mbox series

[00/32] staging: wfx: rework the Tx queue

Message ID 20200401110405.80282-1-Jerome.Pouiller@silabs.com
Headers show
Series staging: wfx: rework the Tx queue | expand

Message

Jérôme Pouiller April 1, 2020, 11:03 a.m. UTC 
From: Jérôme Pouiller <jerome.pouiller@silabs.com>

This current implementation of the Tx queue is far more complex than
necessary and its behavior is dubious on some corner cases. This series
rework the Tx queue:
  - Simplify support for QoS (since device already do the job)
  - Simplify support for Power Saving stations (since mac80211 already
    do the job)
  - Improve support of Content After DTIM Beacon (CAB) when device is
    AP and some station use power save.
  - Improve robustness of flushing


Jérôme Pouiller (32):
  staging: wfx: add sanity checks to hif_join()
  staging: wfx: do not stop mac80211 queueing during tx_policy upload
  staging: wfx: take advantage of ieee80211_{stop/start}_queues
  staging: wfx: remove "burst" mechanism
  staging: wfx: uniformize queue_id retrieval
  staging: wfx: drop useless queue_id field
  staging: wfx: avoid useless wake_up
  staging: wfx: simplify hif_handle_tx_data()
  staging: wfx: simplify wfx_tx_queues_empty()
  staging: wfx: drop unused argument in wfx_get_prio_queue()
  staging: wfx: simplify wfx_tx_queue_mask_get()
  staging: wfx: drop useless sta_asleep_mask
  staging: wfx: drop argument tx_allowed_mask since it is constant now
  staging: wfx: do not use link_map_cache to track CAB
  staging: wfx: drop useless link_map_cache
  staging: wfx: do not rely anymore on link_id to choose packet in queue
  staging: wfx: drop unused link_id field
  staging: wfx: drop unused raw_link_id field
  staging: wfx: rename wfx_tx_get_raw_link_id()
  staging: wfx: replace wfx_tx_queues_get_after_dtim() by
    wfx_tx_queues_has_cab()
  staging: wfx: introduce a counter of pending frames
  staging: wfx: change the way to choose frame to send
  staging: wfx: drop now useless field edca_params
  staging: wfx: drop struct wfx_queue_stats
  staging: wfx: simplify usage of wfx_tx_queues_put()
  staging: wfx: improve interface between data_tx.c and queue.c
  staging: wfx: relocate wfx_skb_dtor() prior its callers
  staging: wfx: repair wfx_flush()
  staging: wfx: wfx_flush() did not ensure that frames are processed
  staging: wfx: fix potential deadlock in wfx_tx_flush()
  staging: wfx: fix case where AP stop with CAB traffic pending
  staging: wfx: remove hack about tx_rate policies

 drivers/staging/wfx/bh.c      |   4 +-
 drivers/staging/wfx/data_tx.c | 196 +++++--------
 drivers/staging/wfx/data_tx.h |   5 +-
 drivers/staging/wfx/hif_tx.c  |   2 +
 drivers/staging/wfx/main.c    |   2 -
 drivers/staging/wfx/queue.c   | 529 +++++++++++-----------------------
 drivers/staging/wfx/queue.h   |  36 +--
 drivers/staging/wfx/sta.c     |  87 ++----
 drivers/staging/wfx/sta.h     |   2 -
 drivers/staging/wfx/wfx.h     |   7 +-
 10 files changed, 281 insertions(+), 589 deletions(-)

Comments

Dan Carpenter April 3, 2020, 8:03 a.m. UTC | #1 
I didn't quite finish reviewing these pathches last night.  Looks good.
You will need a check on "ssidlen" to prevent memory corruption, as
discussed in patch 1, but that's not a bug which was introduced by this
patchset.  None of my other comments really applied to the patchset
itself, just to the surrounding code.

Looks good.

Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>

regards,
dan carpenter