);
+	nlmsg_free(rskb);
+	ethnl_free_get_data(ops, req_info);
+err_rtnl:
+	rtnl_unlock();
+err_dev:
+	if (req_info->dev)
+		dev_put(req_info->dev);
+	return ret;
+}
+
+static int ethnl_get_dump_one(struct sk_buff *skb,
+			      struct net_device *dev,
+			      const struct get_request_ops *ops,
+			      struct common_req_info *req_info)
+{
+	int ret;
+
+	ethnl_init_reply_data(req_info, ops, dev);
+	rtnl_lock();
+	ret = ops->prepare_data(req_info, NULL);
+	if (ret < 0)
+		return ret;
+	ret = ethnl_fill_dev(skb, dev, ops->dev_attrtype);
+	if (ret < 0)
+		return ret;
+	ret = ops->fill_reply(skb, req_info);
+	rtnl_unlock();
+
+	req_info->reply_data->dev = NULL;
+	return ret;
+}
+
+/* generic ->dumpit() handler for GET requests; device iteration copied from
+ * rtnl_dump_ifinfo()
+ * cb->args[0]: pointer to struct get_request_ops
+ * cb->args[1]: pointer to request data
+ * cb->args[2]: iteration position - hashbucket
+ * cb->args[3]: iteration position - ifindex
+ */
+int ethnl_get_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
+{
+	struct net *net = sock_net(skb->sk);
+	struct common_req_info *req_info;
+	const struct get_request_ops *ops;
+	int h, s_h, idx = 0, s_idx;
+	struct hlist_head *head;
+	struct net_device *dev;
+	int ret = 0;
+	void *ehdr;
+
+	ops = (const struct get_request_ops *)cb->args[0];
+	req_info = (struct common_req_info *)cb->args[1];
+	s_h = cb->args[2];
+	s_idx = cb->args[3];
+
+	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
+		idx = 0;
+		head = &net->dev_index_head[h];
+		hlist_for_each_entry(dev, head, index_hlist) {
+			if (idx < s_idx)
+				goto cont;
+			ehdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid,
+					   cb->nlh->nlmsg_seq,
+					   &ethtool_genl_family, 0,
+					   ops->reply_cmd);
+			ret = ethnl_get_dump_one(skb, dev, ops, req_info);
+			if (ret < 0) {
+				genlmsg_cancel(skb, ehdr);
+				if (ret == -EOPNOTSUPP)
+					goto cont;
+				if (likely(skb->len))
+					goto out;
+				goto out_err;
+			}
+			genlmsg_end(skb, ehdr);
+cont:
+			idx++;
+		}
+	}
+out:
+	ret = skb->len;
+out_err:
+	cb->args[2] = h;
+	cb->args[3] = idx;
+	cb->seq = net->dev_base_seq;
+	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
+
+	return ret;
+}
+
+/* generic ->start() handler for GET requests */
+static int ethnl_get_start(struct netlink_callback *cb)
+{
+	struct common_req_info *req_info;
+	const struct get_request_ops *ops;
+	struct genlmsghdr *ghdr;
+	int ret;
+
+	ghdr = nlmsg_data(cb->nlh);
+	ops = get_requests[ghdr->cmd];
+	if (WARN_ONCE(!ops, "cmd %u has no get_request_ops\n", ghdr->cmd))
+		return -EOPNOTSUPP;
+	req_info = ethnl_alloc_get_data(ops);
+	if (!req_info)
+		return -ENOMEM;
+
+	ret = ops->parse_request(req_info, cb->skb, NULL, cb->nlh);
+	if (req_info->dev) {
+		/* We ignore device specification in dump requests but as the
+		 * same parser as for non-dump (doit) requests is used, it
+		 * would take reference to the device if it finds one
+		 */
+		dev_put(req_info->dev);
+		req_info->dev = NULL;
+	}
+	if (ret < 0)
+		return ret;
+
+	cb->args[0] = (long)ops;
+	cb->args[1] = (long)req_info;
+	cb->args[2] = 0;
+	cb->args[3] = 0;
+
+	return 0;
+}
+
+/* generic ->done() handler for GET requests */
+static int ethnl_get_done(struct netlink_callback *cb)
+{
+	ethnl_free_get_data((const struct get_request_ops *)cb->args[0],
+			    (struct common_req_info *)cb->args[1]);
+
+	return 0;
+}
+
+/* generic notification handler */
+static void ethnl_std_notify(struct net_device *dev,
+			     struct netlink_ext_ack *extack, unsigned int cmd,
+			     u32 req_mask, const void *data)
+{
+	struct common_req_info *req_info;
+	const struct get_request_ops *ops;
+	struct sk_buff *skb;
+	void *reply_payload;
+	int reply_len;
+	int ret;
+
+	ops = get_requests[cmd - 1];
+	if (WARN_ONCE(!ops, "cmd %u has no get_request_ops\n", cmd - 1))
+		return;
+	/* when ethnl_std_notify() is used as notify handler, command id of
+	 * corresponding GET request must be one less than cmd argument passed
+	 * to ethnl_std_notify()
+	 */
+	if (WARN_ONCE(ops->reply_cmd != cmd,
+		      "reply_cmd for %u is %u, expected %u\n", cmd - 1,
+		      ops->reply_cmd, cmd))
+		return;
+
+	req_info = ethnl_alloc_get_data(ops);
+	if (!req_info)
+		return;
+	req_info->dev = dev;
+	req_info->req_mask = req_mask;
+	req_info->compact = true;
+
+	ethnl_init_reply_data(req_info, ops, dev);
+	ret = ops->prepare_data(req_info, NULL);
+	if (ret < 0)
+		goto err_data;
+	reply_len = ops->reply_size(req_info);
+	if (reply_len < 0)
+		goto err_data;
+	skb = genlmsg_new(reply_len, GFP_KERNEL);
+	if (!skb)
+		goto err_data;
+	reply_payload = genlmsg_put(skb, 0, ++ethnl_bcast_seq,
+				    &ethtool_genl_family, 0, ops->reply_cmd);
+	if (!reply_payload)
+		goto err_skb;
+
+	ret = ethnl_fill_dev(skb, dev, ops->dev_attrtype);
+	if (ret < 0)
+		goto err_skb;
+	ret = ops->fill_reply(skb, req_info);
+	if (ret < 0)
+		goto err_skb;
+	ethnl_free_get_data(ops, req_info);
+	genlmsg_end(skb, reply_payload);
+
+	genlmsg_multicast(&ethtool_genl_family, skb, 0, ETHNL_MCGRP_MONITOR,
+			  GFP_KERNEL);
+	return;
+
+err_skb:
+	nlmsg_free(skb);
+err_data:
+	ethnl_free_get_data(ops, req_info);
+}
+
 /* notifications */
 
 typedef void (*ethnl_notify_handler_t)(struct net_device *dev,
net/ethtool/netlink.h +111 -0

115 lines not on this page · raw message

Signature · 1 line
2.21.0
#56 lore raw

Re: [PATCH net-next v5 10/22] ethtool: generic handlers for GET requests

Quoted 12 lines · Mon, Mar 25, 2019 at 06:08:24PM CET, [email protected] wrote:

Some parts of processing GET type requests and related notifications are independent of a command. Provide universal functions so that only four callbacks need to be defined for each command type:

parse_request() - parse incoming message

 prepare_data()  - retrieve data from driver or NIC
 reply_size()    - estimate reply message size
 fill_reply()    - compose reply message

These callback are defined in an instance of struct get_request_ops.

[...]

Quoted 23 lines
net/ethtool/netlink.h +23 -0
+/**
+ * struct get_request_ops - unified handling of GET requests
+ * @request_cmd:    command id for request (GET)
+ * @reply_cmd:      command id for reply (SET)
+ * @dev_attr:       attribute type for device specification
+ * @data_size:      total length of data structure
+ * @repdata_offset: offset of "reply data" part (struct common_reply_data)
+ * @allow_nodev_do: do not fail if device is not specified for non-dump request
+ * @parse_request:
+ *	parse request message and fill request info; request info is zero
+ *	initialized on entry except reply_data pointer (which is initialized)
+ * @prepare_data:
+ *	retrieve data needed to compose a reply message; reply data are zero
+ *	initialized on entry except for @dev
+ * @reply_size:
+ *	return size of reply message payload without device specification;
+ *	returned size may be bigger than actual reply size but it must suffice
+ *	to hold the reply
+ * @fill_reply:
+ *	fill reply message payload using the data prepared by @prepare_data()
+ * @cleanup
+ *	(optional) called when data are no longer needed; use e.g. to free
+ *	any additional data structures allocated in prepare_data() which are

net/ethtool/netlink.h static inline unsigned int dev_ident_size(void) · line 259

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+ *	not part of the main structure
+ *
+ * Description of variable parts of GET request handling when using the unified
+ * infrastructure. When used, a pointer to an instance of this structure is to
+ * be added to &get_requests array, generic handlers ethnl_get_doit(),
+ * ethnl_get_dumpit(), ethnl_get_start() and ethnl_get_done() used in
+ * @ethnl_genl_ops and (optionally) ethnl_std_notify() as notification handler
+ * in &ethnl_notify_handlers.
+ */
+struct get_request_ops {

First of all, please maintain the ethnl prefix. Not only here, but also for other structs and functions (common_req_info, parse_*, etc).

But I must ask, why do we need this wrappers of ops around ops? I believe it would be much nicer to use genl ops directly and do the common work in pre_doit and and post_doit. Please see devlink.c for examples.

Wrappers are unfortunately quite common in this patchset. Most of the time, they do things much harder to follow and understand :(

Quoted 22 lines
net/ethtool/netlink.h +18 -0
+	u8			request_cmd;
+	u8			reply_cmd;
+	u16			dev_attrtype;
+	unsigned int		data_size;
+	unsigned int		repdata_offset;
+	bool			allow_nodev_do;
+
+	int (*parse_request)(struct common_req_info *req_info,
+			     struct sk_buff *skb, struct genl_info *info,
+			     const struct nlmsghdr *nlhdr);
+	int (*prepare_data)(struct common_req_info *req_info,
+			    struct genl_info *info);
+	int (*reply_size)(const struct common_req_info *req_info);
+	int (*fill_reply)(struct sk_buff *skb,
+			  const struct common_req_info *req_info);
+	void (*cleanup)(struct common_req_info *req_info);
+};
+

#endif /* _NET_ETHTOOL_NETLINK_H */

Signature · 1 line
2.21.0
#57 lore raw

Re: [PATCH net-next v5 10/22] ethtool: generic handlers for GET requests

Quoted 53 lines · Jiri Pirko
Quoted 12 lines · Mon, Mar 25, 2019 at 06:08:24PM CET, [email protected] wrote:

Some parts of processing GET type requests and related notifications are independent of a command. Provide universal functions so that only four callbacks need to be defined for each command type:

parse_request() - parse incoming message

 prepare_data()  - retrieve data from driver or NIC
 reply_size()    - estimate reply message size
 fill_reply()    - compose reply message

These callback are defined in an instance of struct get_request_ops.

[...]

Quoted 23 lines
net/ethtool/netlink.h +23 -0
+/**
+ * struct get_request_ops - unified handling of GET requests
+ * @request_cmd:    command id for request (GET)
+ * @reply_cmd:      command id for reply (SET)
+ * @dev_attr:       attribute type for device specification
+ * @data_size:      total length of data structure
+ * @repdata_offset: offset of "reply data" part (struct common_reply_data)
+ * @allow_nodev_do: do not fail if device is not specified for non-dump request
+ * @parse_request:
+ *	parse request message and fill request info; request info is zero
+ *	initialized on entry except reply_data pointer (which is initialized)
+ * @prepare_data:
+ *	retrieve data needed to compose a reply message; reply data are zero
+ *	initialized on entry except for @dev
+ * @reply_size:
+ *	return size of reply message payload without device specification;
+ *	returned size may be bigger than actual reply size but it must suffice
+ *	to hold the reply
+ * @fill_reply:
+ *	fill reply message payload using the data prepared by @prepare_data()
+ * @cleanup
+ *	(optional) called when data are no longer needed; use e.g. to free
+ *	any additional data structures allocated in prepare_data() which are

net/ethtool/netlink.h static inline unsigned int dev_ident_size(void) · line 259

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+ *	not part of the main structure
+ *
+ * Description of variable parts of GET request handling when using the unified
+ * infrastructure. When used, a pointer to an instance of this structure is to
+ * be added to &get_requests array, generic handlers ethnl_get_doit(),
+ * ethnl_get_dumpit(), ethnl_get_start() and ethnl_get_done() used in
+ * @ethnl_genl_ops and (optionally) ethnl_std_notify() as notification handler
+ * in &ethnl_notify_handlers.
+ */
+struct get_request_ops {

First of all, please maintain the ethnl prefix. Not only here, but also for other structs and functions (common_req_info, parse_*, etc).

Quoted 7 lines

But I must ask, why do we need this wrappers of ops around ops? I believe it would be much nicer to use genl ops directly and do the common work in pre_doit and and post_doit. Please see devlink.c for examples.

Wrappers are unfortunately quite common in this patchset. Most of the time, they do things much harder to follow and understand :(

I'm a bit surprised by this position because so far my experience with linux networking code seemed to suggest that using simple wrappers and helpers is how things are supposed to be done. And while following a chain of such wrappers (often each in a different file) in a code I was reading for the first time could be frustrating at times, mostly I had to admit that this style has its merits. After all, genetlink itself is full of simple wrappers around netlink functions.

Let me point out one thing: most of these helpers and wrappers are not artificial, they haven't been written in advance with an idea that they might be useful (the patch series does not, of course, reflect the development history); most of them were written when I realized I'm writing the same or almost the same code again and again.

So when I caught myself writing

  ... = nla_nest_start(skb, ... | NLA_F_NESTED);

for the third or fourth time and I realized that every nla_nest_start() call in the code will have this bitwise or, I felt it would deserve a helper. (If I expected some objection, it was rather the optical asymmetry of ethnl_nest_start() being closed with nla_nest_end().) It would be much nicer to have it in nla_nest_start() but unfortunately it's too late for that.

And it's exactly the same case with get_request_ops. For quite long (until after RFC v2), this framework didn't exist and code for get request processing (both doit and dumpit) and notifications was written separately for each message type. Realizing that big part of each new file is in fact an exact copy of the previous one with some string replacements and that it's going to be like that for most of the future files, that led me to identifying which parts are specific to message type and which are generic.

If I have to get rid of get_request_ops, it will only result in having multiple copies of functions which would replace ethnl_get_doit(), ethnl_get_dumpit() and ethnl_std_notify(). They would be slightly simpler but would look the same except for "info" in one being replaced by "strset" in second, "settings" in third etc. Later, there would be one more copy for stats, one for tunables etc.

I don't think the generic code can be handled just by pre_doit and post_doit as the generic and message specific part are interleaved and the generic parts are also different for do requests, dump requests and notifications.

Michal

#58 lore raw

Re: [PATCH net-next v5 10/22] ethtool: generic handlers for GET requests

Quoted 78 lines · Wed, Mar 27, 2019 at 10:53:42PM CET, [email protected] wrote:
Quoted 53 lines · Jiri Pirko
Quoted 12 lines · Mon, Mar 25, 2019 at 06:08:24PM CET, [email protected] wrote:

Some parts of processing GET type requests and related notifications are independent of a command. Provide universal functions so that only four callbacks need to be defined for each command type:

parse_request() - parse incoming message

 prepare_data()  - retrieve data from driver or NIC
 reply_size()    - estimate reply message size
 fill_reply()    - compose reply message

These callback are defined in an instance of struct get_request_ops.

[...]

Quoted 23 lines
net/ethtool/netlink.h +23 -0
+/**
+ * struct get_request_ops - unified handling of GET requests
+ * @request_cmd:    command id for request (GET)
+ * @reply_cmd:      command id for reply (SET)
+ * @dev_attr:       attribute type for device specification
+ * @data_size:      total length of data structure
+ * @repdata_offset: offset of "reply data" part (struct common_reply_data)
+ * @allow_nodev_do: do not fail if device is not specified for non-dump request
+ * @parse_request:
+ *	parse request message and fill request info; request info is zero
+ *	initialized on entry except reply_data pointer (which is initialized)
+ * @prepare_data:
+ *	retrieve data needed to compose a reply message; reply data are zero
+ *	initialized on entry except for @dev
+ * @reply_size:
+ *	return size of reply message payload without device specification;
+ *	returned size may be bigger than actual reply size but it must suffice
+ *	to hold the reply
+ * @fill_reply:
+ *	fill reply message payload using the data prepared by @prepare_data()
+ * @cleanup
+ *	(optional) called when data are no longer needed; use e.g. to free
+ *	any additional data structures allocated in prepare_data() which are

net/ethtool/netlink.h static inline unsigned int dev_ident_size(void) · line 259

Quoted 10 lines
net/ethtool/netlink.h +10 -0
+ *	not part of the main structure
+ *
+ * Description of variable parts of GET request handling when using the unified
+ * infrastructure. When used, a pointer to an instance of this structure is to
+ * be added to &get_requests array, generic handlers ethnl_get_doit(),
+ * ethnl_get_dumpit(), ethnl_get_start() and ethnl_get_done() used in
+ * @ethnl_genl_ops and (optionally) ethnl_std_notify() as notification handler
+ * in &ethnl_notify_handlers.
+ */
+struct get_request_ops {

First of all, please maintain the ethnl prefix. Not only here, but also for other structs and functions (common_req_info, parse_*, etc).

Quoted 7 lines

But I must ask, why do we need this wrappers of ops around ops? I believe it would be much nicer to use genl ops directly and do the common work in pre_doit and and post_doit. Please see devlink.c for examples.

Wrappers are unfortunately quite common in this patchset. Most of the time, they do things much harder to follow and understand :(

I'm a bit surprised by this position because so far my experience with linux networking code seemed to suggest that using simple wrappers and helpers is how things are supposed to be done. And while following a chain of such wrappers (often each in a different file) in a code I was reading for the first time could be frustrating at times, mostly I had to admit that this style has its merits. After all, genetlink itself is full of simple wrappers around netlink functions.

Let me point out one thing: most of these helpers and wrappers are not artificial, they haven't been written in advance with an idea that they might be useful (the patch series does not, of course, reflect the development history); most of them were written when I realized I'm writing the same or almost the same code again and again.

So when I caught myself writing

Quoted 8 lines

... = nla_nest_start(skb, ... | NLA_F_NESTED);

for the third or fourth time and I realized that every nla_nest_start() call in the code will have this bitwise or, I felt it would deserve a helper. (If I expected some objection, it was rather the optical asymmetry of ethnl_nest_start() being closed with nla_nest_end().) It would be much nicer to have it in nla_nest_start() but unfortunately it's too late for that.

Okay, this wrapper I can understand, but it certainly isn't ethnl specific wrapper. It is generic, many others might use it:

$ git grep NLA_F_NESTED |grep nla_nest_start
Quoted 17 lines

And it's exactly the same case with get_request_ops. For quite long (until after RFC v2), this framework didn't exist and code for get request processing (both doit and dumpit) and notifications was written separately for each message type. Realizing that big part of each new file is in fact an exact copy of the previous one with some string replacements and that it's going to be like that for most of the future files, that led me to identifying which parts are specific to message type and which are generic.

If I have to get rid of get_request_ops, it will only result in having multiple copies of functions which would replace ethnl_get_doit(), ethnl_get_dumpit() and ethnl_std_notify(). They would be slightly simpler but would look the same except for "info" in one being replaced by "strset" in second, "settings" in third etc. Later, there would be one more copy for stats, one for tunables etc.

Quoted 4 lines

I don't think the generic code can be handled just by pre_doit and post_doit as the generic and message specific part are interleaved and the generic parts are also different for do requests, dump requests and notifications.

What do you mean by "interleaved"? I guess you can make the attrs to be formatted in the way it is not interleaved. Like what I suggested, to have top-level generic attr enum and one of the generic attrs would nest the cmd-specific attrs. That way, pre_doit can work with the generic attrs, doit op can work with cmd-specific attrs. Similar to devlink code.

Quoted 2 lines

Michal

#59 lore raw

PATCH v5 11/22 ethtool: move string arrays into common file

In reply to Michal Kubecek

Introduce file net/ethtool/common.c for code shared by ioctl and netlink ethtool interface. Move name tables of features, RSS hash functions, tunables and PHY tunables into this file.

---
 net/ethtool/Makefile |  2 +-
 net/ethtool/common.c | 83 ++++++++++++++++++++++++++++++++++++++++++++
 net/ethtool/common.h | 17 +++++++++
 net/ethtool/ioctl.c  | 82 ++-----------------------------------------
 4 files changed, 103 insertions(+), 81 deletions(-)
 create mode 100644 net/ethtool/common.c
 create mode 100644 net/ethtool/common.h
net/ethtool/Makefile +1 -1
index 482fdb9380fa..11782306593b 100644
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0
 
-obj-y				+= ioctl.o
+obj-y				+= ioctl.o common.o
 
 obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
 
net/ethtool/common.c new file +83 -0

83 lines not on this page · raw message

net/ethtool/common.h new file +17 -0
new file mode 100644
index 000000000000..41b2efc1e4e1
@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+
+#ifndef _ETHTOOL_COMMON_H
+#define _ETHTOOL_COMMON_H
+
+#include <linux/ethtool.h>
+
+extern const char
+netdev_features_strings[NETDEV_FEATURE_COUNT][ETH_GSTRING_LEN];
+extern const char
+rss_hash_func_strings[ETH_RSS_HASH_FUNCS_COUNT][ETH_GSTRING_LEN];
+extern const char
+tunable_strings[__ETHTOOL_TUNABLE_COUNT][ETH_GSTRING_LEN];
+extern const char
+phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
+
+#endif /* _ETHTOOL_COMMON_H */
net/ethtool/ioctl.c +2 -80

94 lines not on this page · raw message

Signature · 1 line
2.21.0
Reviewed-by on v5 11/22
#60 lore raw

Re: [PATCH net-next v5 11/22] ethtool: move string arrays into common file

Quoted 5 lines · Michal Kubecek

Introduce file net/ethtool/common.c for code shared by ioctl and netlink ethtool interface. Move name tables of features, RSS hash functions, tunables and PHY tunables into this file.

Signature · 1 line
Florian
#61 lore raw

PATCH v5 12/22 ethtool: provide string sets with GET_STRSET request

In reply to Michal Kubecek

Requests a contents of one or more string sets, i.e. indexed arrays of strings; this information is provided by ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS commands of ioctl interface. There are three types of requests:

- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices

It's possible to request all string sets of given type or only specific sets. With ETHA_STRSET_COUNTS flag, only set sizes (number of strings) are returned.

---
 Documentation/networking/ethtool-netlink.txt |  46 +-
 include/uapi/linux/ethtool.h                 |   2 +
 include/uapi/linux/ethtool_netlink.h         |  43 ++
 net/ethtool/Makefile                         |   2 +-
 net/ethtool/netlink.c                        |   8 +
 net/ethtool/netlink.h                        |   4 +
 net/ethtool/strset.c                         | 447 +++++++++++++++++++
 7 files changed, 549 insertions(+), 3 deletions(-)
 create mode 100644 net/ethtool/strset.c
Documentation/networking/ethtool-netlink.txt +44 -2
index 5e5d785fe215..1508c16a236e 100644
@@ -127,6 +127,8 @@ List of message types
 ---------------------
 
     ETHNL_CMD_EVENT			notification only
+    ETHNL_CMD_GET_STRSET
+    ETHNL_CMD_SET_STRSET		response only
 
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
@@ -167,6 +169,46 @@ and also multiple events of the same type (e.g. two or more newly registered
 devices).
 
 
+GET_STRSET
+----------
+
+Requests contents of a string set as provided by ioctl commands
+ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS. String sets are not user writeable so
+that the corresponding SET_STRSET message is only used in kernel replies.
+There are two types of string sets: global (independent of a device, e.g.
+device feature names) and device specific (e.g. device private flags).
+
+Request contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_COUNTS		(flag)		request only string counts
+    ETHA_STRSET_STRINGSET	(nested)	string set to request
+        ETHA_STRINGSET_ID		(u32)		set id
+
+Kernel response contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_STRINGSET	(nested)	string set to request
+        ETHA_STRINGSET_ID		(u32)		set id
+        ETHA_STRINGSET_COUNT		(u32)		number of strings
+        ETHA_STRINGSET_STRINGS		(nested)	array of strings
+            ETHA_STRING_INDEX			(u32)		string index
+            ETHA_STRING_VALUE			(string)	string value
+
+ETHA_STRSET_DEV, if present, identifies the device to request device specific
+string sets for. Depending on its presence a and NLM_F_DUMP flag, there are
+three type of GET_STRSET requests:
+
+ - no NLM_F_DUMP, no device: get "global" stringsets
+ - no NLM_F_DUMP, with device: get string sets related to the device
+ - NLM_F_DUMP, no device: get device related string sets for all devices
+
+If there is no ETHA_STRSET_STRINGSET attribute, all string sets of requested
+type are returned, otherwise only those specified in the request. Flag
+ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not
+the actual strings.
+
+
 Request translation
 -------------------
 
@@ -201,7 +243,7 @@ ETHTOOL_STXCSUM			n/a
 ETHTOOL_GSG			n/a
 ETHTOOL_SSG			n/a
 ETHTOOL_TEST			n/a
-ETHTOOL_GSTRINGS		n/a
+ETHTOOL_GSTRINGS		ETHNL_CMD_GET_STRSET
 ETHTOOL_PHYS_ID			n/a
 ETHTOOL_GSTATS			n/a
 ETHTOOL_GTSO			n/a
@@ -229,7 +271,7 @@ ETHTOOL_FLASHDEV		n/a
 ETHTOOL_RESET			n/a
 ETHTOOL_SRXNTUPLE		n/a
 ETHTOOL_GRXNTUPLE		n/a
-ETHTOOL_GSSET_INFO		n/a
+ETHTOOL_GSSET_INFO		ETHNL_CMD_GET_STRSET
 ETHTOOL_GRXFHINDIR		n/a
 ETHTOOL_SRXFHINDIR		n/a
 ETHTOOL_GFEATURES		n/a
include/uapi/linux/ethtool.h +2 -0
index 3652b239dad1..eaea804972f6 100644
@@ -574,6 +574,8 @@ enum ethtool_stringset {
 	ETH_SS_TUNABLES,
 	ETH_SS_PHY_STATS,
 	ETH_SS_PHY_TUNABLES,
+
+	ETH_SS_COUNT
 };
 
 /**
include/uapi/linux/ethtool_netlink.h +43 -0

55 lines not on this page · raw message

net/ethtool/Makefile +1 -1
index 11782306593b..11ceb00821b3 100644
@@ -4,4 +4,4 @@ obj-y				+= ioctl.o common.o
 
 obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
 
-ethtool_nl-y	:= netlink.o bitset.o
+ethtool_nl-y	:= netlink.o bitset.o strset.o
net/ethtool/netlink.c +8 -0
index 7371cdf98ce9..63ee0f846b50 100644
@@ -154,6 +154,7 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 /* GET request helpers */
 
 const struct get_request_ops *get_requests[__ETHNL_CMD_CNT] = {
+	[ETHNL_CMD_GET_STRSET]		= &strset_request_ops,
 };
 
 /**
@@ -560,6 +561,13 @@ static struct notifier_block ethnl_netdev_notifier = {
 /* genetlink setup */
 
 static const struct genl_ops ethtool_genl_ops[] = {
+	{
+		.cmd	= ETHNL_CMD_GET_STRSET,
+		.doit	= ethnl_get_doit,
+		.start	= ethnl_get_start,
+		.dumpit	= ethnl_get_dumpit,
+		.done	= ethnl_get_done,
+	},
 };
 
 static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
net/ethtool/netlink.h +4 -0
index 625f912144b1..32d85bb5c49a 100644
@@ -275,4 +275,8 @@ struct get_request_ops {
 	void (*cleanup)(struct common_req_info *req_info);
 };
 
+/* request handlers */
+
+extern const struct get_request_ops strset_request_ops;
+
 #endif /* _NET_ETHTOOL_NETLINK_H */
net/ethtool/strset.c new file +447 -0

447 lines not on this page · raw message

Signature · 1 line
2.21.0
#62 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 23 lines · Mon, Mar 25, 2019 at 06:08:30PM CET, [email protected] wrote:

Requests a contents of one or more string sets, i.e. indexed arrays of strings; this information is provided by ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS commands of ioctl interface. There are three types of requests:

- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices

It's possible to request all string sets of given type or only specific sets. With ETHA_STRSET_COUNTS flag, only set sizes (number of strings) are returned.

---

Documentation/networking/ethtool-netlink.txt |  46 +-
include/uapi/linux/ethtool.h                 |   2 +
include/uapi/linux/ethtool_netlink.h         |  43 ++
net/ethtool/Makefile                         |   2 +-
net/ethtool/netlink.c                        |   8 +
net/ethtool/netlink.h                        |   4 +
net/ethtool/strset.c                         | 447 +++++++++++++++++++
7 files changed, 549 insertions(+), 3 deletions(-)
Quoted 1 line

create mode 100644 net/ethtool/strset.c

First of all, the code is hard to follow. For reasons I mentioned in other replies (lack of prefixes, wrappers, etc).

More importantly, why do we need this? This concept of having strings in kernel for various things and features and sending them to userspace is weird. Certainly not common for Netlink interface. I believe these strings should be avoided and all should be communicated to userspace and back in form of well-defined Netlink attributes. We are introducing new Netlink API, lets do it properly and don't bring baggage from past.

#63 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 29 lines · Jiri Pirko
Quoted 23 lines · Mon, Mar 25, 2019 at 06:08:30PM CET, [email protected] wrote:

Requests a contents of one or more string sets, i.e. indexed arrays of strings; this information is provided by ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS commands of ioctl interface. There are three types of requests:

- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices

It's possible to request all string sets of given type or only specific sets. With ETHA_STRSET_COUNTS flag, only set sizes (number of strings) are returned.

---

Documentation/networking/ethtool-netlink.txt |  46 +-
include/uapi/linux/ethtool.h                 |   2 +
include/uapi/linux/ethtool_netlink.h         |  43 ++
net/ethtool/Makefile                         |   2 +-
net/ethtool/netlink.c                        |   8 +
net/ethtool/netlink.h                        |   4 +
net/ethtool/strset.c                         | 447 +++++++++++++++++++
7 files changed, 549 insertions(+), 3 deletions(-)
Quoted 1 line

create mode 100644 net/ethtool/strset.c

First of all, the code is hard to follow. For reasons I mentioned in other replies (lack of prefixes, wrappers, etc).

Quoted 6 lines

More importantly, why do we need this? This concept of having strings in kernel for various things and features and sending them to userspace is weird. Certainly not common for Netlink interface. I believe these strings should be avoided and all should be communicated to userspace and back in form of well-defined Netlink attributes. We are introducing new Netlink API, lets do it properly and don't bring baggage from past.

For some of the global string sets where the values have fixed meaning and new values are only appended to, it would be possible. But even for those, keeping the list in sync between kernel and ethtool is often less than perfect. And ethtool is only one of the tools using the interface. So even in this case, I don't see string identifiers (or tags or names or whatever we call them) as a baggage from past, rather as a solution to a problem some of the existing interfaces have.

Then e.g. netdev features are not fixed in this sense: the same bit index represents different feature in different kernels. I guess we could introduce some fixed numeric identifiers for them and map between those and actual indices but I don't see an advantage of such approach.

But the really important question is how would you handle what is currently described by "per device" string sets, i.e. private flags, (ethtool) statistics, tests, ...? For these, the list depends on the driver or even device.

Michal

#64 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 45 lines · Wed, Mar 27, 2019 at 11:56:08PM CET, [email protected] wrote:
Quoted 29 lines · Jiri Pirko
Quoted 23 lines · Mon, Mar 25, 2019 at 06:08:30PM CET, [email protected] wrote:

Requests a contents of one or more string sets, i.e. indexed arrays of strings; this information is provided by ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS commands of ioctl interface. There are three types of requests:

- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices

It's possible to request all string sets of given type or only specific sets. With ETHA_STRSET_COUNTS flag, only set sizes (number of strings) are returned.

---

Documentation/networking/ethtool-netlink.txt |  46 +-
include/uapi/linux/ethtool.h                 |   2 +
include/uapi/linux/ethtool_netlink.h         |  43 ++
net/ethtool/Makefile                         |   2 +-
net/ethtool/netlink.c                        |   8 +
net/ethtool/netlink.h                        |   4 +
net/ethtool/strset.c                         | 447 +++++++++++++++++++
7 files changed, 549 insertions(+), 3 deletions(-)
Quoted 1 line

create mode 100644 net/ethtool/strset.c

First of all, the code is hard to follow. For reasons I mentioned in other replies (lack of prefixes, wrappers, etc).

Quoted 6 lines

More importantly, why do we need this? This concept of having strings in kernel for various things and features and sending them to userspace is weird. Certainly not common for Netlink interface. I believe these strings should be avoided and all should be communicated to userspace and back in form of well-defined Netlink attributes. We are introducing new Netlink API, lets do it properly and don't bring baggage from past.

For some of the global string sets where the values have fixed meaning and new values are only appended to, it would be possible. But even for those, keeping the list in sync between kernel and ethtool is often less than perfect. And ethtool is only one of the tools using the interface. So even in this case, I don't see string identifiers (or tags or names or whatever we call them) as a baggage from past, rather as a solution to a problem some of the existing interfaces have.

Quoted 2 lines

Then e.g. netdev features are not fixed in this sense: the same bit index represents different feature in different kernels. I guess we

Should not be exposed in bits in first place. See devlink params for example.

Quoted 7 lines

could introduce some fixed numeric identifiers for them and map between those and actual indices but I don't see an advantage of such approach.

But the really important question is how would you handle what is currently described by "per device" string sets, i.e. private flags, (ethtool) statistics, tests, ...? For these, the list depends on the driver or even device.

Yes, those per-driver things have to be strings. That's what we did for devlink params.

Tests also could be handled in similar way as devlink params.

For stats, I believe that uint->string per-driver mapping needs to be exposed.

#65 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

In reply to Michal Kubecek
Quoted 15 lines · Michal Kubecek

Requests a contents of one or more string sets, i.e. indexed arrays of strings; this information is provided by ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS commands of ioctl interface. There are three types of requests:

- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices

It's possible to request all string sets of given type or only specific sets. With ETHA_STRSET_COUNTS flag, only set sizes (number of strings) are returned.

---

[snip]

Quoted 20 lines
Documentation/networking/ethtool-netlink.txt +5 -0
index 5e5d785fe215..1508c16a236e 100644
@@ -127,6 +127,8 @@ List of message types
 ---------------------
 
     ETHNL_CMD_EVENT			notification only
+    ETHNL_CMD_GET_STRSET
+    ETHNL_CMD_SET_STRSET		response only
 
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
@@ -167,6 +169,46 @@ and also multiple events of the same type (e.g. two or more newly registered
 devices).
 
 
+GET_STRSET
+----------
+

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 184

Quoted 10 lines
Documentation/networking/ethtool-netlink.txt +10 -0
+Requests contents of a string set as provided by ioctl commands
+ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS. String sets are not user writeable so
+that the corresponding SET_STRSET message is only used in kernel replies.
+There are two types of string sets: global (independent of a device, e.g.
+device feature names) and device specific (e.g. device private flags).
+
+Request contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_COUNTS		(flag)		request only string counts

Should not that be part of the nested attribute under ETHA_STRSET_STRINGSET. We should probably think about adding another flag which indicates that we want to get the stringset associated data, see below why.

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 191

Quoted 7 lines
Documentation/networking/ethtool-netlink.txt +7 -0
+    ETHA_STRSET_STRINGSET	(nested)	string set to request
+        ETHA_STRINGSET_ID		(u32)		set id
+
+Kernel response contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_STRINGSET	(nested)	string set to request

string set requested?

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 196

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt +5 -0
+        ETHA_STRINGSET_ID		(u32)		set id
+        ETHA_STRINGSET_COUNT		(u32)		number of strings
+        ETHA_STRINGSET_STRINGS		(nested)	array of strings
+            ETHA_STRING_INDEX			(u32)		string index
+            ETHA_STRING_VALUE			(string)	string value

This is one of the areas where the legacy ethtool ioctl() is painful because we need to request a string set to know how many of those exist to allocate space for those in both kernel and user space.

If we could find a way to have a single command that allows us to dump stringset (count, values) and associated data, then we save ourselves a context switch and having to pre-allocate memory accordingly.

Signature · 1 line
Florian
#66 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 35 lines · Florian Fainelli

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 184

Quoted 13 lines
Documentation/networking/ethtool-netlink.txt +13 -0
+GET_STRSET
+----------
+
+Requests contents of a string set as provided by ioctl commands
+ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS. String sets are not user writeable so
+that the corresponding SET_STRSET message is only used in kernel replies.
+There are two types of string sets: global (independent of a device, e.g.
+device feature names) and device specific (e.g. device private flags).
+
+Request contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_COUNTS		(flag)		request only string counts

Should not that be part of the nested attribute under ETHA_STRSET_STRINGSET. We should probably think about adding another flag which indicates that we want to get the stringset associated data, see below why.

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 191

Quoted 7 lines
Documentation/networking/ethtool-netlink.txt +7 -0
+    ETHA_STRSET_STRINGSET	(nested)	string set to request
+        ETHA_STRINGSET_ID		(u32)		set id
+
+Kernel response contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_STRINGSET	(nested)	string set to request

string set requested?

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 196

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt +5 -0
+        ETHA_STRINGSET_ID		(u32)		set id
+        ETHA_STRINGSET_COUNT		(u32)		number of strings
+        ETHA_STRINGSET_STRINGS		(nested)	array of strings
+            ETHA_STRING_INDEX			(u32)		string index
+            ETHA_STRING_VALUE			(string)	string value
Quoted 7 lines

This is one of the areas where the legacy ethtool ioctl() is painful because we need to request a string set to know how many of those exist to allocate space for those in both kernel and user space.

If we could find a way to have a single command that allows us to dump stringset (count, values) and associated data, then we save ourselves a context switch and having to pre-allocate memory accordingly.

The way the proposed interface is designed, we can get both without an extra roundtrip but it's done the other way around, i.e. passing the strings with the request for data. The API allows two modes of operation:

1. One shot requests, e.g. typical ethtool use. We use "verbose" mode (no ETHA_*_COMPACT flag in the request) and data is provided together with the names. E.g. "ethtool eth2" request would look like

ETHA_SETTINGS_DEV

      ETHA_DEV_NAME = "eth2"

ETHA_SETTINGS_INFO_MASK = ... | ETH_SETTINGS_IM_LINKMODES | ...

and the reply would be

  ETHA_SETTINGS_DEV = {
      ETHA_DEV_INDEX = 4
      ETHA_DEV_NAME = "eth2"
  }
  ETHA_SETTINGS_LINK_INFO = {
      ...
  }
  ETHA_SETTRINGS_LINK_MODES = {
      ETHA_LINKMODES_AUTONEG = 1 (AUTONEG_ENABLE)
      ETHA_LINKMODES_OURS = {
          ETHA_BITSET_SIZE = 67
          ETHA_BITSET_BITS = {
              ETHA_BITS_BIT = {
                  ETHA_BIT_INDEX = 0 (ETHTOOL_LINK_MODE_10baseT_Half_BIT)
                  ETHA_BIT_NAME = "10baseT/Half"
              }
              ETHA_BITS_BIT = {
                  ETHA_BIT_INDEX = 1 (ETHTOOL_LINK_MODE_10baseT_Full_BIT)
                  ETHA_BIT_NAME = "10baseT/Full"
              }
              ETHA_BITS_BIT = {
                  ETHA_BIT_INDEX = 2 (ETHTOOL_LINK_MODE_100baseT_Half_BIT)
                  ETHA_BIT_NAME = "100baseT/Half"
                  ETHA_BIT_VALUE
              }
              ETHA_BITS_BIT = {
                  ETHA_BIT_INDEX = 3 (ETHTOOL_LINK_MODE_100baseT_Full_BIT)
                  ETHA_BIT_NAME = "100baseT/Full"
                  ETHA_BIT_VALUE
              }
              ETHA_BITS_BIT = {
                  ETHA_BIT_INDEX = 5 (ETHTOOL_LINK_MODE_1000baseT_Full_BIT)
                  ETHA_BIT_NAME = "1000baseT/Full"
                  ETHA_BIT_VALUE
              }
          }
      }
      ETHA_LINKMODES_PEER = {
          ...
      }
      ETHA_LINKMODES_SPEED = 1000 (SPEED_1000)
      ETHA_LINKMODES_DUPLEX = 1 (DUPLEX_FULL)
  }
  ...

2. Long time running applications, e.g. "ethtool --monitor", wicked, systemd-networkd, ... For these, it would make sense to cache the strings and either retrieve them in advance (on start and when new device appears) or when they are needed for the first time. The request would then include the ETHA_SETTINGS_COMPACT flag and reply would be

  ...
  ETHA_LINKMODES_OURS = {
      ETHA_BITSET_SIZE = 67
      ETHA_BITSET_VALUE = 0x0000002c 0x00000000 0x00000000
      ETHA_BITSET_MASK = 0x0000002f 0x00000000 0x00000000
  }

For "set" type requests, it's up to the application if it uses verbose or compact form. The verbose form is most useful when we want e.g. to set the values of one or two bits which may be identified by their names (on command line or in config file).

Michal

#67 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 85 lines · Thu, Mar 28, 2019 at 08:18:53AM CET, [email protected] wrote:
Quoted 35 lines · Florian Fainelli

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 184

Quoted 13 lines
Documentation/networking/ethtool-netlink.txt +13 -0
+GET_STRSET
+----------
+
+Requests contents of a string set as provided by ioctl commands
+ETHTOOL_GSSET_INFO and ETHTOOL_GSTRINGS. String sets are not user writeable so
+that the corresponding SET_STRSET message is only used in kernel replies.
+There are two types of string sets: global (independent of a device, e.g.
+device feature names) and device specific (e.g. device private flags).
+
+Request contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_COUNTS		(flag)		request only string counts

Should not that be part of the nested attribute under ETHA_STRSET_STRINGSET. We should probably think about adding another flag which indicates that we want to get the stringset associated data, see below why.

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 191

Quoted 7 lines
Documentation/networking/ethtool-netlink.txt +7 -0
+    ETHA_STRSET_STRINGSET	(nested)	string set to request
+        ETHA_STRINGSET_ID		(u32)		set id
+
+Kernel response contents:
+
+    ETHA_STRSET_DEV		(nested)	device identification
+    ETHA_STRSET_STRINGSET	(nested)	string set to request

string set requested?

Documentation/networking/ethtool-netlink.txt and also multiple events of the same type (e.g. two or more newly registered · line 196

Quoted 5 lines
Documentation/networking/ethtool-netlink.txt +5 -0
+        ETHA_STRINGSET_ID		(u32)		set id
+        ETHA_STRINGSET_COUNT		(u32)		number of strings
+        ETHA_STRINGSET_STRINGS		(nested)	array of strings
+            ETHA_STRING_INDEX			(u32)		string index
+            ETHA_STRING_VALUE			(string)	string value
Quoted 7 lines

This is one of the areas where the legacy ethtool ioctl() is painful because we need to request a string set to know how many of those exist to allocate space for those in both kernel and user space.

If we could find a way to have a single command that allows us to dump stringset (count, values) and associated data, then we save ourselves a context switch and having to pre-allocate memory accordingly.

The way the proposed interface is designed, we can get both without an extra roundtrip but it's done the other way around, i.e. passing the strings with the request for data. The API allows two modes of operation:

1. One shot requests, e.g. typical ethtool use. We use "verbose" mode (no ETHA_*_COMPACT flag in the request) and data is provided together with the names. E.g. "ethtool eth2" request would look like

ETHA_SETTINGS_DEV

     ETHA_DEV_NAME = "eth2"

ETHA_SETTINGS_INFO_MASK = ... | ETH_SETTINGS_IM_LINKMODES | ...

and the reply would be

ETHA_SETTINGS_DEV = {
ETHA_DEV_INDEX = 4
ETHA_DEV_NAME = "eth2"
}

 ETHA_SETTINGS_LINK_INFO = {
     ...
 }
 ETHA_SETTRINGS_LINK_MODES = {
     ETHA_LINKMODES_AUTONEG = 1 (AUTONEG_ENABLE)
     ETHA_LINKMODES_OURS = {
         ETHA_BITSET_SIZE = 67
         ETHA_BITSET_BITS = {
             ETHA_BITS_BIT = {
                 ETHA_BIT_INDEX = 0 (ETHTOOL_LINK_MODE_10baseT_Half_BIT)
                 ETHA_BIT_NAME = "10baseT/Half"
             }
             ETHA_BITS_BIT = {
                 ETHA_BIT_INDEX = 1 (ETHTOOL_LINK_MODE_10baseT_Full_BIT)
                 ETHA_BIT_NAME = "10baseT/Full"
             }
             ETHA_BITS_BIT = {
                 ETHA_BIT_INDEX = 2 (ETHTOOL_LINK_MODE_100baseT_Half_BIT)
                 ETHA_BIT_NAME = "100baseT/Half"
                 ETHA_BIT_VALUE
             }
             ETHA_BITS_BIT = {
Quoted 8 lines
                 ETHA_BIT_INDEX = 3 (ETHTOOL_LINK_MODE_100baseT_Full_BIT)
                 ETHA_BIT_NAME = "100baseT/Full"
                 ETHA_BIT_VALUE
             }
             ETHA_BITS_BIT = {
                 ETHA_BIT_INDEX = 5 (ETHTOOL_LINK_MODE_1000baseT_Full_BIT)
                 ETHA_BIT_NAME = "1000baseT/Full"
                 ETHA_BIT_VALUE

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};

and then there should be 2 attrs:
ETHTOOL_A_LINK_MODE_LIST_OUR /* nest */
ETHTOOL_A_LINK_MODE_LIST_PEER /* nest */
ETHTOOL_A_LINK_MODE /* u32 */

and then the message should look like:

ETHTOOL_A_LINK_MODE_LIST_OUR start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_OUR end nest ETHTOOL_A_LINK_MODE_LIST_PEER start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_PEER end nest

Nice and simple. No bits, no strings.

Quoted 30 lines
             }
         }
     }
     ETHA_LINKMODES_PEER = {
         ...
     }
     ETHA_LINKMODES_SPEED = 1000 (SPEED_1000)
     ETHA_LINKMODES_DUPLEX = 1 (DUPLEX_FULL)

}
...

2. Long time running applications, e.g. "ethtool --monitor", wicked, systemd-networkd, ... For these, it would make sense to cache the strings and either retrieve them in advance (on start and when new device appears) or when they are needed for the first time. The request would then include the ETHA_SETTINGS_COMPACT flag and reply would be

...

 ETHA_LINKMODES_OURS = {
     ETHA_BITSET_SIZE = 67
     ETHA_BITSET_VALUE = 0x0000002c 0x00000000 0x00000000
     ETHA_BITSET_MASK = 0x0000002f 0x00000000 0x00000000

}

For "set" type requests, it's up to the application if it uses verbose or compact form. The verbose form is most useful when we want e.g. to set the values of one or two bits which may be identified by their names (on command line or in config file).

Michal

#68 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in sync with kernel and (2) maintain their our tables of names. Experience shows we are not very good and satisfying these conditions even for the one which should be best at keeping up.

Quoted 22 lines

and then there should be 2 attrs:
ETHTOOL_A_LINK_MODE_LIST_OUR /* nest */
ETHTOOL_A_LINK_MODE_LIST_PEER /* nest */
ETHTOOL_A_LINK_MODE /* u32 */

and then the message should look like:

ETHTOOL_A_LINK_MODE_LIST_OUR start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_OUR end nest ETHTOOL_A_LINK_MODE_LIST_PEER start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_PEER end nest

Quoted 1 line

Nice and simple. No bits, no strings.

A bit too simple, actually. You would need third nest to distinguish supported and advertised modes. And for setting, you would also need two arrays if you want to set only some of the modes (unless you introduce something that would be similar to mine except for omitting the names).

More important: you still didn't explain how is your "no strings" approach supposed to work for bit sets where userspace cannot possibly know the set of available flags (e.g. the private flags).

Michal

#69 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 13 lines · Thu, Mar 28, 2019 at 03:04:28PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};
Quoted 2 lines

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in

That is how it is usually done. UAPI defines ATTRS and values, userspace assigns appropriate strings.

Quoted 3 lines

sync with kernel and (2) maintain their our tables of names. Experience shows we are not very good and satisfying these conditions even for the one which should be best at keeping up.

Moreover, I think that speed, duplex and type should be sent separatelly:

ETHTOOL_A_LINK_MODE_LIST_OUR start nest
ETHTOOL_A_LINK_MODE start nest
ETHTOOL_A_LINK_MODE_SPEED = 100 /* this should be u64 */ ETHTOOL_A_LINK_MODE_DUPLEX = ETHTOOL_LINK_MODE_DUPLEX_FULL ETHTOOL_A_LINK_MODE_TYPE = ETHTOOL_LINK_MODE_TYPE_BASET
ETHTOOL_A_LINK_MODE start end
ETHTOOL_A_LINK_MODE start nest
ETHTOOL_A_LINK_MODE_SPEED = 10 ETHTOOL_A_LINK_MODE_DUPLEX = ETHTOOL_LINK_MODE_DUPLEX_HALF ETHTOOL_A_LINK_MODE_TYPE = ETHTOOL_LINK_MODE_TYPE_BASET
ETHTOOL_A_LINK_MODE start end
ETHTOOL_A_LINK_MODE_LIST_PEER end nest

Does not really make sense to combine those 3 attributes together.

Quoted 23 lines
Quoted 23 lines

and then there should be 2 attrs:
ETHTOOL_A_LINK_MODE_LIST_OUR /* nest */
ETHTOOL_A_LINK_MODE_LIST_PEER /* nest */
ETHTOOL_A_LINK_MODE /* u32 */

and then the message should look like:

ETHTOOL_A_LINK_MODE_LIST_OUR start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_OUR end nest ETHTOOL_A_LINK_MODE_LIST_PEER start nest
ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_10baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Half ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_100baseT_Full ETHTOOL_A_LINK_MODE = ETHTOOL_LINK_MODE_1000baseT_Full
ETHTOOL_A_LINK_MODE_LIST_PEER end nest

Quoted 3 lines
Quoted 1 line

Nice and simple. No bits, no strings.

A bit too simple, actually. You would need third nest to distinguish

It is just an example.

Quoted 7 lines

supported and advertised modes. And for setting, you would also need two arrays if you want to set only some of the modes (unless you introduce something that would be similar to mine except for omitting the names).

More important: you still didn't explain how is your "no strings" approach supposed to work for bit sets where userspace cannot possibly know the set of available flags (e.g. the private flags).

I know, that is a different part of this thread.

Quoted 2 lines

Michal

#70 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 14 lines · Jiri Pirko
Quoted 13 lines · Thu, Mar 28, 2019 at 03:04:28PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};  
Quoted 5 lines
Quoted 2 lines

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in

That is how it is usually done. UAPI defines ATTRS and values, userspace assigns appropriate strings.

+1 FWIW, I'm with Jiri on the string situation.

#71 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 15 lines · Jakub Kicinski
Quoted 14 lines · Jiri Pirko
Quoted 13 lines · Thu, Mar 28, 2019 at 03:04:28PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};  
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in

That is how it is usually done. UAPI defines ATTRS and values, userspace assigns appropriate strings.

+1 FWIW, I'm with Jiri on the string situation.

And I'm still waiting for any of you to tell me how would you handle private flags, stats, tests etc. without the string sets.

Ditching the verbose form of bit sets would be a nuisance for userspace using the interface but compared to e.g. having to mix three different kernel interfaces, it's just a minor problem. Ditching the static string sets would mean giving up the opportunity to get rid of having to sync all kinds of tables with every userspace consumer whenever a new flag is introduced. Pity... but still doable. But how do you want to do without the string sets which are provided by drivers?

Michal

#72 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 27 lines · Michal Kubecek
Quoted 15 lines · Jakub Kicinski
Quoted 14 lines · Jiri Pirko
Quoted 13 lines · Thu, Mar 28, 2019 at 03:04:28PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};    
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in

That is how it is usually done. UAPI defines ATTRS and values, userspace assigns appropriate strings.

+1 FWIW, I'm with Jiri on the string situation.

And I'm still waiting for any of you to tell me how would you handle private flags, stats, tests etc. without the string sets.

Quoted 7 lines

Ditching the verbose form of bit sets would be a nuisance for userspace using the interface but compared to e.g. having to mix three different kernel interfaces, it's just a minor problem. Ditching the static string sets would mean giving up the opportunity to get rid of having to sync all kinds of tables with every userspace consumer whenever a new flag is introduced. Pity... but still doable. But how do you want to do without the string sets which are provided by drivers?

I don't think there is a disagreement about stuff which is private to the driver. The strings have to be exposed there.

#73 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

Quoted 36 lines · Thu, Mar 28, 2019 at 10:06:07PM CET, [email protected] wrote:
Quoted 27 lines · Michal Kubecek
Quoted 15 lines · Jakub Kicinski
Quoted 14 lines · Jiri Pirko
Quoted 13 lines · Thu, Mar 28, 2019 at 03:04:28PM CET, [email protected] wrote:
Quoted 11 lines · Jiri Pirko

I don't like this. This should not be bitfield/set. This should be simply nested array of enum values:

enum ethtool_link_mode {
	ETHTOOL_LINK_MODE_10baseT_Half,
	ETHTOOL_LINK_MODE_10baseT_Full,
	ETHTOOL_LINK_MODE_100baseT_Half,
	ETHTOOL_LINK_MODE_100baseT_Full,
	ETHTOOL_LINK_MODE_1000baseT_Full,
};    
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

We already have such enum. The problem with your "no string" approach is that it requires all userspace applications to (1) keep this enum in

That is how it is usually done. UAPI defines ATTRS and values, userspace assigns appropriate strings.

+1 FWIW, I'm with Jiri on the string situation.

And I'm still waiting for any of you to tell me how would you handle private flags, stats, tests etc. without the string sets.

Quoted 7 lines

Ditching the verbose form of bit sets would be a nuisance for userspace using the interface but compared to e.g. having to mix three different kernel interfaces, it's just a minor problem. Ditching the static string sets would mean giving up the opportunity to get rid of having to sync all kinds of tables with every userspace consumer whenever a new flag is introduced. Pity... but still doable. But how do you want to do without the string sets which are provided by drivers?

Quoted 2 lines

I don't think there is a disagreement about stuff which is private to the driver. The strings have to be exposed there.

Yeah, there is no other way of doing that. However, please check if some driver-specific setting cannot be moved to devlink with use of devlink params (we have both devlink driver specific params and per-port devlink driver specific params).

#74 lore raw

Re: [PATCH net-next v5 12/22] ethtool: provide string sets with GET_STRSET request

In reply to Jiri Pirko
Quoted 17 lines · Jiri Pirko

Moreover, I think that speed, duplex and type should be sent separatelly:

ETHTOOL_A_LINK_MODE_LIST_OUR start nest
ETHTOOL_A_LINK_MODE start nest
ETHTOOL_A_LINK_MODE_SPEED = 100 /* this should be u64 */ ETHTOOL_A_LINK_MODE_DUPLEX = ETHTOOL_LINK_MODE_DUPLEX_FULL ETHTOOL_A_LINK_MODE_TYPE = ETHTOOL_LINK_MODE_TYPE_BASET
ETHTOOL_A_LINK_MODE start end
ETHTOOL_A_LINK_MODE start nest
ETHTOOL_A_LINK_MODE_SPEED = 10 ETHTOOL_A_LINK_MODE_DUPLEX = ETHTOOL_LINK_MODE_DUPLEX_HALF ETHTOOL_A_LINK_MODE_TYPE = ETHTOOL_LINK_MODE_TYPE_BASET
ETHTOOL_A_LINK_MODE start end
ETHTOOL_A_LINK_MODE_LIST_PEER end nest

Quoted 1 line

Does not really make sense to combine those 3 attributes together.

This helped me to realize the primary source of misunderstanding: as I'm working on kernel and ethtool implementation simultaneously, I don't look only at the API itself and don't consider only if its design is clean and logical. I always think about the actual userspace programs wanting to use the interface and try to imagine what the design means for them.

So when you say e.g. "this belongs to rtnetlink", I have to admit that from strictly logical point of view you are right but at the same time, chain of thoughts starts: "that means two sockets; we need a table which command needs which, some might need both, monitor will have to listen at two sockets, that's fork/pthread/poll..."

From this point of view, the scheme above which, on its own, makes perfect sense (there might be a bit of a problem with 10000baseR_FEC mode but that one is a trouble in any scheme), means that one side will translate the link mode number to the three parameters and the other will look the triplet out in its own table to get the link mode number. On userspace side, there will be another translation between link mode number and name. What exactly is the gain from such representation? No idea.

There is one crucial difference between ethtool and devlink. You are building devlink from scratch so you define the logic, define the API based on that logic and make both NIC drivers and userspace conform to it. With ethtool, the situation is exactly the opposite: on one side, there are ethtool_ops as a constraint, on the other, it's ethtool with its feature set and users used to it. Changing ethtool_ops is going to be slow and painful process. Changing the users and their habits... we already know how that works.

Michal

#75 lore raw

PATCH v5 13/22 ethtool: provide driver/device information in GET_INFO request

In reply to Michal Kubecek

Implement GET_INFO request to get basic driver and device information as provided by ETHTOOL_GDRVINFO ioct command. The information is read only so that the corresponding SET_INFO message is only used in kernel replies.

Move most of ethtool_get_drvinfo() int common.c so that the code can be shared by both ioctl and netlink interface.

---
 Documentation/networking/ethtool-netlink.txt |  43 ++++-
 include/uapi/linux/ethtool_netlink.h         |  30 ++++
 net/ethtool/Makefile                         |   2 +-
 net/ethtool/common.c                         |  52 ++++++
 net/ethtool/common.h                         |   2 +
 net/ethtool/info.c                           | 158 +++++++++++++++++++
 net/ethtool/ioctl.c                          |  50 +-----
 net/ethtool/netlink.c                        |   8 +
 net/ethtool/netlink.h                        |   1 +
 9 files changed, 299 insertions(+), 47 deletions(-)
 create mode 100644 net/ethtool/info.c
Documentation/networking/ethtool-netlink.txt +42 -1
index 1508c16a236e..cffa508ca6c6 100644
@@ -129,6 +129,8 @@ List of message types
     ETHNL_CMD_EVENT			notification only
     ETHNL_CMD_GET_STRSET
     ETHNL_CMD_SET_STRSET		response only
+    ETHNL_CMD_GET_INFO
+    ETHNL_CMD_SET_INFO			response only
 
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
@@ -209,6 +211,45 @@ ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not
 the actual strings.
 
 
+GET_INFO
+--------
+
+GET_INFO requests information provided by ioctl commands ETHTOOL_GDRVINFO,
+ETHTOOL_GPERMADDR and ETHTOOL_GET_TS_INFO to provide basic device information.
+Common pattern is that all information is read only so that SET_INFO message
+exists but is only used by kernel for replies to GET_INFO requests. There is
+also no corresponding notification.
+
+Request contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_INFOMASK		(u32)		info mask
+    ETHA_INFO_COMPACT		(flag)		request compact bitsets
+
+Info mask bits meaning:
+
+    ETH_INFO_IM_DRVINFO			driver info (GDRVINFO)
+    ETH_INFO_IM_PERMADDR		permanent HW address (GPERMADDR)
+    ETH_INFO_IM_TSINFO			timestamping info (GET_TS_INFO)
+
+Kernel response contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_DRVINFO		(nested)	driver information
+        ETHA_DRVINFO_DRIVER		(string)	driver name
+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version
+
+The meaning of DRVINFO attributes follows the corresponding fields of
+ETHTOOL_GDRVINFO response. Second part with various counts and sizes is
+omitted as these are not really needed (and if they are, they can be easily
+found by different means). Driver version is also omitted as it is rather
+misleading in most cases.
+
+GET_INFO requests allow dumps.
+
+
 Request translation
 -------------------
 
@@ -220,7 +261,7 @@ ioctl command			netlink command
 ---------------------------------------------------------------------
 ETHTOOL_GSET			n/a
 ETHTOOL_SSET			n/a
-ETHTOOL_GDRVINFO		n/a
+ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
 ETHTOOL_GWOL			n/a
 ETHTOOL_SWOL			n/a
include/uapi/linux/ethtool_netlink.h +30 -0
index 66aeb436b822..386d4273ac44 100644
@@ -16,6 +16,8 @@ enum {
 	ETHNL_CMD_EVENT,		/* only for notifications */
 	ETHNL_CMD_GET_STRSET,
 	ETHNL_CMD_SET_STRSET,		/* only for reply */
+	ETHNL_CMD_GET_INFO,
+	ETHNL_CMD_SET_INFO,		/* only for reply */
 
 	__ETHNL_CMD_CNT,
 	ETHNL_CMD_MAX = (__ETHNL_CMD_CNT - 1)
@@ -141,6 +143,34 @@ enum {
 	ETHA_STRSET_MAX = (__ETHA_STRSET_CNT - 1)
 };
 
+/* GET_INFO / SET_INFO */
+
+enum {
+	ETHA_INFO_UNSPEC,
+	ETHA_INFO_DEV,				/* nest - ETHA_DEV_* */
+	ETHA_INFO_INFOMASK,			/* u32 */
+	ETHA_INFO_COMPACT,			/* flag */
+	ETHA_INFO_DRVINFO,			/* nest - ETHA_DRVINFO_* */
+
+	__ETHA_INFO_CNT,
+	ETHA_INFO_MAX = (__ETHA_INFO_CNT - 1)
+};
+
+#define ETH_INFO_IM_DRVINFO			(1U << 0)
+
+#define ETH_INFO_IM_ALL (ETH_INFO_IM_DRVINFO)
+
+enum {
+	ETHA_DRVINFO_UNSPEC,
+	ETHA_DRVINFO_DRIVER,			/* string */
+	ETHA_DRVINFO_FWVERSION,			/* string */
+	ETHA_DRVINFO_BUSINFO,			/* string */
+	ETHA_DRVINFO_EROM_VER,			/* string */
+
+	__ETHA_DRVINFO_CNT,
+	ETHA_DRVINFO_MAX = (__ETHA_DRVINFO_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/Makefile +1 -1
index 11ceb00821b3..96d41dc45d4f 100644
@@ -4,4 +4,4 @@ obj-y				+= ioctl.o common.o
 
 obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
 
-ethtool_nl-y	:= netlink.o bitset.o strset.o
+ethtool_nl-y	:= netlink.o bitset.o strset.o info.o
net/ethtool/common.c +52 -0

60 lines not on this page · raw message

net/ethtool/common.h +2 -0
index 41b2efc1e4e1..e87e58b3a274 100644
@@ -3,6 +3,7 @@
 #ifndef _ETHTOOL_COMMON_H
 #define _ETHTOOL_COMMON_H
 
+#include <linux/netdevice.h>
 #include <linux/ethtool.h>
 
 extern const char
@@ -14,4 +15,5 @@ tunable_strings[__ETHTOOL_TUNABLE_COUNT][ETH_GSTRING_LEN];
 extern const char
 phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
 
+int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info);
 #endif /* _ETHTOOL_COMMON_H */
net/ethtool/info.c new file +158 -0

158 lines not on this page · raw message

net/ethtool/ioctl.c +5 -45

59 lines not on this page · raw message

net/ethtool/netlink.c +8 -0
index 63ee0f846b50..7f9597c0c7a7 100644
@@ -155,6 +155,7 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 
 const struct get_request_ops *get_requests[__ETHNL_CMD_CNT] = {
 	[ETHNL_CMD_GET_STRSET]		= &strset_request_ops,
+	[ETHNL_CMD_GET_INFO]		= &info_request_ops,
 };
 
 /**
@@ -568,6 +569,13 @@ static const struct genl_ops ethtool_genl_ops[] = {
 		.dumpit	= ethnl_get_dumpit,
 		.done	= ethnl_get_done,
 	},
+	{
+		.cmd	= ETHNL_CMD_GET_INFO,
+		.doit	= ethnl_get_doit,
+		.start	= ethnl_get_start,
+		.dumpit	= ethnl_get_dumpit,
+		.done	= ethnl_get_done,
+	},
 };
 
 static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
net/ethtool/netlink.h +1 -0
index 32d85bb5c49a..48980ae9e963 100644
@@ -278,5 +278,6 @@ struct get_request_ops {
 /* request handlers */
 
 extern const struct get_request_ops strset_request_ops;
+extern const struct get_request_ops info_request_ops;
 
 #endif /* _NET_ETHTOOL_NETLINK_H */
Signature · 1 line
2.21.0
#76 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 57 lines · Mon, Mar 25, 2019 at 06:08:33PM CET, [email protected] wrote:

Implement GET_INFO request to get basic driver and device information as provided by ETHTOOL_GDRVINFO ioct command. The information is read only so that the corresponding SET_INFO message is only used in kernel replies.

Move most of ethtool_get_drvinfo() int common.c so that the code can be shared by both ioctl and netlink interface.

---

Documentation/networking/ethtool-netlink.txt |  43 ++++-
include/uapi/linux/ethtool_netlink.h         |  30 ++++
net/ethtool/Makefile                         |   2 +-
net/ethtool/common.c                         |  52 ++++++
net/ethtool/common.h                         |   2 +
net/ethtool/info.c                           | 158 +++++++++++++++++++
net/ethtool/ioctl.c                          |  50 +-----
net/ethtool/netlink.c                        |   8 +
net/ethtool/netlink.h                        |   1 +
9 files changed, 299 insertions(+), 47 deletions(-)

create mode 100644 net/ethtool/info.c

Documentation/networking/ethtool-netlink.txt +2 -0
index 1508c16a236e..cffa508ca6c6 100644
@@ -129,6 +129,8 @@ List of message types
    ETHNL_CMD_EVENT			notification only
    ETHNL_CMD_GET_STRSET
    ETHNL_CMD_SET_STRSET		response only
+    ETHNL_CMD_GET_INFO
+    ETHNL_CMD_SET_INFO			response only
 

All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT" to indicate the type.

Documentation/networking/ethtool-netlink.txt +0 -0
@@ -209,6 +211,45 @@ ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not

the actual strings.

Documentation/networking/ethtool-netlink.txt +19 -0
+GET_INFO
+--------
+
+GET_INFO requests information provided by ioctl commands ETHTOOL_GDRVINFO,
+ETHTOOL_GPERMADDR and ETHTOOL_GET_TS_INFO to provide basic device information.
+Common pattern is that all information is read only so that SET_INFO message
+exists but is only used by kernel for replies to GET_INFO requests. There is
+also no corresponding notification.
+
+Request contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_INFOMASK		(u32)		info mask
+    ETHA_INFO_COMPACT		(flag)		request compact bitsets
+
+Info mask bits meaning:
+
+    ETH_INFO_IM_DRVINFO			driver info (GDRVINFO)
+    ETH_INFO_IM_PERMADDR		permanent HW address (GPERMADDR)

Documentation/networking/ethtool-netlink.txt ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not · line 242

Quoted 10 lines
Documentation/networking/ethtool-netlink.txt +10 -0
+    ETH_INFO_IM_TSINFO			timestamping info (GET_TS_INFO)
+
+Kernel response contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_DRVINFO		(nested)	driver information
+        ETHA_DRVINFO_DRIVER		(string)	driver name
+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version

These are already very nicely supported in devlink. No need to duplicate here.

#77 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 13 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not · line 242

Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:33PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt +9 -0
+
+Kernel response contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_DRVINFO		(nested)	driver information
+        ETHA_DRVINFO_DRIVER		(string)	driver name
+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version

These are already very nicely supported in devlink. No need to duplicate here.

They are supported by devlink as an interface. But devlink itself is only supported by few NIC drivers at the moment:

mike@unicorn:~/work/git/net-next> grep -r devlink_ops drivers/net/ drivers/net/ethernet/mellanox/mlx4/main.c:static const struct devlink_ops mlx4_devlink_ops = {
drivers/net/ethernet/mellanox/mlx4/main.c: devlink = devlink_alloc(&mlx4_devlink_ops, sizeof(*priv));
drivers/net/ethernet/mellanox/mlxsw/core.c:static const struct devlink_ops mlxsw_devlink_ops = {
drivers/net/ethernet/mellanox/mlxsw/core.c: devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
drivers/net/ethernet/mellanox/mlx5/core/main.c:static const struct devlink_ops mlx5_devlink_ops = { drivers/net/ethernet/mellanox/mlx5/core/main.c: devlink = devlink_alloc(&mlx5_devlink_ops, sizeof(*dev)); drivers/net/ethernet/cavium/liquidio/lio_main.c:static const struct devlink_ops liquidio_devlink_ops = {
drivers/net/ethernet/cavium/liquidio/lio_main.c: devlink = devlink_alloc(&liquidio_devlink_ops,
drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c:static const struct devlink_ops bnxt_dl_ops = { drivers/net/ethernet/netronome/nfp/nfp_main.c: devlink = devlink_alloc(&nfp_devlink_ops, sizeof(*pf)); drivers/net/ethernet/netronome/nfp/nfp_main.h:extern const struct devlink_ops nfp_devlink_ops; drivers/net/ethernet/netronome/nfp/nfp_devlink.c:const struct devlink_ops nfp_devlink_ops = { drivers/net/netdevsim/devlink.c:static const struct devlink_ops nsim_devlink_ops = {
drivers/net/netdevsim/devlink.c: devlink = devlink_alloc(&nsim_devlink_ops, 0);

That's 6 drivers from 4 vendors (if I don't count netdevsim). And I did not check if all of them do actually provide the information shown above. On the other hand:

mike@unicorn:~/work/git/net-next> egrep -r '\.get_drvinfo' drivers/net/ | wc -l 240

Some of these 240 lines assign the same handler but not enough to make me optimistic about being able to implement "ethtool -i <dev>" using devlink interface in near future (say few months or one year).

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Michal

#78 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 15 lines · Michal Kubecek
Quoted 13 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not · line 242

Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:33PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt +9 -0
+
+Kernel response contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_DRVINFO		(nested)	driver information
+        ETHA_DRVINFO_DRIVER		(string)	driver name
+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version

These are already very nicely supported in devlink. No need to duplicate here.

Quoted 4 lines

They are supported by devlink as an interface. But devlink itself is only supported by few NIC drivers at the moment:

mike@unicorn:~/work/git/net-next> grep -r devlink_ops drivers/net/

You might want to include net/ in your list too.

Quoted 26 lines

drivers/net/ethernet/mellanox/mlx4/main.c:static const struct devlink_ops mlx4_devlink_ops = {
drivers/net/ethernet/mellanox/mlx4/main.c: devlink = devlink_alloc(&mlx4_devlink_ops, sizeof(*priv));
drivers/net/ethernet/mellanox/mlxsw/core.c:static const struct devlink_ops mlxsw_devlink_ops = {
drivers/net/ethernet/mellanox/mlxsw/core.c: devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
drivers/net/ethernet/mellanox/mlx5/core/main.c:static const struct devlink_ops mlx5_devlink_ops = { drivers/net/ethernet/mellanox/mlx5/core/main.c: devlink = devlink_alloc(&mlx5_devlink_ops, sizeof(*dev)); drivers/net/ethernet/cavium/liquidio/lio_main.c:static const struct devlink_ops liquidio_devlink_ops = {
drivers/net/ethernet/cavium/liquidio/lio_main.c: devlink = devlink_alloc(&liquidio_devlink_ops,
drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c:static const struct devlink_ops bnxt_dl_ops = { drivers/net/ethernet/netronome/nfp/nfp_main.c: devlink = devlink_alloc(&nfp_devlink_ops, sizeof(*pf)); drivers/net/ethernet/netronome/nfp/nfp_main.h:extern const struct devlink_ops nfp_devlink_ops; drivers/net/ethernet/netronome/nfp/nfp_devlink.c:const struct devlink_ops nfp_devlink_ops = { drivers/net/netdevsim/devlink.c:static const struct devlink_ops nsim_devlink_ops = {
drivers/net/netdevsim/devlink.c: devlink = devlink_alloc(&nsim_devlink_ops, 0);

That's 6 drivers from 4 vendors (if I don't count netdevsim). And I did not check if all of them do actually provide the information shown above. On the other hand:

mike@unicorn:~/work/git/net-next> egrep -r '\.get_drvinfo' drivers/net/ | wc -l 240

Some of these 240 lines assign the same handler but not enough to make me optimistic about being able to implement "ethtool -i <dev>" using devlink interface in near future (say few months or one year).

Quoted 5 lines

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Can't we just interrogate devlink first if there is a devlink <-> net_device mapping and fallback to ethtool_ops::get_drvinfo if that did not work? Maybe issue a (rate-limited) warning "use devlink instead".

Signature · 1 line
Florian
#79 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

In reply to Michal Kubecek
Quoted 45 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:
Quoted 13 lines · Jiri Pirko

Documentation/networking/ethtool-netlink.txt ETHA_STRSET_COUNTS tells kernel to only return string counts of the sets, not · line 242

Quoted 9 lines · Mon, Mar 25, 2019 at 06:08:33PM CET, [email protected] wrote:
Documentation/networking/ethtool-netlink.txt +9 -0
+
+Kernel response contents:
+
+    ETHA_INFO_DEV		(nested)	device identification
+    ETHA_INFO_DRVINFO		(nested)	driver information
+        ETHA_DRVINFO_DRIVER		(string)	driver name
+        ETHA_DRVINFO_FWVERSION		(string)	firmware version
+        ETHA_DRVINFO_BUSINFO		(string)	device bus address
+        ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version

These are already very nicely supported in devlink. No need to duplicate here.

They are supported by devlink as an interface. But devlink itself is only supported by few NIC drivers at the moment:

mike@unicorn:~/work/git/net-next> grep -r devlink_ops drivers/net/ drivers/net/ethernet/mellanox/mlx4/main.c:static const struct devlink_ops mlx4_devlink_ops = {
drivers/net/ethernet/mellanox/mlx4/main.c: devlink = devlink_alloc(&mlx4_devlink_ops, sizeof(*priv));
drivers/net/ethernet/mellanox/mlxsw/core.c:static const struct devlink_ops mlxsw_devlink_ops = {
drivers/net/ethernet/mellanox/mlxsw/core.c: devlink = devlink_alloc(&mlxsw_devlink_ops, alloc_size);
drivers/net/ethernet/mellanox/mlx5/core/main.c:static const struct devlink_ops mlx5_devlink_ops = { drivers/net/ethernet/mellanox/mlx5/core/main.c: devlink = devlink_alloc(&mlx5_devlink_ops, sizeof(*dev)); drivers/net/ethernet/cavium/liquidio/lio_main.c:static const struct devlink_ops liquidio_devlink_ops = {
drivers/net/ethernet/cavium/liquidio/lio_main.c: devlink = devlink_alloc(&liquidio_devlink_ops,
drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c:static const struct devlink_ops bnxt_dl_ops = { drivers/net/ethernet/netronome/nfp/nfp_main.c: devlink = devlink_alloc(&nfp_devlink_ops, sizeof(*pf)); drivers/net/ethernet/netronome/nfp/nfp_main.h:extern const struct devlink_ops nfp_devlink_ops; drivers/net/ethernet/netronome/nfp/nfp_devlink.c:const struct devlink_ops nfp_devlink_ops = { drivers/net/netdevsim/devlink.c:static const struct devlink_ops nsim_devlink_ops = {
drivers/net/netdevsim/devlink.c: devlink = devlink_alloc(&nsim_devlink_ops, 0);

That's 6 drivers from 4 vendors (if I don't count netdevsim). And I did not check if all of them do actually provide the information shown above. On the other hand:

mike@unicorn:~/work/git/net-next> egrep -r '\.get_drvinfo' drivers/net/ | wc -l 240

Some of these 240 lines assign the same handler but not enough to make me optimistic about being able to implement "ethtool -i <dev>" using devlink interface in near future (say few months or one year).

Quoted 5 lines

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

Quoted 2 lines

Michal

#80 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 8 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Quoted 6 lines

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Michal

#81 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 16 lines · Thu, Mar 28, 2019 at 10:53:47AM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Quoted 6 lines

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

Quoted 6 lines

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Yep. Plan to do that next week.

Quoted 2 lines

Michal

#82 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 17 lines · Jiri Pirko
Quoted 16 lines · Thu, Mar 28, 2019 at 10:53:47AM CET, [email protected] wrote:
Quoted 14 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

Quoted 8 lines
Quoted 6 lines

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Yep. Plan to do that next week.

The problem with the main part of dev info - fw_version - is that it is often overloaded in drivers and becomes impossible to parse for users. I'd rather we didn't dump that nasty chaos in devlink info and let it die with ethtool IOCTL.

Flashing can also be handled at user space tool level.

#83 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 25 lines · Jakub Kicinski
Quoted 24 lines · Jiri Pirko
Quoted 22 lines · Thu, Mar 28, 2019 at 10:53:47AM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Quoted 6 lines

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Quoted 6 lines
Quoted 1 line

Yep. Plan to do that next week.

The problem with the main part of dev info - fw_version - is that it is often overloaded in drivers and becomes impossible to parse for users. I'd rather we didn't dump that nasty chaos in devlink info and let it die with ethtool IOCTL.

If I understand you correctly, what you are suggesting would result in "ethtool -i" through netlink not showing firmware version for NICs without actual devlink implementation (which is vast majority of NICs at the moment). I'm afraid doing that would make sure ioctl ethtool wouldn't die any time soon.

Quoted 1 line

Flashing can also be handled at user space tool level.

I'm not sure how to understand this.

Michal

#84 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

Quoted 33 lines · Michal Kubecek
Quoted 25 lines · Jakub Kicinski
Quoted 24 lines · Jiri Pirko
Quoted 22 lines · Thu, Mar 28, 2019 at 10:53:47AM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Quoted 6 lines

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Quoted 6 lines
Quoted 1 line

Yep. Plan to do that next week.

The problem with the main part of dev info - fw_version - is that it is often overloaded in drivers and becomes impossible to parse for users. I'd rather we didn't dump that nasty chaos in devlink info and let it die with ethtool IOCTL.

Quoted 5 lines

If I understand you correctly, what you are suggesting would result in "ethtool -i" through netlink not showing firmware version for NICs without actual devlink implementation (which is vast majority of NICs at the moment). I'm afraid doing that would make sure ioctl ethtool wouldn't die any time soon.

I was trying to say that'd be opposed to showing the conglomerate fw_version from ethtool in devlink. I'm not opposed to showing it in ethnl or having ethnl call into devlink to produce the conglomerate, but I'd rather not see the legacy of ethtool creeping back into devlink.

Quoted 3 lines
Quoted 1 line

Flashing can also be handled at user space tool level.

I'm not sure how to understand this.

If we want ethnl user space command to provide flashing, it can just send a devlink CMD behind the scenes.

#85 lore raw

Re: [PATCH net-next v5 13/22] ethtool: provide driver/device information in GET_INFO request

In reply to Jakub Kicinski
Quoted 25 lines · Jakub Kicinski
Quoted 24 lines · Jiri Pirko
Quoted 22 lines · Thu, Mar 28, 2019 at 10:53:47AM CET, [email protected] wrote:
Quoted 8 lines · Jiri Pirko
Quoted 6 lines · Wed, Mar 27, 2019 at 11:25:54PM CET, [email protected] wrote:

I'm all for implementing new features which are are related to physical device (ASIC) rather than network interface only in devlink (at the level of kernel-userspace interface). But for features already provided by ethtool (userspace utility) I can't help seeing the state of devlink support in NIC drivers as a serious blocker.

Quoted 6 lines

What I'm thinking about at for some time now would be en explicit default devlink and devlink_port registration for every netdev for drivers that does not support devlink themselves. I need to think about it more, but it seems doable. Then we can hang appropriate things there and make the ethtoolnl/devlink separation clear. I believe we need to do it.

That sounds great, such "generic devlink" implementation would be a way around. Kernel could then emulate features which are not implemented by an actual devlink handler (i.e. "generic devlink" is used or particular handler is missing) by falling back to ethtool_ops handler so that userspace could rely on devlink API for things like device information, various dumps, flashing etc. without losing anything.

Quoted 9 lines
Quoted 1 line

Yep. Plan to do that next week.

The problem with the main part of dev info - fw_version - is that it is often overloaded in drivers and becomes impossible to parse for users. I'd rather we didn't dump that nasty chaos in devlink info and let it die with ethtool IOCTL.

Flashing can also be handled at user space tool level.

Yes, I really like how sfc does it which is to expose MTD devices and certainly have a custom flashing tool, that should be the way to go IMHO.

Signature · 1 line
Florian
#86 lore raw

PATCH v5 14/22 ethtool: provide timestamping information in GET_INFO request

In reply to Michal Kubecek

Add timestamping information as provided by ETHTOOL_GET_TS_INFO ioctl command in GET_INFO reply if ETH_INFO_IM_TSINFO flag is set in the request.

Add constants for counts of HWTSTAMP_TX_* and HWTSTAM_FILTER_* constants and provide symbolic names for timestamping related values so that they can be retrieved in GET_STRSET and GET_INFO requests.

---
 Documentation/networking/ethtool-netlink.txt |  10 +-
 include/uapi/linux/ethtool.h                 |   6 +
 include/uapi/linux/ethtool_netlink.h         |  16 ++-
 include/uapi/linux/net_tstamp.h              |  13 ++
 net/ethtool/common.c                         |  24 ++++
 net/ethtool/common.h                         |   2 +
 net/ethtool/info.c                           | 138 +++++++++++++++++++
 net/ethtool/ioctl.c                          |  20 +--
 net/ethtool/netlink.h                        |   9 ++
 net/ethtool/strset.c                         |  18 +++
 10 files changed, 236 insertions(+), 20 deletions(-)
Documentation/networking/ethtool-netlink.txt +9 -1
index cffa508ca6c6..74c62f11810c 100644
@@ -240,6 +240,11 @@ Kernel response contents:
         ETHA_DRVINFO_FWVERSION		(string)	firmware version
         ETHA_DRVINFO_BUSINFO		(string)	device bus address
         ETHA_DRVINFO_EROM_VER		(string)	expansion ROM version
+    ETHA_INFO_TSINFO		(nested)	timestamping information
+        ETHA_TSINFO_TIMESTAMPING	(bitset)	supported flags
+        ETHA_TSINFO_PHC_INDEX		(u32)		associated PHC
+        ETHA_TSINFO_TX_TYPES		(bitset)	Tx types
+        ETHA_TSINFO_RX_FILTERS		(bitset)	Rx filters
 
 The meaning of DRVINFO attributes follows the corresponding fields of
 ETHTOOL_GDRVINFO response. Second part with various counts and sizes is
@@ -247,6 +252,9 @@ omitted as these are not really needed (and if they are, they can be easily
 found by different means). Driver version is also omitted as it is rather
 misleading in most cases.
 
+ETHA_TSINFO_PHC_INDEX can be unsigned as there is no need for special value;
+if no PHC is associated, the attribute is not present.
+
 GET_INFO requests allow dumps.
 
 
@@ -322,7 +330,7 @@ ETHTOOL_SCHANNELS		n/a
 ETHTOOL_SET_DUMP		n/a
 ETHTOOL_GET_DUMP_FLAG		n/a
 ETHTOOL_GET_DUMP_DATA		n/a
-ETHTOOL_GET_TS_INFO		n/a
+ETHTOOL_GET_TS_INFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GMODULEINFO		n/a
 ETHTOOL_GMODULEEEPROM		n/a
 ETHTOOL_GEEE			n/a
include/uapi/linux/ethtool.h +6 -0
index eaea804972f6..882a542eaaa9 100644
@@ -563,6 +563,9 @@ struct ethtool_pauseparam {
  * @ETH_SS_RSS_HASH_FUNCS: RSS hush function names
  * @ETH_SS_PHY_STATS: Statistic names, for use with %ETHTOOL_GPHYSTATS
  * @ETH_SS_PHY_TUNABLES: PHY tunable names
+ * @ETH_SS_TSTAMP_SOF: timestamping flag names
+ * @ETH_SS_TSTAMP_TX_TYPE: timestamping Tx type names
+ * @ETH_SS_TSTAMP_RX_FILTER: timestamping Rx filter names
  */
 enum ethtool_stringset {
 	ETH_SS_TEST		= 0,
@@ -574,6 +577,9 @@ enum ethtool_stringset {
 	ETH_SS_TUNABLES,
 	ETH_SS_PHY_STATS,
 	ETH_SS_PHY_TUNABLES,
+	ETH_SS_TSTAMP_SOF,
+	ETH_SS_TSTAMP_TX_TYPE,
+	ETH_SS_TSTAMP_RX_FILTER,
 
 	ETH_SS_COUNT
 };
include/uapi/linux/ethtool_netlink.h +15 -1
index 386d4273ac44..e7a6e813150b 100644
@@ -151,14 +151,17 @@ enum {
 	ETHA_INFO_INFOMASK,			/* u32 */
 	ETHA_INFO_COMPACT,			/* flag */
 	ETHA_INFO_DRVINFO,			/* nest - ETHA_DRVINFO_* */
+	ETHA_INFO_TSINFO,			/* nest - ETHA_TSINFO_* */
 
 	__ETHA_INFO_CNT,
 	ETHA_INFO_MAX = (__ETHA_INFO_CNT - 1)
 };
 
 #define ETH_INFO_IM_DRVINFO			(1U << 0)
+#define ETH_INFO_IM_TSINFO			(1U << 1)
 
-#define ETH_INFO_IM_ALL (ETH_INFO_IM_DRVINFO)
+#define ETH_INFO_IM_ALL (ETH_INFO_IM_DRVINFO | \
+			 ETH_INFO_IM_TSINFO)
 
 enum {
 	ETHA_DRVINFO_UNSPEC,
@@ -171,6 +174,17 @@ enum {
 	ETHA_DRVINFO_MAX = (__ETHA_DRVINFO_CNT - 1)
 };
 
+enum {
+	ETHA_TSINFO_UNSPEC,
+	ETHA_TSINFO_TIMESTAMPING,		/* bitset */
+	ETHA_TSINFO_PHC_INDEX,			/* u32 */
+	ETHA_TSINFO_TX_TYPES,			/* bitset */
+	ETHA_TSINFO_RX_FILTERS,			/* bitset */
+
+	__ETHA_TSINFO_CNT,
+	ETHA_TSINFO_MAX = (__ETHA_TSINFO_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
include/uapi/linux/net_tstamp.h +13 -0
index e5b39721c6e4..e5b0472c4416 100644
@@ -30,6 +30,9 @@ enum {
 	SOF_TIMESTAMPING_OPT_STATS = (1<<12),
 	SOF_TIMESTAMPING_OPT_PKTINFO = (1<<13),
 	SOF_TIMESTAMPING_OPT_TX_SWHW = (1<<14),
+	/* when adding a flag, please update so_timestamping_labels array
+	 * in net/ethtool/info.c
+	 */
 
 	SOF_TIMESTAMPING_LAST = SOF_TIMESTAMPING_OPT_TX_SWHW,
 	SOF_TIMESTAMPING_MASK = (SOF_TIMESTAMPING_LAST - 1) |
@@ -90,6 +93,11 @@ enum hwtstamp_tx_types {
 	 * queue.
 	 */
 	HWTSTAMP_TX_ONESTEP_SYNC,
+	/* when adding a value, please update tstamp_tx_type_labels array
+	 * in net/ethtool/info.c
+	 */
+
+	HWTSTAMP_TX_LAST = HWTSTAMP_TX_ONESTEP_SYNC
 };
 
 /* possible values for hwtstamp_config->rx_filter */
@@ -132,6 +140,11 @@ enum hwtstamp_rx_filters {
 
 	/* NTP, UDP, all versions and packet modes */
 	HWTSTAMP_FILTER_NTP_ALL,
+	/* when adding a value, please update tstamp_rx_filter_labels array
+	 * in net/ethtool/info.c
+	 */
+
+	HWTSTAMP_FILTER_LAST = HWTSTAMP_FILTER_NTP_ALL
 };
 
 /* SCM_TIMESTAMPING_PKTINFO control message */
net/ethtool/common.c +24 -0
index 8da992129321..7e846f4151f9 100644
@@ -1,6 +1,8 @@
 // SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
 
 #include <linux/rtnetlink.h>
+#include <linux/phy.h>
+#include <linux/net_tstamp.h>
 #include <net/devlink.h>
 #include "common.h"
 
@@ -133,3 +135,25 @@ int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 
 	return 0;
 }
+
+int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info)
+{
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	struct phy_device *phydev = dev->phydev;
+	int err = 0;
+
+	memset(info, 0, sizeof(*info));
+	info->cmd = ETHTOOL_GET_TS_INFO;
+
+	if (phydev && phydev->drv && phydev->drv->ts_info) {
+		err = phydev->drv->ts_info(phydev, info);
+	} else if (ops->get_ts_info) {
+		err = ops->get_ts_info(dev, info);
+	} else {
+		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
+					SOF_TIMESTAMPING_SOFTWARE;
+		info->phc_index = -1;
+	}
+
+	return err;
+}
net/ethtool/common.h +2 -0
index e87e58b3a274..02cbee79da35 100644
@@ -16,4 +16,6 @@ extern const char
 phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
 
 int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info);
+int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info);
+
 #endif /* _ETHTOOL_COMMON_H */
net/ethtool/info.c +138 -0

189 lines not on this page · raw message

net/ethtool/ioctl.c +2 -18
index 844a4f4b1552..18721b5aa353 100644
@@ -2027,28 +2027,12 @@ static int ethtool_get_dump_data(struct net_device *dev,
 
 static int ethtool_get_ts_info(struct net_device *dev, void __user *useraddr)
 {
-	int err = 0;
+	int err;
 	struct ethtool_ts_info info;
-	const struct ethtool_ops *ops = dev->ethtool_ops;
-	struct phy_device *phydev = dev->phydev;
-
-	memset(&info, 0, sizeof(info));
-	info.cmd = ETHTOOL_GET_TS_INFO;
-
-	if (phydev && phydev->drv && phydev->drv->ts_info) {
-		err = phydev->drv->ts_info(phydev, &info);
-	} else if (ops->get_ts_info) {
-		err = ops->get_ts_info(dev, &info);
-	} else {
-		info.so_timestamping =
-			SOF_TIMESTAMPING_RX_SOFTWARE |
-			SOF_TIMESTAMPING_SOFTWARE;
-		info.phc_index = -1;
-	}
 
+	err = __ethtool_get_ts_info(dev, &info);
 	if (err)
 		return err;
-
 	if (copy_to_user(useraddr, &info, sizeof(info)))
 		err = -EFAULT;
 
net/ethtool/netlink.h +9 -0
index 48980ae9e963..4ad9a3eee3e5 100644
@@ -7,14 +7,23 @@
 #include <linux/netdevice.h>
 #include <net/genetlink.h>
 #include <net/sock.h>
+#include <linux/net_tstamp.h>
 
 #define ETHNL_SET_ERRMSG(info, msg) \
 	do { if (info) GENL_SET_ERR_MSG(info, msg); } while (0)
 
+#define __SOF_TIMESTAMPING_COUNT (const_ilog2(SOF_TIMESTAMPING_LAST) + 1)
+#define __HWTSTAMP_TX_COUNT (HWTSTAMP_TX_LAST + 1)
+#define __HWTSTAMP_FILTER_COUNT (HWTSTAMP_FILTER_LAST + 1)
+
 extern u32 ethnl_bcast_seq;
 
 extern struct genl_family ethtool_genl_family;
 
+extern const char *const so_timestamping_labels[];
+extern const char *const tstamp_tx_type_labels[];
+extern const char *const tstamp_rx_filter_labels[];
+
 struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest);
 int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype);
 
net/ethtool/strset.c +18 -0
index 808ab6b4b387..e1b831097ca7 100644
@@ -67,6 +67,24 @@ static const struct strset_info info_template[] = {
 		.count		= ARRAY_SIZE(phy_tunable_strings),
 		.data		= { .legacy = phy_tunable_strings },
 	},
+	[ETH_SS_TSTAMP_SOF] = {
+		.type		= ETH_SS_TYPE_SIMPLE,
+		.per_dev	= false,
+		.count		= __SOF_TIMESTAMPING_COUNT,
+		.data		= { .simple = so_timestamping_labels },
+	},
+	[ETH_SS_TSTAMP_TX_TYPE] = {
+		.type		= ETH_SS_TYPE_SIMPLE,
+		.per_dev	= false,
+		.count		= __HWTSTAMP_TX_COUNT,
+		.data		= { .simple = tstamp_tx_type_labels },
+	},
+	[ETH_SS_TSTAMP_RX_FILTER] = {
+		.type		= ETH_SS_TYPE_SIMPLE,
+		.per_dev	= false,
+		.count		= __HWTSTAMP_FILTER_COUNT,
+		.data		= { .simple = tstamp_rx_filter_labels },
+	},
 };
 
 struct strset_data {
Signature · 1 line
2.21.0
#87 lore raw

Re: [PATCH net-next v5 14/22] ethtool: provide timestamping information in GET_INFO request

Quoted 9 lines · Michal Kubecek

Add timestamping information as provided by ETHTOOL_GET_TS_INFO ioctl command in GET_INFO reply if ETH_INFO_IM_TSINFO flag is set in the request.

Add constants for counts of HWTSTAMP_TX_* and HWTSTAM_FILTER_* constants and provide symbolic names for timestamping related values so that they can be retrieved in GET_STRSET and GET_INFO requests.

---

[snip]

Quoted 8 lines
net/ethtool/common.c +8 -0
+int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info)
+{
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	struct phy_device *phydev = dev->phydev;
+	int err = 0;
+
+	memset(info, 0, sizeof(*info));
+	info->cmd = ETHTOOL_GET_TS_INFO;

net/ethtool/common.c int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) · line 156

Quoted 10 lines
net/ethtool/common.c +10 -0
+
+	if (phydev && phydev->drv && phydev->drv->ts_info) {
+		err = phydev->drv->ts_info(phydev, info);
+	} else if (ops->get_ts_info) {
+		err = ops->get_ts_info(dev, info);
+	} else {
+		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
+					SOF_TIMESTAMPING_SOFTWARE;
+		info->phc_index = -1;
+	}

We don't have to solve this just yet, but this is the type of logic that we want to be able to change in the future, there are a number of things in the current ethtool code where, say, we always prefer PHY time stamping over MAC time stamping when there should be a choice if both are supported.

I don't know if you maintain a TODO for after when this patch series get merged, but we may want to start one :)

Signature · 1 line
Florian
#88 lore raw

Re: [PATCH net-next v5 14/22] ethtool: provide timestamping information in GET_INFO request

Quoted 31 lines · Florian Fainelli
Quoted 9 lines · Michal Kubecek

Add timestamping information as provided by ETHTOOL_GET_TS_INFO ioctl command in GET_INFO reply if ETH_INFO_IM_TSINFO flag is set in the request.

Add constants for counts of HWTSTAMP_TX_* and HWTSTAM_FILTER_* constants and provide symbolic names for timestamping related values so that they can be retrieved in GET_STRSET and GET_INFO requests.

---

[snip]

Quoted 16 lines
net/ethtool/common.c +16 -0
+int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info)
+{
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	struct phy_device *phydev = dev->phydev;
+	int err = 0;
+
+	memset(info, 0, sizeof(*info));
+	info->cmd = ETHTOOL_GET_TS_INFO;
+
+	if (phydev && phydev->drv && phydev->drv->ts_info) {
+		err = phydev->drv->ts_info(phydev, info);
+	} else if (ops->get_ts_info) {
+		err = ops->get_ts_info(dev, info);
+	} else {
+		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
+					SOF_TIMESTAMPING_SOFTWARE;
Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		info->phc_index = -1;
+	}

We don't have to solve this just yet, but this is the type of logic that we want to be able to change in the future, there are a number of things in the current ethtool code where, say, we always prefer PHY time stamping over MAC time stamping when there should be a choice if both are supported.

For the record, this is just a code block carved out of existing ethtool_ts_info() where the only difference is that info is a pointer rather than a structure itself.

Quoted 2 lines

I don't know if you maintain a TODO for after when this patch series get merged, but we may want to start one :)

I have some informal TODO locally; some ideas are in the cover letter, I also mentioned some in my talk last week. Perhaps it is time to make it a publicly visible document and start collecting ideas.

Michal

#89 lore raw

PATCH v5 15/22 ethtool: provide link mode names as a string set

In reply to Michal Kubecek

Add table of ethernet link mode names and make it available as a string set to userspace GET_STRSET requests.

---
 include/linux/ethtool.h      |  4 ++
 include/uapi/linux/ethtool.h |  2 +
 net/ethtool/netlink.c        | 81 ++++++++++++++++++++++++++++++++++++
 net/ethtool/netlink.h        |  1 +
 net/ethtool/strset.c         |  6 +++
 5 files changed, 94 insertions(+)
include/linux/ethtool.h +4 -0
index e6ebc9761822..ac3612599450 100644
@@ -102,6 +102,10 @@ static inline u32 ethtool_rxfh_indir_default(u32 index, u32 n_rx_rings)
 #define __ETHTOOL_DECLARE_LINK_MODE_MASK(name)		\
 	DECLARE_BITMAP(name, __ETHTOOL_LINK_MODE_MASK_NBITS)
 
+/* compose link mode index from speed, type and duplex */
+#define ETHTOOL_LINK_MODE(speed, type, duplex) \
+	ETHTOOL_LINK_MODE_ ## speed ## base ## type ## _ ## duplex ## _BIT
+
 /* drivers must ignore base.cmd and base.link_mode_masks_nwords
  * fields, but they are allowed to overwrite them (will be ignored).
  */
include/uapi/linux/ethtool.h +2 -0
index 882a542eaaa9..ef1288eb737c 100644
@@ -566,6 +566,7 @@ struct ethtool_pauseparam {
  * @ETH_SS_TSTAMP_SOF: timestamping flag names
  * @ETH_SS_TSTAMP_TX_TYPE: timestamping Tx type names
  * @ETH_SS_TSTAMP_RX_FILTER: timestamping Rx filter names
+ * @ETH_SS_LINK_MODES: link mode names
  */
 enum ethtool_stringset {
 	ETH_SS_TEST		= 0,
@@ -580,6 +581,7 @@ enum ethtool_stringset {
 	ETH_SS_TSTAMP_SOF,
 	ETH_SS_TSTAMP_TX_TYPE,
 	ETH_SS_TSTAMP_RX_FILTER,
+	ETH_SS_LINK_MODES,
 
 	ETH_SS_COUNT
 };
net/ethtool/netlink.c +81 -0

93 lines not on this page · raw message

net/ethtool/netlink.h +1 -0
index 4ad9a3eee3e5..ce60d7e18651 100644
@@ -23,6 +23,7 @@ extern struct genl_family ethtool_genl_family;
 extern const char *const so_timestamping_labels[];
 extern const char *const tstamp_tx_type_labels[];
 extern const char *const tstamp_rx_filter_labels[];
+extern const char *const link_mode_names[];
 
 struct net_device *ethnl_dev_get(struct genl_info *info, struct nlattr *nest);
 int ethnl_fill_dev(struct sk_buff *msg, struct net_device *dev, u16 attrtype);
net/ethtool/strset.c +6 -0
index e1b831097ca7..577d1309ba6f 100644
@@ -85,6 +85,12 @@ static const struct strset_info info_template[] = {
 		.count		= __HWTSTAMP_FILTER_COUNT,
 		.data		= { .simple = tstamp_rx_filter_labels },
 	},
+	[ETH_SS_LINK_MODES] = {
+		.type		= ETH_SS_TYPE_SIMPLE,
+		.per_dev	= false,
+		.count		= __ETHTOOL_LINK_MODE_MASK_NBITS,
+		.data		= { .simple = link_mode_names },
+	},
 };
 
 struct strset_data {
Signature · 1 line
2.21.0
#91 lore raw

PATCH v5 16/22 ethtool: provide link settings and link modes in GET_SETTINGS request

In reply to Michal Kubecek

Implement GET_SETTINGS netlink request to get link settings and link mode information provided by ETHTOOL_GLINKSETTINGS ioctl command.

The information in SET_SETTINGS message sent as reply is divided into two
parts: autonegotiation, speed, duplex and supported, advertised and peer advertised link modes when ETH_SETTINGS_IM_LINKMODES flag is set in the request and other settings when ETH_SETTINGS_IM_LINKINFO is set.

Send notification in the same format as the reply message when relevant fields are modified using the ioctl interface (ETHTOOL_SLINKSETTINGS or ETHTOOL_SSET command).

---
 Documentation/networking/ethtool-netlink.txt |  50 +++-
 include/linux/ethtool_netlink.h              |   3 +
 include/uapi/linux/ethtool_netlink.h         |  46 +++
 net/ethtool/Makefile                         |   2 +-
 net/ethtool/common.c                         |  48 ++++
 net/ethtool/common.h                         |   4 +
 net/ethtool/ioctl.c                          |  64 +----
 net/ethtool/netlink.c                        |   9 +
 net/ethtool/netlink.h                        |   1 +
 net/ethtool/settings.c                       | 277 +++++++++++++++++++
 10 files changed, 451 insertions(+), 53 deletions(-)
 create mode 100644 net/ethtool/settings.c
Documentation/networking/ethtool-netlink.txt +48 -2
index 74c62f11810c..d723b3537c46 100644
@@ -131,6 +131,8 @@ List of message types
     ETHNL_CMD_SET_STRSET		response only
     ETHNL_CMD_GET_INFO
     ETHNL_CMD_SET_INFO			response only
+    ETHNL_CMD_GET_SETTINGS
+    ETHNL_CMD_SET_SETTINGS		response only (for now)
 
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
@@ -258,6 +260,50 @@ if no PHC is associated, the attribute is not present.
 GET_INFO requests allow dumps.
 
 
+GET_SETTINGS
+------------
+
+GET_SETTINGS request retrieves information provided by ETHTOOL_GLINKSETTINGS,
+ETHTOOL_GWOL, ETHTOOL_GMSGLVL and ETHTOOL_GLINK ioctl requests. The request
+doesn't use any attributes.
+
+Request attributes:
+
+    ETHA_SETTINGS_DEV		(nested)	device identification
+    ETHA_SETTINGS_INFOMASK	(u32)		info mask
+    ETHA_SETTINGS_COMPACT	(flag)		request compact bitsets
+
+Info mask bits meaning:
+
+    ETH_SETTINGS_IM_LINKINFO		link settings
+    ETH_SETTINGS_IM_LINKMODES		link modes and related
+
+Response contents:
+
+    ETHA_SETTINGS_DEV		(nested)	device identification
+    ETHA_SETTINGS_LINK_INFO	(nested)	link settings
+        ETHA_LINKINFO_PORT		(u8)		physical port
+        ETHA_LINKINFO_PHYADDR		(u8)		MDIO address of phy
+        ETHA_LINKINFO_TP_MDIX		(u8)		MDI(-X) status
+        ETHA_LINKINFO_TP_MDIX_CTRL	(u8)		MDI(-X) control
+        ETHA_LINKINFO_TRANSCEIVER	(u8)		transceiver
+    ETHA_SETTINGS_LINK_MODES	(nested)	link modes
+        ETHA_LINKMODES_AUTONEG		(u8)		autoneotiation status
+        ETHA_LINKMODES_OURS		(bitset)	advertised link modes
+        ETHA_LINKMODES_PEER		(bitset)	partner link modes
+        ETHA_LINKMODES_SPEED		(u32)		link speed (Mb/s)
+        ETHA_LINKMODES_DUPLEX		(u8)		duplex mode
+
+Most of the attributes and their values have the same meaning as matching
+members of the corresponding ioctl structures. For ETHA_LINKMODES_OURS,
+value represents advertised modes and mask represents supported modes.
+ETHA_LINKMODES_PEER in the reply is a bit list.
+
+GET_SETTINGS requests allow dumps and messages in the same format as response
+to them are broadcasted as notifications on change of these settings using
+netlink or ioctl ethtool interface.
+
+
 Request translation
 -------------------
 
@@ -267,7 +313,7 @@ have their netlink replacement yet.
 
 ioctl command			netlink command
 ---------------------------------------------------------------------
-ETHTOOL_GSET			n/a
+ETHTOOL_GSET			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SSET			n/a
 ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
@@ -341,7 +387,7 @@ ETHTOOL_GTUNABLE		n/a
 ETHTOOL_STUNABLE		n/a
 ETHTOOL_GPHYSTATS		n/a
 ETHTOOL_PERQUEUE		n/a
-ETHTOOL_GLINKSETTINGS		n/a
+ETHTOOL_GLINKSETTINGS		ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SLINKSETTINGS		n/a
 ETHTOOL_PHY_GTUNABLE		n/a
 ETHTOOL_PHY_STUNABLE		n/a
include/linux/ethtool_netlink.h +3 -0
index 2a15e64a16f3..e770e6e9acca 100644
@@ -7,6 +7,9 @@
 #include <linux/ethtool.h>
 #include <linux/netdevice.h>
 
+#define __ETHTOOL_LINK_MODE_MASK_NWORDS \
+	DIV_ROUND_UP(__ETHTOOL_LINK_MODE_MASK_NBITS, 32)
+
 enum ethtool_multicast_groups {
 	ETHNL_MCGRP_MONITOR,
 };
include/uapi/linux/ethtool_netlink.h +46 -0

58 lines not on this page · raw message

net/ethtool/Makefile +1 -1
index 96d41dc45d4f..6a7a182e1568 100644
@@ -4,4 +4,4 @@ obj-y				+= ioctl.o common.o
 
 obj-$(CONFIG_ETHTOOL_NETLINK)	+= ethtool_nl.o
 
-ethtool_nl-y	:= netlink.o bitset.o strset.o info.o
+ethtool_nl-y	:= netlink.o bitset.o strset.o info.o settings.o
net/ethtool/common.c +48 -0

51 lines not on this page · raw message

net/ethtool/common.h +4 -0
index 02cbee79da35..7a3e0b10e69a 100644
@@ -18,4 +18,8 @@ phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
 int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info);
 int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info);
 
+bool convert_legacy_settings_to_link_ksettings(
+	struct ethtool_link_ksettings *link_ksettings,
+	const struct ethtool_cmd *legacy_settings);
+
 #endif /* _ETHTOOL_COMMON_H */
net/ethtool/ioctl.c +14 -50

94 lines not on this page · raw message

net/ethtool/netlink.c +9 -0
index f0385c30169b..9b4ad09799c5 100644
@@ -234,6 +234,7 @@ struct sk_buff *ethnl_reply_init(size_t payload, struct net_device *dev, u8 cmd,
 const struct get_request_ops *get_requests[__ETHNL_CMD_CNT] = {
 	[ETHNL_CMD_GET_STRSET]		= &strset_request_ops,
 	[ETHNL_CMD_GET_INFO]		= &info_request_ops,
+	[ETHNL_CMD_GET_SETTINGS]	= &settings_request_ops,
 };
 
 /**
@@ -558,6 +559,7 @@ typedef void (*ethnl_notify_handler_t)(struct net_device *dev,
 				       const void *data);
 
 ethnl_notify_handler_t ethnl_notify_handlers[] = {
+	[ETHNL_CMD_SET_SETTINGS]	= ethnl_std_notify,
 };
 
 void ethtool_notify(struct net_device *dev, struct netlink_ext_ack *extack,
@@ -654,6 +656,13 @@ static const struct genl_ops ethtool_genl_ops[] = {
 		.dumpit	= ethnl_get_dumpit,
 		.done	= ethnl_get_done,
 	},
+	{
+		.cmd	= ETHNL_CMD_GET_SETTINGS,
+		.doit	= ethnl_get_doit,
+		.start	= ethnl_get_start,
+		.dumpit	= ethnl_get_dumpit,
+		.done	= ethnl_get_done,
+	},
 };
 
 static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
net/ethtool/netlink.h +1 -0
index ce60d7e18651..fd7a362d79fa 100644
@@ -289,5 +289,6 @@ struct get_request_ops {
 
 extern const struct get_request_ops strset_request_ops;
 extern const struct get_request_ops info_request_ops;
+extern const struct get_request_ops settings_request_ops;
 
 #endif /* _NET_ETHTOOL_NETLINK_H */
net/ethtool/settings.c new file +277 -0

277 lines not on this page · raw message

Signature · 1 line
2.21.0
#92 lore raw

Re: [PATCH net-next v5 16/22] ethtool: provide link settings and link modes in GET_SETTINGS request

Quoted 14 lines · Michal Kubecek

Implement GET_SETTINGS netlink request to get link settings and link mode information provided by ETHTOOL_GLINKSETTINGS ioctl command.

The information in SET_SETTINGS message sent as reply is divided into two
parts: autonegotiation, speed, duplex and supported, advertised and peer advertised link modes when ETH_SETTINGS_IM_LINKMODES flag is set in the request and other settings when ETH_SETTINGS_IM_LINKINFO is set.

Send notification in the same format as the reply message when relevant fields are modified using the ioctl interface (ETHTOOL_SLINKSETTINGS or ETHTOOL_SSET command).

---

[snip]

Documentation/networking/ethtool-netlink.txt if no PHC is associated, the attribute is not present. · line 291

Quoted 9 lines
Documentation/networking/ethtool-netlink.txt +9 -0
+    ETHA_SETTINGS_DEV		(nested)	device identification
+    ETHA_SETTINGS_LINK_INFO	(nested)	link settings
+        ETHA_LINKINFO_PORT		(u8)		physical port
+        ETHA_LINKINFO_PHYADDR		(u8)		MDIO address of phy
+        ETHA_LINKINFO_TP_MDIX		(u8)		MDI(-X) status
+        ETHA_LINKINFO_TP_MDIX_CTRL	(u8)		MDI(-X) control
+        ETHA_LINKINFO_TRANSCEIVER	(u8)		transceiver
+    ETHA_SETTINGS_LINK_MODES	(nested)	link modes
+        ETHA_LINKMODES_AUTONEG		(u8)		autoneotiation status

Typo: auto-negotiation status.

Signature · 1 line
Florian
#93 lore raw

Re: [PATCH net-next v5 16/22] ethtool: provide link settings and link modes in GET_SETTINGS request

Quoted 30 lines · Florian Fainelli
Quoted 14 lines · Michal Kubecek

Implement GET_SETTINGS netlink request to get link settings and link mode information provided by ETHTOOL_GLINKSETTINGS ioctl command.

The information in SET_SETTINGS message sent as reply is divided into two
parts: autonegotiation, speed, duplex and supported, advertised and peer advertised link modes when ETH_SETTINGS_IM_LINKMODES flag is set in the request and other settings when ETH_SETTINGS_IM_LINKINFO is set.

Send notification in the same format as the reply message when relevant fields are modified using the ioctl interface (ETHTOOL_SLINKSETTINGS or ETHTOOL_SSET command).

---

[snip]

Documentation/networking/ethtool-netlink.txt if no PHC is associated, the attribute is not present. · line 291

Quoted 9 lines
Documentation/networking/ethtool-netlink.txt +9 -0
+    ETHA_SETTINGS_DEV		(nested)	device identification
+    ETHA_SETTINGS_LINK_INFO	(nested)	link settings
+        ETHA_LINKINFO_PORT		(u8)		physical port
+        ETHA_LINKINFO_PHYADDR		(u8)		MDIO address of phy
+        ETHA_LINKINFO_TP_MDIX		(u8)		MDI(-X) status
+        ETHA_LINKINFO_TP_MDIX_CTRL	(u8)		MDI(-X) control
+        ETHA_LINKINFO_TRANSCEIVER	(u8)		transceiver
+    ETHA_SETTINGS_LINK_MODES	(nested)	link modes
+        ETHA_LINKMODES_AUTONEG		(u8)		autoneotiation status
Quoted 1 line

Typo: auto-negotiation status.

Will fix, thanks.

Michal

#94 lore raw

PATCH v5 17/22 ethtool: set link settings and link modes with SET_SETTINGS request

In reply to Michal Kubecek

Implement SET_SETTINGS netlink request allowing to set link settings and advertised link modes as an alternative to ETHTOOL_SLINKSETTINGS ioctl command.

ETHA_SETTINGS_LINK_MODES attribute is used to set (or modify) advertised link modes and related settings (autonegotiation, speed and duplex) and ETHA_SETTINGS_LINK_INFO to set other link settings.

---
 Documentation/networking/ethtool-netlink.txt |  34 ++-
 net/ethtool/netlink.c                        |   5 +
 net/ethtool/netlink.h                        |   2 +
 net/ethtool/settings.c                       | 305 +++++++++++++++++++
 4 files changed, 343 insertions(+), 3 deletions(-)
Documentation/networking/ethtool-netlink.txt +31 -3
index d723b3537c46..e79ae9fe01be 100644
@@ -132,7 +132,7 @@ List of message types
     ETHNL_CMD_GET_INFO
     ETHNL_CMD_SET_INFO			response only
     ETHNL_CMD_GET_SETTINGS
-    ETHNL_CMD_SET_SETTINGS		response only (for now)
+    ETHNL_CMD_SET_SETTINGS
 
 All constants use ETHNL_CMD_ prefix, usually followed by "GET", "SET" or "ACT"
 to indicate the type.
@@ -304,6 +304,34 @@ to them are broadcasted as notifications on change of these settings using
 netlink or ioctl ethtool interface.
 
 
+SET_SETTINGS
+------------
+
+SET_SETTINGS request allows setting some of the data reported by GET_SETTINGS.
+Request flags, info_mask and index are ignored. These attributes are allowed
+to be passed with SET_SETTINGS request:
+
+    ETHA_SETTINGS_DEV		(nested)	device identification
+    ETHA_SETTINGS_LINK_INFO	(nested)	link settings
+        ETHA_LINKINFO_PORT		(u8)		physical port
+        ETHA_LINKINFO_PHYADDR		(u8)		MDIO address of phy
+        ETHA_LINKINFO_TP_MDIX_CTRL	(u8)		MDI(-X) control
+    ETHA_SETTINGS_LINK_MODES	(nested)	link modes
+        ETHA_LINKMODES_AUTONEG		(u8)		autonegotiation
+        ETHA_LINKMODES_OURS		(bitset)	advertised link modes
+        ETHA_LINKMODES_SPEED		(u32)		link speed (Mb/s)
+        ETHA_LINKMODES_DUPLEX		(u8)		duplex mode
+
+ETHA_LINKMODES_OURS bit set allows setting advertised link modes. If
+autonegotiation is on (either set now or kept from before), advertised modes
+are not changed (no ETHA_LINKMODES_OURS attribute) and at least one of speed
+and duplex is specified, kernel adjusts advertised modes to all supported
+modes matching speed, duplex or both (whatever is specified). This
+autoselection is done on ethtool side with ioctl interface, netlink interface
+is supposed to allow requesting changes without knowing what exactly kernel
+supports.
+
+
 Request translation
 -------------------
 
@@ -314,7 +342,7 @@ have their netlink replacement yet.
 ioctl command			netlink command
 ---------------------------------------------------------------------
 ETHTOOL_GSET			ETHNL_CMD_GET_SETTINGS
-ETHTOOL_SSET			n/a
+ETHTOOL_SSET			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
 ETHTOOL_GWOL			n/a
@@ -388,7 +416,7 @@ ETHTOOL_STUNABLE		n/a
 ETHTOOL_GPHYSTATS		n/a
 ETHTOOL_PERQUEUE		n/a
 ETHTOOL_GLINKSETTINGS		ETHNL_CMD_GET_SETTINGS
-ETHTOOL_SLINKSETTINGS		n/a
+ETHTOOL_SLINKSETTINGS		ETHNL_CMD_SET_SETTINGS
 ETHTOOL_PHY_GTUNABLE		n/a
 ETHTOOL_PHY_STUNABLE		n/a
 ETHTOOL_GFECPARAM		n/a
net/ethtool/netlink.c +5 -0
index 9b4ad09799c5..50ed82d5f18b 100644
@@ -663,6 +663,11 @@ static const struct genl_ops ethtool_genl_ops[] = {
 		.dumpit	= ethnl_get_dumpit,
 		.done	= ethnl_get_done,
 	},
+	{
+		.cmd	= ETHNL_CMD_SET_SETTINGS,
+		.flags	= GENL_UNS_ADMIN_PERM,
+		.doit	= ethnl_set_settings,
+	},
 };
 
 static const struct genl_multicast_group ethtool_nl_mcgrps[] = {
net/ethtool/netlink.h +2 -0
index fd7a362d79fa..853cd5b23415 100644
@@ -291,4 +291,6 @@ extern const struct get_request_ops strset_request_ops;
 extern const struct get_request_ops info_request_ops;
 extern const struct get_request_ops settings_request_ops;
 
+int ethnl_set_settings(struct sk_buff *skb, struct genl_info *info);
+
 #endif /* _NET_ETHTOOL_NETLINK_H */
net/ethtool/settings.c +305 -0

314 lines not on this page · raw message

Signature · 1 line
2.21.0
#95 lore raw

PATCH v5 18/22 ethtool: provide link state in GET_SETTINGS request

In reply to Michal Kubecek

Add information about device link state (as provided by ETHTOOL_GLINK ioctl command) into the GET_SETTINGS reply when ETH_SETTINGS_IM_LINKSTATE flag is set in the request.

Note: we cannot use NLA_FLAG for link state as we need three states: off, on and unknown. The attribute is also encapsulated in a nest to allow future extensions.

---
 Documentation/networking/ethtool-netlink.txt |  5 ++-
 include/uapi/linux/ethtool_netlink.h         | 13 ++++++-
 net/ethtool/common.c                         |  8 +++++
 net/ethtool/common.h                         |  1 +
 net/ethtool/ioctl.c                          |  8 ++---
 net/ethtool/settings.c                       | 37 ++++++++++++++++++++
 6 files changed, 66 insertions(+), 6 deletions(-)
Documentation/networking/ethtool-netlink.txt +4 -1
index e79ae9fe01be..ebd1a0404828 100644
@@ -277,6 +277,7 @@ Info mask bits meaning:
 
     ETH_SETTINGS_IM_LINKINFO		link settings
     ETH_SETTINGS_IM_LINKMODES		link modes and related
+    ETH_SETTINGS_IM_LINKSTATE		link state
 
 Response contents:
 
@@ -293,6 +294,8 @@ Response contents:
         ETHA_LINKMODES_PEER		(bitset)	partner link modes
         ETHA_LINKMODES_SPEED		(u32)		link speed (Mb/s)
         ETHA_LINKMODES_DUPLEX		(u8)		duplex mode
+    ETHA_SETTINGS_LINK_STATE	(nested)	link state
+        ETHA_LINKSTATE_LINK		(u8)		link on/off/unknown
 
 Most of the attributes and their values have the same meaning as matching
 members of the corresponding ioctl structures. For ETHA_LINKMODES_OURS,
@@ -350,7 +353,7 @@ ETHTOOL_SWOL			n/a
 ETHTOOL_GMSGLVL			n/a
 ETHTOOL_SMSGLVL			n/a
 ETHTOOL_NWAY_RST		n/a
-ETHTOOL_GLINK			n/a
+ETHTOOL_GLINK			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_GEEPROM			n/a
 ETHTOOL_SEEPROM			n/a
 ETHTOOL_GCOALESCE		n/a
include/uapi/linux/ethtool_netlink.h +12 -1
index 895ba9cd8924..45c27c7291d0 100644
@@ -196,6 +196,7 @@ enum {
 	ETHA_SETTINGS_COMPACT,			/* flag */
 	ETHA_SETTINGS_LINK_INFO,		/* nest - ETHA_LINKINFO_* */
 	ETHA_SETTINGS_LINK_MODES,		/* nest - ETHA_LINKMODES_* */
+	ETHA_SETTINGS_LINK_STATE,		/* nest - ETHA_LINKSTATE_* */
 
 	__ETHA_SETTINGS_CNT,
 	ETHA_SETTINGS_MAX = (__ETHA_SETTINGS_CNT - 1)
@@ -203,9 +204,11 @@ enum {
 
 #define ETH_SETTINGS_IM_LINKINFO		(1U << 0)
 #define ETH_SETTINGS_IM_LINKMODES		(1U << 1)
+#define ETH_SETTINGS_IM_LINKSTATE		(1U << 2)
 
 #define ETH_SETTINGS_IM_ALL (ETH_SETTINGS_IM_LINKINFO | \
-			     ETH_SETTINGS_IM_LINKMODES)
+			     ETH_SETTINGS_IM_LINKMODES | \
+			     ETH_SETTINGS_IM_LINKSTATE)
 
 enum {
 	ETHA_LINKINFO_UNSPEC,
@@ -231,6 +234,14 @@ enum {
 	ETHA_LINKMODES_MAX = (__ETHA_LINKMODES_CNT - 1)
 };
 
+enum {
+	ETHA_LINKSTATE_UNSPEC,
+	ETHA_LINKSTATE_LINK,			/* u8 */
+
+	__ETHA_LINKSTATE_CNT,
+	ETHA_LINKSTATE_MAX = (__ETHA_LINKSTATE_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/common.c +8 -0
index a91a4f00d275..dc907d8b6e43 100644
@@ -205,3 +205,11 @@ convert_legacy_settings_to_link_ksettings(
 		= legacy_settings->eth_tp_mdix_ctrl;
 	return retval;
 }
+
+int __ethtool_get_link(struct net_device *dev)
+{
+	if (!dev->ethtool_ops->get_link)
+		return -EOPNOTSUPP;
+
+	return netif_running(dev) && dev->ethtool_ops->get_link(dev);
+}
net/ethtool/common.h +1 -0
index 7a3e0b10e69a..a5ddd7f5cfce 100644
@@ -17,6 +17,7 @@ phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
 
 int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info);
 int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info);
+int __ethtool_get_link(struct net_device *dev);
 
 bool convert_legacy_settings_to_link_ksettings(
 	struct ethtool_link_ksettings *link_ksettings,
net/ethtool/ioctl.c +4 -4
index be3e34ac8cc7..58669dafeaf9 100644
@@ -1318,12 +1318,12 @@ static int ethtool_nway_reset(struct net_device *dev)
 static int ethtool_get_link(struct net_device *dev, char __user *useraddr)
 {
 	struct ethtool_value edata = { .cmd = ETHTOOL_GLINK };
+	int link = __ethtool_get_link(dev);
 
-	if (!dev->ethtool_ops->get_link)
-		return -EOPNOTSUPP;
-
-	edata.data = netif_running(dev) && dev->ethtool_ops->get_link(dev);
+	if (link < 0)
+		return link;
 
+	edata.data = link;
 	if (copy_to_user(useraddr, &edata, sizeof(edata)))
 		return -EFAULT;
 	return 0;
net/ethtool/settings.c +37 -0
index 2e0f425029ef..7a2729f01f79 100644
@@ -11,6 +11,7 @@ struct settings_data {
 	struct common_reply_data	repdata_base;
 	struct ethtool_link_ksettings	ksettings;
 	struct ethtool_link_settings	*lsettings;
+	int				link;
 	bool				lpm_empty;
 };
 
@@ -112,6 +113,7 @@ static const struct nla_policy get_settings_policy[ETHA_SETTINGS_MAX + 1] = {
 	[ETHA_SETTINGS_COMPACT]		= { .type = NLA_FLAG },
 	[ETHA_SETTINGS_LINK_INFO]	= { .type = NLA_REJECT },
 	[ETHA_SETTINGS_LINK_MODES]	= { .type = NLA_REJECT },
+	[ETHA_SETTINGS_LINK_STATE]	= { .type = NLA_REJECT },
 };
 
 static int parse_settings(struct common_req_info *req_info,
@@ -168,6 +170,7 @@ static int prepare_settings(struct common_req_info *req_info,
 
 	data->lsettings = &data->ksettings.base;
 	data->lpm_empty = true;
+	data->link = -EOPNOTSUPP;
 
 	ret = ethnl_before_ops(dev);
 	if (ret < 0)
@@ -189,6 +192,8 @@ static int prepare_settings(struct common_req_info *req_info,
 		ethnl_bitmap_to_u32(data->ksettings.link_modes.lp_advertising,
 				    __ETHTOOL_LINK_MODE_MASK_NWORDS);
 	}
+	if (req_mask & ETH_SETTINGS_IM_LINKSTATE)
+		data->link = __ethtool_get_link(dev);
 	ethnl_after_ops(dev);
 
 	data->repdata_base.info_mask = req_mask;
@@ -237,6 +242,13 @@ static int link_modes_size(const struct ethtool_link_ksettings *ksettings,
 	return nla_total_size(len);
 }
 
+static int link_state_size(int link)
+{
+	if (link < 0)
+		return nla_total_size(0);
+	return nla_total_size(nla_total_size(sizeof(u8)));
+}
+
 /* To keep things simple, reserve space for some attributes which may not
  * be added to the message (e.g. ETHA_SETTINGS_SOPASS); therefore the length
  * returned may be bigger than the actual length of the message sent
@@ -258,6 +270,8 @@ static int settings_size(const struct common_req_info *req_info)
 			return ret;
 		len += ret;
 	}
+	if (info_mask & ETH_SETTINGS_IM_LINKSTATE)
+		len += link_state_size(data->link);
 
 	return len;
 }
@@ -330,6 +344,23 @@ static int fill_link_modes(struct sk_buff *skb,
 	return -EMSGSIZE;
 }
 
+static int fill_link_state(struct sk_buff *skb, u8 link)
+{
+	struct nlattr *nest;
+
+	nest = ethnl_nest_start(skb, ETHA_SETTINGS_LINK_STATE);
+	if (!nest)
+		return -EMSGSIZE;
+	if (link >=0 && nla_put_u8(skb, ETHA_LINKSTATE_LINK, link))
+		goto err;
+	nla_nest_end(skb, nest);
+	return 0;
+
+err:
+	nla_nest_cancel(skb, nest);
+	return -EMSGSIZE;
+}
+
 static int fill_settings(struct sk_buff *skb,
 			 const struct common_req_info *req_info)
 {
@@ -350,6 +381,11 @@ static int fill_settings(struct sk_buff *skb,
 		if (ret < 0)
 			return ret;
 	}
+	if (info_mask & ETH_SETTINGS_IM_LINKSTATE) {
+		ret = fill_link_state(skb, data->link);
+		if (ret < 0)
+			return ret;
+	}
 
 	return 0;
 }
@@ -394,6 +430,7 @@ static const struct nla_policy set_settings_policy[ETHA_SETTINGS_MAX + 1] = {
 	[ETHA_SETTINGS_COMPACT]		= { .type = NLA_FLAG },
 	[ETHA_SETTINGS_LINK_INFO]	= { .type = NLA_NESTED },
 	[ETHA_SETTINGS_LINK_MODES]	= { .type = NLA_NESTED },
+	[ETHA_SETTINGS_LINK_STATE]	= { .type = NLA_REJECT },
 };
 
 static int ethnl_set_link_ksettings(struct genl_info *info,
Signature · 1 line
2.21.0
#96 lore raw

PATCH v5 19/22 ethtool: provide WoL information in GET_SETTINGS request

In reply to Michal Kubecek

Add information about supported and enabled wake on LAN modes into the GET_SETTINGS reply when ETH_SETTINGS_IM_WOL flag is set in the request.

The GET_SETTINGS request can be still sent by unprivileged users but in such case the SecureOn password (if any) is not included in the reply.

Send notification in the same format as reply SET_SETTINGS message when wake on LAN settings are modified using ioctl interface (ETHTOOL_SWOL command).

---
 Documentation/networking/ethtool-netlink.txt | 13 +++-
 include/uapi/linux/ethtool_netlink.h         | 14 ++++-
 net/ethtool/common.c                         | 10 ++++
 net/ethtool/common.h                         |  1 +
 net/ethtool/ioctl.c                          | 10 ++--
 net/ethtool/settings.c                       | 63 ++++++++++++++++++++
 6 files changed, 105 insertions(+), 6 deletions(-)
Documentation/networking/ethtool-netlink.txt +12 -1
index ebd1a0404828..bc7f28f0182f 100644
@@ -278,6 +278,7 @@ Info mask bits meaning:
     ETH_SETTINGS_IM_LINKINFO		link settings
     ETH_SETTINGS_IM_LINKMODES		link modes and related
     ETH_SETTINGS_IM_LINKSTATE		link state
+    ETH_SETTINGS_IM_WOL			struct ethtool_wolinfo
 
 Response contents:
 
@@ -296,12 +297,22 @@ Response contents:
         ETHA_LINKMODES_DUPLEX		(u8)		duplex mode
     ETHA_SETTINGS_LINK_STATE	(nested)	link state
         ETHA_LINKSTATE_LINK		(u8)		link on/off/unknown
+    ETHA_SETTINGS_WOL		(nested)	wake on LAN settings
+        ETHA_WOL_MODES			(bitfield32)	wake on LAN modes
+        ETHA_WOL_SOPASS			(binary)	SecureOn(tm) password
 
 Most of the attributes and their values have the same meaning as matching
 members of the corresponding ioctl structures. For ETHA_LINKMODES_OURS,
 value represents advertised modes and mask represents supported modes.
 ETHA_LINKMODES_PEER in the reply is a bit list.
 
+For ETHA_WOL_MODES, selector reports wake on LAN modes supported by the
+device and value enabled modes.
+
+GET_SETTINGS request is allowed for unprivileged user but ETHA_WOL_SOPASS
+is only provided by kernel in response to privileged (netns CAP_NET_ADMIN)
+requests.
+
 GET_SETTINGS requests allow dumps and messages in the same format as response
 to them are broadcasted as notifications on change of these settings using
 netlink or ioctl ethtool interface.
@@ -348,7 +359,7 @@ ETHTOOL_GSET			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SSET			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
-ETHTOOL_GWOL			n/a
+ETHTOOL_GWOL			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SWOL			n/a
 ETHTOOL_GMSGLVL			n/a
 ETHTOOL_SMSGLVL			n/a
include/uapi/linux/ethtool_netlink.h +13 -1
index 45c27c7291d0..532ad11ae87c 100644
@@ -197,6 +197,7 @@ enum {
 	ETHA_SETTINGS_LINK_INFO,		/* nest - ETHA_LINKINFO_* */
 	ETHA_SETTINGS_LINK_MODES,		/* nest - ETHA_LINKMODES_* */
 	ETHA_SETTINGS_LINK_STATE,		/* nest - ETHA_LINKSTATE_* */
+	ETHA_SETTINGS_WOL,			/* nest - ETHA_WOL_* */
 
 	__ETHA_SETTINGS_CNT,
 	ETHA_SETTINGS_MAX = (__ETHA_SETTINGS_CNT - 1)
@@ -205,10 +206,12 @@ enum {
 #define ETH_SETTINGS_IM_LINKINFO		(1U << 0)
 #define ETH_SETTINGS_IM_LINKMODES		(1U << 1)
 #define ETH_SETTINGS_IM_LINKSTATE		(1U << 2)
+#define ETH_SETTINGS_IM_WOL			(1U << 3)
 
 #define ETH_SETTINGS_IM_ALL (ETH_SETTINGS_IM_LINKINFO | \
 			     ETH_SETTINGS_IM_LINKMODES | \
-			     ETH_SETTINGS_IM_LINKSTATE)
+			     ETH_SETTINGS_IM_LINKSTATE | \
+			     ETH_SETTINGS_IM_WOL)
 
 enum {
 	ETHA_LINKINFO_UNSPEC,
@@ -242,6 +245,15 @@ enum {
 	ETHA_LINKSTATE_MAX = (__ETHA_LINKSTATE_CNT - 1)
 };
 
+enum {
+	ETHA_WOL_UNSPEC,
+	ETHA_WOL_MODES,				/* bitfield32 */
+	ETHA_WOL_SOPASS,			/* binary */
+
+	__ETHA_WOL_CNT,
+	ETHA_WOL_MAX = (__ETHA_WOL_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/common.c +10 -0
index dc907d8b6e43..4b2f08da910a 100644
@@ -213,3 +213,13 @@ int __ethtool_get_link(struct net_device *dev)
 
 	return netif_running(dev) && dev->ethtool_ops->get_link(dev);
 }
+
+int __ethtool_get_wol(struct net_device *dev, struct ethtool_wolinfo *wol)
+{
+	if (!dev->ethtool_ops->get_wol)
+		return -EOPNOTSUPP;
+
+	dev->ethtool_ops->get_wol(dev, wol);
+
+	return 0;
+}
net/ethtool/common.h +1 -0
index a5ddd7f5cfce..bbe3e51f7308 100644
@@ -18,6 +18,7 @@ phy_tunable_strings[__ETHTOOL_PHY_TUNABLE_COUNT][ETH_GSTRING_LEN];
 int __ethtool_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info);
 int __ethtool_get_ts_info(struct net_device *dev, struct ethtool_ts_info *info);
 int __ethtool_get_link(struct net_device *dev);
+int __ethtool_get_wol(struct net_device *dev, struct ethtool_wolinfo *wol);
 
 bool convert_legacy_settings_to_link_ksettings(
 	struct ethtool_link_ksettings *link_ksettings,
net/ethtool/ioctl.c +6 -4
index 58669dafeaf9..bba02a218eea 100644
@@ -1242,11 +1242,11 @@ static int ethtool_reset(struct net_device *dev, char __user *useraddr)
 static int ethtool_get_wol(struct net_device *dev, char __user *useraddr)
 {
 	struct ethtool_wolinfo wol = { .cmd = ETHTOOL_GWOL };
+	int rc;
 
-	if (!dev->ethtool_ops->get_wol)
-		return -EOPNOTSUPP;
-
-	dev->ethtool_ops->get_wol(dev, &wol);
+	rc = __ethtool_get_wol(dev, &wol);
+	if (rc < 0)
+		return rc;
 
 	if (copy_to_user(useraddr, &wol, sizeof(wol)))
 		return -EFAULT;
@@ -1269,6 +1269,8 @@ static int ethtool_set_wol(struct net_device *dev, char __user *useraddr)
 		return ret;
 
 	dev->wol_enabled = !!wol.wolopts;
+	ethtool_notify(dev, NULL, ETHNL_CMD_SET_SETTINGS, ETH_SETTINGS_IM_WOL,
+		       NULL);
 
 	return 0;
 }
net/ethtool/settings.c +63 -0

130 lines not on this page · raw message

Signature · 1 line
2.21.0
#97 lore raw

Re: [PATCH net-next v5 19/22] ethtool: provide WoL information in GET_SETTINGS request

Quoted 9 lines · Michal Kubecek

Add information about supported and enabled wake on LAN modes into the GET_SETTINGS reply when ETH_SETTINGS_IM_WOL flag is set in the request.

The GET_SETTINGS request can be still sent by unprivileged users but in such case the SecureOn password (if any) is not included in the reply.

Send notification in the same format as reply SET_SETTINGS message when wake on LAN settings are modified using ioctl interface (ETHTOOL_SWOL command).

Looks good to me, just one question below:

Quoted 3 lines

---

[snip]

net/ethtool/settings.c static const struct nla_policy get_settings_policy[ETHA_SETTINGS_MAX + 1] = { · line 131

Quoted 2 lines
net/ethtool/settings.c +1 -0
+	data->privileged = ethnl_is_privileged(skb);

Is not this potentially creating a time of check/time of use issue? Can you check for ethnl_is_privileged() at the time where you fill in the reply skb?

Signature · 1 line
Florian
#98 lore raw

Re: [PATCH net-next v5 19/22] ethtool: provide WoL information in GET_SETTINGS request

Quoted 7 lines · Florian Fainelli

net/ethtool/settings.c static const struct nla_policy get_settings_policy[ETHA_SETTINGS_MAX + 1] = { · line 131

Quoted 2 lines · Michal Kubecek
net/ethtool/settings.c +1 -0
+	data->privileged = ethnl_is_privileged(skb);

Is not this potentially creating a time of check/time of use issue? Can you check for ethnl_is_privileged() at the time where you fill in the reply skb?

I'm not sure what kind of race you have in mind but if this is a problem, we might have even bigger one for any "SET" request where the privilege check is performed even earlier by the genetlink dispatcher (by setting GENL_UNS_ADMIN_PERM or GENL_ADMIN_PERM flag in genl_ops).

Michal

#99 lore raw

PATCH v5 20/22 ethtool: set WoL settings with SET_SETTINGS request

In reply to Michal Kubecek

Allow enabling and disabling wake on LAN modes using SET_SETTINGS request.

ETHA_SETTINGS_WOL nested attribute is used to set or modify enabled WoL modes and SecureOn(tm) password.

---
 Documentation/networking/ethtool-netlink.txt | 12 +++-
 net/ethtool/settings.c                       | 60 +++++++++++++++++++-
 2 files changed, 70 insertions(+), 2 deletions(-)
Documentation/networking/ethtool-netlink.txt +11 -1
index bc7f28f0182f..3568d35ad7ec 100644
@@ -335,6 +335,9 @@ to be passed with SET_SETTINGS request:
         ETHA_LINKMODES_OURS		(bitset)	advertised link modes
         ETHA_LINKMODES_SPEED		(u32)		link speed (Mb/s)
         ETHA_LINKMODES_DUPLEX		(u8)		duplex mode
+    ETHA_SETTINGS_WOL		(nested)	wake on LAN settings
+        ETHA_WOL_MODES			(bitfield32)	wake on LAN modes
+        ETHA_WOL_SOPASS			(binary)	SecureOn(tm) password
 
 ETHA_LINKMODES_OURS bit set allows setting advertised link modes. If
 autonegotiation is on (either set now or kept from before), advertised modes
@@ -345,6 +348,13 @@ autoselection is done on ethtool side with ioctl interface, netlink interface
 is supposed to allow requesting changes without knowing what exactly kernel
 supports.
 
+ETHA_WOL_MODES bitfield is interpreted in the usual way, i.e. bits set in the
+selector are set to 0 or 1 according to value. To allow the semantics of the
+ioctl interface where the whole bitmap is set rather than only modified,
+selectors may have also bits not supported by device set and an error is only
+issued if any of them is also set in the value (i.e. if userspace tries to
+enable mode not supported by device).
+
 
 Request translation
 -------------------
@@ -360,7 +370,7 @@ ETHTOOL_SSET			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
 ETHTOOL_GWOL			ETHNL_CMD_GET_SETTINGS
-ETHTOOL_SWOL			n/a
+ETHTOOL_SWOL			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_GMSGLVL			n/a
 ETHTOOL_SMSGLVL			n/a
 ETHTOOL_NWAY_RST		n/a
net/ethtool/settings.c +59 -1

90 lines not on this page · raw message

Signature · 1 line
2.21.0
Reviewed-by on v5 20/22
#100 lore raw

Re: [PATCH net-next v5 20/22] ethtool: set WoL settings with SET_SETTINGS request

Quoted 7 lines · Michal Kubecek

Allow enabling and disabling wake on LAN modes using SET_SETTINGS request.

ETHA_SETTINGS_WOL nested attribute is used to set or modify enabled WoL modes and SecureOn(tm) password.

Signature · 1 line
Florian
#101 lore raw

PATCH v5 21/22 ethtool: provide message level in GET_SETTINGS request

In reply to Michal Kubecek

Add information about supported and enabled message levels to the GET_SETTINGS reply when ETH_SETTINGS_IM_DEBUG flag is set in the request.

Unlike in the ioctl interface, "message level" is called "message mask" as it is in fact interpreted as a bit mask. It is put into a nested attribute ETHA_SETTINGS_DEBUG to allow future extensions (e.g. an actual verbosity level).

Send notification in the same format as reply when message level is modified using the ioctl interface (ETHTOOL_SMSGLVL command).

---
 Documentation/networking/ethtool-netlink.txt |  9 ++++-
 include/linux/netdevice.h                    |  2 +
 include/uapi/linux/ethtool_netlink.h         | 13 ++++++-
 net/ethtool/ioctl.c                          |  3 ++
 net/ethtool/settings.c                       | 40 ++++++++++++++++++++
 5 files changed, 65 insertions(+), 2 deletions(-)
Documentation/networking/ethtool-netlink.txt +8 -1
index 3568d35ad7ec..603acfbefe29 100644
@@ -279,6 +279,7 @@ Info mask bits meaning:
     ETH_SETTINGS_IM_LINKMODES		link modes and related
     ETH_SETTINGS_IM_LINKSTATE		link state
     ETH_SETTINGS_IM_WOL			struct ethtool_wolinfo
+    ETH_SETTINGS_IM_DEBUG		debugging
 
 Response contents:
 
@@ -300,6 +301,8 @@ Response contents:
     ETHA_SETTINGS_WOL		(nested)	wake on LAN settings
         ETHA_WOL_MODES			(bitfield32)	wake on LAN modes
         ETHA_WOL_SOPASS			(binary)	SecureOn(tm) password
+    ETHA_SETTINGS_DEBUG		(nested)	debugging
+        ETHA_DEBUG_MSG_MASK		(bitfield32)	debug message mask
 
 Most of the attributes and their values have the same meaning as matching
 members of the corresponding ioctl structures. For ETHA_LINKMODES_OURS,
@@ -309,6 +312,10 @@ ETHA_LINKMODES_PEER in the reply is a bit list.
 For ETHA_WOL_MODES, selector reports wake on LAN modes supported by the
 device and value enabled modes.
 
+ETHA_DEBUG_MSG_MASK corresponds to message level (which is actually a bitfield)
+as reported by ETHTOOL_GMSGLVL. The selector reports all message types
+recognized by kernel and value types enabled for the device.
+
 GET_SETTINGS request is allowed for unprivileged user but ETHA_WOL_SOPASS
 is only provided by kernel in response to privileged (netns CAP_NET_ADMIN)
 requests.
@@ -371,7 +378,7 @@ ETHTOOL_GDRVINFO		ETHNL_CMD_GET_INFO
 ETHTOOL_GREGS			n/a
 ETHTOOL_GWOL			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SWOL			ETHNL_CMD_SET_SETTINGS
-ETHTOOL_GMSGLVL			n/a
+ETHTOOL_GMSGLVL			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SMSGLVL			n/a
 ETHTOOL_NWAY_RST		n/a
 ETHTOOL_GLINK			ETHNL_CMD_GET_SETTINGS
include/linux/netdevice.h +2 -0
index bc761511edb4..23ebb37769c9 100644
@@ -3859,6 +3859,8 @@ enum {
 	NETIF_MSG_PKTDATA	= 0x1000,
 	NETIF_MSG_HW		= 0x2000,
 	NETIF_MSG_WOL		= 0x4000,
+
+	NETIF_MSG_ALL		= 0x7fff,
 };
 
 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
include/uapi/linux/ethtool_netlink.h +12 -1
index 532ad11ae87c..924f7059f944 100644
@@ -198,6 +198,7 @@ enum {
 	ETHA_SETTINGS_LINK_MODES,		/* nest - ETHA_LINKMODES_* */
 	ETHA_SETTINGS_LINK_STATE,		/* nest - ETHA_LINKSTATE_* */
 	ETHA_SETTINGS_WOL,			/* nest - ETHA_WOL_* */
+	ETHA_SETTINGS_DEBUG,			/* nest - ETHA_DEBUG_* */
 
 	__ETHA_SETTINGS_CNT,
 	ETHA_SETTINGS_MAX = (__ETHA_SETTINGS_CNT - 1)
@@ -207,11 +208,13 @@ enum {
 #define ETH_SETTINGS_IM_LINKMODES		(1U << 1)
 #define ETH_SETTINGS_IM_LINKSTATE		(1U << 2)
 #define ETH_SETTINGS_IM_WOL			(1U << 3)
+#define ETH_SETTINGS_IM_DEBUG			(1U << 4)
 
 #define ETH_SETTINGS_IM_ALL (ETH_SETTINGS_IM_LINKINFO | \
 			     ETH_SETTINGS_IM_LINKMODES | \
 			     ETH_SETTINGS_IM_LINKSTATE | \
-			     ETH_SETTINGS_IM_WOL)
+			     ETH_SETTINGS_IM_WOL | \
+			     ETH_SETTINGS_IM_DEBUG)
 
 enum {
 	ETHA_LINKINFO_UNSPEC,
@@ -254,6 +257,14 @@ enum {
 	ETHA_WOL_MAX = (__ETHA_WOL_CNT - 1)
 };
 
+enum {
+	ETHA_DEBUG_UNSPEC,
+	ETHA_DEBUG_MSG_MASK,			/* bitfield32 */
+
+	__ETHA_DEBUG_CNT,
+	ETHA_DEBUG_MAX = (__ETHA_DEBUG_CNT - 1)
+};
+
 /* generic netlink info */
 #define ETHTOOL_GENL_NAME "ethtool"
 #define ETHTOOL_GENL_VERSION 1
net/ethtool/ioctl.c +3 -0
index bba02a218eea..e94fcd947c87 100644
@@ -2464,6 +2464,9 @@ int dev_ethtool(struct net *net, struct ifreq *ifr)
 	case ETHTOOL_SMSGLVL:
 		rc = ethtool_set_value_void(dev, useraddr,
 				       dev->ethtool_ops->set_msglevel);
+		if (rc >= 0)
+			ethtool_notify(dev, NULL, ETHNL_CMD_SET_SETTINGS,
+				       ETH_SETTINGS_IM_DEBUG, NULL);
 		break;
 	case ETHTOOL_GEEE:
 		rc = ethtool_get_eee(dev, useraddr);
net/ethtool/settings.c +40 -0
index ea5279dad826..53409dd8af34 100644
@@ -14,6 +14,7 @@ struct settings_data {
 	struct ethtool_wolinfo		wolinfo;
 	struct ethtool_link_settings	*lsettings;
 	int				link;
+	u32				msglevel;
 	bool				lpm_empty;
 };
 
@@ -123,6 +124,7 @@ static const struct nla_policy get_settings_policy[ETHA_SETTINGS_MAX + 1] = {
 	[ETHA_SETTINGS_LINK_MODES]	= { .type = NLA_REJECT },
 	[ETHA_SETTINGS_LINK_STATE]	= { .type = NLA_REJECT },
 	[ETHA_SETTINGS_WOL]		= { .type = NLA_REJECT },
+	[ETHA_SETTINGS_DEBUG]		= { .type = NLA_REJECT },
 };
 
 static int parse_settings(struct common_req_info *req_info,
@@ -188,6 +190,7 @@ static int prepare_settings(struct common_req_info *req_info,
 	struct settings_data *data =
 		container_of(req_info, struct settings_data, reqinfo_base);
 	struct net_device *dev = data->repdata_base.dev;
+	const struct ethtool_ops *eops = dev->ethtool_ops;
 	u32 req_mask = req_info->req_mask;
 	int ret;
 
@@ -222,6 +225,12 @@ static int prepare_settings(struct common_req_info *req_info,
 		if (ret < 0)
 			req_mask &= ~ETH_SETTINGS_IM_WOL;
 	}
+	if (req_mask & ETH_SETTINGS_IM_DEBUG) {
+		if (eops->get_msglevel)
+			data->msglevel = eops->get_msglevel(dev);
+		else
+			req_mask &= ~ETH_SETTINGS_IM_DEBUG;
+	}
 	ethnl_after_ops(dev);
 
 	data->repdata_base.info_mask = req_mask;
@@ -283,6 +292,11 @@ static int wol_size(void)
 			      nla_total_size(SOPASS_MAX));
 }
 
+static int debug_size(void)
+{
+	return nla_total_size(nla_total_size(sizeof(struct nla_bitfield32)));
+}
+
 /* To keep things simple, reserve space for some attributes which may not
  * be added to the message (e.g. ETHA_SETTINGS_SOPASS); therefore the length
  * returned may be bigger than the actual length of the message sent
@@ -308,6 +322,8 @@ static int settings_size(const struct common_req_info *req_info)
 		len += link_state_size(data->link);
 	if (info_mask & ETH_SETTINGS_IM_WOL)
 		len += wol_size();
+	if (info_mask & ETH_SETTINGS_IM_DEBUG)
+		len += debug_size();
 
 	return len;
 }
@@ -424,6 +440,24 @@ static int fill_wolinfo(struct sk_buff *skb,
 	return -EMSGSIZE;
 }
 
+static int fill_debug(struct sk_buff *skb, u32 msglevel)
+{
+	struct nlattr *nest;
+
+	nest = ethnl_nest_start(skb, ETHA_SETTINGS_DEBUG);
+	if (!nest)
+		return -EMSGSIZE;
+	if (nla_put_bitfield32(skb, ETHA_DEBUG_MSG_MASK, msglevel,
+			       NETIF_MSG_ALL))
+		goto err;
+	nla_nest_end(skb, nest);
+	return 0;
+
+err:
+	nla_nest_cancel(skb, nest);
+	return -EMSGSIZE;
+}
+
 static int fill_settings(struct sk_buff *skb,
 			 const struct common_req_info *req_info)
 {
@@ -454,6 +488,11 @@ static int fill_settings(struct sk_buff *skb,
 		if (ret < 0)
 			return ret;
 	}
+	if (info_mask & ETH_SETTINGS_IM_DEBUG) {
+		ret = fill_debug(skb, data->msglevel);
+		if (ret < 0)
+			return ret;
+	}
 
 	return 0;
 }
@@ -508,6 +547,7 @@ static const struct nla_policy set_settings_policy[ETHA_SETTINGS_MAX + 1] = {
 	[ETHA_SETTINGS_LINK_MODES]	= { .type = NLA_NESTED },
 	[ETHA_SETTINGS_LINK_STATE]	= { .type = NLA_REJECT },
 	[ETHA_SETTINGS_WOL]		= { .type = NLA_NESTED },
+	[ETHA_SETTINGS_DEBUG]		= { .type = NLA_REJECT },
 };
 
 static int ethnl_set_link_ksettings(struct genl_info *info,
Signature · 1 line
2.21.0
Reviewed-by on v5 21/22
#102 lore raw

Re: [PATCH net-next v5 21/22] ethtool: provide message level in GET_SETTINGS request

Quoted 9 lines · Michal Kubecek

Add information about supported and enabled message levels to the GET_SETTINGS reply when ETH_SETTINGS_IM_DEBUG flag is set in the request.

Unlike in the ioctl interface, "message level" is called "message mask" as it is in fact interpreted as a bit mask. It is put into a nested attribute ETHA_SETTINGS_DEBUG to allow future extensions (e.g. an actual verbosity level).

Quoted 4 lines

Send notification in the same format as reply when message level is modified using the ioctl interface (ETHTOOL_SMSGLVL command).

Signature · 1 line
Florian
#103 lore raw

PATCH v5 22/22 ethtool: set message level with SET_SETTINGS request

In reply to Michal Kubecek

Allow setting device message level using ETHA_SETTINGS_DEBUG nested attribute.

Unlike in ioctl interface "message level" is called "message mask" (as it is in fact used as a bit mask) and put inside a nested attribute to allow future extensions.

---
 Documentation/networking/ethtool-netlink.txt |  8 +++-
 net/ethtool/settings.c                       | 45 +++++++++++++++++++-
 2 files changed, 50 insertions(+), 3 deletions(-)
Documentation/networking/ethtool-netlink.txt +6 -2
index 603acfbefe29..20e43d42afdd 100644
@@ -345,6 +345,8 @@ to be passed with SET_SETTINGS request:
     ETHA_SETTINGS_WOL		(nested)	wake on LAN settings
         ETHA_WOL_MODES			(bitfield32)	wake on LAN modes
         ETHA_WOL_SOPASS			(binary)	SecureOn(tm) password
+    ETHA_SETTINGS_DEBUG		(nested)	debugging
+        ETHA_DEBUG_MSG_MASK		(bitfield32)	message mask
 
 ETHA_LINKMODES_OURS bit set allows setting advertised link modes. If
 autonegotiation is on (either set now or kept from before), advertised modes
@@ -360,7 +362,9 @@ selector are set to 0 or 1 according to value. To allow the semantics of the
 ioctl interface where the whole bitmap is set rather than only modified,
 selectors may have also bits not supported by device set and an error is only
 issued if any of them is also set in the value (i.e. if userspace tries to
-enable mode not supported by device).
+enable mode not supported by device). ETHA_SETTINGS_MSGLEVEL bitfield also
+allows bits not recognized by kernel in selector as long as the request does
+not attempt to enable them.
 
 
 Request translation
@@ -379,7 +383,7 @@ ETHTOOL_GREGS			n/a
 ETHTOOL_GWOL			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_SWOL			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_GMSGLVL			ETHNL_CMD_GET_SETTINGS
-ETHTOOL_SMSGLVL			n/a
+ETHTOOL_SMSGLVL			ETHNL_CMD_SET_SETTINGS
 ETHTOOL_NWAY_RST		n/a
 ETHTOOL_GLINK			ETHNL_CMD_GET_SETTINGS
 ETHTOOL_GEEPROM			n/a
net/ethtool/settings.c +44 -1

69 lines not on this page · raw message

Signature · 1 line
2.21.0
Reviewed-by on v5 22/22
#104 lore raw

Re: [PATCH net-next v5 22/22] ethtool: set message level with SET_SETTINGS request

Quoted 8 lines · Michal Kubecek

Allow setting device message level using ETHA_SETTINGS_DEBUG nested attribute.

Unlike in ioctl interface "message level" is called "message mask" (as it is in fact used as a bit mask) and put inside a nested attribute to allow future extensions.

Signature · 1 line
Florian
#106 lore raw
Quoted 1 line · Jiri Pirko

Could you please perhaps split the patchset so it contains max ~15 patches?

Would 1-17 still be acceptable? The thing is that patch 16 is the first where some "normal" notification (i.e. not EVENT one) is sent and patch 17 is the first implementation of a request modifying data. I believe having at least one example of both querying and setting data would help the review discussion.

Or perhaps I could leave GET_INFO related patches (at least second and third) for later.

Michal

#107 lore raw
Quoted 11 lines · Wed, Mar 27, 2019 at 03:28:06PM CET, [email protected] wrote:
Quoted 1 line · Jiri Pirko

Could you please perhaps split the patchset so it contains max ~15 patches?

Would 1-17 still be acceptable? The thing is that patch 16 is the first where some "normal" notification (i.e. not EVENT one) is sent and patch 17 is the first implementation of a request modifying data. I believe having at least one example of both querying and setting data would help the review discussion.

Or perhaps I could leave GET_INFO related patches (at least second and third) for later.

I think 17 is okay. Please split if possible.

Quoted 2 lines

Michal

maintainer
#108 lore raw
In reply to Michal Kubecek

From: Michal Kubecek <[email protected]>
Date: Wed, 27 Mar 2019 15:28:06 +0100

Quoted 4 lines · Jiri Pirko
Quoted 1 line

Could you please perhaps split the patchset so it contains max ~15 patches?

Would 1-17 still be acceptable?

I'm OK with ~17 especially considering the scope of this work.