users February 2021

users@lists.genode.org

19 participants
24 discussions

Help with nested signal context warnings
by Colin Parker 17 Feb '21

17 Feb '21

Hello Genodians, I'm hoping someone can help or point me to a good reference for the error that pops up occasionally saying "attempt to handle the same signal context twice (nested)". This happens within some Sculpt components (decorator, depot_query), but also happens in this USB wifi driver I'm trying to make. Mostly I want to understand it - my long explanation is below. My basic understanding is Genode gets upset when waiting for a signal within another signal handler. However, the "natural" way I would like this driver to behave is to be signal driven - all code is responding to some signal or other. However, sometimes one needs to do synchronous IO operations. If it is truly forbidden ever to wait within a signal handler, I considered a few options, but found all to be non-ideal. 1) Use stub signal handlers to insert elements into a work queue and drive everything from an event loop. This seems like reinventing the signal mechanism, but might be what's required. 2) Build state machines to break up synchronous IO operations with no waiting at all. This makes simple-looking operations complicated and increases debugging complexity. 3) Busy-wait for synchronous operations. Probably a big performance penalty, and obviously wasteful. So I proceeded, because I do notice that sometimes it is OK to wait within a handler (i.e. the warning does not occur). But I still get the warning other times, and I can't quite figure out when or why it happens. Initially, I thought that if I separated things out, so that, say, Signal_context A waits, but can only receive signals for Signal_context B or C, it would be OK. Now, I am not totally sure I've actually achieved this, but I think that I have, and I've become suspicious that I don't really understand the "rules." So, is there a way to understand when one can wait safely wait within a signal handler? Is it really as simple as "Signal A cannot be generated while waiting if a handler for A is on the stack?" Does the App vs IO signal distinction come into play (I have both). Also, is there a way to get a Genode app to output a stack trace - I can patch that warning mesage to output a stack trace and at least see what actually happened? Thanks as always to the Genode community for sharing your knowledge with me. Best regards, Colin

3 5

golang runtime
by Sergey Platonov 15 Feb '21

15 Feb '21

Hello all We have invested some time analyzing golang and how much it will take to make it production-ready. The build system requires some effort, and it's not fully clear if it's a one-time effort. Different golang projects use other build systems (bazel and so on). Networking is crucial for us, and we invested most of our time making it work. Unfortunately, we failed to do so: it seems the goroutines scheduler makes some assumptions that we overlooked. So, at the end of the day, we have a network that sometimes works. I have proposed a patch [1] that will fix some issues and hopefully help someone later on. Unfortunately, we will not invest more resources in porting golang since it seems too resource-consuming for our small team. [1] https://github.com/tor-m6/genode/pull/1 Regards, Sergey -- gapfruit AG Baarerstrasse 135 6300 Zug +41 762 444 560 sergey.platonov(a)gapfruit.com

2 3

general improvements for Sculpt
by Edoardo Mantovani 09 Feb '21

09 Feb '21

Hello for everyone, Some days ago I've used your os (Sculpt v.20.08) from bootable usb and I remained genuinely impressed regarding its style and its security base idea, after having read some informations on it, I was able to develop some tips which I would like to share with the community: ======================= Language Compatibility ======================= *) there is no support for Italian keyboard, I was able to find only german, french and english .chargen files, for a next update would be great to add other charger files. *) I can give my full support for translating document pages in Italian, this could (apparently) be great for increasing the possible audience, but Honestly I doubt that an Italian community could be created (we are quite far in the security-technological field, No comment on that, please). ======================= BEGINNER FRIENDLY OS? ======================= *) I know that this operating system is not developed to be user friendly, but it could be great to put some helps for beginners, like a graphical instruction box when clicking on certain things in Leitzentrale, a graphical example could be the following: |------------| |-------| | depot | (when clicking on it appears an info box) --> | info | |_______| |-------| (sorry for the bad ASCII image) ======================= GRAPHICS ======================= *) Probably the best point for your os consists on its graph graphics, the idea of Leitzentrale is innovative if we confront it to Windows os or similar, a small hint could be to create the same style for the "file" part, so something which could be similar to graphviz2 output instead of "block representation". { direct link to my idea: https://graphviz-perl.github.io/ see the graph under the "Input file # 17 - t/gen.parse.stt.t" example, s/other_names/folder_names/g } *) small consideration: I find some graphical glitch in the "motif decoration" 's node in Leitzentrale (it flew up and down without any input) [Still studying this..] ======================== DOCUMENTATIONS ======================== *)Your site is pretty cool and interesting, documentation is great but I haven't seen anything on Sculpt, could be great to add like UNIX's manpages system. ========================= WIRELESS SUBSYSTEM ========================= *) I asked for the documentation on it before speaking (Thanks Alexander), it is extremely similar to Linux wireless subsystem, except for a modified libnl support for user space, could I have the source code for it? I would like to study the modification you've apported for libnl. I am pretty fascinated by those modified subsystems. *) It would be great to extend the range for other devices, I saw you implemented only the iwlwifi driver, would be great to implement also brcmf (both HardMac and SoftMac) and even Qualcomm's wireless card, I can give my full support for that. /* *) I have some ideas for controlling the iwlwifi blob, for now I would like to test it on my own, in general it would be great to ""customize"" the "blob loading". */ ======================== Custom Patching system ======================== *) I've seen the potential of the software based hardware management, my biggest hint should be to implement a security patching system with a similar interface to Depot (something like: choose customization part, insert code patch [like linux's kernel system] and process it) | V could be a dangerous door for local-cyber-attacks ======================== TERMINAL DOWNSIDE ======================== *) Using the inspect shell I wasn't able to scroll up, do you know if it's implemented? It should be better to have a scroll bar, ======================= DEPOT SYSTEM ======================= As I understood, depot is like a package manager and the various repositories contains various applications, Here I'll insert my ideas on it: *) Many of those repos are inaccessible even if I am connected to the wireless network: { I am not able to download/fetch the following repos: [-] ehmry [-] blarson [-] rite } *) in general there isn't much variety in the repos, many packages are the same, it would be great to extend the "candidates" list, I can contribute to the development of several CLI apps. *) I can work for inserting some scripting languages interpreter packages inside Genode { [-] Minised [-] lightweight Perl languages (microperl, miniperl or my customized languages) } and maybe adapt them for the OS needs and tasks. *) A quite interesting idea would be to add local depots, In general: -->) We have 2 devices, (featured phones or similar) -->) The first is the client, the second is the depot server -->) through Bluetooth, BLE or even other standards (nfc??) could be cool to use the repository stored in the second phone into the first. [ Obviously this is extremely complex to do: we should set a correct cryptographic key exchange and set an authentication rule, this could lead an ∞ of problems: *) Hijacking or other Bluetooth attacks are possible, the worst case scenario is that an untrusted depot server can inject the device, bypassing the security authentication restriction. If this will be implemented, it MUST be well structured. *) Not every device has the same bluetooth version, with some works it could be implemented a workaround which act like a switch statement: { example: switch(bluetooth_version){ case (version == 5.0 ) {do this} case (version == 4.1 ) {do that} } } *) Obviously it must be implemented a tiny bluetooth subsystem, I suggest using Linux's one. (I have also the 2008-2009 documentation on it). *) This idea could be a great topic for your blog, I can help with the formal verifications (pi calculus?) exposition and with other implementation details. ] I have many other ideas, These are some "fast" considerations for improve the OS quality, my "knowledge CV" is here: (https://github.com/Baseband-processor) Let me know if you have some doubt, Regards and Thanks for you work, Edoardo Mantovani, 2021 -- Edoardo Mantovani Independent security researcher email: Baseband(a)cpan.org Urbino, Italy

5 9

scp, shell and noux
by Stefan Thöni 09 Feb '21

09 Feb '21

Hello Genodians We would like to port scp to Genode to securely copy files to and from a target. The scp binary works by calling it's counterpart on the remote system through the SSH channel using a shell. The obvious way would be to create a noux package and make use of bash. However it came to our attention that noux is to be deprecated. Therefore our question: What is the future of the shell in Genode? How should we go about using SSH and porting scp? Best regards Stefan -- Freundliche Grüsse Stefan Thöni Chairman of the Board Senior Security Architect +41 79 610 64 95 gapfruit AG Baarerstrasse 135 6300 Zug https://gapfruit.com

2 1

Re: Cmus issue
by Dima Sokolov 08 Feb '21

08 Feb '21

Hello Josef, > The new driver version is marked with “(fix)” I installed this version per your instructions, and I got the same cmus behavior as before, but with this LOG output: [runtime -> bsd_audio_drv] Error: LD: jump slot relocation failed for symbol: 'Genode::Timeout::Timeout(Timer::Connection&)' [runtime -> bsd_audio_drv] Error: deadlock ahead, mutex=0x161a40, return ip=0xebbb0 > For adjusting volume levels let me point you to Alex' [1] Genodians post. Thank you for additional info. Best regards, Dima Sokolov

2 1

Re: Cmus issue
by Dima Sokolov 08 Feb '21

08 Feb '21

Hello Josef, Yes, I used your packages from the experimental section for Sculpt 20.08. If I noticed correctly, new package version installs replace the old version without the need to manually uninstall it? After adding the packages once again using the "+" menu, I found only these (new) versions listed @ /depot/cnuke/pkg: bsd audio drv 2020-09-24-b cmus 2020-09-11-a BSD audio driver LOG output: [runtime] child "bsd_audio_drv" [runtime] RAM quota: 16136K [runtime] cap quota: 222 [runtime] ELF binary: pci_audio_drv [runtime] priority: 0 [runtime] provides service Audio_out [runtime] provides service Audio_in Mixer LOG output: [runtime] child "mixer" [runtime] RAM quota: 61192K [runtime] cap quota: 66 [runtime] ELF binary: mixer [runtime] priority: 0 [runtime] provides service Audio_out Upon adding the cmus component, the cmus window is still stuck waiting. [runtime] priority: 0 is still the last shown line in the LOG output. If I start cmus and then remove the bsd audio drv component, the cmus terminal window finally shows the cmus interface. Naturally, the LOG output shows these messages since the audio driver is missing: [runtime -> cmus -> cmus] Audio_out-session creation failed [runtime] Warning: cmus: Lookup for service "Audio_out" failed And inside cmus: Error opening audio device: cannot initialize required output plugin If I add the bsd audio drv once again and restart cmus, it gets stuck waiting again, and the window is black. Should I compile Sculpt 20.11 from git and work with that version instead (if cmus is known to work there)? Best regards, Dima Sokolov

2 2

Clipboard domain creation
by Dima Sokolov 06 Feb '21

06 Feb '21

Hello, What would be the proper config for custom clipboard domains? Let's say I want to add a new domain called "risky" and to assign falkon to it: /config/nitpicker <domain name="risky" layer="5" content="client" label="no" hover="always" focus="click" /> <policy label="falkon" domain="risky"/> And let's say I want to test my new domain by allowing the user to copy config snippets from Falkon to leitzentrale, but not the other way around: /config/clipboard <flow from="risky" to="leitzentrale" /> I also don't want to allow any copying nor pasting between the "desktop" domain and my new "risky" domain. But I'm still able to copy text from the file browser to Falkon. And from qt5-textedit to Falkon. So, something is amiss in my config. "Files" and "Inspect" belong to the "leitzentrale" domain, right? And apps like system shell and qt5-textedit belong to the "desktop" domain? Have I added the necessary config lines to the right config files? This is on Sculpt 20.08. What about different configurations of the same app, are they identified by their launcher name (label)? For example: <policy label="falkon-banking" domain="banking"/> "falkon-banking" would be a Falkon instance with access to private data, in contrast to the regular Falkon instance. Best regards, Dima Sokolov

1 0

Re: Cmus issue
by Dima Sokolov 06 Feb '21

06 Feb '21

Hello Josef, I'm using Sculpt 20.08 (the officially released image). Here's the LOG output: [runtime] child "cmus" [runtime] RAM quota: 65288K [runtime] cap quota: 466 [runtime] ELF binary: init [runtime] priority: 0 All the lines you've mentioned after [runtime] priority: 0 are not shown. > the Terminal change, where it now uses the gui session directly but in the past used a framebuffer session I guess this could be the problem. Best regards, Dima Sokolov

2 1

Cmus issue
by Dima Sokolov 02 Feb '21

02 Feb '21

Hello, I'm experiencing an issue with cmus. Namely, the cmus terminal window is completely black and I cannot use the app. Routing has been done as follows: BSD audio drv: Device access -> audio hardware Region maps -> custom virtual memory objects Mixer: Audio output -> bsd audio driver Report -> system reports Cmus: File system (media) -> ahci-1.fs File system (home) -> ahci-1.fs File system (fonts) -> fonts fs GUI -> themed wm Audio output -> mixer ROM (clipboard) -> themed wm Report (clipboard) -> themed wm Real-time clock -> system clock Region maps -> custom virtual memory objects What would be the proper configuration? Best regards, Dima Sokolov

2 1

Re: Network issue
by Dima Sokolov 02 Feb '21

02 Feb '21

Hello Christian, I made some tests, and Cloudflare 1.1.1.1 works fine. Test 1: Setting the router's primary DNS server to Cloudflare 1.1.1.1 with Sculpt 20.08 default network settings in Virtualbox Using Falkon to browse the Internet The relevant /config/nic_router: <config verbose_domain_state="yes"> <report interval_sec="5" bytes="yes" config="yes" config_triggers="yes"/> <default-policy domain="default"/> <uplink label="wired" domain="uplink"/> <domain name="uplink"> <nat domain="default" tcp-ports="1000" udp-ports="1000" icmp-ids="1000"/> </domain> <domain name="default" interface="10.0.1.1/24"> <dhcp-server ip_first="10.0.1.2" ip_last="10.0.1.200" dns_server_from="uplink"/> <tcp dst="0.0.0.0/0"> <permit-any domain="uplink"/> </tcp> <udp dst="0.0.0.0/0"> <permit-any domain="uplink"/> </udp> <icmp dst="0.0.0.0/0" domain="uplink"/> </domain> </config> Test 1 result: Network is fully operational. Test 2: Setting the primary DNS server to 1.1.1.1 from Sculpt 20.08 in Virtualbox while the router has a different primary DNS server (Cloudflare 1.0.0.1) Using Falkon to browse the Internet The relevant /config/nic_router: <config verbose_domain_state="yes"> <report interval_sec="5" bytes="yes" config="yes" config_triggers="yes"/> <default-policy domain="default"/> <uplink label="wired" domain="uplink"/> <domain name="uplink"> <nat domain="default" tcp-ports="1000" udp-ports="1000" icmp-ids="1000"/> </domain> <domain name="default" interface="10.0.1.1/24"> <dhcp-server ip_first="10.0.1.2" ip_last="10.0.1.200" dns_server="1.1.1.1"/> <tcp dst="0.0.0.0/0"> <permit-any domain="uplink"/> </tcp> <udp dst="0.0.0.0/0"> <permit-any domain="uplink"/> </udp> <icmp dst="0.0.0.0/0" domain="uplink"/> </domain> </config> Test 2 result: Network is fully operational. Best regards, Dima Sokolov

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

users February 2021