Skip to contentSkip to conversation

iwlwifi: iwl-dbg: Use del_timer_sync() before freeing

RFC PATCH In review Guenter Roeck 2 participants 4 messages Apr 6, 2022 – Apr 8, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 iwlwifi: iwl-dbg: Use del_timer_sync() before freeing

In Chrome OS, a large number of crashes is observed due to corrupted timer lists. Steven Rostedt pointed out that this usually happens when a timer is freed while still active, and that the problem is often triggered by code calling del_timer() instead of del_timer_sync() just before freeing.

Steven also identified the iwlwifi driver as one of the possible culprits since it does exactly that.

Cc Steven Rostedt <[email protected]>
Cc Shahar S Matityahu <[email protected]>
Cc Johannes Berg <[email protected]>
Fixes 60e8abd9d3e91 ("iwlwifi: dbg_ini: add periodic trigger new API support")

---
RFC:
Maybe there was a reason to use del_timer() instead of del_timer_sync(). Also, I am not sure if the change is sufficient since I don't see any obvious locking that would prevent timers from being added and then modified in iwl_dbg_tlv_set_periodic_trigs() while being removed in iwl_dbg_tlv_del_timers().

 drivers/net/wireless/intel/iwlwifi/iwl-dbg-tlv.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/wireless/intel/iwlwifi/iwl-dbg-tlv.c +1 -1
index 866a33f49915..3237d4b528b5 100644
@@ -371,7 +371,7 @@ void iwl_dbg_tlv_del_timers(struct iwl_trans *trans)
 	struct iwl_dbg_tlv_timer_node *node, *tmp;
 
 	list_for_each_entry_safe(node, tmp, timer_list, list) {
-		del_timer(&node->timer);
+		del_timer_sync(&node->timer);
 		list_del(&node->list);
 		kfree(node);
 	}
Signature · 1 line
2.35.1
#1 lore raw

Hi,

Quoted 15 lines · Guenter Roeck

In Chrome OS, a large number of crashes is observed due to corrupted timer lists. Steven Rostedt pointed out that this usually happens when a timer is freed while still active, and that the problem is often triggered by code calling del_timer() instead of del_timer_sync() just before freeing.

Steven also identified the iwlwifi driver as one of the possible culprits since it does exactly that.

Cc Steven Rostedt <[email protected]>
Cc Shahar S Matityahu <[email protected]>
Cc Johannes Berg <[email protected]>
Fixes 60e8abd9d3e91 ("iwlwifi: dbg_ini: add periodic trigger new API support")
Quoted 8 lines

---
RFC:
Maybe there was a reason to use del_timer() instead of del_timer_sync(). Also, I am not sure if the change is sufficient since I don't see any obvious locking that would prevent timers from being added and then modified in iwl_dbg_tlv_set_periodic_trigs() while being removed in iwl_dbg_tlv_del_timers().

I prepared a new version of this patch, introducing a mutex to protect changes to periodic_trig_list. I'd like to get some feedback before sending it out, though, so I'll wait until next week before sending it.

If you have any feedback/thoughts/comments, please let me know.

Thanks,
Guenter

Quoted 16 lines
  drivers/net/wireless/intel/iwlwifi/iwl-dbg-tlv.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/wireless/intel/iwlwifi/iwl-dbg-tlv.c +1 -1
index 866a33f49915..3237d4b528b5 100644
@@ -371,7 +371,7 @@ void iwl_dbg_tlv_del_timers(struct iwl_trans *trans)
  	struct iwl_dbg_tlv_timer_node *node, *tmp;
 
  	list_for_each_entry_safe(node, tmp, timer_list, list) {
-		del_timer(&node->timer);
+		del_timer_sync(&node->timer);
  		list_del(&node->list);
  		kfree(node);
  	}
#2 lore raw
Quoted 27 lines · Guenter Roeck

Hi,

Quoted 15 lines · Guenter Roeck

In Chrome OS, a large number of crashes is observed due to corrupted timer lists. Steven Rostedt pointed out that this usually happens when a timer is freed while still active, and that the problem is often triggered by code calling del_timer() instead of del_timer_sync() just before freeing.

Steven also identified the iwlwifi driver as one of the possible culprits since it does exactly that.

Cc Steven Rostedt <[email protected]>
Cc Shahar S Matityahu <[email protected]>
Cc Johannes Berg <[email protected]>
Fixes 60e8abd9d3e91 ("iwlwifi: dbg_ini: add periodic trigger new API support")
Quoted 8 lines

---
RFC:
Maybe there was a reason to use del_timer() instead of del_timer_sync(). Also, I am not sure if the change is sufficient since I don't see any obvious locking that would prevent timers from being added and then modified in iwl_dbg_tlv_set_periodic_trigs() while being removed in iwl_dbg_tlv_del_timers().

Quoted 5 lines

I prepared a new version of this patch, introducing a mutex to protect changes to periodic_trig_list. I'd like to get some feedback before sending it out, though, so I'll wait until next week before sending it.

If you have any feedback/thoughts/comments, please let me know.

Hi Guenter,

Thanks for your proposal!

I recently moved from the Intel WiFi team to the Graphics team, so I'm adding Gregory, who has taken over my duties, to the discussion.

I don't recall any specific reasons for using del_timer() instead of del_timer_sync() here. So your patch does look correct to me.

--
Cheers,
Luca.

#3 lore raw

Hi Luca,

Quoted 36 lines · Luciano
Quoted 27 lines · Guenter Roeck

Hi,

Quoted 15 lines · Guenter Roeck

In Chrome OS, a large number of crashes is observed due to corrupted timer lists. Steven Rostedt pointed out that this usually happens when a timer is freed while still active, and that the problem is often triggered by code calling del_timer() instead of del_timer_sync() just before freeing.

Steven also identified the iwlwifi driver as one of the possible culprits since it does exactly that.

Cc Steven Rostedt <[email protected]>
Cc Shahar S Matityahu <[email protected]>
Cc Johannes Berg <[email protected]>
Fixes 60e8abd9d3e91 ("iwlwifi: dbg_ini: add periodic trigger new API support")
Quoted 8 lines

---
RFC:
Maybe there was a reason to use del_timer() instead of del_timer_sync(). Also, I am not sure if the change is sufficient since I don't see any obvious locking that would prevent timers from being added and then modified in iwl_dbg_tlv_set_periodic_trigs() while being removed in iwl_dbg_tlv_del_timers().

Quoted 5 lines

I prepared a new version of this patch, introducing a mutex to protect changes to periodic_trig_list. I'd like to get some feedback before sending it out, though, so I'll wait until next week before sending it.

If you have any feedback/thoughts/comments, please let me know.

Hi Guenter,

Quoted 8 lines

Thanks for your proposal!

I recently moved from the Intel WiFi team to the Graphics team, so I'm adding Gregory, who has taken over my duties, to the discussion.

I don't recall any specific reasons for using del_timer() instead of del_timer_sync() here. So your patch does look correct to me.

Thanks a lot for the feedback. I spent some time trying to determine if a mutex to protect the periodic timer list is needed, but concluded that it is not necessary because the code adding the timer list and the code removing it are never executed in parallel. Of course, I may be missing something, so I'd be happy to be corrected.

Thanks,
Guenter