Re: [PATCH] adding per sb inode list to make invalidate_inodes() faster

[Andrew Morton <akpm@osdl.org>]

Kirill Korotaev <dev@sw.ru> wrote:
>
> Well for sure this bug can be triggered only on really big servers with
>  a huge amount of memory and cache size.
>  It's up to you whether to apply it or not. I understand your position 
>  about 8 bytes, but probably it's just a question of using kernel, 
>  whether it's a user or server system.
>  Probably we can introduce some config option which would trigger 
>  features such as this one for enterprise systems.

I am paralysed by indecision!

It would be nice if we had evidence that more than one site in the world
was affected by this :(

I can't see an less space-consuming alternative here (apart from per-sb lru)

>  >> Also, the additional sizeof(struct list_head) is only a requirement
>  >> while the global inode LRU is maintained. I believed it would have
>  >> been beneficial to have localized the LRU to the sb also, which would
>  >> have maintained sizeof(struct inode0 at parity with current mainline.
>  > 
>  > Could be.  We would give each superblock its own shrinker callback and
>  > everything should balance out nicely (hah).
>
>  heh, and how do you plan to make per-sb LRU to be fair?

Good point.