On Thu, Oct 23, 2014 at 8:57 AM, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > On Thu, Oct 23, 2014 at 07:14:20AM -0700, Dan Williams wrote: >> On Thu, Oct 23, 2014 at 6:52 AM, Roman Gushchin <klamm@xxxxxxxxxxxxxx> wrote: >> >> That find by Roman discovered that even though NET_DMA was marked >> >> broken we were still attempting to pin pages for dma offload. >> > >> > Not really. Actually I had NET_DMA in my config by mistake. >> >> Well, that is material to whether this patch should be applied. I had >> another report from someone stating that changing the get_user_pages() >> call to get_user_pages_fast() improved a network performance benchmark >> by a large margin. We shouldn't be pinning pages at all, so with >> those two reports I proceeded with deleting rather than mess around >> with more surgical disabling. Let me take a closer look and put >> together a smaller patch for -stable. > > Ok, especially as your patch didn't even apply to 3.17-stable :) So, it seems more people turn on CONFIG_BROKEN than I assumed as both Roman's report and the bad network performance reports I have seen should be negated by NET_DMA=n. Sorry for the noise, -stable can stick with "NET_DMA depends on BROKEN". -- Dan -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html