Hello Dima,
welcome to the Genode mailing list.
From your issue description below I suspect your network configuration
is more complex non-standard than our NIC router component currently supports. For example, we had issues in the past (and this is still a TODO item) with DHCP offers that contained multiple DNS servers (e.g., site-local and internet name resolution).
Beyond my suspicion, we fixed several issues (see below) in the 20.05 release of Genode. You may cherry-pick the commits to try if your issue is fixed already.
I hope Martin Stein (the author of the NIC router) has some hints too and jumps in to help you as I'm on vacation next week.
[1] https://github.com/genodelabs/genode/issues/3716 [2] https://github.com/genodelabs/genode/issues/3717 [3] https://github.com/genodelabs/genode/issues/3686 [4] https://github.com/genodelabs/genode/issues/3681
Regards
Unfortunately, I sent my original response to Dima's private address and not to the mailing list. So, here again for the list:
Hi Dima,
On which commit have you compiled when speaking of 20.03 ? I ask because there are some commits to the NIC Router that entered mainline between 20.02 and 20.05 and that affect DHCP:
dc9074054 nic_router: accept DHCP reply with missing option ba55409c8 nic_router: keep dyn IP when updating domain obj c1b600024 nic_router: don't call DHCP client on requests 0a468a07a nic_router: fix lifetime of DHCP clients
If you don't have them in your 20.03 build you might want to try cherrypicking them.
Furthermore, you might want to use the verbosity/report features of the NIC Router to inspect the problem at runtime. They provide you with deep packet inspection, interface-state tracking, traffic statistics, and routing decisions.
As soon as you have activated your networking in Sculpt you can see the NIC Router in the "Components" graph of the Leitzentrale. Then you can activate "Storage -> [HARD_DISC] -> Inspect" in the graph and go to the "Inspect" tab of the Leitzentrale. In this shell you can copy the NIC Router configuration '/config/managed/nic_router' to '/config/nic_router' to prevent that it is managed by the Sculpt Manager (note that the content of the "Network" menu changes when doing so). Then open '/config/nic_router' with Vim to adapt the configuration to your needs.
The configuration changes are applied as soon as you save the file and without interrupting the NIC Router services. Please have a look at 'os/src/server/nic_router/README' for details about the configuration. I think you're especially interested in all attributes that start with 'verbose' in the chapter "Verbosity" and the <report> tag in the chapter "Configuring reporting functionality". The output generated by the 'verbose' attributes can be found in '/report/log' (intermixed with the rest of the system log) and is normally prefixed with the [name] of the affected routing domain. The output generated by the <report> tag can be found in '/report/runtime/nic_router/state'.
I hope this helps you!? If you have any questions regarding this, please don't hesitate to ask ;)
Cheers, Martin
El 13.08.20 a las 11:00, Dima Sokolov escribió:
Hello. I am experiencing a network issue in Sculpt OS. Wired connection is established, and the computer gets an IP address assigned dynamically (e.g. 192.168.1.6), RAM FS is activated and in use, but I cannot download anything from the depot (using any source). Fetchurl keeps trying, and these messages can be seen in the log: Libc RNG not configured Fetchurl got stuck, respawning (it’s shown multiple times) Packet operation = 0 failed (finally, I get this one) I’m testing on bare metal. I have tried various network card combinations (Intel / Realtek) as well as various different motherboard / CPU configurations – the result is still the same. I tried Sculpt 20.02 and compiled 20.03 from the sources – again, the result is the same. Also, I tried disconnecting and reconnecting – still the same. I have these network cards, that should be compatible as stated in the manual (E1000), but they yield the same results: Intel PRO/1000 MT (PCI) Intel PRO/1000 GT DESKTOP (PCI) Motherboard / CPU combinations: Asus P5W64 WS Pro (classic BIOS, AHCI, VT-x) + Intel Xeon X5460 Asus Sabertooth X58 (classic BIOS, AHCI, VT-x, VT-d) + Intel Xeon X5650 Asus M5A97 LE R2.0 (UEFI BIOS, AHCI, IOMMU) + AMD Phenom II X6 1055T What should I do to fix the problem? Is there anything I missed in configuration? On a side note, if I use bridged networking inside Virtualbox on Ubuntu (trying to rule out motherboard incompatibility), the results are exactly the same. I downloaded Arora using NAT networking and hard drive FS depot, and then tried bridge networking to see if the culprit is in fetchurl, but Arora also cannot access websites. Since I want to test Virtualbox and Seoul on Sculpt OS, and nested virtualization results in a black screen for either of them, I need to go bare metal. Best regards, Dima Sokolov
Genode users mailing list users@lists.genode.org https://lists.genode.org/listinfo/users
Hello Dima,
can you still produce the DNS errors with Cloudflare 1.1.1.1 and Sculpt 20.08?
Greets