| Message ID | 20250204164240.370153-1-Alexander@wetzel-home.de (mailing list archive) |
|---|---|
| State | New |
| Delegated to: | Johannes Berg |
| Headers | show |
| Series | wifi: mac80211: remove debugfs dir for virtual monitor | expand |
On Tue, 2025-02-04 at 17:42 +0100, Alexander Wetzel wrote: > > My iwlmvm card is not taking it well when I bring down the monitor interface > while TXing on it. The firmware crashes for my "Alder Lake-S PCH CNVi WiFi" > card. > > But that also happens with unpatched kernels and so I don't care > much about that. I can make a bug report, if that's news and want you to > look into that. If you have it now anyway, send it so we can take a look? But I guess not the most important thing. If you want to reproduce with "trace-cmd record -e iwlwifi" running that might help explain more, but I guess we can anyway do the same if we want to. johannes
diff --git a/net/mac80211/driver-ops.c b/net/mac80211/driver-ops.c index 299d38e9e863..2fc60e1e77a5 100644 --- a/net/mac80211/driver-ops.c +++ b/net/mac80211/driver-ops.c @@ -116,8 +116,14 @@ void drv_remove_interface(struct ieee80211_local *local, sdata->flags &= ~IEEE80211_SDATA_IN_DRIVER; - /* Remove driver debugfs entries */ - ieee80211_debugfs_recreate_netdev(sdata, sdata->vif.valid_links); + /* + * Remove driver debugfs entries. + * The virtual monitor interface doesn't get a debugfs + * entry, so it's exempt here. + */ + if (sdata != local->monitor_sdata) + ieee80211_debugfs_recreate_netdev(sdata, + sdata->vif.valid_links); trace_drv_remove_interface(local, sdata); local->ops->remove_interface(&local->hw, &sdata->vif); diff --git a/net/mac80211/iface.c b/net/mac80211/iface.c index 7d3ebfcb8c2b..e627e2df3038 100644 --- a/net/mac80211/iface.c +++ b/net/mac80211/iface.c @@ -1203,16 +1203,17 @@ void ieee80211_del_virtual_monitor(struct ieee80211_local *local) return; } - RCU_INIT_POINTER(local->monitor_sdata, NULL); - mutex_unlock(&local->iflist_mtx); - - synchronize_net(); - + clear_bit(SDATA_STATE_RUNNING, &sdata->state); ieee80211_link_release_channel(&sdata->deflink); if (ieee80211_hw_check(&local->hw, WANT_MONITOR_VIF)) drv_remove_interface(local, sdata); + RCU_INIT_POINTER(local->monitor_sdata, NULL); + mutex_unlock(&local->iflist_mtx); + + synchronize_net(); + kfree(sdata); }
Don't call ieee80211_debugfs_recreate_netdev() for virtual monitor interface when deleting it. The virtual monitor interface shouldn't have debugfs entries and trying to update them will *create* them on deletion. And when the virtual monitor interface is created/destroyed multiple times we'll get warnings about debugfs name conflicts. Signed-off-by: Alexander Wetzel <Alexander@wetzel-home.de> --- That was a bit more tricky than expected: local->monitor_sdata is normally already deleted when we call drv_remove_interface. And we can't simply skip the call for all monitor interfaces... And one additional remark: My iwlmvm card is not taking it well when I bring down the monitor interface while TXing on it. The firmware crashes for my "Alder Lake-S PCH CNVi WiFi" card. But that also happens with unpatched kernels and so I don't care much about that. I can make a bug report, if that's news and want you to look into that. Alexander --- net/mac80211/driver-ops.c | 10 ++++++++-- net/mac80211/iface.c | 11 ++++++----- 2 files changed, 14 insertions(+), 7 deletions(-)