* [PATCH 0/3] roaming/disassoc fixes
@ 2009-08-06 18:41 Johannes Berg
2009-08-06 18:41 ` [PATCH 1/3] mac80211: stay authenticated after disassoc Johannes Berg
` (3 more replies)
0 siblings, 4 replies; 5+ messages in thread
From: Johannes Berg @ 2009-08-06 18:41 UTC (permalink / raw)
To: John Linville; +Cc: linux-wireless, Pavel Roskin
Pavel reported a bug, and it was hard to fix :)
Here's the fix patchset, including another fix
for a bug I noticed while testing.
This requires Pavel's patch, of course, but applies
without it in the tree too.
johannes
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/3] mac80211: stay authenticated after disassoc
2009-08-06 18:41 [PATCH 0/3] roaming/disassoc fixes Johannes Berg
@ 2009-08-06 18:41 ` Johannes Berg
2009-08-06 18:41 ` [PATCH 2/3] cfg80211: fix nl80211 disconnected events Johannes Berg
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Johannes Berg @ 2009-08-06 18:41 UTC (permalink / raw)
To: John Linville; +Cc: linux-wireless, Pavel Roskin
After being disassociated by the AP, mac80211 currently
reports this to cfg80211, and then goes to delete the
association. That's fine, but cfg80211 assumes that it's
still authenticated, however, mac80211 throws away all
state.
This fixes mac80211 to keep track of the authentication
in that case so that cfg80211 can request a deauth or
new association properly.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
net/mac80211/ieee80211_i.h | 1 +
net/mac80211/mlme.c | 31 ++++++++++++++++++++++---------
2 files changed, 23 insertions(+), 9 deletions(-)
--- wireless-testing.orig/net/mac80211/ieee80211_i.h 2009-08-06 20:34:45.000000000 +0200
+++ wireless-testing/net/mac80211/ieee80211_i.h 2009-08-06 20:34:49.000000000 +0200
@@ -284,6 +284,7 @@ struct ieee80211_if_managed {
struct mutex mtx;
struct ieee80211_bss *associated;
+ struct ieee80211_mgd_work *old_associate_work;
struct list_head work_list;
u8 bssid[ETH_ALEN];
--- wireless-testing.orig/net/mac80211/mlme.c 2009-08-06 20:34:45.000000000 +0200
+++ wireless-testing/net/mac80211/mlme.c 2009-08-06 20:34:49.000000000 +0200
@@ -880,10 +880,11 @@ static u32 ieee80211_handle_bss_capabili
}
static void ieee80211_set_associated(struct ieee80211_sub_if_data *sdata,
- struct ieee80211_bss *bss,
+ struct ieee80211_mgd_work *wk,
u32 bss_info_changed)
{
struct ieee80211_local *local = sdata->local;
+ struct ieee80211_bss *bss = wk->bss;
bss_info_changed |= BSS_CHANGED_ASSOC;
/* set timing information */
@@ -896,6 +897,7 @@ static void ieee80211_set_associated(str
bss->cbss.capability, bss->has_erp_value, bss->erp_value);
sdata->u.mgd.associated = bss;
+ sdata->u.mgd.old_associate_work = wk;
memcpy(sdata->u.mgd.bssid, bss->cbss.bssid, ETH_ALEN);
/* just to be sure */
@@ -1010,7 +1012,8 @@ ieee80211_authenticate(struct ieee80211_
return RX_MGMT_NONE;
}
-static void ieee80211_set_disassoc(struct ieee80211_sub_if_data *sdata)
+static void ieee80211_set_disassoc(struct ieee80211_sub_if_data *sdata,
+ bool deauth)
{
struct ieee80211_if_managed *ifmgd = &sdata->u.mgd;
struct ieee80211_local *local = sdata->local;
@@ -1028,6 +1031,16 @@ static void ieee80211_set_disassoc(struc
ifmgd->associated = NULL;
memset(ifmgd->bssid, 0, ETH_ALEN);
+ if (deauth) {
+ kfree(ifmgd->old_associate_work);
+ ifmgd->old_associate_work = NULL;
+ } else {
+ struct ieee80211_mgd_work *wk = ifmgd->old_associate_work;
+
+ wk->state = IEEE80211_MGD_STATE_IDLE;
+ list_add(&wk->list, &ifmgd->work_list);
+ }
+
/*
* we need to commit the associated = NULL change because the
* scan code uses that to determine whether this iface should
@@ -1345,7 +1358,7 @@ ieee80211_rx_mgmt_deauth(struct ieee8021
sdata->dev->name, bssid, reason_code);
if (!wk) {
- ieee80211_set_disassoc(sdata);
+ ieee80211_set_disassoc(sdata, true);
} else {
list_del(&wk->list);
kfree(wk);
@@ -1378,7 +1391,7 @@ ieee80211_rx_mgmt_disassoc(struct ieee80
printk(KERN_DEBUG "%s: disassociated (Reason: %u)\n",
sdata->dev->name, reason_code);
- ieee80211_set_disassoc(sdata);
+ ieee80211_set_disassoc(sdata, false);
return RX_MGMT_CFG80211_DISASSOC;
}
@@ -1581,7 +1594,8 @@ ieee80211_rx_mgmt_assoc_resp(struct ieee
* ieee80211_set_associated() will tell the driver */
bss_conf->aid = aid;
bss_conf->assoc_capability = capab_info;
- ieee80211_set_associated(sdata, wk->bss, changed);
+ /* this will take ownership of wk */
+ ieee80211_set_associated(sdata, wk, changed);
/*
* Start timer to probe the connection to the AP now.
@@ -1590,7 +1604,6 @@ ieee80211_rx_mgmt_assoc_resp(struct ieee
ieee80211_sta_rx_notify(sdata, (struct ieee80211_hdr *)mgmt);
mod_beacon_timer(sdata);
- kfree(wk);
return RX_MGMT_CFG80211_ASSOC;
}
@@ -2096,7 +2109,7 @@ static void ieee80211_sta_work(struct wo
printk(KERN_DEBUG "No probe response from AP %pM"
" after %dms, disconnecting.\n",
bssid, (1000 * IEEE80211_PROBE_WAIT)/HZ);
- ieee80211_set_disassoc(sdata);
+ ieee80211_set_disassoc(sdata, true);
mutex_unlock(&ifmgd->mtx);
/*
* must be outside lock due to cfg80211,
@@ -2500,7 +2513,7 @@ int ieee80211_mgd_deauth(struct ieee8021
if (ifmgd->associated && &ifmgd->associated->cbss == req->bss) {
bssid = req->bss->bssid;
- ieee80211_set_disassoc(sdata);
+ ieee80211_set_disassoc(sdata, true);
} else list_for_each_entry(wk, &ifmgd->work_list, list) {
if (&wk->bss->cbss == req->bss) {
bssid = req->bss->bssid;
@@ -2552,7 +2565,7 @@ int ieee80211_mgd_disassoc(struct ieee80
return -ENOLINK;
}
- ieee80211_set_disassoc(sdata);
+ ieee80211_set_disassoc(sdata, false);
mutex_unlock(&ifmgd->mtx);
--
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 2/3] cfg80211: fix nl80211 disconnected events
2009-08-06 18:41 [PATCH 0/3] roaming/disassoc fixes Johannes Berg
2009-08-06 18:41 ` [PATCH 1/3] mac80211: stay authenticated after disassoc Johannes Berg
@ 2009-08-06 18:41 ` Johannes Berg
2009-08-06 18:41 ` [PATCH 3/3] cfg80211: fix SME association after disassociation Johannes Berg
2009-08-08 3:34 ` [PATCH 0/3] roaming/disassoc fixes Pavel Roskin
3 siblings, 0 replies; 5+ messages in thread
From: Johannes Berg @ 2009-08-06 18:41 UTC (permalink / raw)
To: John Linville; +Cc: linux-wireless, Pavel Roskin
When reporting a disconnection to userspace, we try
to report whether it was from the AP or by our own
choice. However, we misreported a broadcast deauth
or disassoc as being by own choice, which is wrong.
Fix this by checking the sender address instead of
the destination address.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
net/wireless/mlme.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
--- wireless-testing.orig/net/wireless/mlme.c 2009-08-06 20:36:05.000000000 +0200
+++ wireless-testing/net/wireless/mlme.c 2009-08-06 20:40:09.000000000 +0200
@@ -149,7 +149,7 @@ static void __cfg80211_send_deauth(struc
reason_code = le16_to_cpu(mgmt->u.deauth.reason_code);
- from_ap = memcmp(mgmt->da, dev->dev_addr, ETH_ALEN) == 0;
+ from_ap = memcmp(mgmt->sa, dev->dev_addr, ETH_ALEN) != 0;
__cfg80211_disconnected(dev, NULL, 0, reason_code, from_ap);
} else if (wdev->sme_state == CFG80211_SME_CONNECTING) {
__cfg80211_connect_result(dev, mgmt->bssid, NULL, 0, NULL, 0,
@@ -215,7 +215,7 @@ static void __cfg80211_send_disassoc(str
reason_code = le16_to_cpu(mgmt->u.disassoc.reason_code);
- from_ap = memcmp(mgmt->da, dev->dev_addr, ETH_ALEN) == 0;
+ from_ap = memcmp(mgmt->sa, dev->dev_addr, ETH_ALEN) != 0;
__cfg80211_disconnected(dev, NULL, 0, reason_code, from_ap);
}
--
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 3/3] cfg80211: fix SME association after disassociation
2009-08-06 18:41 [PATCH 0/3] roaming/disassoc fixes Johannes Berg
2009-08-06 18:41 ` [PATCH 1/3] mac80211: stay authenticated after disassoc Johannes Berg
2009-08-06 18:41 ` [PATCH 2/3] cfg80211: fix nl80211 disconnected events Johannes Berg
@ 2009-08-06 18:41 ` Johannes Berg
2009-08-08 3:34 ` [PATCH 0/3] roaming/disassoc fixes Pavel Roskin
3 siblings, 0 replies; 5+ messages in thread
From: Johannes Berg @ 2009-08-06 18:41 UTC (permalink / raw)
To: John Linville; +Cc: linux-wireless, Pavel Roskin
When an AP disassociates us, we currently go into a weird
state because the SME doesn't handle authenticated but not
associated well unless it's within its own state machine,
it can't recover from that. However, it shouldn't need to,
since we don't do any decisions in it really -- so when we
get disconnected, simply deauthenticate too.
Reported-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
net/wireless/sme.c | 20 ++++++++++++++++++++
1 file changed, 20 insertions(+)
--- wireless-testing.orig/net/wireless/sme.c 2009-08-06 20:40:09.000000000 +0200
+++ wireless-testing/net/wireless/sme.c 2009-08-06 20:40:19.000000000 +0200
@@ -570,10 +570,30 @@ void __cfg80211_disconnected(struct net_
wdev->ssid_len = 0;
if (wdev->conn) {
+ const u8 *bssid;
+ int ret;
+
kfree(wdev->conn->ie);
wdev->conn->ie = NULL;
kfree(wdev->conn);
wdev->conn = NULL;
+
+ /*
+ * If this disconnect was due to a disassoc, we
+ * we might still have an auth BSS around. For
+ * the userspace SME that's currently expected,
+ * but for the kernel SME (nl80211 CONNECT or
+ * wireless extensions) we want to clear up all
+ * state.
+ */
+ for (i = 0; i < MAX_AUTH_BSSES; i++) {
+ if (!wdev->auth_bsses[i])
+ continue;
+ bssid = wdev->auth_bsses[i]->pub.bssid;
+ ret = __cfg80211_mlme_deauth(rdev, dev, bssid, NULL, 0,
+ WLAN_REASON_DEAUTH_LEAVING);
+ WARN(ret, "deauth failed: %d\n", ret);
+ }
}
nl80211_send_disconnected(rdev, dev, reason, ie, ie_len, from_ap);
--
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH 0/3] roaming/disassoc fixes
2009-08-06 18:41 [PATCH 0/3] roaming/disassoc fixes Johannes Berg
` (2 preceding siblings ...)
2009-08-06 18:41 ` [PATCH 3/3] cfg80211: fix SME association after disassociation Johannes Berg
@ 2009-08-08 3:34 ` Pavel Roskin
3 siblings, 0 replies; 5+ messages in thread
From: Pavel Roskin @ 2009-08-08 3:34 UTC (permalink / raw)
To: Johannes Berg; +Cc: John Linville, linux-wireless
On Thu, 2009-08-06 at 20:41 +0200, Johannes Berg wrote:
> Pavel reported a bug, and it was hard to fix :)
>
> Here's the fix patchset, including another fix
> for a bug I noticed while testing.
>
> This requires Pavel's patch, of course, but applies
> without it in the tree too.
I'm getting a reasonable behavior in the managed mode now. Thank you!
--
Regards,
Pavel Roskin
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2009-08-08 3:34 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-08-06 18:41 [PATCH 0/3] roaming/disassoc fixes Johannes Berg
2009-08-06 18:41 ` [PATCH 1/3] mac80211: stay authenticated after disassoc Johannes Berg
2009-08-06 18:41 ` [PATCH 2/3] cfg80211: fix nl80211 disconnected events Johannes Berg
2009-08-06 18:41 ` [PATCH 3/3] cfg80211: fix SME association after disassociation Johannes Berg
2009-08-08 3:34 ` [PATCH 0/3] roaming/disassoc fixes Pavel Roskin
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).