diff options
| author | Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp> | 2012-11-09 15:21:10 +0900 |
|---|---|---|
| committer | Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp> | 2012-11-22 17:23:29 +0900 |
| commit | 4104adf71c3ca3a31dde30deff6874b411ad38a4 (patch) | |
| tree | 98768abc5fb6a1c8cc82959ceb83daae50e98397 /makedumpfile.8 | |
| parent | 6067dfed0d84fa2305bbe1b9033b17edf1b7f1d2 (diff) | |
[PATCH v2 2/2] Calculate the size of cyclic buffer automatically.
The size of cyclic buffer is set to DEFAULT_BUFSIZE_CYCLIC(1MB) when
the size isn't specified. But 1MB is too small for large memory machine
and the number of cycle will be so large. As a result, a lot of time
will be spend for dump filtering.
To resolve the issue above, I add calculate_cyclic_buffer_size() to
calculate the buffer size as needed.
Concretely, the logic is to choose the lesser value of the two below
as the size of cyclic buffer:
a. the size enough for storing the 1st/2nd bitmap for the whole of vmcore
b. 80% of free memory (as safety limit)
calculate_cyclic_buffer_size() uses max_mapnr, so it needs to be done
after get_max_mapnr().
Signed-off-by: Atsushi Kumagai <kumagai-atsushi@mxc.nes.nec.co.jp>
Diffstat (limited to 'makedumpfile.8')
| -rw-r--r-- | makedumpfile.8 | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/makedumpfile.8 b/makedumpfile.8 index 9f394f0..dfc6ec5 100644 --- a/makedumpfile.8 +++ b/makedumpfile.8 @@ -356,7 +356,8 @@ In the cyclic mode, the number of cycles is represented as: num_of_cycles = system_memory / (\fIbuffer_size\fR * 1024 * bit_per_bytes * page_size ) The lesser number of cycles, the faster working speed is expected. -Default \fIbuffer_size\fR is 1024. +By default, \fIbuffer_size\fR will be calculated automatically depending on system memory +size, so ordinary users don't need to specify this option. .br .B Example: |