Mount

At the beginning of the mount process EBS allocates memory for c->summary and for a special buffer (c->summary->sum_buf). EBS allocates c->sector_size memory for c->summary->sum_buf. The first difference between normal mount and mount with EBS is in the scan method. If EBS support is compiled in then it tries to read the end of each eraseblock (8 bytes) to find a special marker for EBS (jffs2_sum_marker) instead of scaning every eraseblock. There are two cases of EBS operation mode:

Figure 4: EB status (closed, not closed)

• jeb has valid jffs2_sum_marker
  It means that the jeb is full and closed by jffs2_sum_marker. jffs2_sum_marker has an offset field and this field shows the offset from the beginning of the jeb whence we can load the summary information for this jeb. EBS allocates c->sector_size - offset (size of summary info) memory to load the summary info. EBS has node and data CRC checksums and if the checksums are correct then EBS begins to process the summary info. This process allocates as many memory for jffs2_raw_node_refs, jffs2_full_dirents and jffs2_inode_caches as the normal scan process but with EBS the fs doesn't have to scan the whole jeb for nodes. After the processing of the summary node EBS allocates a plus jffs2_raw_node_ref for the summary node it is necessary because the summary info on flash is reported as USED_SPACE and jeb->used_size have to be equal to the sum of the totlens of the rawrefs in the jeb. At the end of the summary info processing EBS frees the allocated c->sector_size - offset memory.
• jeb doesn't have valid jffs2_sum_marker
  It means the jeb is not full and closed that is to say it hasn't got summary information and normal scan process must be executed with the extension of EBS information collecting. This extension means the following if the scan process finds a node of:
  • JFFS2_NODETYPE_INODE then it allocates jffs2_sum_inode_mem structure and copies the necessary fields from the inode to the jffs2_sum_inode_mem, adds the structure into the end of summary list (sum_list_tail) and increases sum_size and sum_num.
  • JFFS2_NODETYPE_DIRENT then it allocates jffs2_sum_dirent_mem structure and copies the necessary fields and the name of the dirent from the dirent to the jffs2_sum_dirent_mem, adds the structure into the end of summary list (sum_list_tail) and increases sum_size and sum_num.
  • JFFS2_NODETYPE_CLEANMARKER then it does nothing, because the summary node has a field (cln_mkr) to store the size of the cleanmarker.
  • JFFS2_NODETYPE_PADDING then it increases the sum_padded variable by the size of the padding node.
  At the end of the scanning of the current jeb JFFS2 classifies the jeb according to which kind of list it will belong to. Everything is the same as without EBS except the case when the jeb classification is BLK_STATE_PARTDIRTY. In this case the current jeb has more free space than the previously selected c->nextblock then this jeb will be the new c->nextblock, the previously collected summary information will be deleted and the summary information of the new c->nextblock will be moved to c->summary.