The Implementation And Improvement Of μClinux FILE System

μClinux is one of the most popular Embedded OS in the world. It is designed for Non-MMU processor device. It is famous with its stabilization and strong functions, and also can transplant easily. File system is one of important part ofμClinux OS, When it is used on embedded devices with large volume Flash storage, large data exchanging request must be satisfied.μClinux file system in Flash storage device is always consisted by 4 main parts: VFS (Virtual Files system Switch ), Concrete File System, MTD (Memory Technology Device) and hardware drive. In these 4 parts, VFS, MTD and hardware drive's structures are almost firmed, that is to say, the key is Concrete File System. At present, JFFS2 (Journaling Flash File System 2)is most popular Concrete File System onμClinux. New storage structure and special garbage collection are applied in JFFS2, Consume Balance and Break Protection Function are also supported by it. But at the beginning of JFFS2 Design, the volume of storage in the device at that time is not as large as nowadays. If import it directly into embedded OS today, error calibration efficiency and loading time will be big problems. These two questions are the very key points of my paper.To the first question, recalibration Hamming coding is a more efficient resolving method. Compare with other methods, hamming code has higher speed and better Fault Tolerance ability. Data classification is also available to this question, if store the ordinary data by ordinary structure, and store data with high precision requirement by coding storage structure, much time and storage space will be saved. The last step is to improve the JFFS2 Documentation system. In order to strengthen the fault tolerance calibration ability, node structure and input/output function are redesigned.To the other question, reducing the operation times of flash read/write action is the only way to go. To realize it, redefine the structure of nodes will be helpful because before the read/write action being carried out, device can find the block memory image first just like the function of index. In the other hand, building nodes distribution map can be always set before read/write action because in ordinary operation, the frequency of read/write action is lower than other actions.