well seems it cannot manage wpa connection http://svnweb.mageia.org/soft/drakx/trunk/mdk-stage1/wireless.c?view=markup

Summary: Network installer stage 1 asks for WEP key on WPA2 encrypted wifi. => Network installer stage 1 doesn't support WPA2 encrypted wifi
Source RPM: (none) => drakx-installer-images

*** Bug 12071 has been marked as a duplicate of this bug. ***

patch welcome :-)

Hardware: i586 => All

I suppose this bug is still valid. Can you confirm?

Severity: normal => major
Keywords: (none) => NEEDINFO, Triaged
Assignee: bugsquad => thierry.vignaud

Yes bug is still valid not supportted still in stage1 netinstall.

CC: (none) => ozkyster

Stage1 only support WEP for now

There seem to be problem even for WEP now using ipw2200 : bug 21042

*** Bug 23390 has been marked as a duplicate of this bug. ***

CC: (none) => hamnisdude

*** Bug 23752 has been marked as a duplicate of this bug. ***

still valid :(

CC: (none) => mageia

And this is still valid for Netinstall nonfree ISO for Cauldron (M8).

Workaround:
1) use WEP on your Router during install... not satisfying from a security point of view.
2) use a RJ45-cable during install and use WiFi WPA2 after in Desktop Environment.
3) use Classic ISO for installing offline and go online after with Desktop environment and do updates.

CC: (none) => ouaurelien

Please try the latest cauldron netinstall. You'll need to use the nonfree version for most wireless devices.

Tested and working with my access point using:

  - No security
  - WEP Open Key (40-bit)
  - WEP Shared Key (40-bit)
  - WPA-Personal with TKIP encryption
  - WPA2-Personal with TKIP encryption
  - WPA2-Personal with AES encryption

Assignee: mageiatools => mageia
CC: (none) => mageia
Status: NEW => ASSIGNED

Using network install, nonfree i586 fetched last hour.

Machine: Thinkpad T43 with Intel PRO/Wireless 2200; [Calexico2]; ipw2200

Tried to connect to two routers using WPA2-Personal (TKIP or AES i have forgotten).  I am entering the same ESSID and key as i do in my mobile phone at the same time.

Error popup say
 unable to start wpa_supplicant daemon for interface "wlp11s2": 65280

/var/log/wpa_supplicant say:
 Driver does not support authentication/association or connect commands

Alt-F4 for kernel messages do not work (Bug 21041).

Same 32 bit nonfree netinstall iso works with same wifi router using WPA2-Personal in Dell Dimension M4400; different wifi hardware.   :)

Well done!

So we just have to figure out why it do not work on ipw2200 (and maybe others) - that same T43 works great with that card and Mageia for then years +


Sidenote: My preferred mirror http://ftp.acc.umu.se/ was not listed.  It is a predefined list that need to be updated or is it some automagics that glitched?

(In reply to Morgan Leijström from comment #14)
> So we just have to figure out why it do not work on ipw2200 (and maybe
> others) - that same T43 works great with that card and Mageia for then years
> +

Could be there's a kernel module missing on the ISO. In the installer, use Alt-F2 to switch to the debug shell and type 'dmesg | less' to view  the kernel messages and see if there are any hints there.

Also in the debug shell, try running

  /usr/sbin/wpa_supplicant -Dwext -i wlp11s2 -c /etc/wpa_supplicant.conf

to see if that works any better.

Also attach the output from 'lsmod' when running on that machine with working wireless.

> Sidenote: My preferred mirror http://ftp.acc.umu.se/ was not listed.  It is
> a predefined list that need to be updated or is it some automagics that
> glitched?

The mirror list is fetched from mirrors.mageia.org.

(In reply to Martin Whitaker from comment #15)
> Could be there's a kernel module missing on the ISO. In the installer, use
> Alt-F2 to switch to the debug shell and type 'dmesg | less' to view  the
> kernel messages and see if there are any hints there.

No hints i can see.  Even not after exercising the commands per below i see anything regarding driver.  Possibly this, rather early in boot:

platform regulatory.0: Direct firmware load for regulatory.db failed with error -2


> Also in the debug shell, try running
> 
>   /usr/sbin/wpa_supplicant -Dwext -i wlp11s2 -c /etc/wpa_supplicant.conf

sh-5.0# /usr/sbin/wpa_supplicant -Dwext -i wlp11s2 -c /etc/wpa_supplicant.conf
Successfully initialized wpa_supplicant
Line 9: unknown global field 'Successfully initialized wpa_supplicant'.
Line 9: Invalid configuration line 'Successfully initialized wpa_supplicant'.
Failed to read or parse configuration '/etc/wpa_supplicant.conf'.
sh-5.0#

I saw in Alt-F3 terminal what command it executed, and i tried the same:

sh-5.0# /usr/sbin/wpa_supplicant -B -i wlp11s2 -c /etc/wpa_supplicant.conf -f /var/log/wpa_supplicant.log -P /var/run/wpa_supplicant.pid


/var/log/wpa_supplicant.log now consist of same four lines as output of previous command, and before them:

nl80211: Driver does not support authentication/association or connect commands
nl80211: deinit ifname=wlp11s2 disabled_11b_rates=0
wlp11s2: Failed to initialize driver interface


> Also attach the output from 'lsmod' when running on that machine with
> working wireless.

Coming.

I realise my instructions weren't precise enough - you need to run the command I specified after entering your wireless details and getting the failure message, otherwise /etc/wpa_supplicant.conf won't have been created.

In the debug shell you can mount one of the partitions on your hard disk, e.g. something like

  mkdir /mnt
  mount /dev/sda1 /mnt

(replace sda1 with a suitable partition). Then you can copy debug info to /mnt. Best to manually

  umount /mnt

when you are done.

Can you save a copy of /etc/wpa_supplicant.conf, /var/log/wpa_supplicant.log,  and the output from dmesg and attach them here (redact your passphrase in /etc/wpa_supplicant.conf)

Created attachment 12039 [details]
Output of lsmod from my T43 booted on 7.1 Live i586 and wifi working.

I didn't need to look far down the list to see the answer :-)
Those lib80211_crypt_* modules aren't on the netinstall ISO.

Good  :)

So no more test needed until next spin?

thanks martin for working on this.

When fully done, this should also be added to release notes for Mageia 8

Keywords: (none) => FOR_RELEASENOTES8

Morgan, new build in progress, so please test again when it reaches the mirrors.

(In reply to Thomas Backlund from comment #21)
> thanks martin for working on this.

+1

> When fully done, this should also be added to release notes for Mageia 8

Actually this is already done

The "bad" things are:
- I'ven't explained it will mostly work with all-nonfree.rdz or with Mageia-Cauldron-netinstall-x86_64.iso due to lack of firmwares
- that classic ISOs is based off !nonfree boot.iso and thus suffer the same issue (not as important as adding extra repos over wifi would be less common)

Keywords: FOR_RELEASENOTES8 => (none)

We could maybe use sg like ../perl-install/list_firmware.pm & http://gitweb.mageia.org/software/drakx/tree/perl-install/install/any.pm#n397 in order to report that a firmware is needed and it won't work w/o boot-nonfree.iso ?

@Thierry, when/where would you report that? At least with my hardware, loading the wireless module fails if the firmware isn't found, so no network device is added, and stage1 doesn't ask which device you want to use. We'd have to insert an extra dialogue, which no doubt would lead to complaints.

The lack of working wireless on the classic ISOs is a perennial source of complaints from our users. We could add an extra boot menu option to enable use of nonfree drivers/firmware (like what we have on the Live ISOs to enable the nvidia drivers). Something to consider for Mageia 9 perhaps.

@Morgan, the new images are now built and have reached distrib-coffee.

Yes, even umu.se. already.

And well done Martin!  The same setup as in Comment 13 now works.

One little oddity: I have two wifi routers, and tested sitting in the same spot where i know it normally have best signal strength and performance to connect to when i used Live Mageia 7.1, but the installer got timeout during DHCP.  Then I tried the more remote router, and that worked.  I will try to check up if they differ in TKIP/AES.

Is it time to call in other testers?

BTW, one thing that could be improved - easily i believe - is to sort the list of mirrors alphabetically.

( Another improvement if it could search and list found wifi ESSIDs, but that is more work, and user normally knows his networks names. )

Problem with TKIP:

Using router D-Link DIR-868L Firmware Version : 1.09
Selecting "WPA Personal" (WPA-PSK), "WPA2 Only", "TKIP" (only): do not work :(
After a while Error dialogue "No DHCP reply received".
Last line from dmesg:
 lib80211_crypt: registered algorithm 'TKIP'
So it seem to detect correct protocol, but ??

Changing TKIP to AES, current netinstaller nonfree i586 works :)
This is when using wifi adapter ipw2200, in Thinkpad T43.

An updated Mageia 7 Live persistence USB stick booted on same machine works with both TKIP and AES.

Other settings used in both Live and installer: DHCP, no proxy.
Tested TKIP again just to be sure. Same negative result.
All tests performed four metres from the router.

So some bit for TKIP is still missing in netinstaller iso, I believe.

Can you attach
1. dmesg output and /var/log/wpa_supplicant.log from installer when using TKIP
   (see comment 17 for how).
2. dmesg and lsmod output from Live when using TKIP.

Created attachment 12045 [details]
dmesg from installer, TKIP fail

Created attachment 12046 [details]
wpa_supplicant.log from installer, TKIP fail

Created attachment 12047 [details]
wpa_supplicant.conf from installer, TKIP fail

Created attachment 12048 [details]
dmesg from mga7 Live, TKIP works

Created attachment 12049 [details]
lsmod from mga7 Live, TKIP works

(mga7.1 Live Xfce i586 persistent, configured and rebooted several times but system is not updated)

I think I've identified the missing module. Please try the new build when it becomes available.

Yes!  Good work.  It works with TKIP now.

I am calling other testers from QA to test it more, because this is a significant step ahead and as tmb say should be in release notes, so we should test it thoroughly.

What I understand, for this bug testers only need to see that stage 2 get loaded.

To test: different wifi hardware and protocols.

Right Martin?

BTW should we need to test WEP more?   Bug 21042

Target Milestone: --- => Mageia 8

Tested Mageia-Cauldron-netinstall-nonfree-x86_64.iso from 2020-Dec-03 02:01:34.

Dumped this to a USB drive.
Rebooted the machine.
Menus goes through selecting Network type connection.

BUT this always try to connect via Ethernet card (on the motherboard) than try to setup the WiFi present on a PCIE card. Not that Ethernet card has no wire in it!
(RJ45 was unplugged while testing this).

My WiFi is well functioning under an installed M8/M7 system.

Hardware as see by inxi -nn command:

Device-1: Intel Ethernet I219-V driver: e1000e 
IF: enp0s31f6 state: down

Device-2: Intel Wi-Fi 6 AX200 driver: iwlwifi 
IF: wlp5s0 state: up

stage1 never proposes to setup WiFI.

Stop testing this.

Good you found that.
Back to Martin then.

---

When Cauldron netinstaller is tested enough, would it be much work to update mga7 netinstaller?

Could be something to blog about, especially as mga7 isos are getting old.

@Aurelian, please attach output from 'lsmod' on Mageia 8 system with wireless working.

(In reply to Martin Whitaker from comment #25)
> @Thierry, when/where would you report that? At least with my hardware,
> loading the wireless module fails if the firmware isn't found, so no network
> device is added, and stage1 doesn't ask which device you want to use. We'd
> have to insert an extra dialogue, which no doubt would lead to complaints.

Since we have the list of the modules that need a firmware, we could check first to see if the firmware exists and at least log one line about the situation and a 2nd line advising to use boot-nonfree instead, in order to help advanced users…

(In reply to Aurelien Oudelet from comment #41)
> Here is the requested output:

When as ask you attach, please attach a file instead of pasting hundred of lines that makes the bug report unreadable

All tests were done with a home network on a Linksys EA7400 router, using WPA2-PSK encryption.

Tested the 32-bit non-free iso on a Dell inspiron 5100 with an Atheros AR2413/AR2414 Wireless Network Adapter. It worked perfectly, detecting both Ethernet and wifi devices, with no Ethernet cable attached.

Tested the 64-bit non-free iso on a Dell Dimension e520 with a Realtec rtl8192cu usb dongle. It worked perfectly, detecting both Ethernet and wifi devices, with no Ethernet cable attached.

Tested the 64-bit non-free iso on an AMD Phenom II X4 desktop with an Atheros AR92BX PCI Express card and an Ethernet cable connected to a different, older router. It worked perfectly, detecting both devices, connecting with wifi even with an Ethernet cable attached.

Tested both non-free isos on an HP Probook 6550b, with an Intel Centrino Advanced N 6200 wifi adapter, and no Ethernet cable attached. Both failed, apparently not detecting the wifi adapter at all. I was not asked which device to use. In contrast, using the M8 Beta1 Live Xfce iso in Live mode, the net_applet has no problem detecting the wifi adapter, and connecting to the network.

Tested the Probook again, this time plugging a Ralink mt7601u dongle into a usb port. That worked perfectly, with the usb dongle and Ethernet devices detected, but still not the internal Intel wifi adapter.

Thanks for the tests TJ. The failure with the Probook is most likely a missing kernel driver or subsidiary kernel module. Please could you attach the output of 'lsmod' when running in Live mode with the Intel wireless working.

Is there a way to add all wifi related modules/drivers in one go instead of finding one missing at a time...?  Maybe sneak peaking on what is included on the live systems?

Interesting that, so far, both of the wifi devices that have failed are Intel devices...

Created attachment 12053 [details]
Output of lsmod from an M8 Live Plasma with wifi working on a Probook

I can't see any likely candidates for missing kernel modules in TJ's lsmod output.
The cmac crypto module is a possibility for Aurelien.

Looking at the contents of iwlwifi-firmware-20201118-1.mga8.nonfree.noarch.rpm, we only include a subset of the available firmware.

@Thierry, is there some policy on that?

@Aurelien and TJ, in the installer, when you get to the "Please select your network connection type" dialogue, press Alt-F2 to get to the debug console and type "dmesg | grep iwl". There should be an error message if the problem is missing firmware.

Created attachment 12054 [details]
output of lsmod with a working Cauldron with WiFi

Obsoleting Comment 41
Added lsmod.txt file here.

$ inxi -n
Network:   Device-1: Intel Ethernet I219-V driver: e1000e 
           IF: enp0s31f6 state: down mac: 1c:1b:0d:66:1d:12 
           Device-2: Intel Wi-Fi 6 AX200 driver: iwlwifi 
           IF: wlp5s0 state: up mac: 14:f6:d8:66:c6:1e

Output of "dmesg | grep iwl" while booting Netinstall nonfree Cauldron ISO from 3-12-2020 here: https://i.imgur.com/xDB7qX5.jpg (Too large to fit bugzilla).

Below is same output from a working Cauldron installation on SAME hardware.

$ dmesg | grep iwl
[    2.137387] iwlwifi 0000:05:00.0: enabling device (0000 -> 0002)
[    2.146153] iwlwifi 0000:05:00.0: Direct firmware load for iwlwifi-cc-a0-56.ucode failed with error -2
[    2.147335] iwlwifi 0000:05:00.0: api flags index 2 larger than supported by driver
[    2.147343] iwlwifi 0000:05:00.0: TLV_FW_FSEQ_VERSION: FSEQ Version: 89.3.35.22
[    2.147344] iwlwifi 0000:05:00.0: Found debug destination: EXTERNAL_DRAM
[    2.147345] iwlwifi 0000:05:00.0: Found debug configuration: 0
[    2.147497] iwlwifi 0000:05:00.0: loaded firmware version 55.d9698065.0 cc-a0-55.ucode op_mode iwlmvm
[    2.198926] iwlwifi 0000:05:00.0: Detected Intel(R) Wi-Fi 6 AX200 160MHz, REV=0x340
[    2.379708] iwlwifi 0000:05:00.0: base HW address: 14:f6:d8:66:c6:1e
[    2.395883] iwlwifi 0000:05:00.0 wlp5s0: renamed from wlan0

Very strange, isn't it? Same packages, same hardware,...

Created attachment 12055 [details]
Output of "dmesg | grep iwl" on a Probook

(Used The Gimp to crop and scale the photo to make the file small enough)

Note that there are fewer messages on mine, but that the actual fault seems to be the same - the proper firmware wasn't found.

One relatively minor niggle: when the non-free iso is dumped to a USB stick by Isodumper, the stick label is the same as the one for the free version. I think "non-free" should be added to that label, so one might know the difference at a later date if one neglected to put a physical label on the device.

Following Aurelien's lead, I booted into Live mode from the Beta1 Plasma iso, used the net_applet to configure wifi and connect, then immediately used the "dmesg | grep iwl" command. This is the result:

live@localhost ~]$ dmesg | grep iwl
[   17.302004] iwlwifi 0000:43:00.0: can't disable ASPM; OS doesn't have ASPM control
[   17.318667] iwlwifi 0000:43:00.0: Direct firmware load for iwlwifi-6000-6.ucode failed with error -2
[   17.318799] iwlwifi 0000:43:00.0: Direct firmware load for iwlwifi-6000-5.ucode failed with error -2
[   17.464843] iwlwifi 0000:43:00.0: loaded firmware version 9.221.4.1 build 25532 6000-4.ucode op_mode iwldvm
[   17.465227] iwlwifi 0000:43:00.0: Direct firmware load for iwl-debug-yoyo.bin failed with error -2
[   17.585425] iwlwifi 0000:43:00.0: CONFIG_IWLWIFI_DEBUG enabled
[   17.585428] iwlwifi 0000:43:00.0: CONFIG_IWLWIFI_DEBUGFS disabled
[   17.585430] iwlwifi 0000:43:00.0: CONFIG_IWLWIFI_DEVICE_TRACING disabled
[   17.585432] iwlwifi 0000:43:00.0: Detected Intel(R) Centrino(R) Advanced-N 6200 AGN, REV=0x74
[   17.619235] ieee80211 phy0: Selected rate control algorithm 'iwl-agn-rs'
[   17.880348] iwlwifi 0000:43:00.0 wlo1: renamed from wlan0
[   37.655136] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[   37.890798] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[   38.040251] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[   38.274632] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[  187.465531] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[  187.700937] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[  187.811130] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1
[  188.046220] iwlwifi 0000:43:00.0: Radio type=0x1-0x3-0x1

(In reply to Martin Whitaker from comment #49)
> Looking at the contents of
> iwlwifi-firmware-20201118-1.mga8.nonfree.noarch.rpm, we only include a
> subset of the available firmware.
> 
> @Thierry, is there some policy on that?

In the old days we only filtered ou some:
See http://gitweb.mageia.org/software/drakx/commit/images/make_boot_img?id=77964056ffe63d87bee4e4afe8ddec541b0b5ce2 for the logic

But now this is handled by dracut

Humm, I just added a BR on iwlwifi-firmware in the spec file.

It might be interested to look at any other missing firmware package as BR

(In reply to Thomas Andrews from comment #53)
> One relatively minor niggle: when the non-free iso is dumped to a USB stick
> by Isodumper, the stick label is the same as the one for the free version. I
> think "non-free" should be added to that label, so one might know the
> difference at a later date if one neglected to put a physical label on the
> device.

# blkid /dev/sdf*
/dev/sdf: UUID="2020-01-30-05-36-22-00" LABEL="Mageia-8-x86_64-netinstall" TYPE="iso9660" PTUUID="365ad177" PTTYPE="dos"
/dev/sdf1: UUID="2020-01-30-05-36-22-00" LABEL="Mageia-8-x86_64-netinstall" TYPE="iso9660" PTUUID="365ad177" PTTYPE="dos" PARTUUID="365ad177-01"
/dev/sdf2: SEC_TYPE="msdos" UUID="B569-A60E" TYPE="vfat" PARTUUID="365ad177-02"
/dev/sdf3: UUID="2ce10a8f-8c6d-48ed-9f75-c86aa99976f8" TYPE="crypto_LUKS" PARTUUID="365ad177-03"

From "man genisoimage" ...
-V volid
              Specifies the volume ID (volume name or label) to be written into the master block.  There is space for 32 characters.

The current label is 26 characters. I don't know if the 32 character limit
includes the trailing null in the string or not. Either way there is non enough
room to add "-nonfree", so perhaps just add "-nf" so the label becomes
Mageia-8-x86_64-netinstall-nf

CC: (none) => davidwhodgins

Created attachment 12060 [details]
Missing firmware reported by dracut when building installer images

The attached list was generated by adding --debug to the dracut command and grepping for "missing firmware".

Looking at the iwlwifi driver in particular, the problem appears to be that the driver will accept a range of firmware versions, but modinfo only reports a single (latest?) version. So for example, TJ's screenshot shows the driver searching for iwlwifi-6000-4.ucode .. iwlwifi-6006-6.ucode. But

% /sbin/modinfo iwlwifi | grep iwlwifi-6000-
firmware:       iwlwifi-6000-6.ucode

% urpmq -l iwlwifi-firmware-20201118-1.mga8.nonfree.noarch.rpm | grep iwlwifi-6000-
/lib/firmware/iwlwifi-6000-4.ucode

hence dracut doesn't find a match.

It appears the iwlwifi driver supports firmware API versions they have never publicly released, and the driver only reports the latest API version for each firmware type. To my mind, this is a bug in the iwlwifi driver, but I guess the chances of getting that fixed are slim. Google finds me references to iwlwifi-6000-6 not being found going back years.

Without hacking dracut to treat iwlwifi as a special case, I think our only option is to explicitly add /lib/firmware/iwlwifi*.

The latest netinstall ISO should include the missing Intel firmware.

Both nonfree versions are now working as expected on my Probook.

\o/ :)

Thanks for testing TJ. I'm going to set this to fixed. If there are still missing drivers or firmware, they should be separate bug reports.

Status: ASSIGNED => RESOLVED
Resolution: (none) => FIXED

commit 1c9033ec7da14fcbf0377362a9e1ce3f6cba4635
Author: Martin Whitaker <mageia@...>
Date:   Sat Nov 28 13:58:47 2020 +0000

    mgainstaller: add wpa_supplicant to support WPA/WPA2 (mga#9541)
---
 Commit Link:
   https://gitweb.mageia.org/software/mageia-dracut-conf/commit/?id=1c9033ec7da14fcbf0377362a9e1ce3f6cba4635