[PATCH] makedumpfile: Add description of incomplete dump file

["Zhou, Wenjian/周文剑"]

On 11/13/2014 05:01 PM, Atsushi Kumagai wrote:
> Hello Zhou,
>
>> Add description of incomplete elf and kdump-compressed dump file
>> in IMPLEMENTATION.
>
> That's a good idea, thanks Zhou.
> I have some comments.
>
>> Signed-off-by: Qiao Nuohan<qiaonuohan at cn.fujitsu.com>
>> Signed-off-by: Zhou Wenjian<zhouwj-fnst at cn.fujitsu.com>
>> ---
>> IMPLEMENTATION |   76 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 files changed, 76 insertions(+), 0 deletions(-)
>>
>> diff --git a/IMPLEMENTATION b/IMPLEMENTATION
>> index 2f4cfd6..8f8a945 100644
>> --- a/IMPLEMENTATION
>> +++ b/IMPLEMENTATION
>> @@ -112,3 +112,79 @@
>>          unsigned long long      page_flags;     /* page flags */
>>      } page_desc_t;
>>
>> +* The incomplete DUMPFILE
>> +  When generating DUMPFILE, if ENOSPACE error happens, the DUMPFILE will be
>> +  incomplete.
>> +
>> +  - The incomplete kdump-compressed DUMPFILE
>> +
>> +           the complete                  the incomplete
>> +    +-----------------------+    +-----------------------+
>> +    |      main header      |    |      main header      |  have incomplete flag
>> +    |-----------------------+    |-----------------------+
>> +    |      sub header       |    |      sub header       |
>> +    |-----------------------+    |-----------------------+
>> +    |      1st-bitmap       |    |      1st-bitmap       |
>> +    |-----------------------+    |-----------------------+
>> +    |      2nd-bitmap       |    |      2nd-bitmap       |
>> +    |-----------------------+    |-----------------------+
>> +    | page header for pfn 0 |    | page header for pfn 0 |
>> +    | page header for pfn 1 |    | page header for pfn 1 |
>> +    |                  :    |    |                  :    |
>> +    | page header for pfn N |    | page header for pfn N |  The page headers after
>> +    |                  :    |    |                       |  N don't exist. The
>> +    | page header for pfn Z |    |                       |  value of it is zero,
>> +    |-----------------------|    |-----------------------|  when try to read it.
>> +    |   page data (pfn 0)   |    |   page data (pfn 0)   |
>> +    |   page data (pfn 1)   |    |   page data (pfn 1)   |
>> +    |                  :    |    |                  :    |
>> +    |   page data (pfn N)   |    |   page data (pfn N)   |
>> +    |                  :    |    +-----------------------+
>> +    |   page data (pfn Z)   |
>> +    +-----------------------+
>> +
>> +    The incomplete flag is set into status of disk_dump_header by
>> +                          status |= DUMP_DH_COMPRESSED_INCOMPLETE
>> +    DUMP_DH_COMPRESSED_INCOMPLETE : 0x8
>> +
>> +    The page header and page data are written in pairs. When writting page data
>
> s/writting/writing/
>
>> +    (pfn N+1), if ENOSPACE error happens, the page headers after N won't be
>> +    written either.
>> +
>> +    If there is no page data dumped into the DUMPFILE, the DUMPFILE can't be
>> +    analysed by crash.
>> +
>> +  - The incomplete elf DUMPFILE
>
> There is no description of the ELF format itself in this documentation
> unlike the kdump-compressed format, I would be glad if you could write it.
>
>> +
>> +      the complete          the incomplete
>> +    +-------------+         +-------------+
>> +    |  elf header |         |  elf header |  have incomplete flag
>> +    |-------------+         |-------------+
>> +    |  PT_NOTE    |         |  PT_NOTE    |
>> +    |  PT_LOAD(1) |         |  PT_LOAD(1) |
>> +    |         :   |         |         :   |
>> +    |  PT_LOAD(N) |         |  PT_LOAD(N) |  The PT_LOAD after N don't exist.
>> +    |         :   |         |             |  The value of it is zero, when try
>> +    |  PT_LOAD(Z) |         |             |  to read it.
>> +    |-------------+         |-------------+
>
> The whole program header must exist since reserve_diskspace()
> pre-allocates enough space for it, the description above isn't right.
>
The space exists, but since PT_LOAD and segment are written in pairs,
if ENOSPACE happens, the left PT_LOAD won't be generated and won't be
written as well. The space reserved is for the PT_LOAD still in buffer.

-- 
Thanks
Zhou Wenjian