On 23.06.20 11:40, John Crispin wrote:
The reconf patch breaks wifi down under certain conditions. The root cause
is that during reload the wdev state gets flushed. This has the effect, that
the phy is lost from the state resulting in teardown breaking.
Fix this by changing the lifetime of wdev->data.
Signed-off-by: John Crispin
---
Changes in V2
* free() call was in the wrong place
wireless.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/wireless.c b/wireless.c
index efb7992..e295f28 100644
--- a/wireless.c
+++ b/wireless.c
@@ -275,8 +275,6 @@ wireless_device_free_state(struct wireless_device *wdev)
uloop_timeout_cancel(&wdev->script_check);
uloop_timeout_cancel(&wdev->timeout);
wireless_complete_kill_request(wdev);
- free(wdev->data);
- wdev->data = NULL;
vlist_for_each_element(&wdev->interfaces, vif, node) {
free(vif->data);
vif->data = NULL;
@@ -460,6 +458,7 @@ wireless_device_free(struct wireless_device *wdev)
vlist_flush_all(&wdev->vlans);
vlist_flush_all(&wdev->stations);
avl_delete(&wireless_devices.avl, &wdev->node.avl);
+ free(wdev->data);
free(wdev->config);
free(wdev->prev_config);
free(wdev);
@@ -1414,6 +1413,8 @@ wireless_device_notify(struct wireless_device *wdev,
struct blob_attr *data,
if (*pdata)
return UBUS_STATUS_INVALID_ARGUMENT;
+ if (*pdata)
+ free(*pdata);
*pdata = blob_memdup(cur);
if (vif)
wireless_interface_set_data(vif);
thx, looks like the rest of the patch fixes the issue, been stress testing
this today, will have a closer look tomorrow ..
John
___
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
http://lists.infradead.org/mailman/listinfo/openwrt-devel
