It's with some hesitation I send this mail to what looks like a developer list. According https://wireless.wiki.kernel.org/en/users/support it should be ok to send end user questions here though. If it's inappropriate just let me know and I'll go away. My setup is a home server which among other things acts as an AP. Recently, some devices have had problems connecting to the network. It works some of the time, but fails other times. The frequency differs between devices, mainly different mobile phones. The message in the phone is that it can't get any IP address. Authentication is via WPA-PSK. As I understand it, the the authentication itself works; at least I get a different behaviour in the phone if I intentionally type the wrong password. But as far as I can tell, the attempt from the phone to ask the DHCP server for an address never arrives to dhcpd. The problems started around the time the server was upgraded with new core hardware components in the machine, including a new wireless network card, which makes me suspect it is somehow related to the driver. The rtl8188ee driver gets selected for the card, and it identifies to "lspci" thus: 05:00.0 Network controller [0280]: Realtek Semiconductor Co., Ltd. RTL8188EE Wireless Network Adapter [10ec:8179] (rev 01) The OS is a Fedora installation using hostapd-2.9-2.fc31.x86_64 kernel-5.3.15-300.fc31.x86_64 In my attempts to debug, I have listened on the interface with "tcpdump". It sees only EAPOL packages. 11:44:44.362818 EAPOL key (3) v2, len 95 11:44:44.389702 EAPOL key (3) v1, len 117 11:44:44.389951 EAPOL key (3) v2, len 151 11:44:44.408127 EAPOL key (3) v1, len 95 11:44:44.408324 14:1f:78:3e:76:5c (oui Unknown) > Broadcast Null Unnumbered, xid, Flags [Response], length 6: 01 00 11:45:20.771775 EAPOL key (3) v2, len 95 11:45:21.772681 EAPOL key (3) v2, len 95 11:45:21.788413 EAPOL key (3) v1, len 117 11:45:21.788721 EAPOL key (3) v2, len 151 11:45:21.824727 EAPOL key (3) v1, len 95 11:45:21.825064 14:1f:78:3e:76:5c (oui Unknown) > Broadcast Null Unnumbered, xid, Flags [Response], length 6: 01 00 When it works, these are immediately (well, a tenth of a second later or so) followed by DHCP packets. Looking in the journal, the pattern is the same. jan 14 11:44:44 mimmi hostapd[2287864]: wlo1: STA 14:1f:78:3e:76:5c IEEE 802.11: authenticated jan 14 11:44:44 mimmi hostapd[2287864]: wlo1: STA 14:1f:78:3e:76:5c IEEE 802.11: associated (aid 2) jan 14 11:44:44 mimmi hostapd[2287864]: wlo1: STA 14:1f:78:3e:76:5c RADIUS: starting accounting session D15F3A088F96AE62 jan 14 11:44:44 mimmi hostapd[2287864]: wlo1: STA 14:1f:78:3e:76:5c WPA: pairwise key handshake completed (RSN) jan 14 11:44:44 mimmi kernel: rtlwifi: -----hwsec_cam_bitmap: 0x10 entry_idx=5 Again, when it works, these lines are immediately followed by logs from dhcpd, but nothing like this appears here. At first I was a bit suspicious about the rtlwifi log from the kernel. After googling it, it seems to be a red herring, though. Additionally, such kernel lines appear also when the connection is successful. This would seem to indicate the phone was misbehaving, but why would that happen only some of the time, and to several different devices starting at the same time, if with different frequencies? In addition, when I try to look at logs on that side using "adb logcat", it DOES say it sends a DHCPDISCOVER. 01-14 11:44:48.389 2684 17585 D DhcpClient: Broadcasting DHCPDISCOVER The full log is quite large why I don't attach it, but it is available at ftp://ftp.uddeborg.se/pub/wifi/adb-logcat.xz I also tried to debug using the hostapd_cli command. It doesn't tell me any more than the above, but is available as ftp://ftp.uddeborg.se/pub/wifi/hostapd_cli It seems to me the DHCPDISCOVER package disappears somewhere in the hardware itself, or in the hardware driver before it reaches the more generic network code where wireshark/tcpdump would see it. But my understanding of these things is pretty vague. I haven't been able to find any probable pattern when it works and when it doesn't. Does anyone have any suggestion what could be wrong, how to debug it further, or any other tips on what to do?
