On Thu, Sep 12, 2013 at 5:56 PM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> I'll walk through the code, it looked suspicious. Maybe there's
> something subtle that makes it work, but I don't see it.
Btw, it's not just the DCACHE_LRU_LIST bit. The games with
"nr_dentry_unused" look totally broken too. It's decremented in
dentry_lru_isolate_shrink() for each dentry we remove, and then it is
decremented *again* in shrink_dcache_sb() by the number of dentries we
removed.
Maybe I'm confused, but the code sure looks more confused than I feel.
I would suggest keeping the same semantics for 'nr_dentry_unused'.
Dentries are unused whether they are on the "real" LRU list or have
been tagged with DCACHE_SHRINK_LIST. So moving from one list to the
other does nothing. It's the "list_del_init()" that should trigger
both 'nr_dentry_unused' and DCACHE_LRU_LIST bit-clearing.
In fact, maybe a helper function for _actually_ removing the thing
from all lists, and adding them back. Right now there are
"list_del_init()" and "list_add[_tail]()" calls sprinkled around in
random places, mixed up with the new "list_lru_add()".
Damn, the code is too confused. I have to go to a highschool parent
back-to-school meeting, so I won't get to this until maybe on a plane
tomorrow. Al, can you please give this a look?
Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-next" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
