[RFC PATCH 33/39] ipv6: move address sysctl context to priv

[wen.yang@linux.dev]

From: Wen Yang <wen.yang@linux.dev>

addrconf's handlers need the inet6_dev and the netns to find the devconf
field and to notify after a change. The table carries them in extra1 and
extra2, next to the numeric bounds in the same two fields, so every
handler has to know which meaning applies.

Carry the pair in priv instead, leave extra1/extra2 to the bounds, and
let the handlers use data as the devconf field directly. Registration
only fills priv, so it can no longer overwrite a value pointer or a
range.

Signed-off-by: Wen Yang <wen.yang@linux.dev>
---
 net/ipv6/addrconf.c | 163 +++++++++++++++++++++++++++-----------------
 1 file changed, 99 insertions(+), 64 deletions(-)

diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index c8c0dd04982d..27eaa95f1f21 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -806,6 +806,12 @@ static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 }
 
 #ifdef CONFIG_SYSCTL
+/* Owned by the sysctl allocation; idev (if any) and net outlive it. */
+struct addrconf_sysctl_ctx {
+	struct inet6_dev *idev;
+	struct net *net;
+};
+
 static void dev_forward_change(struct inet6_dev *idev)
 {
 	struct net_device *dev;
@@ -873,9 +879,10 @@ static void addrconf_forward_change(struct net *net, __s32 newf)
 	}
 }
 
-static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
+static int addrconf_fixup_forwarding(const struct addrconf_sysctl_ctx *ctx,
+				     int *p, int newf)
 {
-	struct net *net = (struct net *)table->extra2;
+	struct net *net = ctx->net;
 	int old;
 
 	if (!rtnl_net_trylock(net))
@@ -911,7 +918,7 @@ static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int
 						     NETCONFA_IFINDEX_ALL,
 						     net->ipv6.devconf_all);
 	} else if ((!newf) ^ (!old))
-		dev_forward_change((struct inet6_dev *)table->extra1);
+		dev_forward_change(ctx->idev);
 	rtnl_net_unlock(net);
 
 	if (newf)
@@ -940,9 +947,10 @@ static void addrconf_linkdown_change(struct net *net, __s32 newf)
 	}
 }
 
-static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
+static int addrconf_fixup_linkdown(const struct addrconf_sysctl_ctx *ctx,
+				   int *p, int newf)
 {
-	struct net *net = (struct net *)table->extra2;
+	struct net *net = ctx->net;
 	int old;
 
 	if (!rtnl_net_trylock(net))
@@ -969,7 +977,7 @@ static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int ne
 						     net->ipv6.devconf_all);
 	} else {
 		if (!newf ^ !old) {
-			struct inet6_dev *idev = table->extra1;
+			struct inet6_dev *idev = ctx->idev;
 
 			inet6_netconf_notify_devconf(net,
 						     RTM_NEWNETCONF,
@@ -6389,6 +6397,7 @@ static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
 		void *buffer, size_t *lenp, loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	struct ctl_table lctl;
 	int *valp = ctl->data;
 	int val = *valp;
@@ -6406,20 +6415,20 @@ static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
 		return ret;
 
 	if (write)
-		ret = addrconf_fixup_forwarding(ctl, valp, val);
+		ret = addrconf_fixup_forwarding(ctx, valp, val);
 	return ret;
 }
 
 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
 		void *buffer, size_t *lenp, loff_t *ppos)
 {
-	struct inet6_dev *idev = ctl->extra1;
-	int min_mtu = IPV6_MIN_MTU;
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
+	struct inet6_dev *idev = ctx->idev;
 	struct ctl_table lctl;
 
 	lctl = *ctl;
-	lctl.extra1 = &min_mtu;
-	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
+	lctl.min = IPV6_MIN_MTU;
+	lctl.max = idev ? READ_ONCE(idev->dev->mtu) : INT_MAX;
 
 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
 }
@@ -6455,9 +6464,10 @@ static void addrconf_disable_change(struct net *net, __s32 newf)
 	}
 }
 
-static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
+static int addrconf_disable_ipv6(const struct addrconf_sysctl_ctx *ctx,
+				 int *p, int newf)
 {
-	struct net *net = (struct net *)table->extra2;
+	struct net *net = ctx->net;
 	int old;
 
 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
@@ -6475,7 +6485,7 @@ static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf
 		WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
 		addrconf_disable_change(net, newf);
 	} else if ((!newf) ^ (!old)) {
-		dev_disable_change((struct inet6_dev *)table->extra1);
+		dev_disable_change(ctx->idev);
 	}
 
 	rtnl_net_unlock(net);
@@ -6485,6 +6495,7 @@ static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf
 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
 		void *buffer, size_t *lenp, loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	struct ctl_table lctl;
 	int *valp = ctl->data;
 	int val = *valp;
@@ -6502,14 +6513,15 @@ static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
 		return ret;
 
 	if (write)
-		ret = addrconf_disable_ipv6(ctl, valp, val);
+		ret = addrconf_disable_ipv6(ctx, valp, val);
 	return ret;
 }
 
 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
 		void *buffer, size_t *lenp, loff_t *ppos)
 {
-	struct net *net = ctl->extra2;
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
+	struct net *net = ctx->net;
 	int *valp = ctl->data;
 	int old, new;
 	int ret;
@@ -6533,7 +6545,7 @@ static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
 						     NETCONFA_IFINDEX_ALL,
 						     net->ipv6.devconf_all);
 		} else {
-			struct inet6_dev *idev = ctl->extra1;
+			struct inet6_dev *idev = ctx->idev;
 
 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 						     NETCONFA_PROXY_NEIGH,
@@ -6551,10 +6563,11 @@ static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
 					 void *buffer, size_t *lenp,
 					 loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	int ret = 0;
 	u32 new_val;
-	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
-	struct net *net = (struct net *)ctl->extra2;
+	struct inet6_dev *idev = ctx->idev;
+	struct net *net = ctx->net;
 	struct ctl_table tmp = {
 		.data = &new_val,
 		.maxlen = sizeof(new_val),
@@ -6618,16 +6631,18 @@ static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
 					 void *buffer, size_t *lenp,
 					 loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	int err;
 	struct in6_addr addr;
 	char str[IPV6_MAX_STRLEN];
-	struct ctl_table lctl = *ctl;
-	struct net *net = ctl->extra2;
+	struct ctl_table lctl;
+	struct net *net = ctx->net;
 	struct ipv6_stable_secret *secret = ctl->data;
 
 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
 		return -EIO;
 
+	lctl = *ctl;
 	lctl.maxlen = IPV6_MAX_STRLEN;
 	lctl.data = str;
 
@@ -6669,7 +6684,7 @@ static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
 			}
 		}
 	} else {
-		struct inet6_dev *idev = ctl->extra1;
+		struct inet6_dev *idev = ctx->idev;
 
 		WRITE_ONCE(idev->cnf.addr_gen_mode,
 			   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
@@ -6687,6 +6702,7 @@ int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
 						size_t *lenp,
 						loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	struct ctl_table lctl;
 	int *valp = ctl->data;
 	int val = *valp;
@@ -6703,7 +6719,7 @@ int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
 		return ret;
 
 	if (write)
-		ret = addrconf_fixup_linkdown(ctl, valp, val);
+		ret = addrconf_fixup_linkdown(ctx, valp, val);
 	return ret;
 }
 
@@ -6749,9 +6765,10 @@ void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
 }
 
 static
-int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
+int addrconf_disable_policy(const struct addrconf_sysctl_ctx *ctx,
+			    int *valp, int val)
 {
-	struct net *net = (struct net *)ctl->extra2;
+	struct net *net = ctx->net;
 	struct inet6_dev *idev;
 
 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
@@ -6773,7 +6790,7 @@ int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
 				addrconf_disable_policy_idev(idev, val);
 		}
 	} else {
-		idev = (struct inet6_dev *)ctl->extra1;
+		idev = ctx->idev;
 		addrconf_disable_policy_idev(idev, val);
 	}
 
@@ -6784,6 +6801,7 @@ int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
 				   void *buffer, size_t *lenp, loff_t *ppos)
 {
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
 	struct ctl_table lctl;
 	int *valp = ctl->data;
 	int val = *valp;
@@ -6796,7 +6814,7 @@ static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write
 		return ret;
 
 	if (write && (*valp != val))
-		ret = addrconf_disable_policy(ctl, valp, val);
+		ret = addrconf_disable_policy(ctx, valp, val);
 
 	return ret;
 }
@@ -6823,16 +6841,17 @@ static void addrconf_force_forward_change(struct net *net, __s32 newf)
 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
 					    void *buffer, size_t *lenp, loff_t *ppos)
 {
-	struct inet6_dev *idev = ctl->extra1;
+	const struct addrconf_sysctl_ctx *ctx = ctl->priv;
+	struct inet6_dev *idev = ctx->idev;
 	struct ctl_table tmp_ctl = *ctl;
-	struct net *net = ctl->extra2;
+	struct net *net = ctx->net;
 	int *valp = ctl->data;
 	int new_val = *valp;
 	int old_val = *valp;
 	int ret;
 
-	tmp_ctl.extra1 = SYSCTL_ZERO;
-	tmp_ctl.extra2 = SYSCTL_ONE;
+	tmp_ctl.min = 0;
+	tmp_ctl.max = 1;
 	tmp_ctl.data = &new_val;
 
 	ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
@@ -6867,10 +6886,6 @@ static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int wri
 	return ret;
 }
 
-static int minus_one = -1;
-static const int two_five_five = 255;
-static u32 ioam6_if_id_max = U16_MAX;
-
 static const struct ctl_table addrconf_sysctl[] = {
 	{
 		.procname	= "forwarding",
@@ -6885,8 +6900,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= (void *)SYSCTL_ONE,
-		.extra2		= (void *)&two_five_five,
+		.min		= 1,
+		.max		= 255,
 	},
 	{
 		.procname	= "mtu",
@@ -6929,7 +6944,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= &minus_one,
+		.min		= (unsigned long)(long)-1,
+		.max		= INT_MAX,
 	},
 	{
 		.procname	= "router_solicitation_interval",
@@ -7037,7 +7053,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u32),
 		.mode		= 0644,
 		.proc_handler	= proc_douintvec_minmax,
-		.extra1		= (void *)SYSCTL_ONE,
+		.min		= 1,
+		.max		= UINT_MAX,
 	},
 	{
 		.procname	= "accept_ra_min_hop_limit",
@@ -7066,8 +7083,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u8),
 		.mode		= 0644,
 		.proc_handler	= proc_dou8vec_minmax,
-		.extra1		= SYSCTL_ZERO,
-		.extra2		= SYSCTL_ONE,
+		.min		= 0,
+		.max		= 1,
 	},
 	{
 		.procname	= "ra_honor_pio_pflag",
@@ -7075,8 +7092,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u8),
 		.mode		= 0644,
 		.proc_handler	= proc_dou8vec_minmax,
-		.extra1		= SYSCTL_ZERO,
-		.extra2		= SYSCTL_ONE,
+		.min		= 0,
+		.max		= 1,
 	},
 #ifdef CONFIG_IPV6_ROUTER_PREF
 	{
@@ -7284,8 +7301,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= (void *)SYSCTL_ZERO,
-		.extra2		= (void *)&two_five_five,
+		.min		= 0,
+		.max		= 255,
 	},
 	{
 		.procname	= "rpl_seg_enabled",
@@ -7293,8 +7310,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler   = proc_dointvec_minmax,
-		.extra1         = SYSCTL_ZERO,
-		.extra2         = SYSCTL_ONE,
+		.min		= 0,
+		.max		= 1,
 	},
 	{
 		.procname	= "ioam6_enabled",
@@ -7302,8 +7319,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u8),
 		.mode		= 0644,
 		.proc_handler	= proc_dou8vec_minmax,
-		.extra1		= (void *)SYSCTL_ZERO,
-		.extra2		= (void *)SYSCTL_ONE,
+		.min		= 0,
+		.max		= 1,
 	},
 	{
 		.procname	= "ioam6_id",
@@ -7311,8 +7328,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u32),
 		.mode		= 0644,
 		.proc_handler	= proc_douintvec_minmax,
-		.extra1		= (void *)SYSCTL_ZERO,
-		.extra2		= (void *)&ioam6_if_id_max,
+		.min		= 0,
+		.max		= U16_MAX,
 	},
 	{
 		.procname	= "ioam6_id_wide",
@@ -7327,8 +7344,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(u8),
 		.mode		= 0644,
 		.proc_handler	= proc_dou8vec_minmax,
-		.extra1		= (void *)SYSCTL_ZERO,
-		.extra2		= (void *)SYSCTL_ONE,
+		.min		= 0,
+		.max		= 1,
 	},
 	{
 		.procname	= "accept_untracked_na",
@@ -7336,8 +7353,8 @@ static const struct ctl_table addrconf_sysctl[] = {
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_minmax,
-		.extra1		= SYSCTL_ZERO,
-		.extra2		= SYSCTL_TWO,
+		.min		= 0,
+		.max		= 2,
 	},
 	{
 		.procname	= "force_forwarding",
@@ -7348,27 +7365,45 @@ static const struct ctl_table addrconf_sysctl[] = {
 	},
 };
 
+struct addrconf_sysctl_table {
+	struct ctl_table table[ARRAY_SIZE(addrconf_sysctl)];
+	struct addrconf_sysctl_ctx contexts[ARRAY_SIZE(addrconf_sysctl)];
+};
+
+static bool addrconf_sysctl_uses_context(proc_handler *handler)
+{
+	return handler == addrconf_sysctl_forward ||
+	       handler == addrconf_sysctl_mtu ||
+	       handler == addrconf_sysctl_disable ||
+	       handler == addrconf_sysctl_proxy_ndp ||
+	       handler == addrconf_sysctl_addr_gen_mode ||
+	       handler == addrconf_sysctl_stable_secret ||
+	       handler == addrconf_sysctl_ignore_routes_with_linkdown ||
+	       handler == addrconf_sysctl_disable_policy ||
+	       handler == addrconf_sysctl_force_forwarding;
+}
+
 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
 		struct inet6_dev *idev, struct ipv6_devconf *p)
 {
 	size_t table_size = ARRAY_SIZE(addrconf_sysctl);
 	int i, ifindex;
+	struct addrconf_sysctl_table *t;
 	struct ctl_table *table;
 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
 
-	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
-	if (!table)
+	t = kmalloc(sizeof(*t), GFP_KERNEL_ACCOUNT);
+	if (!t)
 		goto out;
 
+	table = t->table;
+	memcpy(table, addrconf_sysctl, sizeof(addrconf_sysctl));
 	for (i = 0; i < table_size; i++) {
 		table[i].data += (char *)p - (char *)&ipv6_devconf;
-		/* If one of these is already set, then it is not safe to
-		 * overwrite either of them: this makes proc_dointvec_minmax
-		 * usable.
-		 */
-		if (!table[i].extra1 && !table[i].extra2) {
-			table[i].extra1 = idev; /* embedded; no ref */
-			table[i].extra2 = net;
+		if (addrconf_sysctl_uses_context(table[i].proc_handler)) {
+			t->contexts[i].idev = idev;
+			t->contexts[i].net = net;
+			table[i].priv = &t->contexts[i];
 		}
 	}
 
-- 
2.25.1