Mam Alphawave, firmware 1.5.6, hostap 0.0.1 managed, vynikajici signal (-60 dBm, SNR +40 dB). Presto mne karta vytuhava, protoze minimalne jednou za hodinu zahlasi
hfa384x_setup_bap - timeout after
a nasleduje reset, ktery je aspon jednou za den smrtelny. Kdyz jsem udelal zminenou opravu, tak uz k timeoutu v hfa384x_setup_bap() nedochazi (za 10 hodin provozu ani jeden), zrejme se muj problem vyresil. Az bude cas, otestuju, ktera cast toho patche tomu napomohla, delaji se tam totiz dve veci:
- prodlouzeni timeoutu na dvojnasobek
- maskovani preruseni
Protoze to podle mne tim maskovanim neni, mohlo by se mozna odstranit. Ale to budu muset nejprve vyzkouset. Uz jsem mel nekolik konfiguraci s hostapem, ale tohle je prvni slusne reseni. Postup: (1) zajistit slusny signal, aby nevznikaly resety kvuli nedostatku beaconu, (2) vyresit timeout v hfa384x_setup_bap().
Tak jsem se koukal na change log poslední verze v0.0.2 z oficiálního cvs Jouniho a našel jsem:
2003-05-03 - v0.0.2
* fixed iwconfig 'rate' command to change supported rates element
in beacon frames
* added support for setting basic rate set with new prism2_param
'basic_rates' and operational rates with 'oper_rates'
* added support for new AVS (linux-wlan-ng) monitoring header
(patch from Gerald Britton)
* fixed TX rate controlling for Repeater/WDS mode
* fixed power saving mode updating on PS Poll frames
* optimized AP management frame handling by removing not anymore needed
RX packet queue since AP code is not called in hw IRQ context anymore
* added support for wireless extensions v16 iwspy operations (based on
patch from Jean Tourrilhes)
* improved hostapd event logging (stdout and/or syslog)
* fixed Linux 2.2.x compatibility code for monitor mode with Prism2
headers (monitor_type 1)
* fixed CRC-16 (signature stamp) generation in prism2_srec (this is
needed for writing flash upgrade)
* added support for non-volatile download, i.e., flash upgrade
(based on patch from Pavel Roskin)
Note! This has _not_ yet been tested thorougly and the current
version of code should not be used unless you are prepared to
facing corrupted flash and killed cards.. The code is commented out
in default build and can be enabled by defining
PRISM2_NON_VOLATILE_DOWNLOAD.
* fixed IEEE 802.11 defragmentation when using host-based WEP
decryption (this caused a kernel crash in netif_rx())
* fixed host-based WEP decryption not to leave 8 extra octets in the
end of the frame
* added support for IEEE 802.11 Shared Key Authentication when using
hostapd (new hostapd.conf field 'auth_algs' for configuring supported
algorithms)
* fixed race condition in get/set RID operations that seemed to have
caused BAP timeouts and RID read failures (e.g., 'RID len mismatch'
on get_rid and problems with host_roaming)
* fixed automatic WDS addition to schedule a process context task for
registering the new device to avoid crashing kernel
* added automatic WDS link addition on reception of 4-addr WDS frame
from unknown transmitter (if autom_ap_wds is set)
* fixed handling of non-WDS frames from a STA in 'AP client' mode
(do not assume that they are bogus WDS frames if STA has associated)
* added support for Session-Timeout Attribute in RADIUS messages
(both IEEE 802.11 ACL and IEEE 802.1X)
Zaujal mě bod:
* optimized AP management frame handling by removing not anymore needed RX packet queue since AP code is not called in hw IRQ context anymore.
Diskutovalo se tu někde na foru o problému co se stane když přijde frame při resetu karty. Tohle by mělo pomoci - možná je to právě oprava z prvního postu v tomhle threadu.
Otázka zní zkoušel jste ho někdo? Budu ho dnes večer kompilovat do LRP Wisp-Dist distribuce. Takže bych mohl mít zítra alespoň první dojmy. Tedy doufám jestli to vůbec bude chodit :-)
Mala technicka poznamka: dost casto se tu resi, ze je problem, kdyz z karty prijde packet ve chvili, kdyz se resetuje. Jak muze poslat nebo prijmout packet, kdyz se resetuje? Nechapu...mela by byt neaktivni, ne?