On 10/02/17 00:05, Vincent Danjean wrote: > Le 09/02/2017 à 23:20, Andrew Shadura a écrit : >> Hi, >> >> On 9 February 2017 at 21:10, Vincent Danjean <vdanj...@debian.org> wrote: >>> Package: hostapd >>> Version: 1:2.5-2+v2.4-3+b1 >>> Severity: grave >>> Justification: renders package unusable >>> >>> Hi, >>> >>> I just upgraded a machine from jessie to stretch. With the stretch >>> version of hostapd, the deamon does not start: >>> # hostapd /etc/hostapd/hostapd.conf >>> Configuration file: /etc/hostapd/hostapd.conf >>> Using interface wlan0 with hwaddr f4:ec:38:c5:fd:94 and ssid "dino-wpa1" >>> Failed to set beacon parameters >>> Interface initialization failed >>> wlan0: interface state UNINITIALIZED->DISABLED >>> wlan0: AP-DISABLED >>> wlan0: Unable to setup interface. >>> wlan0: interface state DISABLED->DISABLED >>> wlan0: AP-DISABLED >>> hostapd_free_hapd_data: Interface wlan0_2 wasn't started >>> wlan0: AP-DISABLED >>> hostapd_free_hapd_data: Interface wlan0_1 wasn't started >>> wlan0: AP-DISABLED >>> hostapd_free_hapd_data: Interface wlan0 wasn't started >>> nl80211: deinit ifname=wlan0 disabled_11b_rates=0 >>> # >>> >>> If I downgrade the hostapd package to the jessie version >>> (1:2.3-1+deb8u4), then it works perfectly: >>> # apt-get install --reinstall hostapd/jessie >>> [...] >>> Les paquets suivants seront mis à une VERSION INFÉRIEURE : >>> hostapd >>> 0 mis à jour, 0 nouvellement installés, 1 remis à une version inférieure, 0 >>> à enlever et 44 non mis à jour. >>> [...] >>> Préparation du dépaquetage de .../hostapd_1%3a2.3-1+deb8u4_amd64.deb ... >>> Dépaquetage de hostapd (1:2.3-1+deb8u4) sur (1:2.5-2+v2.4-3+b1) ... >>> Paramétrage de hostapd (1:2.3-1+deb8u4) ... >>> [...] >>> # hostapd /etc/hostapd/hostapd.conf >>> Configuration file: /etc/hostapd/hostapd.conf >>> Using interface wlan0 with hwaddr f4:ec:38:c5:fd:94 and ssid "dino-wpa1" >>> Using interface wlan0_1 with hwaddr f4:ec:38:c5:fd:95 and ssid "dino" >>> Using interface wlan0_2 with hwaddr f4:ec:38:c5:fd:96 and ssid >>> "dino-guess" >>> wlan0: interface state UNINITIALIZED->ENABLED >>> wlan0: AP-ENABLED >>> [.. and then authentifications from client...] >>> >>> So, it seems there is a major regression in the hostapd package. >>> Feel free to ask more information if required. >> >> I'm going to ask you to tell me what hardware you use. > > From lspci: > 04:00.0 Network controller: Qualcomm Atheros AR9285 Wireless Network Adapter > (PCI-Express) (rev 01) > If you need more information, just tell me where to look at. > >> Also, please try >> and get more logs please, especially those wpa_supplicant produces with >> -dt (you can add it with a systemd override file). > > I do not use wpa_supplicant. I only run hostapd. I use this computer > (and this wireless card) as an AP at home. > You can find in attachment the output of > # hostapd -dd /etc/hostapd/hostapd.conf > - for the testing version > - for the stable version > - for the unstable version
Here, I meant hostapd. However, wpa_supplicant and hostapd are two parts of the same software package, and they use the same drivers AFAIK. >>> Note that the unstable version works. So I will mark this bug as fixed >>> for the 1:2.6-3 version as soon as I get its number >> >> That's pretty much a bad news, as 2.6 breaks networking for lot more >> people, it seems. >> >> There's a thread about a related issue: >> http://lists.infradead.org/pipermail/hostap/2017-January/037042.html >> >> Please have a look and let me know whether what you're observing is the >> same bug or not, and whether the patch works. > > I'm not using wpa_supplicant, so it does not seem the same bug. > Moreover, the proposed patch in > http://lists.infradead.org/pipermail/hostap/2017-January/037060.html > do not apply at all for the testing sources. > In the testing sources, in the wpa_driver_nl80211_set_ap function, > the "if (beacon_set)" test do not have any "else" clause to patch. That doesn't sound good. Maybe try emailing the hostap mailing list? -- Cheers, Andrew