Hello Hatayama,

I re-run tests in dev env. I took your latest kernel patchset from patchwork for vmcore + devel branch of makedumpfile + fix to open and write to /dev/null. Run this test on 1Tb memory machine with memory used by some user space processes. crashkernel=384M.

Please see my results for makedumpfile process work:
[gzip compression]
-c -d31 /dev/null
real 37.8 m
user 29.51 m
sys 7.12 m

[no compression]
-d31 /dev/null
real 27 m
user 23 m
sys   4 m

[no compression, disable cyclic mode]
-d31 --non-cyclic /dev/null
real 26.25 m
user 23 m
sys 3.13 m

[gzip compression]
-c -d31 /dev/null
% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 54.75   38.840351         110    352717           mmap
 44.55   31.607620          90    352716         1 munmap
  0.70    0.497668           0  25497667           brk
  0.00    0.000356           0    111920           write
  0.00    0.000280           0    111904           lseek
  0.00    0.000025           4         7           open
  0.00    0.000000           0       473           read
  0.00    0.000000           0         7           close
  0.00    0.000000           0         3           fstat
  0.00    0.000000           0         1           getpid
  0.00    0.000000           0         1           execve
  0.00    0.000000           0         1           uname
  0.00    0.000000           0         2           unlink
  0.00    0.000000           0         1           arch_prctl
------ ----------- ----------- --------- --------- ----------------
100.00   70.946300              26427420         1 total


I used 2.6.39 kernel + your patches due to mine machine successfully work with it. I think that kernel version is not sufficient here due to /proc/vmcore is  very isolated.

Is that the same numbers which you have?

Interesting is that makedumpfile almost all time works in user space. And in case without compression  and without disk I/O process time is not significantly reduced. What is the bottleneck in 'copy dump' phase?

Thank you,
Maxim.