Patrick J. McGee
- Establish goals for the summer: develop Embedded Xinu's serial port, Ethernet driver, HDMI driver, keyboard driver, and USB storage driver.
- Look into logistics of a formal XinuPi3 release (separate branch on our GitHub to prevent single-core assumption conflicts)
- Make note of MU courses that may benefit from Xinu (possible changes necessary depending on the course)
- Clean working branch of our repository to exclude unnecessary single-core platforms and #ifdef statements. Fix outstanding compiler warnings. Goal: arm-rpi3 is single platform for new "multicore" branch of formal xinu-os release.
- Complete framebuffer driver port for XinuPi3.
- Evaluate USB subsystem for storage device (file system capability).
- Look into Power-over-Ethernet (PoE) additions for the O/S. Troubleshoot PoE-capable network switch
- Begin cleaning up ethernet driver code (LAN7800)
- Begin docs (Doxygen)
- Begin designing cache maintenance for the kernel, so that DMA buffers will work for transfers such as USB.
- Create an account for our Monday.com page (to stay organized and share progress with the team)
- Continue design and assessment. Close to initializing all devices --> looking into other possible issues.
- Realize that enabling L1 cache also breaks the frame buffer (as well as USB xfers)
- Update our page on Monday.com
- Work on Doxygen
- Update our page on Monday.com
- Assist team members with issues related to USB transfers and Xinu's I/O devices.
- Begin documentation by fixing Doxygen structural issues
- Resolve Doxygen issues, updated all documentation
- Work on README updates to include Pi 3 setup required
- Resolve printf() asynchronous I/O calls so the testsuite works
- Merge testsuite branch into doxygen branch. Will leave doxygen branch alone until all documentation has been verified for the final release
- TODO for the rest of the summer: resolve MMU cache, minor scheduler updates (putting Amy H. in charge of this), documentation, minor cleanup of system changes for multicore release.
- New angle for solving MMU cache coherency: since maintenance operations are not solving our USB transfer over DMA, we will force unaligned buffers so we can use memcpy(). Then, we will keep the DMA buffer area in memory uncached (through 1MB MMU sections), and all other memory space cached. In theory, this will prevent DMA issues with USB xfers.
Specific progress can be viewed on our GitHub repository (commits across multiple branches).
|Week 1: May 30-June 2|
|Began attempts to boot the full version of Embedded Xinu kernel with Raspberry Pi 3|
|Researched a Linaro Cross-Compiler to use for its Cortex A-53 ARM processor with ARMv8 architecture|
|Drew diagrams of the boot sequence and compilation for better understanding of the Pi 3's mechanisms|
|Traced through the dynamics of written Assembly code in the full research version of Xinu|
|Week 2: June 5-9|
|Tested the LED program in updated Assembly code. Began testing the Serial Port Transceiver driver for the Pi 3|
|Read about Raspberry Pi bare metal programming via online sources|
|Read about how the Raspberry Pi 3 controls the frequency of its UART pins|
|Read ARM processor documentation about configuring UART clock frequency|
|Used the oscilloscope in attempt to obtain the clock frequency of the UART pin|
|Week 3: June 12-16|
|Read an online textbook about the ARM processor of the Pi 3.|
|Successfully controlled the PL011 UART with a 48MHz frequency. Began porting full version of Xinu.|
|Read ARM64 documentation on registers to use with process interrupt handling.|
|Began working on the context switch for the 64-bit architecture.|
|Finishing work on the files necessary for context switching and rescheduling.|
|Week 4: June 19-23|
|Continue working on the boot functions of our implementation. Finished the Research Conduct course.|
|Read about register manipulation in the ARM guide. Continue working on process creation.|
|Begin researching physical memory space in the Pi 3.|
|Read guides to porting from ARM32 to ARM64. Digging deep for a solution to a memory allocation problem.|
|Read about ATAGS and what steps need to be taken to ensure the proper accessible RAM amount.|
|Week 5: June 26-30|
|Narrowed down the memory issue to an Aarch64-specific register-load requirement. Working on the fix.|
|Fixed memory issue - now allocating nearly all available memory.|
|Presentation by Dr. Kim Factor on tips to create a good research poster.|
|Began formulating research poster. Working on process queue structure.|
|Continue outlining research paper for submission to EWiLi Embedded Systems conference.|
|Week 6: July 3-7|
|Major change of working state: switch from 64-bit to 32-bit for compatibility and teachability reasons.|
|Migrating all code from non-working 64-bit to working 32-bit.|
|Processes create and context switch among one core. Push to resolve multiple cores.|
|Week 7: July 10-14|
|Expanding startup code to include change in privilege level.|
|Finished outlining poster. Tested ethernet driver and multiple core threading throughout the week.|
|Week 8: July 17-21|
|Studied simple Linux-style multi-core scheduling algorithms using mutual exclusion.|
|Created a method of scheduling across multiple cores with Xinu.|
|Tested multicore synchronization with atomic instructions.|
|Week 9: July 24-28|
|Studied technical documentation on how Raspberry Pi CPU mailboxes have worked in the past.|
|Test core mailboxes and work out kinks corresponding to core setup.|
|Finished final version of research poster: Utilizing Multi-core threading on Embedded Xinu.|
|Finished research paper and submitted it to EWiLi conference.|
|Week 10: July 31-August 4|
|Expanded material of final presentation.|
|Presented my research poster alongside colleagues. Viewed and discussed posters outside my discipline.|
|Watched my fellow researchers' presentations and continued making notes of guidelines.|
|Gave my final formal research presentation on the final day of the REU.|