[linux-lvm] Bugs in DM_DEVICE_STATUS ioctl no_flush flag settings

[Ming-Hung Tsai]

2016-03-31 20:16 GMT+08:00 Zdenek Kabelac <zkabelac at redhat.com>:
>
> Yeah - I've been create lv_cache struct with this in mind.
>
> However the whole issue should be seen more generically in 'OO-way'.
>
> So we do want to have 'status' object with reporting methods,
> and this object should be filled by 'segment' method.
>
> The complexity grows however as we do have some segments being related
> to other segments - that's why we currently have very ugly code hack
> inside to do filling of most status struct in one place - which is totally
> unextendable - but ATM was the quickest way to achieve reasonable goal.

Agree. We should utilize the segtype_handler framework as much as we can.

> For now I'm rather conservative about large code-base changes till Heinz's
> raid branch is fully merged.
> PS: I'm amazed some else is able to read lvm2 code.

Because there are lots of legacy code, hacks, and undocumented stuff? :p
It's painful when I started to read lvm2 code, but now I think I'm able
to write a maintenance guide of lvm2...

Hope that someday LVM code would become well-organized, maybe lvm3...


Ming-Hung Tsai