Hi,
IIRC the logic for choice of name-based shm implementation for APR does
exactly match that of MM - is that not true? I don't know why shmget is
preferred over mmap in this case. For anonymous shm, mmap/MAP_ANON is
indeed preferred over shmget, which does seem inconsistent.
For me an MAP_SHARED used with MAP_ANON isn't correct because an SHM should
be used to sync a file and a memory. Perhaps i am wrong..
But it isn't recommended to use MAP_FIXED (i.e: manpage) in general. How can
process A know the SHM start address of process B ? with a pipe ? that is
dirty. The sys ipc implementation is also not correct and there is something
you can't do like remapping.
I have made a small implementation of a new SHM design :
http://www.badcode.be/~descript/.apache/
FEATURES:
- "pager" allows you to specify a multiple of pages that may be used. It
reduces reallocation so it reduces kernel call.
- Memory auto alignement
- File map auto padding
- Auto remove on bad header
- Auto create if file doesn't exist
- Remap/reallocation possibility.
- Auto padding (mem/file) when reallocating.
- Kernel lock/unlock.
- Memory bridge capatibility.
- 2 lock types.
This implementation was stoped because Apache folks have not been interested
by this code (this is probably because the code isn't "apr ready")
-mv
