[lustre-devel] [PATCH 32/33] lnet: ksocklnd: ksocklnd_ni_get_eth_intf_speed() must use only rtnl lock

[James Simmons]

A kernel with debug enable is reporting the following:

include/linux/inetdevice.h:221 suspicious rcu_dereference_check() usage!
other info that might help us debug this:
rcu_scheduler_active = 2, debug_locks = 1
5 locks held by lctl/42289:
 #0: ffffffffa04263f8 ((libcfs_ioctl_list).rwsem){++++}-{3:3}, at: __blocking_notifier_call_chain+0x44/0xa0
 #1: ffffffffa04fa7b0 (lnet_config_mutex){+.+.}-{3:3}, at: lnet_configure+0x1d/0xc0 [lnet]
 #2: ffffffffa04f06e8 (the_lnet.ln_api_mutex){+.+.}-{3:3}, at: LNetNIInit+0x4c/0x960 [lnet]
 #3: ffffffffa04f0788 (&the_lnet.ln_lnd_mutex){+.+.}-{3:3}, at: lnet_startup_lndnet+0x11e/0xa90 [lnet]
 #4: ffffffff834a0cd0 (rtnl_mutex){+.+.}-{3:3}, at: rtnl_lock+0x1b/0x30

stack backtrace:
CPU: 2 PID: 42289 Comm: lctl Kdump: loaded Tainted: G        W  O     --------- -  - 4.18.0rh8.5-debug #2
Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
  Call Trace:
  ? dump_stack+0x119/0x18e
  ? lockdep_rcu_suspicious+0x141/0x14f
  ? ksocklnd_ni_get_eth_intf_speed.isra.1+0x308/0x360 [ksocklnd]
  ? mark_held_locks+0x6a/0xc0
  ? ktime_get_with_offset+0x233/0x2b0
  ? trace_hardirqs_on+0x4e/0x220
  ? ksocknal_tunables_setup+0xed/0x200 [ksocklnd]
  ? ksocknal_startup+0x4ff/0x1180 [ksocklnd]

The function __ethtool_get_link_ksettings() has a hard requirement
to have the in_dev device protected by the rtnl mutex. At the
same time we are aquiring in_dev using the rcu lock. The in_dev
needs to be stabilized by the same lock. So use rtnl functions
instead.

WC-bug-id: https://jira.whamcloud.com/browse/LU-16807
Lustre-commit: 65d664bff40590366 ("LU-16807 ksocklnd: ksocklnd_ni_get_eth_intf_speed() must use only rtnl lock")
Signed-off-by: James Simmons <jsimmons at infradead.org>
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/51028
Reviewed-by: Oleg Drokin <green at whamcloud.com>
Reviewed-by: Chris Horn <chris.horn at hpe.com>
Reviewed-by: Neil Brown <neilb at suse.de>
---
 net/lnet/klnds/socklnd/socklnd_modparams.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/lnet/klnds/socklnd/socklnd_modparams.c b/net/lnet/klnds/socklnd/socklnd_modparams.c
index 5eb58ca97c06..031fe13d2a38 100644
--- a/net/lnet/klnds/socklnd/socklnd_modparams.c
+++ b/net/lnet/klnds/socklnd/socklnd_modparams.c
@@ -183,11 +183,11 @@ static int ksocklnd_ni_get_eth_intf_speed(struct lnet_ni *ni)
 		if (!(flags & IFF_UP))
 			continue;
 
-		in_dev = __in_dev_get_rcu(dev);
+		in_dev = __in_dev_get_rtnl(dev);
 		if (!in_dev)
 			continue;
 
-		in_dev_for_each_ifa_rcu(ifa, in_dev) {
+		in_dev_for_each_ifa_rtnl(ifa, in_dev) {
 			if (strcmp(ifa->ifa_label, ni->ni_interface) == 0)
 				intf_idx = dev->ifindex;
 		}
-- 
2.39.3