Pending Tasks to Finish Project

[abertschi]

Overview about Pending Tasks

Report

• [x] Memory Management
• [x] Process, Threads, Dispatch
• [x] LMP Message Passing
• [x] Page Fault Handling
• [x] Multicore
• [x] URPC Message Passing
• [x] Shell
• [x] Filesystem
• [x] Networking
• [x] Nameserver
• [x] Development Methodology
• [x] User Guide

Performance Evaluations

• [x] LMP Message Passing
• [x] Multicore simple URPC framework
• [x] URPC Message Passing
• [x] Filesystem
• [x] Networking

Individual Milestones

• [x] Shell #135
• [x] Filesystem #136
• [x] Networking #137
• [x] Nameserver

High Priority Bugs

• [x] Add Support to free memory again #132
• [ ] ~Threading Bugs~

[abertschi]

Paging Bugs

Possible options to fix paging issues

• introduce slabs
• rewrite monitor and drop thread support
• separate aos_malloc into static_malloc and dynamic_malloc such that no copying of Header structs needed
• find bug in current code base and use existing design

[eikendev]

Introducing slabs to the paging system has been started in #133. Note that the branch is unstable.

[abertschi]

before submission: ensure grading functions are called in all services

[abertschi]

Tasks

• [x] Bring up 2nd core and setup nameservice
• [x] reactivate free() in aos_malloc
• [x] spawn domains from filesystem
• [x] benchmarking

[leopoldsedev]

Issues:

• [x] Random out-of-memory errors. Not because there is actually no memory left, but because an absurd amount is requests: mem_alloc:ram_alloc_aligned_handler(-208896, 4096). At least once such request can come from the processserver
  • Reproducible on largemalloc-debug branch by running oncore rpc-test in the shell twice back-to-back
• [x] Processserver more PID higher than 256
• [x] getchar() in rpc-test
• [x] sometimes freeze when spawning process in rpc-test (maybe a problem with the serialserver? because it sometimes also happends after a process exits and sometimes it still reacts but extremely slowly)
• [x] backspace in shell sometimes prints random characters

[abertschi]

• [x] calling rpctest builtin when spawning shell on core 1 (oncore -f -c 1 aosh) causes printf to transmit trash to serialserver

[abertschi]

• [x] ~make sure builtin/rpctest.c spawns domains on core i%2, call rpctest builtin in aosh. rpctest gets stuck while spawning. last output is ram_alloc_aligned_handler~ cannot reproduce

  
  aosh >>> rpctestaosh.0.0: Running RPC tests...
  aosh.0.0: Testing init RPC...

init.0.0: mem_alloc:ram_alloc_aligned_handler(4096, 4096) init.0.0: mem_alloc:ram_alloc_aligned_handler(4096, 4096) aosh.0.0: aos_rpc_lmp_send_number aosh.0.0: aos_rpc_send_string aosh.0.0: aos_rpc_send_string initserver.0.0: handle_number(13) initserver.0.0: handle_string(1234567890abcdefghejklmnopqrstuvwxyz) init.0.0: mem_alloc:ram_alloc_aligned_handler(4096, 4096) initserver.0.0: handle_string(this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long stringthis is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long stringthis is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long stringthis is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string this is a very long string) aosh.0.0: Testing aos_rpc_process_spawn() (spawning 5 processes)... init.0.0: mem_alloc:ram_alloc_aligned_handler(69632, 4096) init.0.0: mem_alloc:ram_alloc_aligned_handler(4096, 4096) dummy.0.0: Dummy spawned dummy.0.0: libc_exit pid 105 aosh.0.0: spawned child: pid 105 init.0.0: mem_alloc:ram_alloc_aligned_handler(4096, 4096)

spawning all on core 0 or all on core 1 works fine.

[eikendev]

@abertschi, I think the paging rework using slabs can be discarded. Threading will not be stable at time of submission, right?

[Liblor]

Yup, threading is not stable :(

[eikendev]

Good to have:

• [x] Benchmarks for LMP
• [x] Illustrate results to report

[leopoldsedev]

• [x] Call all grading functions correctly