On 5/15/21 4:42 AM, Brian Norris wrote:
> We can deadlock when rmmod'ing the driver or going through firmware
> reset, because the cfg80211_unregister_wdev() has to bring down the link
> for us, ... which then grab the same wiphy lock.
>
> nl80211_del_interface() already handles a very similar case, with a nice
> description:
>
> /*
> * We hold RTNL, so this is safe, without RTNL opencount cannot
> * reach 0, and thus the rdev cannot be deleted.
> *
> * We need to do it for the dev_close(), since that will call
> * the netdev notifiers, and we need to acquire the mutex there
> * but don't know if we get there from here or from some other
> * place (e.g. "ip link set ... down").
> */
> mutex_unlock(&rdev->wiphy.mtx);
> ...
>
> Do similarly for mwifiex teardown, by ensuring we bring the link down
> first.
>
> Sample deadlock trace:
>
> [ 247.103516] INFO: task rmmod:2119 blocked for more than 123 seconds.
> [ 247.110630] Not tainted 5.12.4 #5
> [ 247.115796] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 247.124557] task:rmmod state:D stack: 0 pid: 2119 ppid: 2114 flags:0x00400208
> [ 247.133905] Call trace:
> [ 247.136644] __switch_to+0x130/0x170
> [ 247.140643] __schedule+0x714/0xa0c
> [ 247.144548] schedule_preempt_disabled+0x88/0xf4
> [ 247.149714] __mutex_lock_common+0x43c/0x750
> [ 247.154496] mutex_lock_nested+0x5c/0x68
> [ 247.158884] cfg80211_netdev_notifier_call+0x280/0x4e0 [cfg80211]
> [ 247.165769] raw_notifier_call_chain+0x4c/0x78
> [ 247.170742] call_netdevice_notifiers_info+0x68/0xa4
> [ 247.176305] __dev_close_many+0x7c/0x138
> [ 247.180693] dev_close_many+0x7c/0x10c
> [ 247.184893] unregister_netdevice_many+0xfc/0x654
> [ 247.190158] unregister_netdevice_queue+0xb4/0xe0
> [ 247.195424] _cfg80211_unregister_wdev+0xa4/0x204 [cfg80211]
> [ 247.201816] cfg80211_unregister_wdev+0x20/0x2c [cfg80211]
> [ 247.208016] mwifiex_del_virtual_intf+0xc8/0x188 [mwifiex]
> [ 247.214174] mwifiex_uninit_sw+0x158/0x1b0 [mwifiex]
> [ 247.219747] mwifiex_remove_card+0x38/0xa0 [mwifiex]
> [ 247.225316] mwifiex_pcie_remove+0xd0/0xe0 [mwifiex_pcie]
> [ 247.231451] pci_device_remove+0x50/0xe0
> [ 247.235849] device_release_driver_internal+0x110/0x1b0
> [ 247.241701] driver_detach+0x5c/0x9c
> [ 247.245704] bus_remove_driver+0x84/0xb8
> [ 247.250095] driver_unregister+0x3c/0x60
> [ 247.254486] pci_unregister_driver+0x2c/0x90
> [ 247.259267] cleanup_module+0x18/0xcdc [mwifiex_pcie]
>
> Fixes: a05829a7222e ("cfg80211: avoid holding the RTNL when calling the driver")
> Cc: [email protected]
> Link: https://lore.kernel.org/linux-wireless/[email protected]/
> Link: https://lore.kernel.org/linux-wireless/[email protected]/
> Reported-by: Maximilian Luz <[email protected]>
> Reported-by: [email protected]
> Cc: Johannes Berg <[email protected]>
> Signed-off-by: Brian Norris <[email protected]>
Thanks!
Tested-by: Maximilian Luz <[email protected]>
> ---
> drivers/net/wireless/marvell/mwifiex/main.c | 13 ++++++++++---
> 1 file changed, 10 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/wireless/marvell/mwifiex/main.c b/drivers/net/wireless/marvell/mwifiex/main.c
> index 529dfd8b7ae8..17399d4aa129 100644
> --- a/drivers/net/wireless/marvell/mwifiex/main.c
> +++ b/drivers/net/wireless/marvell/mwifiex/main.c
> @@ -1445,11 +1445,18 @@ static void mwifiex_uninit_sw(struct mwifiex_adapter *adapter)
> if (!priv)
> continue;
> rtnl_lock();
> - wiphy_lock(adapter->wiphy);
> if (priv->netdev &&
> - priv->wdev.iftype != NL80211_IFTYPE_UNSPECIFIED)
> + priv->wdev.iftype != NL80211_IFTYPE_UNSPECIFIED) {
> + /*
> + * Close the netdev now, because if we do it later, the
> + * netdev notifiers will need to acquire the wiphy lock
> + * again --> deadlock.
> + */
> + dev_close(priv->wdev.netdev);
> + wiphy_lock(adapter->wiphy);
> mwifiex_del_virtual_intf(adapter->wiphy, &priv->wdev);
> - wiphy_unlock(adapter->wiphy);
> + wiphy_unlock(adapter->wiphy);
> + }
> rtnl_unlock();
> }
>
>