devlink: code split and structured instance walk
Subsystems (4)
Jakub Kicinski
maintainerv2 cover devlink: code split and structured instance walk
Split devlink.c into a handful of files, trying to keep the "core" code away from all the command-specific implementations. The core code has been quite scattered until now. Going forward we can consider using a source file per-subobject, I think that it's quite beneficial to newcomers (based on relative ease with which folks contribute to ethtool vs devlink). But this series doesn't split everything out, yet - partially due to backporting concerns, but mostly due to lack of time. Bulk of the netlink command handling is left in a leftover.c file.
Introduce a context structure for dumps, and use it to store the devlink instance ID of the last dumped devlink instance. This means we don't have to restart the walk from 0 each time.
Finally - introduce a "structured walk". A centralized dump handler in devlink/netlink.c which walks the devlink instances, deals with refcounting/locking, simplifying the per-object implementations quite a bit. Inspired by the ethtool code.
v2:
- various renames
- split patch 2
v1: https://lore.kernel.org/all/[email protected]/
- re-route the locking fix via net
- rename basic.c -> leftover.c
RFC: https://lore.kernel.org/all/[email protected]/
Jakub Kicinski (15):
devlink: move code to a dedicated directory
devlink: rename devlink_netdevice_event ->
devlink_port_netdevice_event
devlink: split out core code
devlink: split out netlink code
netlink: add macro for checking dump ctx size
devlink: use an explicit structure for dump context
devlink: remove start variables from dumps
devlink: drop the filter argument from devlinks_xa_find_get
devlink: health: combine loops in dump
devlink: restart dump based on devlink instance ids (simple)
devlink: restart dump based on devlink instance ids (nested)
devlink: restart dump based on devlink instance ids (function)
devlink: uniformly take the devlink instance lock in the dump loop
devlink: add by-instance dump infra
devlink: convert remaining dumps to the by-instance scheme
include/linux/netlink.h | 4 +
net/Makefile | 1 +
net/core/Makefile | 1 -
net/devlink/Makefile | 3 +
net/devlink/core.c | 345 +++++
net/devlink/devl_internal.h | 207 +++
net/{core/devlink.c => devlink/leftover.c} | 1460 +++++---------------
net/devlink/netlink.c | 242 ++++
net/netfilter/nf_conntrack_netlink.c | 2 +-
9 files changed, 1160 insertions(+), 1105 deletions(-)
create mode 100644 net/devlink/Makefile
create mode 100644 net/devlink/core.c
create mode 100644 net/devlink/devl_internal.h
rename net/{core/devlink.c => devlink/leftover.c} (90%)
create mode 100644 net/devlink/netlink.cSignature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 1/15 devlink: move code to a dedicated directory
The devlink code is hard to navigate with 13kLoC in one file. I really like the way Michal split the ethtool into per-command files and core. It'd probably be too much to split it all up, but we can at least separate the core parts out of the per-cmd implementations and put it in a directory so that new commands can be separate files.
Move the code, subsequent commit will do a partial split.
---
net/Makefile | 1 +
net/core/Makefile | 1 -
net/devlink/Makefile | 3 +++
net/{core/devlink.c => devlink/leftover.c} | 0
4 files changed, 4 insertions(+), 1 deletion(-)
create mode 100644 net/devlink/Makefile
rename net/{core/devlink.c => devlink/leftover.c} (100%)net/Makefile +1 -0
@@ -23,6 +23,7 @@ obj-$(CONFIG_BPFILTER) += bpfilter/ obj-$(CONFIG_PACKET) += packet/ obj-$(CONFIG_NET_KEY) += key/ obj-$(CONFIG_BRIDGE) += bridge/ +obj-$(CONFIG_NET_DEVLINK) += devlink/ obj-$(CONFIG_NET_DSA) += dsa/ obj-$(CONFIG_ATALK) += appletalk/ obj-$(CONFIG_X25) += x25/
net/core/Makefile +0 -1
@@ -33,7 +33,6 @@ obj-$(CONFIG_LWTUNNEL) += lwtunnel.o obj-$(CONFIG_LWTUNNEL_BPF) += lwt_bpf.o obj-$(CONFIG_DST_CACHE) += dst_cache.o obj-$(CONFIG_HWBM) += hwbm.o -obj-$(CONFIG_NET_DEVLINK) += devlink.o obj-$(CONFIG_GRO_CELLS) += gro_cells.o obj-$(CONFIG_FAILOVER) += failover.o obj-$(CONFIG_NET_SOCK_MSG) += skmsg.o
net/devlink/Makefile new file +3 -0
@@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0 + +obj-y := leftover.o
net/devlink/leftover.c renamed from net/core/devlink.c +0 -0
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 01/15] devlink: move code to a dedicated directory
Quoted 11 lines · Thu, Jan 05, 2023 at 05:05:17AM CET, [email protected] wrote:
The devlink code is hard to navigate with 13kLoC in one file. I really like the way Michal split the ethtool into per-command files and core. It'd probably be too much to split it all up, but we can at least separate the core parts out of the per-cmd implementations and put it in a directory so that new commands can be separate files.
Move the code, subsequent commit will do a partial split.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
Jakub Kicinski
maintainerPATCH v2 2/15 devlink: rename devlink_netdevice_event -> devlink_port_netdevice_event
In reply to Jakub KicinskiTo make the upcoming change a pure(er?) code move rename devlink_netdevice_event -> devlink_port_netdevice_event. This makes it clear that it only touches ports and doesn't belong cleanly in the core.
--- net/devlink/leftover.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
net/devlink/leftover.c +5 -5
@@ -9942,8 +9942,8 @@ void devlink_set_features(struct devlink *devlink, u64 features) }EXPORT_SYMBOL_GPL (devlink_set_features); -static int devlink_netdevice_event (struct notifier_block *nb, -unsigned long event,void *ptr); +static int devlink_port_netdevice_event (struct notifier_block *nb, +unsigned long event,void *ptr);/** * devlink_alloc_ns - Allocate new devlink instance resources
@@ -9978,7 +9978,7 @@ struct devlink *devlink_alloc_ns(const struct devlink_ops *ops,if (ret <0 ) goto err_xa_alloc; - devlink->netdevice_nb.notifier_call = devlink_netdevice_event; + devlink->netdevice_nb.notifier_call = devlink_port_netdevice_event; ret =register_netdevice_notifier_net (net, &devlink->netdevice_nb);if (ret) goto err_register_netdevice_notifier;
@@ -10480,8 +10480,8 @@ void devlink_port_type_clear(struct devlink_port *devlink_port) }EXPORT_SYMBOL_GPL (devlink_port_type_clear); -static int devlink_netdevice_event (struct notifier_block *nb, -unsigned long event,void *ptr) +static int devlink_port_netdevice_event (struct notifier_block *nb, +unsigned long event,void *ptr) {struct net_device *netdev =netdev_notifier_info_to_dev (ptr);struct devlink_port *devlink_port = netdev->devlink_port;
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 02/15] devlink: rename devlink_netdevice_event -> devlink_port_netdevice_event
Quoted 6 lines · Thu, Jan 05, 2023 at 05:05:18AM CET, [email protected] wrote:
To make the upcoming change a pure(er?) code move rename
devlink_netdevice_event->devlink_port_netdevice_event. This makes it clear that it only touches ports and doesn't belong cleanly in the core.Signed-off-by Jakub Kicinski <[email protected]>
Jakub Kicinski
maintainerPATCH v2 3/15 devlink: split out core code
In reply to Jakub KicinskiMove core code into a separate file. It's spread around the main file which makes refactoring and figuring out how devlink works harder.
Move the xarray, all the most core devlink instance code out like locking, ref counting, alloc, register, etc. Leave port stuff in leftover.c, if we want to move port code it'd probably be to its own file.
--- net/devlink/Makefile | 2 +- net/devlink/core.c | 347 ++++++++++++++++++++++++++++ net/devlink/devl_internal.h | 117 ++++++++++ net/devlink/leftover.c | 444 +----------------------------------- 4 files changed, 476 insertions(+), 434 deletions(-) create mode 100644 net/devlink/core.c create mode 100644 net/devlink/devl_internal.h
net/devlink/Makefile +1 -1
@@ -1,3 +1,3 @@ # SPDX-License-Identifier: GPL-2.0 -obj-y := leftover.o +obj-y := leftover.o core.o
net/devlink/core.c new file +347 -0
347 lines not on this page · raw message
net/devlink/devl_internal.h new file +117 -0
117 lines not on this page · raw message
net/devlink/leftover.c +11 -433
525 lines not on this page · raw message
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 03/15] devlink: split out core code
Quoted 11 lines · Thu, Jan 05, 2023 at 05:05:19AM CET, [email protected] wrote:
Move core code into a separate file. It's spread around the main file which makes refactoring and figuring out how devlink works harder.
Move the xarray, all the most core devlink instance code out like locking, ref counting, alloc, register, etc. Leave port stuff in
leftover.c, if we want to move port code it'd probably be to its own file.Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
Btw, I don't think it is correct to carry Jacob's review tag around when the patches changed (not only this one).
Jakub Kicinski
maintainerRe: [PATCH net-next v2 03/15] devlink: split out core code
Quoted 2 lines · On Thu, 5 Jan 2023 10:16:40 +0100 Jiri Pirko wrote:
Btw, I don't think it is correct to carry Jacob's review tag around when the patches changed (not only this one).
There's no clear rule for that :( I see more people screaming at submitters for dropping tags than for keeping them so I err on that side myself.
Here I think the case is relatively clear, since I'm only doing renames and bike-shedding stuff, and Jake is not one to bike-shed in my experience.
Jacob Keller
Reviewed-byRe: [PATCH net-next v2 03/15] devlink: split out core code
Quoted 11 lines · Jakub Kicinski
Quoted 2 lines · On Thu, 5 Jan 2023 10:16:40 +0100 Jiri Pirko wrote:
Btw, I don't think it is correct to carry Jacob's review tag around when the patches changed (not only this one).
There's no clear rule for that :( I see more people screaming at submitters for dropping tags than for keeping them so I err on that side myself.
Here I think the case is relatively clear, since I'm only doing renames and bike-shedding stuff, and Jake is not one to bike-shed in my experience.
Now that I'm back from vacation, I can just clarify here:
Jiri Pirko
maintainerRe: [PATCH net-next v2 03/15] devlink: split out core code
In reply to Jakub KicinskiQuoted 11 lines · Thu, Jan 05, 2023 at 07:34:28PM CET, [email protected] wrote:
Quoted 2 lines · On Thu, 5 Jan 2023 10:16:40 +0100 Jiri Pirko wrote:
Btw, I don't think it is correct to carry Jacob's review tag around when the patches changed (not only this one).
There's no clear rule for that :( I see more people screaming at submitters for dropping tags than for keeping them so I err on that side myself.
Here I think the case is relatively clear, since I'm only doing renames and bike-shedding stuff, and Jake is not one to bike-shed in my experience.
Okay.
Jakub Kicinski
maintainerPATCH v2 4/15 devlink: split out netlink code
In reply to Jakub KicinskiMove out the netlink glue into a separate file.
Leave the ops in the old file because we'd have to export a ton of functions. Going forward we should switch to split ops which will let us to put the new ops in the netlink.c file.
Pure code move, no functional changes.
--- net/devlink/Makefile | 2 +- net/devlink/devl_internal.h | 31 ++++++ net/devlink/leftover.c | 212 ++---------------------------------- net/devlink/netlink.c | 195 +++++++++++++++++++++++++++++++++ 4 files changed, 235 insertions(+), 205 deletions(-) create mode 100644 net/devlink/netlink.c
net/devlink/Makefile +1 -1
@@ -1,3 +1,3 @@ # SPDX-License-Identifier: GPL-2.0 -obj-y := leftover.o core.o +obj-y := leftover.o core.o netlink.o
net/devlink/devl_internal.h +31 -0
@@ -97,6 +97,20 @@ devlinks_xa_find_get_next(struct net *net, unsigned long *indexp, xa_mark_t filter);/* Netlink */ +#define DEVLINK_NL_FLAG_NEED_PORT BIT(0) +#define DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT BIT(1) +#define DEVLINK_NL_FLAG_NEED_RATE BIT(2) +#define DEVLINK_NL_FLAG_NEED_RATE_NODE BIT(3) +#define DEVLINK_NL_FLAG_NEED_LINECARD BIT(4) + +enum devlink_multicast_groups { +DEVLINK_MCGRP_CONFIG , +}; + +extern const struct genl_small_ops devlink_nl_ops[56 ]; + +struct devlink *devlink_get_from_attrs (struct net *net,struct nlattr **attrs); +void devlink_notify_unregister (struct devlink *devlink);void devlink_notify_register (struct devlink *devlink);
@@ -104,6 +118,9 @@ void devlink_notify_register(struct devlink *devlink);int devlink_port_netdevice_event (struct notifier_block *nb,unsigned long event,void *ptr); +struct devlink_port * +devlink_port_get_from_info (struct devlink *devlink,struct genl_info *info); +/* Reload */ bool devlink_reload_actions_valid (const struct devlink_ops *ops); int devlink_reload(struct devlink *devlink, struct net *dest_net,
@@ -115,3 +132,17 @@ static inline bool devlink_reload_supported(const struct devlink_ops *ops) {return ops->reload_down && ops->reload_up; } + +/* Line cards */ +struct devlink_linecard ; + +struct devlink_linecard * +devlink_linecard_get_from_info (struct devlink *devlink,struct genl_info *info); +void devlink_linecard_put (struct devlink_linecard *linecard); + +/* Rates */ +struct devlink_rate * +devlink_rate_get_from_info (struct devlink *devlink,struct genl_info *info); +struct devlink_rate * +devlink_rate_node_get_from_info (struct devlink *devlink, +struct genl_info *info);
net/devlink/leftover.c +8 -204
264 lines not on this page · raw message
net/devlink/netlink.c new file +195 -0
195 lines not on this page · raw message
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 04/15] devlink: split out netlink code
Quoted 9 lines · Thu, Jan 05, 2023 at 05:05:20AM CET, [email protected] wrote:
Move out the netlink glue into a separate file.
Leave the ops in the old file because we'd have to export a ton of functions. Going forward we should switch to split ops which will let us to put the new ops in thenetlink.cfile.Pure code move, no functional changes.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
Jakub Kicinski
maintainerPATCH v2 5/15 netlink: add macro for checking dump ctx size
In reply to Jakub KicinskiWe encourage casting struct netlink_callback::ctx to a local struct (in a comment above the field). Provide a convenience macro for checking if the local struct fits into the ctx.
--- include/linux/netlink.h | 4 ++++ net/netfilter/nf_conntrack_netlink.c | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-)
include/linux/netlink.h +4 -0
@@ -263,6 +263,10 @@ struct netlink_callback { }; }; +#define NL_ASSET_DUMP_CTX_FITS (type_name) \ + BUILD_BUG_ON(sizeof(type_name) > \ + sizeof_field(struct netlink_callback, ctx)) +struct netlink_notify {struct net *net;u32 portid;
net/netfilter/nf_conntrack_netlink.c +1 -1
@@ -3866,7 +3866,7 @@ static int __init ctnetlink_init(void) {int ret; - BUILD_BUG_ON(sizeof(struct ctnetlink_list_dump_ctx) > sizeof_field(struct netlink_callback, ctx)); +NL_ASSET_DUMP_CTX_FITS (struct ctnetlink_list_dump_ctx ); ret =nfnetlink_subsys_register (&ctnl_subsys);if (ret <0 ) {
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 05/15] netlink: add macro for checking dump ctx size
Quoted 17 lines · Jakub Kicinski
We encourage casting struct
netlink_callback::ctxto a local struct (in a comment above the field). Provide a convenience macro for checking if the local struct fits into the ctx.Reviewed-by Jacob Keller <[email protected]>Reviewed-by Jiri Pirko <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>--- include/linux/netlink.h | 4 ++++ net/netfilter/nf_conntrack_netlink.c | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-)include/linux/netlink.h +0
-0
include/linux/netlink.h struct netlink_callback { · line 266
Quoted 4 lines
include/linux/netlink.h +1
-0@@ -263,6 +263,10 @@ struct netlink_callback { }; }; +#define NL_ASSET_DUMP_CTX_FITS (type_name) \
Wanted to use this macro, but the name doesn't make sense to me. Should it be NL_ASSERT_DUMP_CTX_FITS() ?
Quoted 22 lines
include/linux/netlink.h +3
-0+BUILD_BUG_ON (sizeof (type_name) > \ + sizeof_field(struct netlink_callback, ctx)) +struct netlink_notify {struct net *net;u32 portid;net/netfilter/nf_conntrack_netlink.c +1
-1@@ -3866,7 +3866,7 @@ static int __init ctnetlink_init(void) {int ret; - BUILD_BUG_ON(sizeof(struct ctnetlink_list_dump_ctx) > sizeof_field(struct netlink_callback, ctx)); +NL_ASSET_DUMP_CTX_FITS (struct ctnetlink_list_dump_ctx ); ret =nfnetlink_subsys_register (&ctnl_subsys);if (ret <0 ) {Signature · 1 line
2.38.1
Jakub Kicinski
maintainerRe: [PATCH net-next v2 05/15] netlink: add macro for checking dump ctx size
Quoted 4 lines · On Mon, 23 Jan 2023 16:20:21 +0200 Ido Schimmel wrote:
Quoted 1 line
+#define
NL_ASSET_DUMP_CTX_FITS(type_name)\Wanted to use this macro, but the name doesn't make sense to me. Should it be
NL_ASSERT_DUMP_CTX_FITS()?
Yes :(
Jakub Kicinski
maintainerPATCH v2 6/15 devlink: use an explicit structure for dump context
In reply to Jakub KicinskiCreate a dump context structure instead of using cb->args as an unsigned long array. This is a pure conversion which is intended to be as much of a noop as possible. Subsequent changes will use this to simplify the code.
The two non-trivial parts are:
- devlink_nl_cmd_health_reporter_dump_get_dumpit() checks args[0] to see if devlink_fmsg_dumpit() has already been called (whether this is the first msg), but doesn't use the exact value, so we can drop the local variable there already
- devlink_nl_cmd_region_read_dumpit() uses args[0] for address but we'll use args[1] now, shouldn't matter
--- net/devlink/devl_internal.h | 23 +++++++++ net/devlink/leftover.c | 98 ++++++++++++++++++++++--------------- 2 files changed, 81 insertions(+), 40 deletions(-)
net/devlink/devl_internal.h +23 -0
@@ -107,6 +107,21 @@ enum devlink_multicast_groups { DEVLINK_MCGRP_CONFIG, }; +/* state held across netlink dumps */ +struct devlink_nl_dump_state { +int idx; +union { +/* DEVLINK_CMD_REGION_READ */ +struct { +u64 start_offset; + }; +/* DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET */ +struct { +u64 dump_ts; + }; + }; +}; +extern const struct genl_small_ops devlink_nl_ops[56 ];struct devlink *devlink_get_from_attrs (struct net *net,struct nlattr **attrs);
@@ -114,6 +129,14 @@ struct devlink *devlink_get_from_attrs(struct net *net, struct nlattr **attrs);void devlink_notify_unregister (struct devlink *devlink);void devlink_notify_register (struct devlink *devlink); +static inline struct devlink_nl_dump_state * +devlink_dump_state (struct netlink_callback *cb) +{ +NL_ASSET_DUMP_CTX_FITS (struct devlink_nl_dump_state ); + +return (struct devlink_nl_dump_state *)cb->ctx; +} +/* Ports */ int devlink_port_netdevice_event (struct notifier_block *nb,unsigned long event,void *ptr);
net/devlink/leftover.c +58 -40
372 lines not on this page · raw message
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 06/15] devlink: use an explicit structure for dump context
Quoted 13 lines · Thu, Jan 05, 2023 at 05:05:22AM CET, [email protected] wrote:
Create a dump context structure instead of using
cb->argsas an unsigned long array. This is a pure conversion which is intended to be as much of a noop as possible. Subsequent changes will use this to simplify the code.The two non-trivial parts are:
-devlink_nl_cmd_health_reporter_dump_get_dumpit()checks args[0] to see ifdevlink_fmsg_dumpit()has already been called (whether this is the first msg), but doesn't use the exact value, so we can drop the local variable there already
-devlink_nl_cmd_region_read_dumpit()uses args[0] for address but we'll use args[1] now, shouldn't matter
Quoted 2 lines
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
Jakub Kicinski
maintainerPATCH v2 7/15 devlink: remove start variables from dumps
In reply to Jakub KicinskiThe start variables made the code clearer when we had to access cb->args[0] directly, as the name args doesn't explain much. Now that we use a structure to hold state this seems no longer needed.
--- net/devlink/leftover.c | 55 +++++++++++++++--------------------------- 1 file changed, 19 insertions(+), 36 deletions(-)
net/devlink/leftover.c +19 -36
@@ -1225,7 +1225,6 @@ static int devlink_nl_cmd_rate_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink_rate *devlink_rate;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;
@@ -1236,7 +1235,7 @@ static int devlink_nl_cmd_rate_get_dumpit(struct sk_buff *msg,enum devlink_command cmd =DEVLINK_CMD_RATE_NEW ;u32 id =NETLINK_CB (cb->skb).portid; -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -1320,13 +1319,12 @@ static int devlink_nl_cmd_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err;devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < start) { +if (idx < state->idx) { idx++;devlink_put (devlink);continue ;
@@ -1377,14 +1375,13 @@ static int devlink_nl_cmd_port_get_dumpit(struct sk_buff *msg,struct devlink *devlink;struct devlink_port *devlink_port;unsigned long index, port_index; -int start = state->idx;int idx =0 ;int err;devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {devl_lock (devlink);xa_for_each (&devlink->ports, port_index, devlink_port) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -2156,7 +2153,6 @@ static int devlink_nl_cmd_linecard_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink_linecard *linecard;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err;
@@ -2164,7 +2160,7 @@ static int devlink_nl_cmd_linecard_get_dumpit(struct sk_buff *msg,devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {mutex_lock (&devlink->linecards_lock);list_for_each_entry (linecard, &devlink->linecard_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -2419,7 +2415,6 @@ static int devlink_nl_cmd_sb_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink;struct devlink_sb *devlink_sb; -int start = state->idx;unsigned long index;int idx =0 ;int err;
@@ -2427,7 +2422,7 @@ static int devlink_nl_cmd_sb_get_dumpit(struct sk_buff *msg,devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {devl_lock (devlink);list_for_each_entry (devlink_sb, &devlink->sb_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -2562,7 +2557,6 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink;struct devlink_sb *devlink_sb; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;
@@ -2573,8 +2567,8 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg,devl_lock (devlink);list_for_each_entry (devlink_sb, &devlink->sb_list, list) { - err =__sb_pool_get_dumpit (msg, start, &idx, devlink, - devlink_sb, + err =__sb_pool_get_dumpit (msg, state->idx, &idx, + devlink, devlink_sb,NETLINK_CB (cb->skb).portid, cb->nlh->nlmsg_seq);if (err == -EOPNOTSUPP ) {
@@ -2778,7 +2772,6 @@ static int devlink_nl_cmd_sb_port_pool_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink;struct devlink_sb *devlink_sb; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;
@@ -2789,7 +2782,7 @@ static int devlink_nl_cmd_sb_port_pool_get_dumpit(struct sk_buff *msg,devl_lock (devlink);list_for_each_entry (devlink_sb, &devlink->sb_list, list) { - err =__sb_port_pool_get_dumpit (msg, start, &idx, + err =__sb_port_pool_get_dumpit (msg, state->idx, &idx, devlink, devlink_sb,NETLINK_CB (cb->skb).portid, cb->nlh->nlmsg_seq);
@@ -3022,7 +3015,6 @@ devlink_nl_cmd_sb_tc_pool_bind_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink;struct devlink_sb *devlink_sb; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;
@@ -3033,9 +3025,8 @@ devlink_nl_cmd_sb_tc_pool_bind_get_dumpit(struct sk_buff *msg,devl_lock (devlink);list_for_each_entry (devlink_sb, &devlink->sb_list, list) { - err =__sb_tc_pool_bind_get_dumpit (msg, start, &idx, - devlink, - devlink_sb, + err =__sb_tc_pool_bind_get_dumpit (msg, state->idx, &idx, + devlink, devlink_sb,NETLINK_CB (cb->skb).portid, cb->nlh->nlmsg_seq);if (err == -EOPNOTSUPP ) {
@@ -4881,13 +4872,12 @@ static int devlink_nl_cmd_selftests_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < start || !devlink->ops->selftest_check) +if (idx < state->idx || !devlink->ops->selftest_check) goto inc;devl_lock (devlink);
@@ -5363,7 +5353,6 @@ static int devlink_nl_cmd_param_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink_param_item *param_item;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;
@@ -5371,7 +5360,7 @@ static int devlink_nl_cmd_param_get_dumpit(struct sk_buff *msg,devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {devl_lock (devlink);list_for_each_entry (param_item, &devlink->param_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -6107,14 +6096,13 @@ static int devlink_nl_cmd_region_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { err =devlink_nl_cmd_region_get_devlink_dumpit (msg, cb, devlink, - &idx, start); + &idx, state->idx);devlink_put (devlink);if (err) goto out;
@@ -6759,13 +6747,12 @@ static int devlink_nl_cmd_info_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err =0 ;devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < start) +if (idx < state->idx) goto inc;devl_lock (devlink);
@@ -7930,7 +7917,6 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,unsigned long index, port_index;struct devlink_port *port;struct devlink *devlink; -int start = state->idx;int idx =0 ;int err;
@@ -7938,7 +7924,7 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,mutex_lock (&devlink->reporters_lock);list_for_each_entry (reporter, &devlink->reporter_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -7962,7 +7948,7 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg,xa_for_each (&devlink->ports, port_index, port) {mutex_lock (&port->reporters_lock);list_for_each_entry (reporter, &port->reporter_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -8513,7 +8499,6 @@ static int devlink_nl_cmd_trap_get_dumpit(struct sk_buff *msg,struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink_trap_item *trap_item;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err;
@@ -8521,7 +8506,7 @@ static int devlink_nl_cmd_trap_get_dumpit(struct sk_buff *msg,devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {devl_lock (devlink);list_for_each_entry (trap_item, &devlink->trap_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -8731,7 +8716,6 @@ static int devlink_nl_cmd_trap_group_get_dumpit(struct sk_buff *msg,struct devlink_trap_group_item *group_item;u32 portid =NETLINK_CB (cb->skb).portid;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err;
@@ -8740,7 +8724,7 @@ static int devlink_nl_cmd_trap_group_get_dumpit(struct sk_buff *msg,devl_lock (devlink);list_for_each_entry (group_item, &devlink->trap_group_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
@@ -9036,7 +9020,6 @@ static int devlink_nl_cmd_trap_policer_get_dumpit(struct sk_buff *msg,struct devlink_trap_policer_item *policer_item;u32 portid =NETLINK_CB (cb->skb).portid;struct devlink *devlink; -int start = state->idx;unsigned long index;int idx =0 ;int err;
@@ -9045,7 +9028,7 @@ static int devlink_nl_cmd_trap_policer_get_dumpit(struct sk_buff *msg,devl_lock (devlink);list_for_each_entry (policer_item, &devlink->trap_policer_list, list) { -if (idx < start) { +if (idx < state->idx) { idx++;continue ; }
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 07/15] devlink: remove start variables from dumps
Quoted 7 lines · Thu, Jan 05, 2023 at 05:05:23AM CET, [email protected] wrote:
The start variables made the code clearer when we had to access cb->args[0] directly, as the name args doesn't explain much. Now that we use a structure to hold state this seems no longer needed.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
Jakub Kicinski
maintainerPATCH v2 8/15 devlink: drop the filter argument from devlinks_xa_find_get
In reply to Jakub KicinskiLooks like devlinks_xa_find_get() was intended to get the mark from the @filter argument. It doesn't actually use @filter, passing DEVLINK_REGISTERED to xa_find_fn() directly. Walking marks other than registered is unlikely so drop @filter argument completely.
--- net/devlink/core.c | 12 +++++------- net/devlink/devl_internal.h | 15 +++++---------- 2 files changed, 10 insertions(+), 17 deletions(-)
net/devlink/core.c +5 -7
@@ -92,7 +92,7 @@ void devlink_put(struct devlink *devlink) }static struct devlink * -devlinks_xa_find_get (struct net *net,unsigned long *indexp,xa_mark_t filter, +devlinks_xa_find_get (struct net *net,unsigned long *indexp,void * (*xa_find_fn)(struct xarray *,unsigned long *,unsigned long ,xa_mark_t )) {
@@ -125,17 +125,15 @@ devlinks_xa_find_get(struct net *net, unsigned long *indexp, xa_mark_t filter, }struct devlink * -devlinks_xa_find_get_first (struct net *net,unsigned long *indexp, -xa_mark_t filter) +devlinks_xa_find_get_first (struct net *net,unsigned long *indexp) { -return devlinks_xa_find_get (net, indexp, filter, xa_find); +return devlinks_xa_find_get (net, indexp, xa_find); }struct devlink * -devlinks_xa_find_get_next (struct net *net,unsigned long *indexp, -xa_mark_t filter) +devlinks_xa_find_get_next (struct net *net,unsigned long *indexp) { -return devlinks_xa_find_get (net, indexp, filter, xa_find_after); +return devlinks_xa_find_get (net, indexp, xa_find_after); }/**
net/devlink/devl_internal.h +5 -10
@@ -81,20 +81,15 @@ extern struct genl_family devlink_nl_family;* devlink_put() needs to be called for each iterated devlink pointer * in loop body in order to release the reference. */ -#define devlinks_xa_for_each_get (net, index, devlink, filter) \ - for (index = 0, \ - devlink = devlinks_xa_find_get_first(net, &index, filter); \ - devlink; devlink = devlinks_xa_find_get_next(net, &index, filter)) -#define devlinks_xa_for_each_registered_get (net, index, devlink) \ - devlinks_xa_for_each_get(net, index, devlink, DEVLINK_REGISTERED) + for (index = 0, \ + devlink = devlinks_xa_find_get_first(net, &index); \ + devlink; devlink = devlinks_xa_find_get_next(net, &index))struct devlink * -devlinks_xa_find_get_first (struct net *net,unsigned long *indexp, -xa_mark_t filter); +devlinks_xa_find_get_first (struct net *net,unsigned long *indexp);struct devlink * -devlinks_xa_find_get_next (struct net *net,unsigned long *indexp, -xa_mark_t filter); +devlinks_xa_find_get_next (struct net *net,unsigned long *indexp);/* Netlink */ #define DEVLINK_NL_FLAG_NEED_PORT BIT(0)
Signature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 9/15 devlink: health: combine loops in dump
In reply to Jakub KicinskiWalk devlink instances only once. Dump the instance reporters and port reporters before moving to the next instance. User space should not depend on ordering of messages.
This will make improving stability of the walk easier.
--- net/devlink/leftover.c | 3 --- 1 file changed, 3 deletions(-)
net/devlink/leftover.c +0 -3
@@ -7940,10 +7940,7 @@ devlink_nl_cmd_health_reporter_get_dumpit(struct sk_buff *msg, idx++; }mutex_unlock (&devlink->reporters_lock); -devlink_put (devlink); - } -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) {devl_lock (devlink);xa_for_each (&devlink->ports, port_index, port) {mutex_lock (&port->reporters_lock);
Signature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 10/15 devlink: restart dump based on devlink instance ids (simple)
In reply to Jakub Kicinskixarray gives each devlink instance an id and allows us to restart walk based on that id quite neatly. This is nice both from the perspective of code brevity and from the stability of the dump (devlink instances disappearing from before the resumption point will not cause inconsistent dumps).
This patch takes care of simple cases where state->idx counts devlink instances only.
--- net/devlink/core.c | 2 +- net/devlink/devl_internal.h | 16 ++++++++++++++++ net/devlink/leftover.c | 36 ++++++++---------------------------- 3 files changed, 25 insertions(+), 29 deletions(-)
net/devlink/core.c +1 -1
@@ -91,7 +91,7 @@ void devlink_put(struct devlink *devlink)call_rcu (&devlink->rcu, __devlink_put_rcu); } -static struct devlink * +struct devlink *devlinks_xa_find_get (struct net *net,unsigned long *indexp,void * (*xa_find_fn)(struct xarray *,unsigned long *,unsigned long ,xa_mark_t ))
net/devlink/devl_internal.h +16 -0
@@ -87,6 +87,10 @@ extern struct genl_family devlink_nl_family; devlink; devlink = devlinks_xa_find_get_next(net, &index))struct devlink * +devlinks_xa_find_get (struct net *net,unsigned long *indexp, +void * (*xa_find_fn)(struct xarray *,unsigned long *, +unsigned long ,xa_mark_t )); +struct devlink *devlinks_xa_find_get_first (struct net *net,unsigned long *indexp);struct devlink *devlinks_xa_find_get_next (struct net *net,unsigned long *indexp);
@@ -104,6 +108,7 @@ enum devlink_multicast_groups {/* state held across netlink dumps */ struct devlink_nl_dump_state { +unsigned long instance;int idx;union {/* DEVLINK_CMD_REGION_READ */
@@ -117,6 +122,17 @@ struct devlink_nl_dump_state { }; }; +/* Iterate over registered devlink instances for devlink dump. +* devlink_put() needs to be called for each iterated devlink pointer +* in loop body in order to release the reference. +* Note: this is NOT a generic iterator, it makes assumptions about the use +* of @state and can only be used once per dumpit implementation. +*/ +#define devlink_dump_for_each_instance_get (msg, state, devlink) \ + for (; (devlink = devlinks_xa_find_get(sock_net(msg->sk), \ + &state->instance, xa_find)); \ + state->instance++) +extern const struct genl_small_ops devlink_nl_ops[56 ];struct devlink *devlink_get_from_attrs (struct net *net,struct nlattr **attrs);
net/devlink/leftover.c +8 -28
@@ -1319,17 +1319,9 @@ static int devlink_nl_cmd_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -unsigned long index; -int idx =0 ;int err; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < state->idx) { - idx++; -devlink_put (devlink); -continue ; - } - +devlink_dump_for_each_instance_get (msg, state, devlink) {devl_lock (devlink); err = devlink_nl_fill(msg, devlink, DEVLINK_CMD_NEW, NETLINK_CB(cb->skb).portid,
@@ -1339,10 +1331,8 @@ static int devlink_nl_cmd_get_dumpit(struct sk_buff *msg,if (err) goto out; - idx++; } out: - state->idx = idx;return msg->len; }
@@ -4872,13 +4862,13 @@ static int devlink_nl_cmd_selftests_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < state->idx || !devlink->ops->selftest_check) - goto inc; +devlink_dump_for_each_instance_get (msg, state, devlink) { +if (!devlink->ops->selftest_check) { +devlink_put (devlink); +continue ; + }devl_lock (devlink); err = devlink_nl_selftests_fill(msg, devlink,
@@ -4890,15 +4880,13 @@ static int devlink_nl_cmd_selftests_get_dumpit(struct sk_buff *msg,devlink_put (devlink);break ; } -inc: - idx++; +devlink_put (devlink); }if (err != -EMSGSIZE )return err; - state->idx = idx;return msg->len; }
@@ -6747,14 +6735,9 @@ static int devlink_nl_cmd_info_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { -if (idx < state->idx) - goto inc; - +devlink_dump_for_each_instance_get (msg, state, devlink) {devl_lock (devlink); err = devlink_nl_info_fill(msg, devlink, DEVLINK_CMD_INFO_GET, NETLINK_CB(cb->skb).portid,
@@ -6767,15 +6750,12 @@ static int devlink_nl_cmd_info_get_dumpit(struct sk_buff *msg,devlink_put (devlink);break ; } -inc: - idx++;devlink_put (devlink); }if (err != -EMSGSIZE )return err; - state->idx = idx;return msg->len; }
Signature · 1 line
2.38.1
Re: [PATCH net-next v2 10/15] devlink: restart dump based on devlink instance ids (simple)
Quoted 11 lines · Thu, Jan 05, 2023 at 05:05:26AM CET, [email protected] wrote:
xarray gives each devlink instance an id and allows us to restart walk based on that id quite neatly. This is nice both from the perspective of code brevity and from the stability of the dump (devlink instances disappearing from before the resumption point will not cause inconsistent dumps).
This patch takes care of simple cases where
state->idxcounts devlink instances only.Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>
[...]
net/devlink/devl_internal.h struct devlink_nl_dump_state { · line 134
Quoted 4 lines
net/devlink/devl_internal.h +4
-0+#define devlink_dump_for_each_instance_get (msg, state, devlink) \ + for (; (devlink = devlinks_xa_find_get(sock_net(msg->sk), \ + &state->instance, xa_find)); \ + state->instance++)
I'm still not happy about this to be honest, but I don't think it matters now as it is cosmetical. I can try to address it in the follow-up patch/set.
Jakub Kicinski
maintainerPATCH v2 11/15 devlink: restart dump based on devlink instance ids (nested)
In reply to Jakub KicinskiUse xarray id for cases of simple sub-object iteration. We'll now use the state->instance for the devlink instances and state->idx for subobject index.
Moving the definition of idx into the inner loop makes sense, so while at it also move other sub-object local variables into the loop.
--- net/devlink/devl_internal.h | 2 +- net/devlink/leftover.c | 93 +++++++++++++++++++------------------ 2 files changed, 49 insertions(+), 46 deletions(-)
net/devlink/devl_internal.h +1 -1
@@ -131,7 +131,7 @@ struct devlink_nl_dump_state {#define devlink_dump_for_each_instance_get (msg, state, devlink) \ for (; (devlink = devlinks_xa_find_get(sock_net(msg->sk), \ &state->instance, xa_find)); \ - state->instance++) + state->instance++, state->idx = 0)extern const struct genl_small_ops devlink_nl_ops[56 ];
net/devlink/leftover.c +48 -45
287 lines not on this page · raw message
Signature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 12/15 devlink: restart dump based on devlink instance ids (function)
In reply to Jakub KicinskiUse xarray id for cases of sub-objects which are iterated in a function.
--- net/devlink/leftover.c | 41 +++++++++++++++++++++-------------------- 1 file changed, 21 insertions(+), 20 deletions(-)
net/devlink/leftover.c +21 -20
@@ -2547,12 +2547,12 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -struct devlink_sb *devlink_sb; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { +devlink_dump_for_each_instance_get (msg, state, devlink) { +struct devlink_sb *devlink_sb; +int idx =0 ; +if (!devlink->ops->sb_pool_get) goto retry;
@@ -2567,6 +2567,7 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg, }else if (err) {devl_unlock (devlink);devlink_put (devlink); + state->idx = idx; goto out; } }
@@ -2578,7 +2579,6 @@ static int devlink_nl_cmd_sb_pool_get_dumpit(struct sk_buff *msg,if (err != -EMSGSIZE )return err; - state->idx = idx;return msg->len; }
@@ -2762,12 +2762,12 @@ static int devlink_nl_cmd_sb_port_pool_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -struct devlink_sb *devlink_sb; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { +devlink_dump_for_each_instance_get (msg, state, devlink) { +struct devlink_sb *devlink_sb; +int idx =0 ; +if (!devlink->ops->sb_port_pool_get) goto retry;
@@ -2782,6 +2782,7 @@ static int devlink_nl_cmd_sb_port_pool_get_dumpit(struct sk_buff *msg, }else if (err) {devl_unlock (devlink);devlink_put (devlink); + state->idx = idx; goto out; } }
@@ -2793,7 +2794,6 @@ static int devlink_nl_cmd_sb_port_pool_get_dumpit(struct sk_buff *msg,if (err != -EMSGSIZE )return err; - state->idx = idx;return msg->len; }
@@ -3005,12 +3005,12 @@ devlink_nl_cmd_sb_tc_pool_bind_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -struct devlink_sb *devlink_sb; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { +devlink_dump_for_each_instance_get (msg, state, devlink) { +struct devlink_sb *devlink_sb; +int idx =0 ; +if (!devlink->ops->sb_tc_pool_bind_get) goto retry;
@@ -3025,6 +3025,7 @@ devlink_nl_cmd_sb_tc_pool_bind_get_dumpit(struct sk_buff *msg, }else if (err) {devl_unlock (devlink);devlink_put (devlink); + state->idx = idx; goto out; } }
@@ -3036,7 +3037,6 @@ devlink_nl_cmd_sb_tc_pool_bind_get_dumpit(struct sk_buff *msg,if (err != -EMSGSIZE )return err; - state->idx = idx;return msg->len; }
@@ -6085,19 +6085,20 @@ static int devlink_nl_cmd_region_get_dumpit(struct sk_buff *msg, {struct devlink_nl_dump_state *state =devlink_dump_state (cb);struct devlink *devlink; -unsigned long index; -int idx =0 ;int err =0 ; -devlinks_xa_for_each_registered_get (sock_net (msg->sk), index, devlink) { +devlink_dump_for_each_instance_get (msg, state, devlink) { +int idx =0 ; + err =devlink_nl_cmd_region_get_devlink_dumpit (msg, cb, devlink, &idx, state->idx);devlink_put (devlink); -if (err) +if (err) { + state->idx = idx; goto out; + } } out: - state->idx = idx;return msg->len; }
Signature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 13/15 devlink: uniformly take the devlink instance lock in the dump loop
In reply to Jakub KicinskiMove the lock taking out of devlink_nl_cmd_region_get_devlink_dumpit(). This way all dumps will take the instance lock in the main iteration loop directly, making refactoring and reading the code easier.
--- net/devlink/leftover.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-)
net/devlink/leftover.c +6 -7
@@ -6050,9 +6050,8 @@ static int devlink_nl_cmd_region_get_devlink_dumpit(struct sk_buff *msg,struct devlink_region *region;struct devlink_port *port;unsigned long port_index; -int err =0 ; +int err; -devl_lock (devlink);list_for_each_entry (region, &devlink->region_list, list) {if (*idx < start) { (*idx)++;
@@ -6064,7 +6063,7 @@ static int devlink_nl_cmd_region_get_devlink_dumpit(struct sk_buff *msg, cb->nlh->nlmsg_seq, NLM_F_MULTI, region);if (err) - goto out; +return err; (*idx)++; }
@@ -6072,12 +6071,10 @@ static int devlink_nl_cmd_region_get_devlink_dumpit(struct sk_buff *msg, err =devlink_nl_cmd_region_get_port_dumpit (msg, cb, port, idx, start);if (err) - goto out; +return err; } -out: -devl_unlock (devlink); -return err; +return 0 ; } static int devlink_nl_cmd_region_get_dumpit(struct sk_buff *msg,
@@ -6090,8 +6087,10 @@ static int devlink_nl_cmd_region_get_dumpit(struct sk_buff *msg,devlink_dump_for_each_instance_get (msg, state, devlink) {int idx =0 ; +devl_lock (devlink); err =devlink_nl_cmd_region_get_devlink_dumpit (msg, cb, devlink, &idx, state->idx); +devl_unlock (devlink);devlink_put (devlink);if (err) { state->idx = idx;
Signature · 1 line
2.38.1
Jakub Kicinski
maintainerPATCH v2 14/15 devlink: add by-instance dump infra
In reply to Jakub KicinskiMost dumpit implementations walk the devlink instances. This requires careful lock taking and reference dropping. Factor the loop out and provide just a callback to handle a single instance dump.
Convert one user as an example, other users converted in the next change.
Slightly inspired by ethtool netlink code.
--- net/devlink/devl_internal.h | 10 +++++++ net/devlink/leftover.c | 55 ++++++++++++++++--------------------- net/devlink/netlink.c | 34 +++++++++++++++++++++++ 3 files changed, 68 insertions(+), 31 deletions(-)
net/devlink/devl_internal.h +10 -0
@@ -122,6 +122,11 @@ struct devlink_nl_dump_state { }; }; +struct devlink_gen_cmd { +int (*dump_one)(struct sk_buff *msg,struct devlink *devlink, +struct netlink_callback *cb); +}; +/* Iterate over registered devlink instances for devlink dump. * devlink_put() needs to be called for each iterated devlink pointer * in loop body in order to release the reference.
@@ -140,6 +145,9 @@ struct devlink *devlink_get_from_attrs(struct net *net, struct nlattr **attrs);void devlink_notify_unregister (struct devlink *devlink);void devlink_notify_register (struct devlink *devlink); +int devlink_nl_instance_iter_dump (struct sk_buff *msg, +struct netlink_callback *cb); +static inline struct devlink_nl_dump_state *devlink_dump_state (struct netlink_callback *cb) {
@@ -175,6 +183,8 @@ devlink_linecard_get_from_info(struct devlink *devlink, struct genl_info *info);void devlink_linecard_put (struct devlink_linecard *linecard);/* Rates */ +extern const struct devlink_gen_cmd devl_gen_rate_get; +struct devlink_rate *devlink_rate_get_from_info (struct devlink *devlink,struct genl_info *info); struct devlink_rate *
net/devlink/leftover.c +24 -31
78 lines not on this page · raw message
net/devlink/netlink.c +34 -0
@@ -5,6 +5,7 @@*/ #include <net/genetlink.h> +#include <net/sock.h> #include "devl_internal.h"
@@ -177,6 +178,39 @@ static void devlink_nl_post_doit(const struct genl_split_ops *ops,devlink_put (devlink); } +static const struct devlink_gen_cmd *devl_gen_cmds[] = { + [DEVLINK_CMD_RATE_GET ] = &devl_gen_rate_get, +}; + +int devlink_nl_instance_iter_dump (struct sk_buff *msg, +struct netlink_callback *cb) +{ +const struct genl_dumpit_info *info =genl_dumpit_info (cb); +struct devlink_nl_dump_state *state =devlink_dump_state (cb); +const struct devlink_gen_cmd *cmd; +struct devlink *devlink; +int err =0 ; + + cmd = devl_gen_cmds[info->op.cmd]; + +devlink_dump_for_each_instance_get (msg, state, devlink) { +devl_lock (devlink); + err = cmd->dump_one (msg, devlink, cb); +devl_unlock (devlink); +devlink_put (devlink); + +if (err) +break ; + +/* restart sub-object walk for the next instance */ + state->idx =0 ; + } + +if (err != -EMSGSIZE ) +return err; +return msg->len; +} +struct genl_family devlink_nl_family__ro_after_init = { .name =DEVLINK_GENL_NAME , .version =DEVLINK_GENL_VERSION ,
Signature · 1 line
2.38.1
Jiri Pirko
maintainerRe: [PATCH net-next v2 14/15] devlink: add by-instance dump infra
Quoted 23 lines · Thu, Jan 05, 2023 at 05:05:30AM CET, [email protected] wrote:
Most dumpit implementations walk the devlink instances. This requires careful lock taking and reference dropping. Factor the loop out and provide just a callback to handle a single instance dump.
Convert one user as an example, other users converted in the next change.
Slightly inspired by ethtool netlink code.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>---
net/devlink/devl_internal.h | 10 +++++++ net/devlink/leftover.c | 55 ++++++++++++++++--------------------- net/devlink/netlink.c | 34 +++++++++++++++++++++++ 3 files changed, 68 insertions(+), 31 deletions(-)net/devlink/devl_internal.h +0
-0
net/devlink/devl_internal.h struct devlink_nl_dump_state { · line 125
Quoted 4 lines
net/devlink/devl_internal.h +1
-0@@ -122,6 +122,11 @@ struct devlink_nl_dump_state { }; }; +struct devlink_gen_cmd {
As I wrote in reply to v1, could this be "genl"?
Quoted 7 lines
net/devlink/devl_internal.h +4
-0+int (*dump_one)(struct sk_buff *msg,struct devlink *devlink, +struct netlink_callback *cb); +}; +/* Iterate over registered devlink instances for devlink dump. * devlink_put() needs to be called for each iterated devlink pointer * in loop body in order to release the reference.
[...]
net/devlink/devl_internal.h const struct genl_small_ops devlink_nl_ops[56] = { · line 9126
Quoted 6 lines
net/devlink/devl_internal.h +1
-1@@ -9130,7 +9123,7 @@ const struct genl_small_ops devlink_nl_ops[56] = { { .cmd = DEVLINK_CMD_RATE_GET, .doit = devlink_nl_cmd_rate_get_doit, - .dumpit = devlink_nl_cmd_rate_get_dumpit, + .dumpit = devlink_nl_instance_iter_dump,
devlink_nl_instance_iter_dumpit to match ".dumpit".
Quoted 3 lines
net/devlink/devl_internal.h +0
-0.internal_flags = DEVLINK_NL_FLAG_NEED_RATE, /* can be retrieved by unprivileged users */ },
[...]
Re: [PATCH net-next v2 14/15] devlink: add by-instance dump infra
Quoted 27 lines · Jiri Pirko
Quoted 26 lines · Thu, Jan 05, 2023 at 05:05:30AM CET, [email protected] wrote:
Most dumpit implementations walk the devlink instances. This requires careful lock taking and reference dropping. Factor the loop out and provide just a callback to handle a single instance dump.
Convert one user as an example, other users converted in the next change.
Slightly inspired by ethtool netlink code.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>---
net/devlink/devl_internal.h | 10 +++++++ net/devlink/leftover.c | 55 ++++++++++++++++--------------------- net/devlink/netlink.c | 34 +++++++++++++++++++++++ 3 files changed, 68 insertions(+), 31 deletions(-)net/devlink/devl_internal.h +0
-0@@ -122,6 +122,11 @@ struct devlink_nl_dump_state { }; };
Quoted 4 lines
net/devlink/devl_internal.h struct devlink_nl_dump_state { · line 125
Quoted 1 line
net/devlink/devl_internal.h +1
-0+struct devlink_gen_cmd {As I wrote in reply to v1, could this be "genl"?
Except Kuba already said this wasn't about "generic netlink" but "generic devlink command" vs "complicated command that can't use the new iterator for whatever reason", so genl feels misleading.
I guess gen is also kind of misleading but I can't think of anything better.
Quoted 27 lines
Quoted 8 lines
net/devlink/devl_internal.h +4
-0+int (*dump_one)(struct sk_buff *msg,struct devlink *devlink, +struct netlink_callback *cb); +}; +/* Iterate over registered devlink instances for devlink dump. * devlink_put() needs to be called for each iterated devlink pointer * in loop body in order to release the reference. [...]
net/devlink/devl_internal.h const struct genl_small_ops devlink_nl_ops[56] = { · line 9126
Quoted 6 lines
net/devlink/devl_internal.h +1
-1@@ -9130,7 +9123,7 @@ const struct genl_small_ops devlink_nl_ops[56] = { { .cmd = DEVLINK_CMD_RATE_GET, .doit = devlink_nl_cmd_rate_get_doit, - .dumpit = devlink_nl_cmd_rate_get_dumpit, + .dumpit = devlink_nl_instance_iter_dump,
devlink_nl_instance_iter_dumpitto match ".dumpit".Quoted 3 lines
net/devlink/devl_internal.h +0
-0.internal_flags = DEVLINK_NL_FLAG_NEED_RATE, /* can be retrieved by unprivileged users */ },[...]
Jiri Pirko
maintainerRe: [PATCH net-next v2 14/15] devlink: add by-instance dump infra
Quoted 30 lines · Fri, Jan 06, 2023 at 01:16:05AM CET, [email protected] wrote:
Quoted 27 lines · Jiri Pirko
Quoted 26 lines · Thu, Jan 05, 2023 at 05:05:30AM CET, [email protected] wrote:
Most dumpit implementations walk the devlink instances. This requires careful lock taking and reference dropping. Factor the loop out and provide just a callback to handle a single instance dump.
Convert one user as an example, other users converted in the next change.
Slightly inspired by ethtool netlink code.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>---
net/devlink/devl_internal.h | 10 +++++++ net/devlink/leftover.c | 55 ++++++++++++++++--------------------- net/devlink/netlink.c | 34 +++++++++++++++++++++++ 3 files changed, 68 insertions(+), 31 deletions(-)net/devlink/devl_internal.h +0
-0@@ -122,6 +122,11 @@ struct devlink_nl_dump_state { }; };
Quoted 7 lines
Quoted 4 lines
net/devlink/devl_internal.h struct devlink_nl_dump_state { · line 125
Quoted 1 line
net/devlink/devl_internal.h +1
-0+struct devlink_gen_cmd {As I wrote in reply to v1, could this be "genl"?
Except Kuba already said this wasn't about "generic netlink" but "generic devlink command" vs "complicated command that can't use the new
Okay, that confuses me. What is supposed to be "generic devlink command"? I don't see anything "generic" about these.
Quoted 31 lines
iterator for whatever reason", so genl feels misleading.
I guess gen is also kind of misleading but I can't think of anything better.
Quoted 27 lines
Quoted 8 lines
net/devlink/devl_internal.h +4
-0+int (*dump_one)(struct sk_buff *msg,struct devlink *devlink, +struct netlink_callback *cb); +}; +/* Iterate over registered devlink instances for devlink dump. * devlink_put() needs to be called for each iterated devlink pointer * in loop body in order to release the reference. [...]
net/devlink/devl_internal.h const struct genl_small_ops devlink_nl_ops[56] = { · line 9126
Quoted 6 lines
net/devlink/devl_internal.h +1
-1@@ -9130,7 +9123,7 @@ const struct genl_small_ops devlink_nl_ops[56] = { { .cmd = DEVLINK_CMD_RATE_GET, .doit = devlink_nl_cmd_rate_get_doit, - .dumpit = devlink_nl_cmd_rate_get_dumpit, + .dumpit = devlink_nl_instance_iter_dump,
devlink_nl_instance_iter_dumpitto match ".dumpit".Quoted 3 lines
net/devlink/devl_internal.h +0
-0.internal_flags = DEVLINK_NL_FLAG_NEED_RATE, /* can be retrieved by unprivileged users */ },[...]
Re: [PATCH net-next v2 14/15] devlink: add by-instance dump infra
Quoted 36 lines · Jiri Pirko
Quoted 35 lines · Fri, Jan 06, 2023 at 01:16:05AM CET, [email protected] wrote:
Quoted 31 lines · Jiri Pirko
Quoted 23 lines · Thu, Jan 05, 2023 at 05:05:30AM CET, [email protected] wrote:
Most dumpit implementations walk the devlink instances. This requires careful lock taking and reference dropping. Factor the loop out and provide just a callback to handle a single instance dump.
Convert one user as an example, other users converted in the next change.
Slightly inspired by ethtool netlink code.
Reviewed-by Jacob Keller <[email protected]>Signed-off-by Jakub Kicinski <[email protected]>---
net/devlink/devl_internal.h | 10 +++++++ net/devlink/leftover.c | 55 ++++++++++++++++--------------------- net/devlink/netlink.c | 34 +++++++++++++++++++++++ 3 files changed, 68 insertions(+), 31 deletions(-)net/devlink/devl_internal.h +0
-0net/devlink/devl_internal.h struct devlink_nl_dump_state { · line 125
Quoted 4 lines
net/devlink/devl_internal.h +1
-0@@ -122,6 +122,11 @@ struct devlink_nl_dump_state { }; }; +struct devlink_gen_cmd {As I wrote in reply to v1, could this be "genl"?
Quoted 6 lines
Quoted 2 lines
Except Kuba already said this wasn't about "generic netlink" but "generic devlink command" vs "complicated command that can't use the new
Okay, that confuses me. What is supposed to be "generic devlink command"? I don't see anything "generic" about these.
I read it as: "uses this new infrastructure for iteration" vs "does not use it". I assume (but haven't double checked) that at least one command still doesn't use this for some reason?
I don't know if there's a simpler name to get that across. I don't think genl is helpful because devlink is always generic netlink.
maybe "devlink_iter_cmd"? but I dunno.
Thanks,
Jake
Jakub Kicinski
maintainerPATCH v2 15/15 devlink: convert remaining dumps to the by-instance scheme
In reply to Jakub KicinskiSoon we'll have to check if a devlink instance is alive after locking it. Convert to the by-instance dumping scheme to make refactoring easier.
Most of the subobject code no longer has to worry about any devlink locking / lifetime rules (the only ones that still do are the two subject types which stubbornly use their own locking). Both dump and do callbacks are given a devlink instance which is already locked and good-to-access (do from the .pre_doit handler, dump from the new dump indirection).
Note that we'll now check presence of an op (e.g. for sb_pool_get) under the devlink instance lock, that will soon be necessary anyway, because we don't hold refs on the driver modules so the memory in which ops live may be gone for a dead instance, after upcoming locking changes.
--- net/devlink/devl_internal.h | 15 + net/devlink/leftover.c | 686 +++++++++++++++--------------------- net/devlink/netlink.c | 13 + 3 files changed, 320 insertions(+), 394 deletions(-)
net/devlink/devl_internal.h +15 -0
@@ -156,6 +156,21 @@ devlink_dump_state(struct netlink_callback *cb)return (struct devlink_nl_dump_state *)cb->ctx; } +/* gen cmds */ +extern const struct devlink_gen_cmd devl_gen_inst; +extern const struct devlink_gen_cmd devl_gen_port; +extern const struct devlink_gen_cmd devl_gen_sb; +extern const struct devlink_gen_cmd devl_gen_sb_pool; +extern const struct devlink_gen_cmd devl_gen_sb_port_pool; +extern const struct devlink_gen_cmd devl_gen_sb_tc_pool_bind; +extern const struct devlink_gen_cmd devl_gen_selftests; +extern const struct devlink_gen_cmd devl_gen_param; +extern const struct devlink_gen_cmd devl_gen_region; +extern const struct devlink_gen_cmd devl_gen_info; +extern const struct devlink_gen_cmd devl_gen_trap; +extern const struct devlink_gen_cmd devl_gen_trap_group; +extern const struct devlink_gen_cmd devl_gen_trap_policer; +/* Ports */ int devlink_port_netdevice_event (struct notifier_block *nb,unsigned long event,void *ptr);
net/devlink/leftover.c +292 -394
957 lines not on this page · raw message
net/devlink/netlink.c +13 -0
@@ -179,7 +179,20 @@ static void devlink_nl_post_doit(const struct genl_split_ops *ops, }static const struct devlink_gen_cmd *devl_gen_cmds[] = { + [DEVLINK_CMD_GET ] = &devl_gen_inst, + [DEVLINK_CMD_PORT_GET ] = &devl_gen_port, + [DEVLINK_CMD_SB_GET ] = &devl_gen_sb, + [DEVLINK_CMD_SB_POOL_GET ] = &devl_gen_sb_pool, + [DEVLINK_CMD_SB_PORT_POOL_GET ] = &devl_gen_sb_port_pool, + [DEVLINK_CMD_SB_TC_POOL_BIND_GET ] = &devl_gen_sb_tc_pool_bind, + [DEVLINK_CMD_PARAM_GET ] = &devl_gen_param, + [DEVLINK_CMD_REGION_GET ] = &devl_gen_region, + [DEVLINK_CMD_INFO_GET ] = &devl_gen_info, [DEVLINK_CMD_RATE_GET ] = &devl_gen_rate_get, + [DEVLINK_CMD_TRAP_GET ] = &devl_gen_trap, + [DEVLINK_CMD_TRAP_GROUP_GET ] = &devl_gen_trap_group, + [DEVLINK_CMD_TRAP_POLICER_GET ] = &devl_gen_trap_policer, + [DEVLINK_CMD_SELFTESTS_GET ] = &devl_gen_selftests, }; int devlink_nl_instance_iter_dump(struct sk_buff *msg,
Signature · 1 line
2.38.1
This series was applied to netdev/net-next.git (master) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (master) by Jakub Kicinski <[email protected]>:
Quoted 11 lines · On Wed, 4 Jan 2023 20:05:16 -0800 you wrote:
Split
devlink.cinto a handful of files, trying to keep the "core" code away from all the command-specific implementations. The core code has been quite scattered until now. Going forward we can consider using a source file per-subobject, I think that it's quite beneficial to newcomers (based on relative ease with which folks contribute to ethtool vs devlink). But this series doesn't split everything out, yet - partially due to backporting concerns, but mostly due to lack of time. Bulk of the netlink command handling is left in aleftover.cfile.[...]
Here is the summary with links:
- [net-next,v2,01/15] devlink: move code to a dedicated directory
(no matching commit)
- [net-next,v2,02/15] devlink: rename devlink_netdevice_event -> devlink_port_netdevice_event
https://git.kernel.org/netdev/net-next/c/e50ef40f9a9a - [net-next,v2,03/15] devlink: split out core code (no matching commit) - [net-next,v2,04/15] devlink: split out netlink code https://git.kernel.org/netdev/net-next/c/623cd13b1654 - [net-next,v2,05/15] netlink: add macro for checking dump ctx size https://git.kernel.org/netdev/net-next/c/2c7bc10d0f7b - [net-next,v2,06/15] devlink: use an explicit structure for dump context https://git.kernel.org/netdev/net-next/c/3015f8224961 - [net-next,v2,07/15] devlink: remove start variables from dumps https://git.kernel.org/netdev/net-next/c/20615659b514 - [net-next,v2,08/15] devlink: drop the filter argument from devlinks_xa_find_get https://git.kernel.org/netdev/net-next/c/8861c0933c78 - [net-next,v2,09/15] devlink: health: combine loops in dump https://git.kernel.org/netdev/net-next/c/a0e13dfdc391 - [net-next,v2,10/15] devlink: restart dump based on devlink instance ids (simple) https://git.kernel.org/netdev/net-next/c/731d69a6bd13 - [net-next,v2,11/15] devlink: restart dump based on devlink instance ids (nested) https://git.kernel.org/netdev/net-next/c/a8f947073f4a - [net-next,v2,12/15] devlink: restart dump based on devlink instance ids (function) https://git.kernel.org/netdev/net-next/c/c9666bac537e - [net-next,v2,13/15] devlink: uniformly take the devlink instance lock in the dump loop https://git.kernel.org/netdev/net-next/c/e4d5015bc11b - [net-next,v2,14/15] devlink: add by-instance dump infra https://git.kernel.org/netdev/net-next/c/07f3af66089e - [net-next,v2,15/15] devlink: convert remaining dumps to the by-instance scheme https://git.kernel.org/netdev/net-next/c/5ce76d78b996
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html