Linux Wireless

High priority:

• cleanup locking
• * review all locking
• * document what each lock is supposed to protect
• * try to reduce the number of locks, semaphores and mutexes
• * convert semaphores to mutexes
• * the DMA bounce buffer should probably be protected by a lock so dueling threads don't corrupt it
• use ath6kl prefix everywhere
• Implement 802.11 ieee disconnect reasons
• high coupling between main.c and init.c
• confusion with header files
• * function prototypes not in corresponding .h files
• * no core.h
• * no init.h
• * no main.h
• * no txrx.h
• merge structs
• * merge struct htc_target to struct ath6kl
• * merge struct ath6kl_device to struct ath6kl
• * merge struct aggr_info to struct ath6kl
• * merge struct wmi to struct ath6kl
• replace all void pointers with proper types
• * for example, use property or unions
• * if not possible for some reason, then document properly what types void pointer might contain
• * * (There are about three void pointers in htc.h, this would result in significant amount of cleanup)
• * remove nl80211_tx_power_setting comment
• * ath6kl_cfg80211_change_iface() needs to change mode before function returns
• * useless memset(ar→ssid, 0, sizeof(ar→ssid))
• * ath6kl_cfg80211_get_key() should not call the callback in the -ENOENT case
• * remove ieee80211com from a comment
• * ath6kl_priv() should return a proper type
• * scat_list should be scat_list[ ATH6KL_SCATTER_ENTRIES_PER_REQ] rather than scat_list[1]
• * memcpy mess in htc_setup_tx_complete()
• * ath6kl_cleanup_amsdu_rxbufs is racy
• * * can't drop lock with list_for_each macros
• * list_empty() in ath6kl_alloc_amsdu_rxbuf() is useless
• * * you can iterate an empty list
• * use ieee80211_data_to_8023()
• * * instead of ath6kl_wmi_dot11_hdr_remove() Low priority:
• * * use align macro in aggr_slice_amsdu()
• * * clean up arEvent usage
• * * clean up wmitimeout usage
• * * remove mdelay()
• * * remove casts from ar6k_priv() users
• * * cleanup wait_event apis
• * * bmi: use struct for commands, not memcpy()
• * * fix 0x%lX debug format
• * * * use %p
• * * * remove (unsigned long) casts
• * * * use decimals when printing lengths
• * * change all read/write functions' buf to a void pointer
• * * move WARN_ON() inside if test
• * * cleanup ath6kl_init_netdev() & co
• * * replace ntohs() with be16_to_cpu()
• * * rename wmip to wmi
• * * ath6kl firmware fetch fails if it's linked to kernel (Y choise)
• * * ATH_COMMON kconfig enables some code which is not needed by ath6kl
• * * fix all FIXMEs
• * * remove ath6kl_cookie (vasanth)
• * * * needs significant code change
• * * inspect structs for alignment and packing appropiateness
• * * * http://marc.info/?l=linux-wireless&m=131053135626622&w=2
• * * switch from cfg80211_inform_bss_frame() to cfg80211_inform_bss()
• * * * was there a reason why we actually need _frame() variant?
• * * logic in htc_tx_from_ep_txq() is convoluted
• * * buffer tx packets
• * * * to improve throughput and avoid excessive stopping of queues
• * * create separate queues for each AC
• * * * so that they can be stopped individually
• * * use of ar→nw_type is not consistent
• * * * some places we test it with '==' even though '&' should be used as it's a bitfield
• * * interface to enable tx uart with hi_serial_enable register
• * * * maybe using debugfs? Then an item is done, please mark it with strike through. If you plan to work on something, please mark the item with “(nick)”.