ath6kl: fix crash when interface is closed but scan is ongoing
authorKalle Valo <kvalo@qca.qualcomm.com>
Mon, 13 Jun 2011 08:54:06 +0000 (11:54 +0300)
committerGreg Kroah-Hartman <gregkh@suse.de>
Tue, 28 Jun 2011 20:14:09 +0000 (13:14 -0700)
When ath6kl module was resumed while a scan was ongoing, for example during
suspend, the driver would crash in ar6k_cfg80211_scanComplete_event():

[26581.586440] Call Trace:
[26581.586440]  [<f99ffeda>] ? ar6k_cfg80211_scanComplete_event+0xaa/0xaa [ath6kl]
[26581.586440]  [<f9a0a020>] wmi_iterate_nodes+0xb/0xd [ath6kl]
[26581.586440]  [<f99ffe78>] ar6k_cfg80211_scanComplete_event+0x48/0xaa [ath6kl]
[26581.586440]  [<f9a038ae>] ar6000_close+0x77/0x7e [ath6kl]
[26581.586440]  [<c139c25d>] __dev_close_many+0x87/0xab
[26581.586440]  [<c139c30a>] dev_close_many+0x54/0xab
[26581.586440]  [<c139c437>] rollback_registered_many+0xa5/0x19e
[26581.586440]  [<c139c595>] rollback_registered+0x23/0x2f
[26581.586440]  [<c139c5ed>] unregister_netdevice_queue+0x4c/0x69
[26581.586440]  [<c139c6b2>] unregister_netdev+0x18/0x1f
[26581.586440]  [<f9a00d4c>] ar6000_destroy+0xf8/0x115 [ath6kl]
[26581.586440]  [<f9a0c765>] ar6k_cleanup_module+0x20/0x29 [ath6kl]
[26581.586440]  [<c1062843>] sys_delete_module+0x181/0x1d9
[26581.586440]  [<c105876b>] ? lock_release_holdtime+0x2b/0xcd
[26581.586440]  [<c10b55dc>] ? sys_munmap+0x3b/0x42
[26581.586440]  [<c14a99dc>] ? restore_all+0xf/0xf
[26581.586440]  [<c14aeb6c>] sysenter_do_call+0x12/0x32
[26581.586440] Code: 89 53 6c 75 07 89 d8 e8 c0 ff ff ff 89 f0 e8 2c f2 a9 c7 5b 5e 5d c3 55 89 e5 57 56 53 89 c3 83 ec 08 89 55 f0 8d 78 04 89 4d ec <8b> b0 b8 00 00 00 46 89 b0 b8 00 00 00 89 f8 e8 ae ed a9 c7 8b

Fix the function not to iterate nodes when the scan is aborted. The nodes
are already freed when the module is being unloaded. Patch "ath6kl: Fix a
kernel panic furing suspend/resume" tried to fix this already but it wasn't
enough as a pointer was still used even after the null check. This patch
removes the null check entirely as the wmi structure is not accessed anymore
during module unload.

Also fix a bug where the status was checked as a bitfield with '&' operator.
But it's not a bitfield, just a regular (enum like) value.

Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Cc: stable <stable@kernel.org> [3.0]
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
drivers/staging/ath6kl/os/linux/cfg80211.c

index d3a774dbb7e8b3cec2faf25bde0270b0caad8772..32e319782f272bfa2cd989fffaedf2a48795f0ff 100644 (file)
@@ -867,26 +867,31 @@ ar6k_cfg80211_scanComplete_event(struct ar6_softc *ar, int status)
 
     AR_DEBUG_PRINTF(ATH_DEBUG_INFO, ("%s: status %d\n", __func__, status));
 
-    if(ar->scan_request)
-    {
-        /* Translate data to cfg80211 mgmt format */
-       if (ar->arWmi)
-               wmi_iterate_nodes(ar->arWmi, ar6k_cfg80211_scan_node, ar->wdev->wiphy);
+    if (!ar->scan_request)
+           return;
+
+    if ((status == A_ECANCELED) || (status == A_EBUSY)) {
+           cfg80211_scan_done(ar->scan_request, true);
+           goto out;
+    }
+
+    /* Translate data to cfg80211 mgmt format */
+    wmi_iterate_nodes(ar->arWmi, ar6k_cfg80211_scan_node, ar->wdev->wiphy);
 
-        cfg80211_scan_done(ar->scan_request,
-            ((status & A_ECANCELED) || (status & A_EBUSY)) ? true : false);
+    cfg80211_scan_done(ar->scan_request, false);
 
-        if(ar->scan_request->n_ssids &&
-           ar->scan_request->ssids[0].ssid_len) {
+    if(ar->scan_request->n_ssids &&
+       ar->scan_request->ssids[0].ssid_len) {
             u8 i;
 
             for (i = 0; i < ar->scan_request->n_ssids; i++) {
-                wmi_probedSsid_cmd(ar->arWmi, i+1, DISABLE_SSID_FLAG,
-                                   0, NULL);
+                   wmi_probedSsid_cmd(ar->arWmi, i+1, DISABLE_SSID_FLAG,
+                                      0, NULL);
             }
-        }
-        ar->scan_request = NULL;
     }
+
+out:
+    ar->scan_request = NULL;
 }
 
 static int