January 2020 - Devel - lists.sel4.systems

答复: some question about the compile of global-components in camkes_arm_vm project
by yadong.li 09 Jan '20

09 Jan '20

I found some example in odroid_vm, I want to make sure my understanding is correct? In CMakeLists.txt DeclareCAmkESComponent(HelloWorld) is just tell me there is a component called “HelloWorld” If you want this component to be compiled to project You must add call in your assembly in .cmakes file like below: assembly { composition { component HelloWorld hello; } Thank you very much 发件人: yadong.li 发送时间: 2019年12月28日 15:32 收件人: 'devel(a)sel4.systems' 主题: some question about the compile of global-components in camkes_arm_vm project hi， I have some question about the compile of global-components in camkes_arm_vm project. 1、In projects/global-components/components, there are some directory include FileServer, SerialServer, TimeServer and the others. I found FileServer had been compiled to project， but the other directory have not been compiled Is there some on-off switch control the compile ？ But I did not find it. 2、The CAmkES of FileServer component did not have a "int run(void) {}, If I want to add an new component with run entry， but it is not rootserver（I think one project only have one rootserver），it run with vm rootserver parallelly？ what should I do ？ Is there some example. Thank you very much

2 1

Error while building in docker
by Edward Sandberg 09 Jan '20

09 Jan '20

This morning I saw this error when building sel4: /bin/sh: 1: PROTOBUF_PROTOC_EXECUTABLE-NOTFOUND: not found I noticed a warning when starting the docker container about trustworthysystems/camkes:latest being over 30 days old. I thought that using an up to date pull of: https://github.com/SEL4PROJ/seL4-CAmkES-L4v-dockerfiles.git would keep the build env up to date but that apparently is either not the case or there is a problem with the repo. I ran: docker pull trustworthysystems/camkes and the warning was resolved and I could build seL4. Is pulling that docker image supposed to be necessary? Thanks! -- Edward Sandberg Adventium Labs 111 3rd Avenue S. Suite #100 Minneapolis, MN 55401 ed.sandberg(a)adventiumlabs.com

2 1

fastpath performance test
by Dd Nirvana 05 Jan '20

05 Jan '20

Hi guys, I have downloaded the lastest seL4 repo and built the seL4 kernel. And I reuse the code in sel4test/projects/sel4test/apps/sel4test-tests/src/tests/ipc.c. To test the fast, I made minor modifications: first, set caller and callee to the same priority; second, caller uses seL4_Call while callee uses seL4_ReplyRecv. However, the fastpath takes more than 2000 cycles (a round trip). Build: use the lastest provided docker and ../init-build.sh - DPLATFORM=x86_64 -DCMAKE_BUILD_TYPE=Release Both hardware and Qemu show similar results. (4.0GHz skylake CPU) BTW. The manifest is the same as https://sel4.systems/About/Performance/home.pml (i.e., bf13e9f673484825a60f9950cf688ff33ac841ad). Any advices about the issue? Best, Jinyu Gu, Institute of Parallel and Distributed Systems (IPADS), School of Software, Shanghai Jiao Tong University

2 1

Understand seL4 x86_64 Kernel ABI calling convention
by Alexandre Mutel 03 Jan '20

03 Jan '20

Hello! While looking at the code base and checking generated code, I'm not sure about the pertinence of having a different ABI from user System V ABI for kernel calls, as it is implying a bit more unnecessary register shuffling, and I was wondering why this choice was made to have r10, r8, r9, r15? instead of keeping System V ABI and just replace rcx by r10 as it is discarded by syscall: SystemVABI_x86_64_user: rdi, rsi, rdx, rcx, r8, r9 SystemVABI_x86_64_Linux_kernel: rdi, rsi, rdx, r10, r8, r9 SystemVABI_x86_64_seL4: rdi, rsi, r10, r8, r9, r15 So on a Linux Kernel, only the register rcx will be transferred to r10 (when transitioning from user to syscall), but in seL4, it has to reshuffle arguments rdx to r9. (Note for context: I'm working on a project where I don't use a standard C compiler to interface with the seL4 Kernel API, and where the register allocator of the compiler might not be as efficient as GCC/CLang after everything is inlined, that's why I'm looking into these details)

1 1