diff mbox series

[ethtool-next,v2,5/6] netlink: use policy dumping to check if stats flag is supported

Message ID 20201006150425.2631432-6-kuba@kernel.org
State New
Headers show
Series pause frame stats | expand

Commit Message

Jakub Kicinski Oct. 6, 2020, 3:04 p.m. UTC
Older kernels don't support statistics, to avoid retries
make use of netlink policy dumps to figure out which
flags kernel actually supports.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 netlink/msgbuff.h |   6 ++
 netlink/netlink.c | 151 ++++++++++++++++++++++++++++++++++++++++++++++
 netlink/netlink.h |   2 +
 3 files changed, 159 insertions(+)

Comments

Ido Schimmel Oct. 6, 2020, 7:39 p.m. UTC | #1
On Tue, Oct 06, 2020 at 08:04:24AM -0700, Jakub Kicinski wrote:
> Older kernels don't support statistics, to avoid retries

> make use of netlink policy dumps to figure out which

> flags kernel actually supports.


Thanks for working on this, Jakub.

Michal, should I try something similar for the legacy flag we have been
discussing earlier this week [1]?

[1] https://lore.kernel.org/netdev/20201005074600.xkbomksbbuliuyft@lion.mk-sys.cz/#t
Michal Kubecek Oct. 6, 2020, 8 p.m. UTC | #2
On Tue, Oct 06, 2020 at 10:39:13PM +0300, Ido Schimmel wrote:
> On Tue, Oct 06, 2020 at 08:04:24AM -0700, Jakub Kicinski wrote:
> > Older kernels don't support statistics, to avoid retries
> > make use of netlink policy dumps to figure out which
> > flags kernel actually supports.
> 
> Thanks for working on this, Jakub.
> 
> Michal, should I try something similar for the legacy flag we have been
> discussing earlier this week [1]?
> 
> [1] https://lore.kernel.org/netdev/20201005074600.xkbomksbbuliuyft@lion.mk-sys.cz/#t

Yes, using the policy dump will be the way to go. I wanted to come with
something myself but we have a virtual conference this week and today
an urgent request took the spare time I had in the morning.

Michal
Michal Kubecek Oct. 12, 2020, 3:44 p.m. UTC | #3
On Tue, Oct 06, 2020 at 08:04:24AM -0700, Jakub Kicinski wrote:
> Older kernels don't support statistics, to avoid retries

> make use of netlink policy dumps to figure out which

> flags kernel actually supports.

> 

> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

> ---

[...]
> +static int family_policy_cb(const struct nlmsghdr *nlhdr, void *data)

> +{

> +	const struct nlattr *tba[NL_POLICY_TYPE_ATTR_MAX + 1] = {};

> +	DECLARE_ATTR_TB_INFO(tba);

> +	const struct nlattr *tb[CTRL_ATTR_MAX + 1] = {};

> +	DECLARE_ATTR_TB_INFO(tb);

> +	const struct nlattr *policy_attr, *attr_attr, *attr;

> +	struct ethtool_op_policy_query_ctx *ctx = data;


I would rather use a different name for this variable as "ctx" is used
for struct cmd_context everywhere else.

> +	unsigned int attr_idx, policy_idx;

> +	int ret;


[...]
> +static int get_flags_policy(struct nl_context *nlctx, struct nl_socket *nlsk,

> +			    unsigned int nlcmd, unsigned int hdrattr)

> +{

> +	struct nl_msg_buff *msgbuff = &nlsk->msgbuff;

> +	struct ethtool_op_policy_query_ctx ctx;


Same here.

> +	int ret;

> +

> +	memset(&ctx, 0, sizeof(ctx));

> +	ctx.nlctx = nlctx;

> +	ctx.op = nlcmd;

> +	ctx.op_hdr_attr = hdrattr;

> +

> +	ret = __msg_init(msgbuff, GENL_ID_CTRL, CTRL_CMD_GETPOLICY,

> +			 NLM_F_REQUEST | NLM_F_ACK | NLM_F_DUMP, 1);

> +	if (ret < 0)

> +		return ret;

> +	ret = -EMSGSIZE;

> +	if (ethnla_put_u16(msgbuff, CTRL_ATTR_FAMILY_ID, nlctx->ethnl_fam))

> +		return ret;

> +	if (ethnla_put_u32(msgbuff, CTRL_ATTR_OP, nlcmd))

> +		return ret;

> +

> +	nlsock_sendmsg(nlsk, NULL);

> +	nlsock_process_reply(nlsk, family_policy_cb, &ctx);

> +

> +	ret = ctx.flag_mask;

> +	return ret;

> +}


The return value is assigned either a negative error code or a u32 flag
mask. If we run this code on a kernel supporting 32 flags, flag in bit
31 would collide with sign and successfully retrieved flag mask would be
interpreted as an error by caller.

Other than this, the series looks good to me.

Michal
diff mbox series

Patch

diff --git a/netlink/msgbuff.h b/netlink/msgbuff.h
index 24b99c5a28d7..7d6731fc24a3 100644
--- a/netlink/msgbuff.h
+++ b/netlink/msgbuff.h
@@ -81,6 +81,12 @@  static inline bool ethnla_put_u32(struct nl_msg_buff *msgbuff, uint16_t type,
 	return ethnla_put(msgbuff, type, sizeof(uint32_t), &data);
 }
 
+static inline bool ethnla_put_u16(struct nl_msg_buff *msgbuff, uint16_t type,
+				  uint16_t data)
+{
+	return ethnla_put(msgbuff, type, sizeof(uint16_t), &data);
+}
+
 static inline bool ethnla_put_u8(struct nl_msg_buff *msgbuff, uint16_t type,
 				 uint8_t data)
 {
diff --git a/netlink/netlink.c b/netlink/netlink.c
index e42d57076a4b..f79da12d3eee 100644
--- a/netlink/netlink.c
+++ b/netlink/netlink.c
@@ -135,6 +135,157 @@  bool netlink_cmd_check(struct cmd_context *ctx, unsigned int cmd,
 	return !(nlctx->ops_flags[cmd] & cap);
 }
 
+struct ethtool_op_policy_query_ctx {
+	struct nl_context *nlctx;
+	unsigned int op;
+	unsigned int op_hdr_attr;
+
+	bool op_policy_found;
+	bool hdr_policy_found;
+	unsigned int op_policy_idx;
+	unsigned int hdr_policy_idx;
+	uint64_t flag_mask;
+};
+
+static int family_policy_find_op(struct ethtool_op_policy_query_ctx *ctx,
+				 const struct nlattr *op_policy)
+{
+	const struct nlattr *attr;
+	unsigned int type;
+	int ret;
+
+	type = ctx->nlctx->is_dump ?
+		CTRL_ATTR_POLICY_DUMP : CTRL_ATTR_POLICY_DO;
+
+	mnl_attr_for_each_nested(attr, op_policy) {
+		const struct nlattr *tb[CTRL_ATTR_POLICY_DUMP_MAX + 1] = {};
+		DECLARE_ATTR_TB_INFO(tb);
+
+		if (mnl_attr_get_type(attr) != ctx->op)
+			continue;
+
+		ret = mnl_attr_parse_nested(attr, attr_cb, &tb_info);
+		if (ret < 0)
+			return ret;
+
+		if (!tb[type])
+			continue;
+
+		ctx->op_policy_found = true;
+		ctx->op_policy_idx = mnl_attr_get_u32(tb[type]);
+		break;
+	}
+
+	return 0;
+}
+
+static int family_policy_cb(const struct nlmsghdr *nlhdr, void *data)
+{
+	const struct nlattr *tba[NL_POLICY_TYPE_ATTR_MAX + 1] = {};
+	DECLARE_ATTR_TB_INFO(tba);
+	const struct nlattr *tb[CTRL_ATTR_MAX + 1] = {};
+	DECLARE_ATTR_TB_INFO(tb);
+	const struct nlattr *policy_attr, *attr_attr, *attr;
+	struct ethtool_op_policy_query_ctx *ctx = data;
+	unsigned int attr_idx, policy_idx;
+	int ret;
+
+	ret = mnl_attr_parse(nlhdr, GENL_HDRLEN, attr_cb, &tb_info);
+	if (ret < 0)
+		return MNL_CB_ERROR;
+
+	if (!ctx->op_policy_found) {
+		if (!tb[CTRL_ATTR_OP_POLICY]) {
+			fprintf(stderr, "Error: op policy map not present\n");
+			return MNL_CB_ERROR;
+		}
+		ret = family_policy_find_op(ctx, tb[CTRL_ATTR_OP_POLICY]);
+		return ret < 0 ? MNL_CB_ERROR : MNL_CB_OK;
+	}
+
+	if (!tb[CTRL_ATTR_POLICY])
+		return MNL_CB_OK;
+
+	policy_attr = mnl_attr_get_payload(tb[CTRL_ATTR_POLICY]);
+	policy_idx = mnl_attr_get_type(policy_attr);
+	attr_attr = mnl_attr_get_payload(policy_attr);
+	attr_idx = mnl_attr_get_type(attr_attr);
+
+	ret = mnl_attr_parse_nested(attr_attr, attr_cb, &tba_info);
+	if (ret < 0)
+		return MNL_CB_ERROR;
+
+	if (policy_idx == ctx->op_policy_idx && attr_idx == ctx->op_hdr_attr) {
+		attr = tba[NL_POLICY_TYPE_ATTR_POLICY_IDX];
+		if (!attr) {
+			fprintf(stderr,	"Error: no policy index in what was expected to be ethtool header attribute\n");
+			return MNL_CB_ERROR;
+		}
+		ctx->hdr_policy_found = true;
+		ctx->hdr_policy_idx = mnl_attr_get_u32(attr);
+	}
+
+	if (ctx->hdr_policy_found && ctx->hdr_policy_idx == policy_idx &&
+	    attr_idx == ETHTOOL_A_HEADER_FLAGS) {
+		attr = tba[NL_POLICY_TYPE_ATTR_MASK];
+		if (!attr) {
+			fprintf(stderr,	"Error: validation mask not reported for ethtool header flags\n");
+			return MNL_CB_ERROR;
+		}
+
+		ctx->flag_mask = mnl_attr_get_u64(attr);
+	}
+
+	return MNL_CB_OK;
+}
+
+static int get_flags_policy(struct nl_context *nlctx, struct nl_socket *nlsk,
+			    unsigned int nlcmd, unsigned int hdrattr)
+{
+	struct nl_msg_buff *msgbuff = &nlsk->msgbuff;
+	struct ethtool_op_policy_query_ctx ctx;
+	int ret;
+
+	memset(&ctx, 0, sizeof(ctx));
+	ctx.nlctx = nlctx;
+	ctx.op = nlcmd;
+	ctx.op_hdr_attr = hdrattr;
+
+	ret = __msg_init(msgbuff, GENL_ID_CTRL, CTRL_CMD_GETPOLICY,
+			 NLM_F_REQUEST | NLM_F_ACK | NLM_F_DUMP, 1);
+	if (ret < 0)
+		return ret;
+	ret = -EMSGSIZE;
+	if (ethnla_put_u16(msgbuff, CTRL_ATTR_FAMILY_ID, nlctx->ethnl_fam))
+		return ret;
+	if (ethnla_put_u32(msgbuff, CTRL_ATTR_OP, nlcmd))
+		return ret;
+
+	nlsock_sendmsg(nlsk, NULL);
+	nlsock_process_reply(nlsk, family_policy_cb, &ctx);
+
+	ret = ctx.flag_mask;
+	return ret;
+}
+
+u32 get_stats_flag(struct nl_context *nlctx, unsigned int nlcmd,
+		   unsigned int hdrattr)
+{
+	int ret;
+
+	if (!nlctx->ctx->show_stats)
+		return 0;
+	if (nlcmd > ETHTOOL_MSG_USER_MAX ||
+	    !(nlctx->ops_flags[nlcmd] & GENL_CMD_CAP_HASPOL))
+		return 0;
+
+	ret = get_flags_policy(nlctx, nlctx->ethnl_socket, nlcmd, hdrattr);
+	if (ret < 0)
+		return 0;
+
+	return ret & ETHTOOL_FLAG_STATS;
+}
+
 /* initialization */
 
 static int genl_read_ops(struct nl_context *nlctx,
diff --git a/netlink/netlink.h b/netlink/netlink.h
index 1012e8e32cd8..799f2556cb31 100644
--- a/netlink/netlink.h
+++ b/netlink/netlink.h
@@ -66,6 +66,8 @@  bool netlink_cmd_check(struct cmd_context *ctx, unsigned int cmd,
 		       bool allow_wildcard);
 const char *get_dev_name(const struct nlattr *nest);
 int get_dev_info(const struct nlattr *nest, int *ifindex, char *ifname);
+u32 get_stats_flag(struct nl_context *nlctx, unsigned int nlcmd,
+		   unsigned int hdrattr);
 
 int linkmodes_reply_cb(const struct nlmsghdr *nlhdr, void *data);
 int linkinfo_reply_cb(const struct nlmsghdr *nlhdr, void *data);