Re: [PATCH] mm: readahead: get back a sensible upper limit

[Linus Torvalds]

On Tue, Feb 24, 2015 at 2:08 PM, Rafael Aquini <aquini@xxxxxxxxxx> wrote:
>
> Would you consider bringing it back, but instead of node memory state,
> utilizing global memory state instead?

Maybe. At least it would be saner than picking random values that make
absolutely no sense.

> People filing bugs complaining their applications that memory map files
> are getting hurt by it.

Show them. And as mentioned, last time this came up (and it has come
up before), it wasn't actually a real load, but some benchmark that
just did the prefetch, and then people were upset because their
benchmark numbers changed.

Which quite frankly doesn't make me care. The benchmark could equally
well just be changed to do prefetching in saner chunks instead.

So I really want to see real numbers from real loads, not some
nebulous "people noticed and complain" that doesn't even specify what
they did.

                         Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/