net_server crashes from marvell_yukon driver

[ArmanHayots]

Haiku hrev57424, x86_64

After installation of Intel AC-6235 module sometimes net_server crashes, network monitor shows wlan as disconnected and no networks available.

Syslog:

KERN: wlan_control: 9235, 15
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9235, 15
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9234, 103
KERN: wlan_open(0xffffffff86c67000)
KERN: [net/iprowifi4965/0] ieee80211_init
KERN: [net/iprowifi4965/0] start running, 0 vaps running
KERN: ieee80211_notify_ifnet_change not implemented, yet.
KERN: [net/iprowifi4965/0] ieee80211_start_locked: up parent 
KERN: [net/iprowifi4965/0] start running, 1 vaps running
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: INIT -> SCAN (nrunning 0 nscanning 0)
KERN: [iprowifi4965] (iwn) iwn_read_firmware: could not read firmware iwn6000g2bfw
KERN: [iprowifi4965] (iwn) iwn_init_locked: could not read firmware, error -2147483643
KERN: wlan_control: 9234, 103
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: pending INIT -> SCAN transition lost
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: INIT -> SCAN (nrunning 0 nscanning 0)
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9235, 15
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9234, 103
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: pending INIT -> SCAN transition lost
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: INIT -> SCAN (nrunning 0 nscanning 0)
KERN: [net/iprowifi4965/0] stop running, 1 vaps running
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: pending INIT -> SCAN transition lost
KERN: [net/iprowifi4965/0] ieee80211_new_state_locked: INIT -> INIT (nrunning 0 nscanning 0)
KERN: ieee80211_notify_ifnet_change not implemented, yet.
KERN: [net/iprowifi4965/0] down parent 
KERN: [net/iprowifi4965/0] ieee80211_newstate_cb: INIT -> INIT arg -1
KERN: [net/iprowifi4965/0] sta_newstate: INIT -> INIT (-1)
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9235, 15
KERN: wlan_control: 9235, 76
KERN: wlan_control: 9235, 15
KERN: wlan_control: 9235, 76
KERN: vm_page_fault: vm_soft_fault returned error 'Bad address' on fault at 0x0, ip 0xef558171e0, write 1, user 1, exec 0, thread 0x20d
KERN: debug_server: Thread 525 entered the debugger: Segment violation
KERN: /dev/net/marvell_yukon/0: link up, media 0x900026 quality 1000 speed 100000000
KERN: vm_page_fault: vm_soft_fault returned error 'Bad address' on fault at 0x0, ip 0xef558171e0, write 1, user 1, exec 0, thread 0x210
KERN: stack trace, current PC 0xef558171e0  </boot/system/lib/libbe.so> _ZN8BMessageC2Ev + 0x10:
KERN:   (0x7fe482c52c30)  0xef55814655  </boot/system/lib/libbe.so> _ZN7BLooper19ReadMessageFromPortEl + 0x35
KERN:   (0x7fe482c52c60)  0xef55815c62  </boot/system/lib/libbe.so> _ZN7BLooper11task_looperEv + 0x62
KERN:   (0x7fe482c52cb0)  0xef55815421  </boot/system/lib/libbe.so> _ZN7BLooper7_task0_EPv + 0x21
KERN:   (0x7fe482c52cd0)  0x14a22b8cdd9  </boot/system/lib/libroot.so> thread_entry + 0x19
DAEMON 'DHCP': /dev/net/marvell_yukon/0: Send DHCP_DISCOVER to 255.255.255.255:67
DAEMON 'DHCP': /dev/net/marvell_yukon/0: Received DHCP_OFFER from 192.168.137.1
DAEMON 'DHCP':   your_address: 192.168.137.34
DAEMON 'DHCP':   server: 192.168.137.1
DAEMON 'DHCP':   subnet: 255.255.255.0
DAEMON 'DHCP':   gateway: 192.168.137.1
DAEMON 'DHCP':   nameserver[0]: 192.168.137.1
DAEMON 'DHCP':   renewal time: 302400 seconds
DAEMON 'DHCP':   rebinding time: 453600 seconds
DAEMON 'DHCP':   lease time: 604800 seconds
DAEMON 'DHCP':   UNKNOWN OPTION 46 (0x2e)
DAEMON 'DHCP':   domain name: "mshome.net"
DAEMON 'DHCP': /dev/net/marvell_yukon/0: Send DHCP_REQUEST for 192.168.137.34 to 255.255.255.255:67
DAEMON 'DHCP': /dev/net/marvell_yukon/0: Received DHCP_ACK from 192.168.137.1
DAEMON 'DHCP':   server: 192.168.137.1
DAEMON 'DHCP':   subnet: 255.255.255.0
DAEMON 'DHCP':   gateway: 192.168.137.1
DAEMON 'DHCP':   nameserver[0]: 192.168.137.1
DAEMON 'DHCP':   renewal time: 302400 seconds
DAEMON 'DHCP':   rebinding time: 453600 seconds
DAEMON 'DHCP':   lease time: 604800 seconds
DAEMON 'DHCP':   UNKNOWN OPTION 46 (0x2e)
DAEMON 'DHCP':   domain name: "mshome.net"
DAEMON 'DHCP': /dev/net/marvell_yukon/0: DHCP status = No error

Additional info:

~> listdev | grep Wireless
      device Wireless (Radio Frequency, Bluetooth) [e0|1|1]
        device 07da: Centrino Bluetooth Wireless Transceiver
~> listdev | grep Network
device Network controller [2|80|0]
device Network controller (Ethernet controller) [2|0|0]
~> listimage | grep dev/net
 2828 0xffffffff81af7000 0xffffffff81b8b000     0          0 /boot/system/add-ons/kernel/drivers/dev/net/iprowifi4965
 3476 0xffffffff85fb0000 0xffffffff85fb4000     0          0 /boot/system/add-ons/kernel/drivers/dev/net/tun_driver
 3498 0xffffffff85fb5000 0xffffffff85fe0000     0          0 /boot/system/add-ons/kernel/drivers/dev/net/marvell_yukon

After installation of intel_wifi package from depot net_server started to regular crash, but WiFi-networks appeared without ability to connect due to "Bad port ID" error. The removal of intel_wifi package over depot requires to uninstall most of installed (and system too) packages, so only manual uninstallation works. After package removal network monitor shows no available networks again.

[korli]

The net_server crashes are probably because of marvell_yukon. I think one of my laptops also has net_server crashes and has a marvell_yukon device. Please try to block the marvell_yukon driver in the boot loader, and see if this helps with net_server crashes.

Please also provide a syslog.

[ArmanHayots]

System succesfully ignored my /boot/settings/packages file, so I've to disable marvell_yukon using safe mode and nothing changed. In other hand, removing bv43_cutter package and after then installation of intel_wifi done the work - system now can see wifi-networks without regular net_server crash (at least not too frequently), but unable to connect to any network due to wrong password error. I'm attaching all 3 syslogs - initial, with marvell disabled and current, and packages file too.

[ArmanHayots]

With disabled marvel_yukon all net_server crashes are gone (I tested for near a half of a hour), but wifi still unable to connect, showing «Bad password». wpa_supplicant log the same.

[waddlesplash]

/dev/net/iprowifi4965/0: Failed to attach pkt_type filter

I can't recall seeing this message before; it might be the problem, or it might be harmless. Can you test with beta4 and see what happens?

[waddlesplash]

korli: Any chance you can debug the crashes a bit further? It looks like an invalid BMessage is trying to be converted; that shouldn't crash applications even if the kernel is sending something wrong.

[ArmanHayots]

Will try. Also sometimes system falls in KDL without ability to use it — I can erase all screen using Backspace and not of any printed commands nor shortcuts working. Syslog shows nothing for this situation. Mouse works and sometimes if Deskbar not busy (like in attached screenshot) I can kill net_server and continue to work, but usually only reboot helps.

[waddlesplash]

Are you sure it's KDL and not an app_server crash?

Maybe the marvell_yukon driver is corrupting memory somehow... that would be very bad.

[waddlesplash]

(Or if you've disabled marvell_yukon entirely then that's not the issue.)

[waddlesplash]

Yes, the screenshot looks like the remnants of a Debugger prompt and not a KDL one. You can try typing "save-report" at the prompt, it should save a report onto the Desktop which should be accessible after rebooting.

[ArmanHayots]

"save-report" created nothing. Attaching link to core dump file on net_server crash, hope it helps. Still unable to connect. ​https://disk.yandex.ru/d/HAzqAn9oUnDUYg

Maybe the marvell_yukon driver is corrupting memory somehow... that would be very bad.

Which started only after module installation. More looks like two drivers (Intel and Marvell) fighting for same memory.

[ArmanHayots]

Performed fresh install of the latest nightly build (57427). All crashes gone and even WiFi successfully connects out of the box, but no Internet access available — all packets lost. After reboot again attacked with wrong WiFi passwords error (and syslog started to overflow with "wlan()" messages again). Connected back to Ethernet and then updated all packages to the latest and then the problem gone — at least for now. WiFi successfully connecting and WebPositive showing sites. After reboot Wifi password error started again, but after disabling Ethernet device all worked.

What I want to say as offtopic.

1. It's will be good to have a tool or script to restore all bundled packages in the state of newly installed system according to build number — even over Depot.
2. Maybe we should provide some default public DNS addresses — like Google/Yandex/OpenDNS?
3. It's strange that driver packages like intel_firmware, realtek or bw43_cutter on installation trying to drag down with them half of system.
4. Package blocking over /boot/settings/packages and save-report sometimes not works (in my case never did) and I have no clue why.

[ArmanHayots]

*bw43_cutter on uninstallation, I mean

WiFi still working, no net_server crashes observed. I should continue testing and will report on the next week.

[ArmanHayots]

Monday report. Wifi only works when RJ45 disabled via Network manager. And sometimes password error starts to bother again so much that only wpa_supplicant kill or system reboot helps. No net_server crash anymore, but one time system stuck at logo screen (will attach syslog.old, but it can be unrelated). Depot app crashed one time, will attach report too. System updated to rev57434.

[ArmanHayots]

1. System not remebering interface state so I need to disable Marvell device on every start.
2. Sometimes it won't connect to my network and just stuck on "Confiuguring" state so I need to kill wpa_supplicant (with net_server sometimes too), connect to another network and then connect back to mine.

[ArmanHayots]

Updated to hrev57458. Marvell shows in interfaces popup even after disabling but in other hand blacklisting over packages started to work again so I'm just disabled marvell_yukon. Password error became permanent and unavoidable so WiFi not usable again. Syslog full again of "KERN: wlan_control: 9234, 16/17/26/103".

[ArmanHayots]

Able to get to work using manual wpa_supplicant start over Terminal (already started wpa_supplicant must be killed). Will attach both syslog and wpa log.