Quoting Roland Dreier <[EMAIL PROTECTED]>:
>  > Allocate memory for large MAD buffers with __get_free_pages,
>  > making it possible to get buffers up to 512KB in size.
> 
> Ugh, why is this an improvement?!

I use this patch mainly as a diagnostic vehicle for various tools
until we get it fixed in a better way.

> What are the chances of an order-9
> allocation succeeding on a system that's been running for a while?

Probably low. But reboot should help ;).

>  > +  if (length >= PAGE_SIZE)
>  > +          buf = (void *)__get_free_pages(gfp_mask, 
> long_log2(roundup_pow_of_two(length)) - PAGE_SHIFT);
>  > +  else
>  > +          buf = kmalloc(length, gfp_mask);
> 
> this is extra-gross too.

I agree here.

> I don't think this is worth it -- much better to fix the MAD API to
> handle gather/scatter lists.

We are looking into that.

-- 
MST
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to