dpll: fix possible deadlock during netlink dump operation

Recently, I've been hitting following deadlock warning during dpll pin
dump:

[52804.637962] ======================================================
[52804.638536] WARNING: possible circular locking dependency detected
[52804.639111] 6.8.0-rc2jiri+ #1 Not tainted
[52804.639529] ------------------------------------------------------
[52804.640104] python3/2984 is trying to acquire lock:
[52804.640581] ffff88810e642678 (nlk_cb_mutex-GENERIC){+.+.}-{3:3}, at: netlink_dump+0xb3/0x780
[52804.641417]
but task is already holding lock:
[52804.642010] ffffffff83bde4c8 (dpll_lock){+.+.}-{3:3}, at: dpll_lock_dumpit+0x13/0x20
[52804.642747]
which lock already depends on the new lock.

[52804.643551]
the existing dependency chain (in reverse order) is:
[52804.644259]
-> #1 (dpll_lock){+.+.}-{3:3}:
[52804.644836] lock_acquire+0x174/0x3e0
[52804.645271] __mutex_lock+0x119/0x1150
[52804.645723] dpll_lock_dumpit+0x13/0x20
[52804.646169] genl_start+0x266/0x320
[52804.646578] __netlink_dump_start+0x321/0x450
[52804.647056] genl_family_rcv_msg_dumpit+0x155/0x1e0
[52804.647575] genl_rcv_msg+0x1ed/0x3b0
[52804.648001] netlink_rcv_skb+0xdc/0x210
[52804.648440] genl_rcv+0x24/0x40
[52804.648831] netlink_unicast+0x2f1/0x490
[52804.649290] netlink_sendmsg+0x36d/0x660
[52804.649742] __sock_sendmsg+0x73/0xc0
[52804.650165] __sys_sendto+0x184/0x210
[52804.650597] __x64_sys_sendto+0x72/0x80
[52804.651045] do_syscall_64+0x6f/0x140
[52804.651474] entry_SYSCALL_64_after_hwframe+0x46/0x4e
[52804.652001]
-> #0 (nlk_cb_mutex-GENERIC){+.+.}-{3:3}:
[52804.652650] check_prev_add+0x1ae/0x1280
[52804.653107] __lock_acquire+0x1ed3/0x29a0
[52804.653559] lock_acquire+0x174/0x3e0
[52804.653984] __mutex_lock+0x119/0x1150
[52804.654423] netlink_dump+0xb3/0x780
[52804.654845] __netlink_dump_start+0x389/0x450
[52804.655321] genl_family_rcv_msg_dumpit+0x155/0x1e0
[52804.655842] genl_rcv_msg+0x1ed/0x3b0
[52804.656272] netlink_rcv_skb+0xdc/0x210
[52804.656721] genl_rcv+0x24/0x40
[52804.657119] netlink_unicast+0x2f1/0x490
[52804.657570] netlink_sendmsg+0x36d/0x660
[52804.658022] __sock_sendmsg+0x73/0xc0
[52804.658450] __sys_sendto+0x184/0x210
[52804.658877] __x64_sys_sendto+0x72/0x80
[52804.659322] do_syscall_64+0x6f/0x140
[52804.659752] entry_SYSCALL_64_after_hwframe+0x46/0x4e
[52804.660281]
other info that might help us debug this:

[52804.661077] Possible unsafe locking scenario:

[52804.661671] CPU0 CPU1
[52804.662129] ---- ----
[52804.662577] lock(dpll_lock);
[52804.662924] lock(nlk_cb_mutex-GENERIC);
[52804.663538] lock(dpll_lock);
[52804.664073] lock(nlk_cb_mutex-GENERIC);
[52804.664490]

The issue as follows: __netlink_dump_start() calls control->start(cb)
with nlk->cb_mutex held. In control->start(cb) the dpll_lock is taken.
Then nlk->cb_mutex is released and taken again in netlink_dump(), while
dpll_lock still being held. That leads to ABBA deadlock when another
CPU races with the same operation.

Fix this by moving dpll_lock taking into dumpit() callback which ensures
correct lock taking order.

Fixes: 9d71b54b65b1 ("dpll: netlink: Add DPLL framework base functions")
Signed-off-by: Jiri Pirko <[e-mail]>
Reviewed-by: Arkadiusz Kubalewski <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>

Documentation/netlink/specs/dpll.yaml

···
             - type
 
       dump:
-        pre: dpll-lock-dumpit
-        post: dpll-unlock-dumpit
         reply: *dev-attrs
 
     -
···
             - fractional-frequency-offset
 
       dump:
-        pre: dpll-lock-dumpit
-        post: dpll-unlock-dumpit
         request:
           attributes:
             - id

drivers/dpll/dpll_netlink.c

···
 	unsigned long i;
 	int ret = 0;
 
+	mutex_lock(&dpll_lock);
 	xa_for_each_marked_start(&dpll_pin_xa, i, pin, DPLL_REGISTERED,
 				 ctx->idx) {
 		if (!dpll_pin_available(pin))
···
 		}
 		genlmsg_end(skb, hdr);
 	}
+	mutex_unlock(&dpll_lock);
+
 	if (ret == -EMSGSIZE) {
 		ctx->idx = i;
 		return skb->len;
···
 	unsigned long i;
 	int ret = 0;
 
+	mutex_lock(&dpll_lock);
 	xa_for_each_marked_start(&dpll_device_xa, i, dpll, DPLL_REGISTERED,
 				 ctx->idx) {
 		hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid,
···
 		}
 		genlmsg_end(skb, hdr);
 	}
+	mutex_unlock(&dpll_lock);
+
 	if (ret == -EMSGSIZE) {
 		ctx->idx = i;
 		return skb->len;
···
 		 struct genl_info *info)
 {
 	mutex_unlock(&dpll_lock);
-}
-
-int dpll_lock_dumpit(struct netlink_callback *cb)
-{
-	mutex_lock(&dpll_lock);
-
-	return 0;
-}
-
-int dpll_unlock_dumpit(struct netlink_callback *cb)
-{
-	mutex_unlock(&dpll_lock);
-
-	return 0;
 }
 
 int dpll_pin_pre_doit(const struct genl_split_ops *ops, struct sk_buff *skb,

drivers/dpll/dpll_nl.c

···
 	},
 	{
 		.cmd	= DPLL_CMD_DEVICE_GET,
-		.start	= dpll_lock_dumpit,
 		.dumpit	= dpll_nl_device_get_dumpit,
-		.done	= dpll_unlock_dumpit,
 		.flags	= GENL_ADMIN_PERM | GENL_CMD_CAP_DUMP,
 	},
 	{
···
 	},
 	{
 		.cmd		= DPLL_CMD_PIN_GET,
-		.start		= dpll_lock_dumpit,
 		.dumpit		= dpll_nl_pin_get_dumpit,
-		.done		= dpll_unlock_dumpit,
 		.policy		= dpll_pin_get_dump_nl_policy,
 		.maxattr	= DPLL_A_PIN_ID,
 		.flags		= GENL_ADMIN_PERM | GENL_CMD_CAP_DUMP,

drivers/dpll/dpll_nl.h

···
 void
 dpll_pin_post_doit(const struct genl_split_ops *ops, struct sk_buff *skb,
 		   struct genl_info *info);
-int dpll_lock_dumpit(struct netlink_callback *cb);
-int dpll_unlock_dumpit(struct netlink_callback *cb);
 
 int dpll_nl_device_id_get_doit(struct sk_buff *skb, struct genl_info *info);
 int dpll_nl_device_get_doit(struct sk_buff *skb, struct genl_info *info);