users

users@lists.genode.org

3 participants
2353 discussions

Init startup sequence
by Steffen Liebergeld 22 May '10

22 May '10

Hi again, I am unsure about the startup sequence of the init task (or any other task other than core, for that matter). The setup looks like this: Task Thread Core main (initializes core, sleeps) activation (used as service provider by core's threads) pager (pager for all(?) threads started by core) init(0) (service thread for init) Init init(1) Now init(1) calls init(0), does some initialization in finally gets stuck in __cxa_guard_acquire. The stack trace looks like this: ========================= 1 0x182818 .... 0x159110 __cxa_guard_acquire(T) + 0x20 = 0x1590f0 2 0x182838 32 0x15d31c Genode::env()(T) + 0x2c = 0x15d2f0 3 0x182868 48 0x1591b5 malloc(t) + 0x25 = 0x159190 4 0x182898 48 0x1596c8 __cxa_allocate_exception(T) + 0x28 = 0x1596a0 5 0x1828d8 64 0x15f00f Genode::Ipc_client::_call()(T) + 0x11f = 0x15eef0 6 0x182918 64 0x15dcb3 Genode::Parent_client::session(char const*, char const*)(W) + 0x183 = 0x15db30 7 0x1829b8 160 0x15cdf5 Genode::Platform_env::Platform_env()(W) + 0xb5 = 0x15cd40 8 0x1829d8 32 0x15d32c Genode::env()(T) + 0x3c = 0x15d2f0 9 0x182a08 48 0x1591b5 malloc(t) + 0x25 = 0x159190 10 0x182a28 32 0x157a2c __register_frame(T) + 0x2c = 0x157a00 11 0x182a38 16 0x1590e0 init_exception_handling()(T) + 0x20 = 0x1590c0 12 0x182a58 32 0x163168 _main(T) + 0x28 = 0x163140 13 0x0 .... 0x1631c0 _start(T) + 0xc = 0x1631b4 =========================== It is unclear to me, why init has to grab a lock. Is it using data structures that reside in shared memory with core? What should happen next? Many thanks in advance, Steffen -- Dipl.-Inf. Steffen Liebergeld <steffen@...37...> Security in Telecommunications TU Berlin/ Deutsche Telekom Laboratories Ernst-Reuter-Platz 7, 10587 Berlin Phone: +30 8353 58 263 Web: http://www.t-labs.tu-berlin.de/sect

3 5

Problems with c++ exceptions
by Steffen Liebergeld 21 May '10

21 May '10

Hello Genoders, I'm having troubles with the C++ exception mechanism: In base/src/platform/_main.cc:_main(), the exception frames are set up in function init_exeption_handling(). Unfortunately, during startup of the init task, this runs into a null pointer. I could trace this back to base/src/base/cxx/exception.cc:__register_frame(__eh_frame_start), which obviously calls some gcc-provided function. I am using the toolchain you provide on your website. Did you have similar experiences during your development? Do you have an idea what is going wrong in my setup? Where should I start my investigations? Many thanks in advance, Steffen -- Dipl.-Inf. Steffen Liebergeld <steffen@...37...> Security in Telecommunications TU Berlin/ Deutsche Telekom Laboratories Ernst-Reuter-Platz 7, 10587 Berlin Phone: +30 8353 58 263 Web: http://www.t-labs.tu-berlin.de/sect

3 4

Server badges in Core
by Steffen Liebergeld 17 May '10

17 May '10

Hello friendly Genode-people, I am having difficulties unterstanding the handling of the server badge (currently in core only) As far as I understand, the badge identifies the receipient-object. Therefore the sender encodes the address of the receiving object's capability in the first slot of the send buffer. On the receiver side, in Server_activation_base::entry(), the badge is used to identify the correct object, whose dispatch() Method is then called. The server however has to insert the capability into the Object_pool first, which is done at Server_entrypoint::manage(). Now I am currently having problems in core's main function. During Ram_session_client(init_ram_session_cap).ref_account(env()->ram_session-cap()), an IPC call is sent to the "activation"-Thread, during which an invalid badge is used. Consequently the activation thread does not find the corresponding object and returns ERR_INVALID_OBJECT. This raises the following questions: a) Which object is the correct handler for this call, and b) where is the capability of the receiver allocated and registered at the Object_pool? I am looking forward to hearing from you. Many thanks in advance. Greetings, Steffen -- Dipl.-Inf. Steffen Liebergeld <steffen@...37...> Security in Telecommunications TU Berlin/ Deutsche Telekom Laboratories Ernst-Reuter-Platz 7, 10587 Berlin Phone: +30 8353 58 263 Web: http://www.t-labs.tu-berlin.de/sect

2 2

Troubles. X86: can't boot bare-metal. ARM: can't compile
by Thomas DuBuisson 04 May '10

04 May '10

Hello all, Below are two separate issues carelessly smashed into one e-mail - if you have any comments on either one then I'm all ears. -------- x86 --------- After building Genode+Fiasco [1] I can run the hello demo and scout demo (from "build/bin/*") on top of Linux, but I can't seem to boot bare metal. I'm copying the binaries from "<compound-build>/genode.fiasco_x86/bin", getting bootstrap and sigma0 from "<compound-build>/l4env/bin/...", and L4 kernel from "<compound-build>/fisco_x86/main". When I boot, the kernel starts up, prints a pc and sp (I notice "Roottask" has a stack address of all zeros). And quickly restarts. I'd paste output but it auto restarts and I don't have a serial port to over which I could collect output. The config (this is one, I've tried lots of alterations): ############## <config> <start> <filename>vesa_drv</filename> <ram_quota>2M</ram_quota> </start> <start> <filename>timer</filename> <ram_quota>0x20000</ram_quota> </start> <start> <filename>nitpicker</filename> <ram_quota>1M</ram_quota> </start> <start> <filename>hello_server</filename> <ram_quota>1M</ram_quota> </start> <start> <filename>hello_client</filename> <ram_quota>1M</ram_quota> </start> </config> ################# And the grub entry: ############# title L4 Fiasco + Genode (10.02) kernel /bootstrap -modaddr=0x03000000 module /fiasco nowait -nokdb -tbuf_entries=16384 -serial_esc module /sigma0 module /core module /init module /config module /vesa_drv module /nitpicker module /timer module /launchpad module /testnit module /scout module /nitlog module /liquid_fb module /hello_server module /hello_client ############### Any help? Let me know if you want more info. [1] http://genode.org/community/wiki/GenodeOnL4Fiasco -------- ARM -------- Separately, I'm having trouble building Fiasco on ARM - basically its missing types.h. Am I missing a step? After altering specs.conf "SPECS = genode fiasco_arm" And after setting tools.conf to point at my cross compiler: [tommd@...36... genode-arm]$ cp tool/builddir/etc/build.conf.fiasco_x86 tool/builddir/etc/build.conf.fiasco_arm [tommd@...36... genode-arm]$ ./tool/builddir/create_builddir fiasco_arm GENODE_DIR=. BUILD_DIR=~/dev/genode-arm/3rd-build/genode-fiasco_arm L4_DIR=~/dev/genode-arm/3rd-build/3rd/l4env successfully created build directory at /home/tommd/dev/genode-arm/3rd-build/genode-fiasco_arm [tommd@...36... genode-arm]$ make -C ~/dev/genode-arm/3rd-build/genode-fiasco_arm/ make: Entering directory `/home/tommd/dev/genode-arm/3rd-build/genode-fiasco_arm' Processing /home/tommd/dev/genode-arm/base-fiasco/src/core/arm/target.mk ------------------------------------------------------------------------ Library avl_tree COMPILE avl_tree_avl_tree.o MERGE avl_tree.lib.a Library slab COMPILE slab_slab.o MERGE slab.lib.a Library allocator_avl COMPILE allocator_avl_allocator_avl.o MERGE allocator_avl.lib.a Library cxx COMPILE cxx_exception.o COMPILE cxx_guard.o COMPILE cxx_malloc_free.o In file included from /home/tommd/dev/genode-arm/base/include/base/typed_capability.h:17, from /home/tommd/dev/genode-arm/base/include/parent/capability.h:17, from /home/tommd/dev/genode-arm/base/include/base/env.h:20, from /home/tommd/dev/genode-arm/base/src/base/cxx/malloc_free.cc:17: /home/tommd/dev/genode-arm/base-fiasco/include/base/capability.h:18: fatal error: l4/sys/types.h: No such file or directory compilation terminated. make[4]: *** [cxx_malloc_free.o] Error 1 make[3]: *** [cxx.tag] Error 2 make[2]: *** [all_sequence] Error 2 make[1]: *** [single_target] Error 2 make: *** [.] Error 2 make: Leaving directory `/home/tommd/dev/genode-arm/3rd-build/genode-fiasco_arm' Any help on either of these would be much appreciated. I'd love to use/contribute to Genode but if I can't build it myself / be self sufficient (perhaps I should be embarrassed) then I'm far from helpful! Cheers, Thomas

2 1

Genode-on-OKL4: Code section properties
by Frank Kaiser 03 May '10

03 May '10

Hello Recently we made an attempt to get Genode-on-OKL4 (v2.1.1.patch9) running on ARM11 (core ARM1136jf-s). The outcome was that the system stalled during initialization, because core waited endlessly for init getting started. OKL4's KDB showed that init's code section was set to read-only but not read-execute with the outrcome that the OKL4 scheduler refused to pass control to init. Interestingly this was not observed on an ARM9 platform, and we think that this just happened by chance, because the main reason of the problem seems to be how the class Dataspace_component maintains properties. In the OKL4 environment the protection domains are packed into the ELF file as modules, being itself ELF objects which are unpacked by function _setup_elf() of file base/src/base/process/process.cc. The function receives from core the capability of the ELF object (init is the first one) and parses the included segments to setup the pager. It checks whether any segment has the writable flag set. For such segments RAM is allocated and initialized, and the associated Dataspace_component gets the boolean attribute _writable set. For all other segments a Dataspace_component with the boolean attribute _writable unset is created, making no distinction between read-only and read-execute segments. The first attempt to start init leads to a page fault, and the methode Rm_client::pager() in file base/src/core/rm_session_component.cc is invoked. The method retrieves the Dataspace_component from the page fault parameters. With that a Mapping object is created to setup the response for the Ipc_pager. The constructor Mapping::Mapping() in file base-okl4/src/base/ipc/pager.cc receives the attribute _writable from the Dataspace_component. If it is set, the page attribute is set to L4_FullyAccessible, otherwise to L4_Readable. The outcome is that the code section is mapped with the X flag unset, and it is not executed while the DATA and BSS sections get the X flag set although this makes little sense. We could solve the problem by setting the page attribute to L4_Readable | L4_eXexcutable, but we still keep wondering why the problem only appears on ARM11 but neither on ARM9 or IA32, although the constructor code is always the same. Finally we think that the implementation of Dataspace_component is a bit short-handed regarding the handling of page attributes: there should be a distinction between read-only and read-execute (and DATA/BSS sections should not have the execute attribute set). At least both the ELF format and OKL4's MapControl interface maintain the full set of RWX flags. Regards Frank Kaiser OpenSynergy GmbH Rotherstr. 9 10245 Berlin Tel.: +49 (30) 20 18 18 35 28 Fax: +49 (30) 20 18 18 35 02 Email: Frank.Kaiser@...24... <mailto:andre.hergenhan@...24...> www.opensynergy.com <http://www.open-synergy.com> Handelsregister: Amtsgericht Charlottenburg, HRB 108616B Geschäftsführer: Frank-Peter Böhm, Dr. Stefaan Sonck Thiebaut, Rolf Morich

2 1

Thread blocking
by Peter Nguyen 28 Apr '10

28 Apr '10

Hi, In messing around a bit further with the http server code, I decided to have a go at making the HTTP server multithreaded. In this case, I wanted the web server to function such that when a HTTP request is received, a new thread is created whose purpose is to handle the HTTP request. In other words, the http_server_serve() function is made a thread, as opposed it being just a function. However, in making the changes, the execution seems to result in the web server hanging, with the HTTP handler thread not executing at all. This seem to be the case because inserting a call to sleep (via the timer connection) in main() resulted in the HTTP handler function executing. I would have thought that as soon as a HTTP request is received, the main thread would block, and from there schedule the HTTP handler thread (I have invoked start() on this thread, by the way). At the moment, it seems that the main thread blocking seems to result in the entire web server hanging. Any ideas as to what's going on? Peter

3 7

message buffer is too small
by Peter Nguyen 28 Apr '10

28 Apr '10

Hi, In moving my sampling code to version 10.02 of genode, and running the sampling functionality, I keep getting the following message during runtime: void copy_utcb_to_msgbuf(Okl4::L4_MsgTag_t, Genode::Msgbuf_base*): receive message buffer too small msg size=2c, buf size=16 Is there a way for me to change the size? Another point I should mention is that this output appears as a result of using L4 IPC. Peter

3 3

Website down
by Peter Nguyen 20 Apr '10

20 Apr '10

Hi, Is it just me, or is the genode website down at the moment? ----------------------------------------------------- Peter Nguyen Jacaranda Research Group

2 1

Weird build error
by Peter Nguyen 19 Apr '10

19 Apr '10

Hi, In using the latest genode version via OKL4 (an unmodified build), and trying to include nic_drv and the http server as part of config, I keep getting the the "no RM attachment" error. This is the output i get when i only include "nic_drv" in the config. int main(): --- create local services --- int main(): --- start init --- int main(): Ok we have 39 MB transfered to init int main(): --- init created, waiting for exit condition --- [init] our quota limit is 41009152 [init] our used quota is 135168 [init] starting timer with quota=524288 [init] using unique child name "timer" [init] starting pci_drv with quota=524288 [init] using unique child name "pci_drv" [init] starting nic_drv with quota=2097152 [init] using unique child name "nic_drv" no RM attachment (READ pf_addr=0 pf_ip=203891c from 04) virtual void Genode::Signal_session_component::submit(Genode::Signal_context_capability, unsigned int): invalid signal-context capability [init -> pci_drv] PCI driver started [init] pci_drv registered service PCI [init] timer registered service Timer I have no idea as to what's going on. The contents of the config file is the following: <config> <start> <filename>timer</filename> <ram_quota>512K</ram_quota> </start> <start> <filename>pci_drv</filename> <ram_quota>512K</ram_quota> </start> <start> <filename>nic_drv</filename> <ram_quota>2M</ram_quota> </start> </config> The quota for nic_drv was originally 512K. Changing it to 2M hasn't helped at all. Attached is the weaver file I am using as well. Can anyone tell me what i've done wrong? Peter

2 3

Rom FS
by Peter Nguyen 14 Apr '10

14 Apr '10

Hi, I noticed that as part of booting up Genode that the contents of the rom FS is printed (provided the variable is set for enabling verbose printing of boot information). Is there a way of having access to this data in my profiling environment? Having access to this information would greatly help with my system profiling. Peter

3 3

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

users