On Fri, 8 Apr 2016 10:26:53 -0700
Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
> On Fri, Apr 08, 2016 at 02:33:40PM +0200, Jesper Dangaard Brouer wrote:
> >
> > On Fri, 8 Apr 2016 12:36:14 +0200 Jesper Dangaard Brouer <brouer@xxxxxxxxxx> wrote:
> >
> > > > +/* user return codes for PHYS_DEV prog type */
> > > > +enum bpf_phys_dev_action {
> > > > + BPF_PHYS_DEV_DROP,
> > > > + BPF_PHYS_DEV_OK,
> > > > +};
> > >
> > > I can imagine these extra return codes:
> > >
> > > BPF_PHYS_DEV_MODIFIED, /* Packet page/payload modified */
> > > BPF_PHYS_DEV_STOLEN, /* E.g. forward use-case */
> > > BPF_PHYS_DEV_SHARED, /* Queue for async processing, e.g. tcpdump use-case */
> > >
> > > The "STOLEN" and "SHARED" use-cases require some refcnt manipulations,
> > > which we can look at when we get that far...
> >
> > I want to point out something which is quite FUNDAMENTAL, for
> > understanding these return codes (and network stack).
> >
> >
> > At driver RX time, the network stack basically have two ways of
> > building an SKB, which is send up the stack.
> >
> > Option-A (fastest): The packet page is writable. The SKB can be
> > allocated and skb->data/head can point directly to the page. And
> > we place/write skb_shared_info in the end/tail-room. (This is done by
> > calling build_skb()).
> >
> > Option-B (slower): The packet page is read-only. The SKB cannot point
> > skb->data/head directly to the page, because skb_shared_info need to be
> > written into skb->end (slightly hidden via skb_shinfo() casting). To
> > get around this, a separate piece of memory is allocated (speedup by
> > __alloc_page_frag) for pointing skb->data/head, so skb_shared_info can
> > be written. (This is done when calling netdev/napi_alloc_skb()).
> > Drivers then need to copy over packet headers, and assign + adjust
> > skb_shinfo(skb)->frags[0] offset to skip copied headers.
> >
> >
> > Unfortunately most drivers use option-B. Due to cost of calling the
> > page allocator. It is only slightly most expensive to get a larger
> > compound page from the page allocator, which then can be partitioned into
> > page-fragments, thus amortizing the page alloc cost. Unfortunately the
> > cost is added later, when constructing the SKB.
> > Another reason for option-B, is that archs with expensive IOMMU
> > requirements (like PowerPC), don't need to dma_unmap on every packet,
> > but only on the compound page level.
> >
> > Side-note: Most drivers have a "copy-break" optimization. Especially
> > for option-B, when copying header data anyhow. For small packet, one
> > might as well free (or recycle) the RX page, if header size fits into
> > the newly allocated memory (for skb_shared_info).
>
> I think you guys are going into overdesign territory, so
> . nack on read-only pages
Unfortunately you cannot just ignore or nack read-only pages. They are
a fact in the current drivers.
Most drivers today (at-least the ones we care about) only deliver
read-only pages. If you don't accept read-only pages day-1, then you
first have to rewrite a lot of drivers... and that will stall the
project! How will you deal with this fact?
The early drop filter use-case in this patchset, can ignore read-only
pages. But ABI wise we need to deal with the future case where we do
need/require writeable pages. A simple need-writable pages in the API
could help us move forward.
> . nack on copy-break approach
Copy-break can be ignored. It sort of happens at a higher-level in the
driver. (Eric likely want/care this happens for local socket delivery).
> . nack on per-ring programs
Hmmm... I don't see it as a lot more complicated to attach the program
to the ring. But maybe we can extend the API later, and thus postpone that
discussion.
> . nack on modified/stolen/shared return codes
>
> The whole thing must be dead simple to use. Above is not simple by any means.
Maybe you missed that the above was a description of how the current
network stack handles this, which is not simple... which is root of the
hole performance issue.
> The programs must see writeable pages only and return codes:
> drop, pass to stack, redirect to xmit.
> If program wishes to modify packets before passing it to stack, it
> shouldn't need to deal with different return values.
> No special things to deal with small or large packets. No header splits.
> Program must not be aware of any such things.
I agree on this. This layer only deals with packets at the page level,
single packets stored in continuous memory.
> Drivers can use DMA_BIDIRECTIONAL to allow received page to be
> modified by the program and immediately sent to xmit.
We just have to verify that DMA_BIDIRECTIONAL does not add extra
overhead (which is explicitly stated that it likely does on the
DMA-API-HOWTO.txt, but I like to verify this with a micro benchmark)
> No dma map/unmap/sync per packet. If some odd architectures/dma setups
> cannot do it, then XDP will not be applicable there.
I do like the idea of rejecting XDP eBPF programs based on the DMA
setup is not compatible, or if the driver does not implement e.g.
writable DMA pages.
Customers wanting this feature will then go buy the NIC which support
this feature. There is nothing more motivating for NIC vendors seeing
customers buying the competitors hardware. And it only require a driver
change to get this market...
> We are not going to sacrifice performance for generality.
Agree.
--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat
Author of http://www.iptv-analyzer.org
LinkedIn: http://www.linkedin.com/in/brouer
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>