On Fri, Aug 26, 2016 at 10:04 AM, Cong Wang <xiyou.wangc...@gmail.com> wrote:
>
> It was correct until...
>
> commit 4cf0b354d92ee2c642532ee39e330f8f580fd985
> Author: Florian Westphal <f...@strlen.de>
> Date: Fri Aug 12 12:03:52 2016 +0200
>
> rhashtable: avoid large lock-array allocations
>
>
> which is:
>
> @@ -83,6 +83,9 @@ static int alloc_bucket_locks(struct rhashtable *ht,
> struct bucket_table *tbl,
> tbl->locks = vmalloc(size * sizeof(spinlock_t));
> else
> #endif
> + if (gfp != GFP_KERNEL)
> + gfp |= __GFP_NOWARN | __GFP_NORETRY;
> +
> tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
> gfp);
Heh.
Yeah, this is a classic case of something we should *not* do.
And I'm not speaking out against that commit 4cf0b354d92e itself: that
isn't the problem. The problem is that #ifdef with the rather subtle
dangling 'else'. Oops.
I even *looked* at that function yesterday, and didn't realize the (in
hindsight) obvious bug because the code had that odd pattern.
I see that Eric sent a patch and fixed the bug, and in the process got
rid of that dangling else thing.
Thanks guys,
Linus
