[dm-devel] [RFC] Naming of dm-multipath hardware handler modules

Mike Christie michaelc at cs.wisc.edu
Wed Aug 1 16:39:13 UTC 2007


Dave Wysochanski wrote:
> Do we have an agreed upon scheme?
> 
> I noticed in 2.6.23-rc1, the RDAC module is named "dm-mpath-rdac.ko",
> but the source file is "dm-rdac.c".  Should we be naming things

Are you sure about that? In linus's tree it is named dm-mpath-rdac.c. 
The module that is created is named dm-rdac.ko.

> "dm-mpath-hwhname" everywhere?  If so, should we rename dm-emc?
> 
> I was not sure exactly what was correct for the HP handler.  Currently
> it follows the dm-emc scheme, which is probably wrong.
> 

Alasdair, initially wanted the modules and file names to have dm-mpath 
prefixes, but userspace for emc is already passing in "emc" and 
dm_get_hw_handler does request_module("dm-%s", name) and 
__find_hw_handler_type does strcmp(name, hwhi->hwht.name).

So we ended up doing a partial conversion leaving userspace alone, but 
changing the filenames to what Alasdair wanted. So rdac got the source 
filename dm-mpath-rdac.c, but its module gets the name dm-rdac.ko.

So yeah, I forgot about that in the review, and instead of

obj-$(CONFIG_DM_MULTIPATH_HP)	+= dm-hp-sw.o

you should do

dm-hw-objs	:= dm-mpath-hw.o

......


obj-$(CONFIG_DM_MULTIPATH_HP_SW) += dm-hp-sw.o

In the Makefile, and the source filename for should be dm-mpath-hp-sw.c.

And dm-emc.c should probably be changed, although we could probably do 
it after the logical unit follow-over cap patch is cleaned up and merged.




More information about the dm-devel mailing list