- 15 May, 2014 40 commits
-
-
Christian Engelmayer authored
Fix a potential leak in the error path of function update_bcn_wps_ie(). Move the affected input verification to the beginning of the function so that it may return directly without leaking already allocated memory. Detected by Coverity - CID 1077718. Signed-off-by:
Christian Engelmayer <cengelma@gmx.at> Signed-off-by:
Jes Sorensen <Jes.Sorensen@redhat.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Jes Sorensen authored
No need to allocate GFP_ATOMIC when we don't need to. Signed-off-by:
-
Jes Sorensen authored
The kernel has a great memory allocator, use it instead of hacking up something in the driver. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
This allows us to use GFP_KERNEL when calling from a work handler. Signed-off-by:
-
Jes Sorensen authored
We only really need to hold pmlmepriv->lock while calling rtw_indicate_connect23a(). rtw_get_stainfo23a() and rtw_alloc_stainfo23a() rely on pstapriv->sta_hash_lock and the non WIFI_AP_STATE path relies on pmlmepvi->scanned_queue.lock, except we need to used clr_fwstate() instead of _clr_fwstate_(). Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
One set of custom error codes for the driver should more than suffice. This allows us to get rid of odm_types.h as well. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Use the knowledge we already have from parsing the TODS/FROMDS bits in hdr->frame_control to obtain the bssid. Note that get_hdr_bssid() would never return NULL as handling 4 combinations of a 2 bit word leaves little space for falling through to the 'default' value. Signed-off-by:
-
Jes Sorensen authored
This is only used in one place - do the work there properly. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Having struct qos_priv containing a single integer, and carrying a dedicated header file for it, is just plain silly. Move the integer into struct mlme_priv, which is the only place qos_priv was used anyway, and get rid of the header file. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
This avoids an ugly cast in mlme_evt_hdl23a() Signed-off-by:
-
Jes Sorensen authored
Every other place uses C2HEvent_Header() for events. Given the struct is endian dependant, use it here too to retrieve data from the parm buffer. Note the length field is not set/read in le order - question is whether it's simply an opaque field? Signed-off-by:
-
Larry Finger authored
This makes sparse happy Signed-off-by:
Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
Split the old work handler into a workqueue for processing usb_read_interrupt_complete() events that require more than ust clearing the event, and leave the old handler to just handle event clearing. This means we can get rid of the hacks with magic pointers to determine what actions needs to be taken in the work handler, and as an extra bonus obsoletes the bizarre cbuf code. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
Jes Sorensen authored
This function no longer does anything, so get rid of it. Signed-off-by:
-
Jes Sorensen authored
Signed-off-by:
-
