Hello, I post this patchset as release candidate of v1.5.0. If there are no comments, I will release them as official version in a week. Explanation: This version introduces the "cyclic mode" as a default running mode. Before this version, makedumpfile uses the amount of memory depending on physical memory size, because it saves the analytical data for whole memory at a time. On the other hand, cyclic mode uses fixed amount of memory regardless of physical memory size, because this mode analyzes only constant region of memory per cycle as below: [v1.4.7 or before] cycle 1 +----------------- -----+ vmcore | ... | +----------------- -----+ [v1.5.0] cycle 1 2 3 4 ... N +----------------- -----+ vmcore | | | | | ... | | +----------------- -----+ And, the number of cycles is represented as: N = physical memory size / (page size * bit per byte(8) * BUFSIZE_CYCLIC) BUFSIZE_CYCLIC is the buffer size for analysis, default size is 1 Mbyte. BTW, in v1.5.0, cyclic mode may cause some slow down of filtering process when N is large. I'm going to improve this issue with new filtering logic (called mem_map array logic) in a future version. At least in v1.5.0, if you feel cyclic mode is slow, you can try 2 workarounds: 1. Use old running mode with "--non-cyclic" option. 2. Decrease the number of cycles by increasing BUFSIZE_CYCLIC with "--cyclic-buffer" option. Please refer to the manual page for how to use these options. I am waiting for your comments. Thanks Atsushi Kumagai