linux-wireless.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Johannes Berg <johannes@sipsolutions.net>
To: John Linville <linville@tuxdriver.com>
Cc: linux-wireless@vger.kernel.org
Subject: [PATCH 15/27] mac80211: fix aggregation timer lockups
Date: Tue, 10 Feb 2009 21:25:51 +0100	[thread overview]
Message-ID: <20090210202555.374514996@sipsolutions.net> (raw)
In-Reply-To: 20090210202536.425266119@sipsolutions.net

As far as I can tell, there are possible lockups because both the RX
session_timer and TX addba_resp_timer are del_timer_sync'ed under
the sta spinlock which both timer functions take. Additionally, the
TX agg code seems to leak memory when TX aggregation is not disabled
before the sta_info is freed.

Fix this by making the free code a little smarter in the RX agg case,
and actually make the sta_info_destroy code free the TX agg info in
the TX agg case. We won't notify the peer, but it'll notice something
is wrong anyway, and normally this only happens after we've told it
in some other way we will no longer talk to it.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 net/mac80211/agg-rx.c   |   11 +++++++++--
 net/mac80211/sta_info.c |   37 +++++++++++++++++++++++++++++++++----
 net/mac80211/sta_info.h |    1 +
 3 files changed, 43 insertions(+), 6 deletions(-)

--- wireless-testing.orig/net/mac80211/agg-rx.c	2009-02-10 20:59:32.000000000 +0100
+++ wireless-testing/net/mac80211/agg-rx.c	2009-02-10 20:59:34.000000000 +0100
@@ -78,11 +78,18 @@ void ieee80211_sta_stop_rx_ba_session(st
 			sta->ampdu_mlme.tid_rx[tid]->reorder_buf[i] = NULL;
 		}
 	}
+
+	spin_lock_bh(&sta->lock);
 	/* free resources */
 	kfree(sta->ampdu_mlme.tid_rx[tid]->reorder_buf);
-	kfree(sta->ampdu_mlme.tid_rx[tid]);
-	sta->ampdu_mlme.tid_rx[tid] = NULL;
+
+	if (!sta->ampdu_mlme.tid_rx[tid]->shutdown) {
+		kfree(sta->ampdu_mlme.tid_rx[tid]);
+		sta->ampdu_mlme.tid_rx[tid] = NULL;
+	}
+
 	sta->ampdu_mlme.tid_state_rx[tid] = HT_AGG_STATE_IDLE;
+	spin_unlock_bh(&sta->lock);
 
 	rcu_read_unlock();
 }
--- wireless-testing.orig/net/mac80211/sta_info.c	2009-02-10 20:59:25.000000000 +0100
+++ wireless-testing/net/mac80211/sta_info.c	2009-02-10 20:59:34.000000000 +0100
@@ -194,12 +194,41 @@ void sta_info_destroy(struct sta_info *s
 		dev_kfree_skb_any(skb);
 
 	for (i = 0; i <  STA_TID_NUM; i++) {
+		struct tid_ampdu_rx *tid_rx;
+		struct tid_ampdu_tx *tid_tx;
+
 		spin_lock_bh(&sta->lock);
-		if (sta->ampdu_mlme.tid_rx[i])
-		  del_timer_sync(&sta->ampdu_mlme.tid_rx[i]->session_timer);
-		if (sta->ampdu_mlme.tid_tx[i])
-		  del_timer_sync(&sta->ampdu_mlme.tid_tx[i]->addba_resp_timer);
+		tid_rx = sta->ampdu_mlme.tid_rx[i];
+		/* Make sure timer won't free the tid_rx struct, see below */
+		if (tid_rx)
+			tid_rx->shutdown = true;
 		spin_unlock_bh(&sta->lock);
+
+		/*
+		 * Outside spinlock - shutdown is true now so that the timer
+		 * won't free tid_rx, we have to do that now. Can't let the
+		 * timer do it because we have to sync the timer outside the
+		 * lock that it takes itself.
+		 */
+		if (tid_rx) {
+			del_timer_sync(&tid_rx->session_timer);
+			kfree(tid_rx);
+		}
+
+		/*
+		 * No need to do such complications for TX agg sessions, the
+		 * path leading to freeing the tid_tx struct goes via a call
+		 * from the driver, and thus needs to look up the sta struct
+		 * again, which cannot be found when we get here. Hence, we
+		 * just need to delete the timer and free the aggregation
+		 * info; we won't be telling the peer about it then but that
+		 * doesn't matter if we're not talking to it again anyway.
+		 */
+		tid_tx = sta->ampdu_mlme.tid_tx[i];
+		if (tid_tx) {
+			del_timer_sync(&tid_tx->addba_resp_timer);
+			kfree(tid_tx);
+		}
 	}
 
 	__sta_info_free(local, sta);
--- wireless-testing.orig/net/mac80211/sta_info.h	2009-02-10 20:59:32.000000000 +0100
+++ wireless-testing/net/mac80211/sta_info.h	2009-02-10 20:59:34.000000000 +0100
@@ -100,6 +100,7 @@ struct tid_ampdu_rx {
 	u16 buf_size;
 	u16 timeout;
 	u8 dialog_token;
+	bool shutdown;
 };
 
 /**

-- 


  parent reply	other threads:[~2009-02-10 20:37 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-02-10 20:25 [PATCH 00/27] mac80211 updates Johannes Berg
2009-02-10 20:25 ` [PATCH 01/27] zd1211rw: do not ratelimit no-ops Johannes Berg
2009-02-10 20:25 ` [PATCH 02/27] mac80211: disable IBSS beacon before join Johannes Berg
2009-02-10 20:25 ` [PATCH 03/27] zd1211rw: honour enable_beacon conf Johannes Berg
2009-02-10 20:25 ` [PATCH 04/27] mac80211: properly validate/translate IW_AUTH_MFP values Johannes Berg
2009-02-10 20:25 ` [PATCH 05/27] mac80211: reject extra IEs for probe request when hw_scan Johannes Berg
2009-02-10 20:25 ` [PATCH 06/27] mac80211: fix beacon enable more Johannes Berg
2009-02-10 20:25 ` [PATCH 07/27] mac80211: remove bssid argument from prepare_for_handlers Johannes Berg
2009-02-10 20:25 ` [PATCH 08/27] mac80211: remove stray aggregation debugfs definition Johannes Berg
2009-02-10 20:25 ` [PATCH 09/27] mac80211: fix RX aggregation timeouts Johannes Berg
2009-02-10 20:25 ` [PATCH 10/27] mac80211: restructure HT code Johannes Berg
2009-02-10 20:25 ` [PATCH 11/27] mac80211: restrict aggregation to supported interface modes Johannes Berg
2009-02-10 20:25 ` [PATCH 12/27] mac80211: hardware should not deny going back to legacy Johannes Berg
2009-02-10 20:25 ` [PATCH 13/27] mac80211: document TX aggregation (and small cleanup) Johannes Berg
2009-02-10 20:25 ` [PATCH 14/27] mac80211: fix race in TX aggregation Johannes Berg
2009-02-10 20:25 ` Johannes Berg [this message]
2009-02-10 20:25 ` [PATCH 16/27] mac80211: clean up BA session teardown Johannes Berg
2009-02-10 20:25 ` [PATCH 17/27] mac80211: RX aggregation: clean up stop session Johannes Berg
2009-02-10 20:25 ` [PATCH 18/27] mac80211: further cleanups to stopping BA sessions Johannes Berg
2009-02-10 20:25 ` [PATCH 19/27] cfg80211/nl80211: scanning (and mac80211 update to use it) Johannes Berg
2009-02-10 20:25 ` [PATCH 20/27] mac80211: dont add BSS when creating IBSS Johannes Berg
2009-02-10 20:25 ` [PATCH 21/27] cfg80211: free_priv for BSS info Johannes Berg
2009-02-10 20:25 ` [PATCH 22/27] cfg80211: allow users to request removing a BSS Johannes Berg
2009-02-10 20:25 ` [PATCH 23/27] cfg80211: add more flexible BSS lookup Johannes Berg
2009-02-10 20:26 ` [PATCH 24/27] mac80211: use cfg80211s BSS infrastructure Johannes Berg
2009-02-10 20:26 ` [PATCH 25/27] mac80211: calculate wstats_flags on the fly Johannes Berg
2009-02-10 20:26 ` [PATCH 26/27] mac80211: fix IBSS auth Johannes Berg
2009-02-10 20:26 ` [PATCH 27/27] mac80211: split managed/ibss code a little more Johannes Berg

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20090210202555.374514996@sipsolutions.net \
    --to=johannes@sipsolutions.net \
    --cc=linux-wireless@vger.kernel.org \
    --cc=linville@tuxdriver.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).