July 2011 - l4-hackers - example.com Site

TLS on amd64
by Daniel Müller 15 Jul '11

15 Jul '11

Hello, I want to use the __thread keyword gcc provides. I was told this was recently adjusted to work (at least in the private repository which I am using). It appears to be working when building for IA32 but when I compile the same program for amd64 I get a page fault when running it (it is a simple program just incrementing a __thread declared int). It faults at 10010f4 (the first access of the variable). 00000000010010f0 <main>: 10010f0: 48 83 ec 08 sub rsp,0x8 10010f4: 64 c7 04 25 fc ff ff mov DWORD PTR fs:0xfffffffffffffffc,0x0 10010fb: ff 00 00 00 00 1001100: bf 60 48 01 01 mov edi,0x1014860 1001105: e8 36 88 00 00 call 1009940 <puts> 100110a: 64 83 04 25 fc ff ff add DWORD PTR fs:0xfffffffffffffffc,0x1 1001111: ff 01 1001113: bf 01 00 00 00 mov edi,0x1 1001118: e8 d3 71 00 00 call 10082f0 <sleep> 100111d: eb e1 jmp 1001100 <main+0x10> 100111f: 90 nop Is TLS for amd64 supposed to work? Any idea on the problem? Greetings, Daniel

2 3

Running the IO server
by Anna Fischer 12 Jul '11

12 Jul '11

I'm trying to run the ARM-based frame buffer configuration from conf/examples/arm-rv-lcd.cfg with pkg/io/config/arm-rv-eb-mc.devs and QEMU (-M realview-pbx-a9). I can boot the L4 Linux kernel on this (without GUI but just serial I/O) configuration, but I get an error when trying to set up the LCD (see below). Perhaps the LCD device configuration from the .devs file does not match what hardware QEMU emulates? How can I actually enumerate in the IO server what hardware it actually runs on? Is there a way to do that? Thanks, Anna L4 Bootstrapper Build: #34 Mon Jul 11 16:17:10 BST 2011, 4.5.2 Scanning up to 256 MB RAM Memory size is 256MB (00000000 - 10000000) RAM: 0000000000000000 - 000000000fffffff: 262144kB Total RAM: 256MB mod11: 012a9000-012cac7c: ex_fb_spectrum mod10: 0121b000-012a846c: mag mod09: 011f1000-0121a5b8: fb-drv mod08: 0119d000-011f01b8: ned mod07: 010c2000-0119cb58: io mod06: 010a8000-010c1510: l4re mod05: 010a7000-010a70e4: arm-rv-lcd.io mod04: 010a6000-010a659d: arm-rv-eb-mc.devs mod03: 010a5000-010a56bb: arm-rv-lcd.cfg mod02: 0106b000-010a45f0: moe mod01: 01061000-0106a33c: sigma0 mod00: 01015000-0106014c: fiasco Moving 12 modules to 1100000 with offset eb000 moving module 12 { 12a9000-12cac7c } -> { 1394000-13b5c7c } moving module 11 { 121b000-12a846c } -> { 1306000-139346c } moving module 10 { 11f1000-121a5b8 } -> { 12dc000-13055b8 } moving module 09 { 119d000-11f01b8 } -> { 1288000-12db1b8 } moving module 08 { 10c2000-119cb58 } -> { 11ad000-1287b58 } moving module 07 { 10a8000-10c1510 } -> { 1193000-11ac510 } moving module 06 { 10a7000-10a70e4 } -> { 1192000-11920e4 } moving module 05 { 10a6000-10a659d } -> { 1191000-119159d } moving module 04 { 10a5000-10a56bb } -> { 1190000-11906bb } moving module 03 { 106b000-10a45f0 } -> { 1156000-118f5f0 } moving module 02 { 1061000-106a33c } -> { 114c000-115533c } moving module 01 { 1015000-106014c } -> { 1100000-114b14c } Scanning fiasco -serial_esc Scanning sigma0 Scanning moe rom/arm-rv-lcd.cfg Relocated mbi to [0x100f000-0x100f19e] Loading fiasco Loading sigma0 Loading moe find kernel info page... found kernel info page at 0x2000 Regions of list regions [ 1000, 19ff] { a00} Kern fiasco [ 2000, 5cfff] { 5b000} Kern fiasco [ 90000, 9e17b] { e17c} Sigma0 sigma0 [ 140000, 17160b] { 3160c} Root moe [ 178000, 18eec7] { 16ec8} Root moe [ 1000000, 10143eb] { 143ec} Boot bootstrap [ 100f000, 100f29b] { 29c} Root Multiboot info [ 1190000, 13b5c7b] { 225c7c} Root Modules Memory API Version: (87) experimental Sigma0 config ip:000900e0 sp:01013044 Roottask config ip:00140178 sp:00000000 Starting kernel fiasco at 00001000 Realview System ID: Rev=0 HBI=178 Build=0 Arch=5 FPGA=00 L2: ID=00000000 Type=00000000 L2 cache enabled Hello from Startup::stage2 Initialize page table Number of IRQs available at this GIC: 96 Vmem_alloc::init() Cache config: ON ID_PFR0: 00001031 ID_PFR1: 00000011 ID_DFR0: 00000000 ID_AFR0: 00000000 ID_MMFR0: 00100103 ID_MMFR1: 20000000 ID_MMFR2: 01230000 ID_MMFR3: 00002111 SERIAL ESC: allocated IRQ 44 for serial uart Not using serial hack in slow timer handler. Welcome to Fiasco.OC (arm)! L4/Fiasco.OC arm microkernel (C) 1998-2011 TU Dresden Rev: r33 compiled with gcc 4.5.2 for Realview [] Build: #1 Tue Jun 21 12:57:15 BST 2011 Calibrating timer loop... done. SIGMA0: Hello! KIP @ 2000 allocated 4KB for maintenance structures SIGMA0: Dump of all resource maps RAM:------------------------ [0:0;fff] [0:5d000;8ffff] [0:9f000;13ffff] [4:140000;171fff] [0:172000;177fff] [4:178000;18efff] [0:18f000;100efff] [4:100f000;100ffff] [0:1010000;118ffff] [4:1190000;13b5fff] [0:13b6000;effffff] IOMEM:---------------------- [0:10000000;ffffffff] MOE: Hello world MOE: found 242832 KByte free memory MOE: found RAM from 5d000 to f000000 MOE: allocated 239 KByte for the page array @0x9f000 MOE: virtual user address space [0-bfffffff] MOE: rom name space cap -> [C:501000] BOOTFS: [1190000-11906bb] [C:503000] arm-rv-lcd.cfg BOOTFS: [1191000-119159d] [C:504000] arm-rv-eb-mc.devs BOOTFS: [1192000-11920e4] [C:505000] arm-rv-lcd.io BOOTFS: [1193000-11ac510] [C:506000] l4re BOOTFS: [11ad000-1287b58] [C:507000] io BOOTFS: [1288000-12db1b8] [C:508000] ned BOOTFS: [12dc000-13055b8] [C:509000] fb-drv BOOTFS: [1306000-139346c] [C:50a000] mag BOOTFS: [1394000-13b5c7c] [C:50b000] ex_fb_spectrum MOE: cmdline: moe rom/arm-rv-lcd.cfg MOE: Starting: rom/ned rom/arm-rv-lcd.cfg MOE: loading 'rom/ned' Ned says: Hi World! Ned: loading file: 'rom/arm-rv-lcd.cfg' IO | Io service mag | libio: Warning: Finding 'icu' in system bus failed with 'Void capability invoked' mag | Hello from MAG fbdrv | libio: Warning: Finding 'icu' in system bus failed with 'Void capability invoked' fbdrv | Could not find LCD. fbdrv | Failed to setup Framebuffer mag | Error: Exception: Void capability invoked: requesting goos info: backtrace ... mag | mag | MAG: destroy FB svc Ned[lua]: Hm, cpapbility 42d000 does not support the meta protocol: -2004 Ned: ERROR: rom/arm-rv-lcd.cfg:62: attempt to call method 'create' (a nil value)

2 2

Fix IO-APIC/IRQ-Routing Bug (affects GSIs 24, 48, 72, ...)
by Christian Ehrhardt 08 Jul '11

08 Jul '11

Hi, there seems to by an off-by-one error in Io_apic::find_apic() that kills IRQ-routing for GSIs that are a multiple of 24. Please consider the patch below (tested on real hardware with 3 IO-APICS). regards Christian diff --git a/src/kernel/fiasco/src/kern/io_apic.cpp b/src/kernel/fiasco/src/kern/io_apic.cpp index 88ffe68..897a1a0 100644 --- a/src/kernel/fiasco/src/kern/io_apic.cpp +++ b/src/kernel/fiasco/src/kern/io_apic.cpp @@ -331,7 +331,7 @@ Io_apic::find_apic(unsigned irqnum) { for (unsigned i = Max_ioapics; i > 0; --i) { - if (_apics[i-1]._apic && _apics[i-1]._offset < irqnum) + if (_apics[i-1]._apic && _apics[i-1]._offset <= irqnum) return i - 1; } return 0;

1 0

Newbie question: L4Linux as a patch
by Pablo Chacin 07 Jul '11

07 Jul '11

Hi all I'm doing a review of options for virtualization in Linux. From the documentation, it seems that L4Linux is a complete port of the Linux Kernel. I'm wondering if there is a patch (or series of patches) that could be used to modify an standard Linux kernel and make it compatible with the L4 virtualization? The question comes from my scenario: I need to deploy a customized linux kernel and want to do so over L4. This kernel is build from the standard kernel applying some patches. I would like to do the same with L4Linux, so the resulting kernel has my patches plus L4. Is this feasible? has been tested? thanks in advance and please consider I'm new to L4Linux. Regards -- ======================================================================== 0 0 0 Pablo Chacin | Dept. d'Arquitectura de Computadors 0 0 0 e-mail: pchacin(a)ac.upc.edu | UPC-Campus Nord. Modul D6-212 0 0 0 phone: +34 +93 405 4059 | Jordi Girona, 1-3 U P C fax: +34 +93 401 7055 | 08034 Barcelona - SPAIN www: http://personals.ac.upc.es/pchacin ========================================================================

2 3

Fiasco.OC Kernel symbolic debugger
by Daniel Waddington 06 Jul '11

06 Jul '11

Hello, Adam, are you likely to get the source-level symbolic debugging working in jdb in the near future? Using objdump is a bit drastic! ;-) Thanks Daniel

2 1

L4 for amd64
by Daniel Müller 05 Jul '11

05 Jul '11

Hi, I have some problems with running L4 for amd64 (i586 is just fine). I am trying to run (qemu-14.1.) the hello world binary and get: --------- L4 Bootstrapper Build: #1 Mi 29. Jun 17:41:45 CEST 2011, x86-64, 4.4.5 RAM: 0000000000000000 - 000000000009f3ff: 637kB RAM: 0000000000100000 - 0000000007ffcfff: 130036kB Total RAM: 127MB Moving 7 modules to 2000000 with offset fa8000 moving module 07 { 1679000-1739a52 } -> { 2621000-26e1a52 } moving module 06 { 1678000-1678039 } -> { 2620000-2620039 } moving module 05 { 1566000-167779e } -> { 250e000-261f79e } moving module 04 { 132d000-1565b9e } -> { 22d5000-250db9e } moving module 03 { 110c000-132cb58 } -> { 20b4000-22d4b58 } moving module 02 { 10be000-110b397 } -> { 2066000-20b3397 } moving module 01 { 1058000-10bdb00 } -> { 2000000-2065b00 } Scanning /fiasco -serial_esc -comspeed 115200 -comport 1 -nokdb This is an invalid binary, fix it (wrong ELF architecture). Key press reboots... Rebooting. --------- This is done with a clean checkout URL: http://svn.tudos.org/repos/oc/tudos/trunk Repository Root: http://svn.tudos.org/repos/oc/tudos Repository UUID: d050ee49-bd90-4346-b210-929a50b99cfc Revision: 33 configured for K8-type CPU. /usr/bin/file tells me: build/obj/bin/amd64_K8/l4f/sigma0: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/l4f/moe: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/l4f/ned: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/l4f/l4re: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/l4f/loader: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/l4f/hello: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped build/obj/bin/amd64_K8/bootstrap: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), statically linked, not stripped ./../kernel/fiasco/build/fiasco: symbolic link to `main' So why is bootstrap not built as 64 bit binary? Daniel

2 3

libc environment for l4linux
by Vasily Sartakov 04 Jul '11

04 Jul '11

Hello Does l4re snapshot has a libc environment for build a applications for l4linux? what toolchains did you use for build busybox? -- Sartakov A. Vasily

2 1

Fiasco.OC
by Daniel Waddington 03 Jul '11

03 Jul '11

Hi, I want to use 32bit x86 Fiasco.OC with lots of SMP CPUs (48+). Can you tell me where the current 32 (for 32bit) and 64 (for 64bit) limitations exist in the code - and suggest whether this could be changed? Thanks Daniel Waddington

2 1

the booting error of l4linux
by cheng guanghui 02 Jul '11

02 Jul '11

Hello: I downloaded the latest l4re-snapshot-2011062216. All the compilation is ok. The hello demo could boot successfully. But when i try to boot L4Linux with this command : make qemu E=L4Linux-mag-x86 O=/home/cheng/TUDOS/Src/l4re-snapshot-2011062216/obj/l4/x86 There are some errors as following: Allocate cpu_mem @ 0xfc54c000 Local APIC[00]: version=11 max_lvt=5 APIC ESR value before/after enabling: 00000000/00000000 Using the PIT (i8254) on IRQ 2 for scheduling Absolute KIP Syscalls using: Sysenter CPU[0:0]: GenuineIntel (6:3:3:0)[00000633] Model: Pentium II (Klamath) at 2395 MHz 32 KB L1 I Cache (8-way associative, 64 bytes per line) 32 KB L1 D Cache (8-way associative, 64 bytes per line) 2048 KB L2 U Cache (8-way associative, 64 bytes per line) Freeing init code/data: 28672 bytes (7 pages) Calibrating timer loop... done. SIGMA0: Hello! KIP @ 400000 Found Fiasco: KIP syscalls: yes allocated 4KB for maintenance structures SIGMA0: Dump of all resource maps RAM:------------------------ [0:1000;9efff] [0:10b000;13ffff] [4:140000;18bfff] [0:18c000;2dcfff] [4:2dd000;2ddfff] [0:2de000;3fffff] [0:465000;136cfff] [4:136d000;166ffff] [0:1670000;1614bfff] IOMEM:---------------------- [0:0;fff] [0:9f000;fffff] [0:17ffd000;febfffff] [0:fec01000;fedfffff] [0:fee01000;ffffffff] IO PORTS-------------------------- [0:0;fffffff] MOE: Hello world MOE: found 357544 KByte free memory MOE: found RAM from 1000 to 1614c000 MOE: allocated 353 KByte for the page array @0x1000 MOE: virtual user address space [0-bfffffff] MOE: rom name space cap -> [C:501000] BOOTFS: [136d000-1470e6d] [C:503000] l4re BOOTFS: [1471000-166e572] [C:504000] ned BOOTFS: [166f000-166f578] [C:505000] l4lx-gfx.cfg MOE: cmdline: /home/cheng/TUDOS/Src/l4re-snapshot-2011062216/obj/l4/x86/bin/x86_586/l4f/moe rom/l4lx-gfx.cfg MOE: Starting: rom/ned rom/l4lx-gfx.cfg MOE: loading 'rom/ned' Ned says: Hi World! Ned: loading file: 'rom/l4lx-gfx.cfg' io | L4Re: FATAL: Exception No such object found: 'rom/io' io | Exception: No such object found: rom/io: backtrace ... fb-drv | L4Re: FATAL: Exception No such object found: 'rom/fb-drv' fb-drv | Exception: No such object found: rom/fb-drv: backtrace ... mag | L4Re: FATAL: Exception No such object found: 'rom/mag' mag | Exception: No such object found: rom/mag: backtrace ... Ned[lua]: Hm, cpapbility 430000 does not support the meta protocol: -2004 Ned: ERROR: rom/l4lx-gfx.cfg:54: attempt to call method 'create' (a nil value) Do i miss something else? Thanks a lot. Warm regards. Cheng Guanghui

2 6