[maint] batman-adv: release station info tidstats

Message ID 20190125080205.10594-1-sven@narfation.org (mailing list archive)
State Accepted, archived
Commit e6150db855210a085feb4c5e4daadb285a668bba
Delegated to: Simon Wunderlich
Headers
Series [maint] batman-adv: release station info tidstats |

Commit Message

Sven Eckelmann Jan. 25, 2019, 8:02 a.m. UTC
  From: Felix Fietkau <nbd@nbd.name>

With the addition of TXQ stats in the per-tid statistics the struct
station_info grew significantly. This resulted in stack size warnings
due to the structure itself being above the limit for the warnings.

To work around this, the TID array was allocated dynamically. Also a
function to free this content was introduced with commit 7ea3e110f2f8
("cfg80211: release station info tidstats where needed") but the necessary
changes were not provided for batman-adv's B.A.T.M.A.N. V implementation.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Fixes: 8689c051a201 ("cfg80211: dynamically allocate per-tid stats for station info")
[sven@narfation.org: add commit message, add compat code]
Signed-off-by: Sven Eckelmann <sven@narfation.org>
---
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Arend van Spriel <aspriel@gmail.com>

Felix Fietkau didn't really send this patch to us and thus also didn't
signed it off. This is just a proposal how we could integrate it - please
inform me how we want to deal with it.

Thanks to Martin Weinelt for discovering this change and forwarding it to
us.
---
 compat-include/net/cfg80211.h | 7 +++++++
 net/batman-adv/bat_v_elp.c    | 3 +++
 2 files changed, 10 insertions(+)
  

Comments

Berg, Johannes Jan. 25, 2019, 8:16 a.m. UTC | #1
Huh, sorry about that.

> From: Felix Fietkau <nbd@nbd.name>

If Felix had nothing to do with the patch, why would you pretend otherwise?

You need two patches - one for backports, this part:

>  compat-include/net/cfg80211.h | 7 +++++++

And for mainline, this part:

>  net/batman-adv/bat_v_elp.c    | 3 +++

So send the two separate patches to the two separate projects, and that's it?

johannes
  
Sven Eckelmann Jan. 25, 2019, 8:23 a.m. UTC | #2
On Friday, 25 January 2019 08.16.46 CET Berg, Johannes wrote:
> Huh, sorry about that.
> 
> > From: Felix Fietkau <nbd@nbd.name>
> 
> If Felix had nothing to do with the patch, why would you pretend otherwise?

This is what he has dropped and Martin Weinelt forwarded to us:
http://nbd.name/batman-adv-memleak.patch

So just the net/batman-adv/bat_v_elp.c part without anything else.

> You need two patches - one for backports, this part:
> 
> >  compat-include/net/cfg80211.h | 7 +++++++

This is not for backports [0] - this is for batman-adv.git [1]

> And for mainline, this part:
> 
> >  net/batman-adv/bat_v_elp.c    | 3 +++
> 
> So send the two separate patches to the two separate projects, and that's it?

The upstream patch (for David) is just stripped from the compat-include part.

Kind regards,
	Sven

[0] https://git.kernel.org/pub/scm/linux/kernel/git/backports/backports.git/
[1] https://git.open-mesh.org/batman-adv.git
[2] https://git.open-mesh.org/linux-merge.git/commit/7d652669b61d702c6e62a39579d17f6881670ab6
  

Patch

diff --git a/compat-include/net/cfg80211.h b/compat-include/net/cfg80211.h
index 79c028af..6b6a0287 100644
--- a/compat-include/net/cfg80211.h
+++ b/compat-include/net/cfg80211.h
@@ -79,4 +79,11 @@  static inline int batadv_cfg80211_get_station(struct net_device *dev,
 
 #endif /* < KERNEL_VERSION(4, 18, 0) && IS_ENABLED(CONFIG_CFG80211) */
 
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 18, 0)
+
+#define cfg80211_sinfo_release_content(sinfo)
+
+#endif /* < KERNEL_VERSION(4, 18, 0) */
+
 #endif	/* _NET_BATMAN_ADV_COMPAT_NET_CFG80211_H_ */
diff --git a/net/batman-adv/bat_v_elp.c b/net/batman-adv/bat_v_elp.c
index e8090f09..ef0dec20 100644
--- a/net/batman-adv/bat_v_elp.c
+++ b/net/batman-adv/bat_v_elp.c
@@ -104,6 +104,9 @@  static u32 batadv_v_elp_get_throughput(struct batadv_hardif_neigh_node *neigh)
 
 		ret = cfg80211_get_station(real_netdev, neigh->addr, &sinfo);
 
+		/* free the TID stats immediately */
+		cfg80211_sinfo_release_content(&sinfo);
+
 		dev_put(real_netdev);
 		if (ret == -ENOENT) {
 			/* Node is not associated anymore! It would be