- l4-hackers - example.com Site

Experimenting the vm-tz example on the Freescale i.mx6 SDB
by Mahdi Aichouch 03 May '15

03 May '15

Hello, I am trying to experiment the "vm-tz" example on the Freescale i.mx6 SDB board. At first, my goal is to understand the example, thus I am interested in running a simple test kernel instead of a full linux kernel. Specifically, I am trying to reproduce the experimentation of the article related to the trustzone implementation on Fiasco.OC L4re, "ARM Trustzone as a Virtualization Technique in Embedded Systems". In particular, it is mentioned in the article at the "Implementation" paragraph that "Multiplexing two or more OS is supported by our design." I would like to investigate this point. I created a new entry to generate the vm-tz binary. I added the following entry to the "src/l4/conf/modules.list" file : entry tz_vm roottask moe --init=rom/vm-tz module l4re module vm-tz Is my configuration entry correct, is there any missing modules that are required by the example? I tested the binary on the board using tftp, and I obtained the following results: L4 Bootstrapper Build: #12 Tue Mar 17 12:10:02 CET 2015, 4.8.3 20140320 (prerelease) Scanning up to 1024 MB RAM, starting at offset 32MB Memory size is 1024MB (10000000 - 4fffffff) RAM: 0000000010000000 - 000000004fffffff: 1048576kB Total RAM: 1024MB Scanning fiasco Scanning sigma0 Scanning moe Moving up to 5 modules behind 11100000 moving module 02 { 110a5000-110de63f } -> { 11195000-111ce63f } [235072] moving module 01 { 1109b000-110a438b } -> { 1118b000-1119438b } [37772] moving module 00 { 11044000-1109a0c7 } -> { 11134000-1118a0c7 } [352456] moving module 04 { 1102a000-11043597 } -> { 1111a000-11133597 } [103832] moving module 03 { 11010000-11029473 } -> { 11100000-11119473 } [103540] Loading fiasco Loading sigma0 Loading moe find kernel info page... found kernel info page at 0x10002000 Regions of list 'regions' [ 10000000, 100000e3] { e4} Root mbi_rt [ 10001000, 10001aff] { b00} Kern fiasco [ 10002000, 10062fff] { 61000} Kern fiasco [ 10090000, 10096edf] { 6ee0} Sigma0 sigma0 [ 10098000, 1009e177] { 6178} Sigma0 sigma0 [ 10140000, 10171d77] { 31d78} Root moe [ 10178000, 101934ab] { 1b4ac} Root moe [ 11000000, 1100f4ff] { f500} Boot bootstrap [ 11100000, 11133fff] { 34000} Root Module API Version: (87) experimental Sigma0 config ip:10090100 sp:00000000 Roottask config ip:10140260 sp:00000000 Starting kernel fiasco at 100012c8 Number of IRQs available at this GIC: 160 FPU0: Arch: VFPv3(3), Part: VFPv3(30), r: 4, v: 9, i: 41, t: hard, p: dbl/sngl L2: ID=410000c7 Type=9e340340 Aux=02070000 WMask=ffff S=0 L2: Type L2C-310 Size = 1024kB Ways=16 Waysize=64 SERIAL ESC: allocated IRQ 58 for serial uart Not using serial hack in slow timer handler. Welcome to Fiasco.OC (arm)! L4/Fiasco.OC arm microkernel (C) 1998-2013 TU Dresden Rev: rexported compiled with gcc 4.8.3 for i.MX6 [] Build: #3 Wed Mar 11 16:43:41 CET 2015 Calibrating timer loop... done. MDB: use page size: 20 MDB: use page size: 12 SIGMA0: Hello! KIP @ 10002000 allocated 4KB for maintenance structures SIGMA0: Dump of all resource maps RAM:------------------------ [4:10000000;10000fff] [0:10063000;1008ffff] [0:10097000;10097fff] [0:1009f000;1013ffff] [4:10140000;10171fff] [0:10172000;10177fff] [4:10178000;10193fff] [0:10194000;110fffff] [4:11100000;11133fff] [0:11134000;4effffff] IOMEM:---------------------- [0:0;fffffff] [0:50000000;ffffffff] MOE: Hello world MOE: found 1031220 KByte free memory MOE: found RAM from 10000000 to 4f000000 MOE: allocated 1008 KByte for the page array @0x10194000 MOE: virtual user address space [0-bfffffff] MOE: rom name space cap -> [C:501000] BOOTFS: [11100000-11119474] [C:503000] l4re BOOTFS: [1111a000-11133598] [C:504000] vm-tz MOE: cmdline: moe --init=rom/vm-tz MOE: Starting: rom/vm-tz MOE: loading 'rom/vm-tz' Vmm started terminate called after throwing an instance of 'N2L413Runtime_errorE' MOE: task 270000679 exited with 127 The code stops at the beginning when it tries to create a VM : L4Re::chksys(env->factory()->create_vm(vm)); After reading through the code, I saw that the code example is expecting a linux-image and an initrd to be provided. Is it possible to tell me where these should be added in the build directory? In my test case, the linux-image will be replaced by a simple kernel image. Could someone help me to correct my test case. Many thanks in advance. Best regards, Mahdi

2 19

context::regs
by watermirror 03 May '15

03 May '15

hi,the implemention of Context::regs() is like this: return reinterpret_cast<Entry_frame*>(Cpu::stack_align(reinterpret_cast<Mword>(this + Size)) -1; how this algorithm came out?i think the memory of the object is allocated by c++ compiler, but why '(this + Size)' could get Entry_frame of a thread?

2 1

the syscall stack frame and Context::regs method
by watermirror 28 Apr '15

28 Apr '15

hi, i am reading the code about system call. after the sys_ipc_wrapper (in file kern/syscalls.cpp) is called.it get the thread's pointer by current_thread and then get the Syscall_frame. the code is like this: ... Thread *curr=current_thread(); Syscall_frame *f = curr->regs(); ... i can not understand the relationship between curr->regs() and the system call's stack frame.

2 1

about sysenter
by watermirror 28 Apr '15

28 Apr '15

hi ,in the file :l4re/pkg/l4sys/include/ARCH-x86/ipc-invoke.h, the comments says: We use ecx to store the user-level espand save eip onto the stack. but where could if find the corresponding code?

2 1

the first instruction in entry_sys_fast_ipc_c
by watermirror 28 Apr '15

28 Apr '15

hi, after entered the entry_sys_fast_ipc_c through sysenter, why the first instruction is 'pop %esp'?

2 1

How to solve the problem about multiple definition of clk_enable?
by Youren Shen 28 Apr '15

28 Apr '15

Hi, L4 hackers, To solve the problem about "undefined reference to `clk_get_rate'", I select COMMON_CLK in config L4_PLATFORM_GENERIC. But when I build the L4Linux this occurs: multiple definition of `clk_enable' l4linux/arch/l4/kernel/arch-arm/l4.c:63: first defined here So, Should I disable L4_CLK_NOOP in L4_PLATFORM_GENERIC? Thank you very much! Best Regards! Youren Shen

2 1

Re: How L4Linux Linked the header file?
by Youren Shen 28 Apr '15

28 Apr '15

Hi, Adam, Sorry to disturb you again. I find the reason of this problem is because the "config GENERIC_CLOCKEVENTS" is removed from file arch/arm/Kconfig after Kernel 3.4. As a result, the GENERIC_CLOCKEVENTS is not defined in the arch/l4/kconfig.arm. And the CONFIG_SMP depends on GENERIC_CLOCKEVENTS. So the CONFIG_GENERIC_CLOCKEVENTS and CONFIG_SMP will be unmask in build processes. Is there any way to get previous version of L4Linux? For instance, If I can get the source code of the L4Linux 3.4, it will be very helpful. Thank you very much for your help! On Sun, Apr 26, 2015 at 12:18 AM, Youren Shen <shenyouren(a)gmail.com> wrote: > Hi, Adam > I'm not so sure how to reply the mail in L4-hackers-list, So I CC to both > the List mail address and thread mail address. > > Thank you for your reply, and that really helps. > > I find that those error occurred because the CONFIG_SMP and > CONFIG_GENERIC_CLOCKEVENTS_BUILD is not set in .config. > > I try to config them by 'make menuconfig' but there is none entry about > CONFIG_SMP or CONFIG_GENERIC_CLOCKEVENTS_BUILD appeared in menuconfig. So > I change the .config file directly, but once I invoke "make", the .config > will be rewrite. > > Would you mind give me any guide about how to set them? > > Thank you very much. > >> >> >> Hi, >> >> On Sat Apr 18, 2015 at 22:26:15 +0800, Youren Shen wrote: >> > I'm trying to port L4Linux 3.14 to Linux Kernel 3.4. But when I complied >> > the kernel, this error happened: >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/main.c: >> > In function 'cpu0_startup': >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/main.c:2124:2: >> > error: implicit declaration of function 'set_my_cpu_offset' >> > [-Werror=implicit-function-declaration] >> > >> > I checked main.c file: >> > >> > #include <asm/unistd.h> >> > #include <asm/percpu.h> >> > #include <asm/generic/l4lib.h> >> > >> > So I add one percpu.h file with definition of 'set_my_cpu_offset' in >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/include/asm/arch-arm/, >> > however this error still happened. >> > >> > Then I add the definition of this function to >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/include/linux/percpu.h, >> > then the error message will not show. >> > >> > Also, discard the change in $(src)/include/linux/percpu.h, then I change >> > the main.c file to include absolute path of percpu.h. >> > >> > #include <asm/unistd.h> >> > #include >> > >> </home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/include/asm/arch-arm/percpu.h> >> > #include <asm/generic/l4lib.h> >> > >> > It passed the complied to. >> >> When you put a header file in arch/l4/include/asm/arch-arm/ including it >> via <asm/foo.h> should work. >> >> > So that is say, the #include <asm/percpu.h> is reference to >> > src/include/linux/percpu.h. That really strange. >> >> So, include <linux/percpu.h> in main.c right away? >> >> The L4 part of L4Linux you're looking at is for 3.14 and you apply it to >> Linux 3.4, please expect that things are different and need adaption. >> >> > I do not change the Makefile in arch/l4, and I really confused about >> this >> > error. >> > Now this problem occurs to me again, when complied the >> > arch/l4/kernel/timer.c file, this error happened: >> > >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/timer.c:153:35: >> > warning: 'struct clock_event_device' declared inside parameter list >> > [enabled by default] >> > >> /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/timer.c:153:35: >> > warning: its scope is only this definition or declaration, which is >> > probably not what you want [enabled by default] >> > >> > but the definition of this to file should be defined in clockchips.h >> which >> > is include in timer.c: >> > >> > #include <linux/timex.h> >> > #include <linux/clockchips.h> >> > #include <linux/irq.h> >> > >> > Thank you for your reading, I just want to know what's the rule of >> L4Linux >> > header file linking, and solve this problem. >> > I'd be so appreciated if anyone can give me some help! >> >> There's a #ifdef CONFIG_GENERIC_CLOCKEVENTS_BUILD in clockchips.h. >> Not defined? >> >> >> >> Adam >> -- >> Adam adam(a)os.inf.tu-dresden.de >> Lackorzynski http://os.inf.tu-dresden.de/~adam/ >> >> -- > Best Regards. > Youren Shen. > -- Best Regards. Youren Shen.

2 1

Re: How L4Linux Linked the header file?
by Youren Shen 25 Apr '15

25 Apr '15

Hi, Adam I'm not so sure how to reply the mail in L4-hackers-list, So I CC to both the List mail address and thread mail address. Thank you for your reply, and that really helps. I find that those error occurred because the CONFIG_SMP and CONFIG_GENERIC_CLOCKEVENTS_BUILD is not set in .config. I try to config them by 'make menuconfig' but there is none entry about CONFIG_SMP or CONFIG_GENERIC_CLOCKEVENTS_BUILD appeared in menuconfig. So I change the .config file directly, but once I invoke "make", the .config will be rewrite. Would you mind give me any guide about how to set them? Thank you very much. > > > Hi, > > On Sat Apr 18, 2015 at 22:26:15 +0800, Youren Shen wrote: > > I'm trying to port L4Linux 3.14 to Linux Kernel 3.4. But when I complied > > the kernel, this error happened: > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/main.c: > > In function 'cpu0_startup': > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/main.c:2124:2: > > error: implicit declaration of function 'set_my_cpu_offset' > > [-Werror=implicit-function-declaration] > > > > I checked main.c file: > > > > #include <asm/unistd.h> > > #include <asm/percpu.h> > > #include <asm/generic/l4lib.h> > > > > So I add one percpu.h file with definition of 'set_my_cpu_offset' in > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/include/asm/arch-arm/, > > however this error still happened. > > > > Then I add the definition of this function to > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/include/linux/percpu.h, > > then the error message will not show. > > > > Also, discard the change in $(src)/include/linux/percpu.h, then I change > > the main.c file to include absolute path of percpu.h. > > > > #include <asm/unistd.h> > > #include > > > </home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/include/asm/arch-arm/percpu.h> > > #include <asm/generic/l4lib.h> > > > > It passed the complied to. > > When you put a header file in arch/l4/include/asm/arch-arm/ including it > via <asm/foo.h> should work. > > > So that is say, the #include <asm/percpu.h> is reference to > > src/include/linux/percpu.h. That really strange. > > So, include <linux/percpu.h> in main.c right away? > > The L4 part of L4Linux you're looking at is for 3.14 and you apply it to > Linux 3.4, please expect that things are different and need adaption. > > > I do not change the Makefile in arch/l4, and I really confused about this > > error. > > Now this problem occurs to me again, when complied the > > arch/l4/kernel/timer.c file, this error happened: > > > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/timer.c:153:35: > > warning: 'struct clock_event_device' declared inside parameter list > > [enabled by default] > > > /home/shenyouren/workspace/kernel/kernel_exynos5250/arch/l4/kernel/timer.c:153:35: > > warning: its scope is only this definition or declaration, which is > > probably not what you want [enabled by default] > > > > but the definition of this to file should be defined in clockchips.h > which > > is include in timer.c: > > > > #include <linux/timex.h> > > #include <linux/clockchips.h> > > #include <linux/irq.h> > > > > Thank you for your reading, I just want to know what's the rule of > L4Linux > > header file linking, and solve this problem. > > I'd be so appreciated if anyone can give me some help! > > There's a #ifdef CONFIG_GENERIC_CLOCKEVENTS_BUILD in clockchips.h. > Not defined? > > > > Adam > -- > Adam adam(a)os.inf.tu-dresden.de > Lackorzynski http://os.inf.tu-dresden.de/~adam/ > > -- Best Regards. Youren Shen.

1 0

(no subject)
by deng.chao1＠zte.com.cn 24 Apr '15

24 Apr '15

Hi: I am reading the code of fiasco kernel.I have a question about the function "Context::schedule". At line 894 in kernel/fiasco/src/kern/context.cpp： "while (EXPECT_FALSE(current_cpu != access_once(&_home_cpu)))" As I know, the variable "_home_cpu" indicates the CPU number which the thread that invokes the “schedue ” as a member function is running on。And the variable “current cpu” indicates the CPU number now this code fragment is running on。 The condition test tells me that it seems the class Context's member function "schedule" will run on a different cpu some time . If it does so, some extra work will need to be done in the "schedule" function. I have searched all of the fiasco kernel sources, but I did not find this special context. So would anybody please tell me whether there has a situation that the "schedule" function will run on different CPU， why does it happen？ Thanks！

2 1

where is the call_bootstrap?
by 水镜 24 Apr '15

24 Apr '15

Hi, in the file main-ia32-32.cpp, there's a line(line 60): call call_bootstrap where could i find the implemention of the call_bootstrap?

2 1