This patch change the API for ndo_xdp_xmit to support bulking
xdp_frames.
When kernel is compiled with CONFIG_RETPOLINE, XDP sees a huge slowdown.
Most of the slowdown is caused by DMA API indirect function calls, but
also the net_device->ndo_xdp_xmit() call.
Benchmarked patch with CONFIG_RETPOLINE, using xdp_redirect_map with
single flow/core test (CPU E5-1650 v4 @ 3.60GHz), showed
performance improved:
for driver ixgbe: 6,042,682 pps -> 6,853,768 pps = +811,086 pps
for driver i40e : 6,187,169 pps -> 6,724,519 pps = +537,350 pps
With frames avail as a bulk inside the driver ndo_xdp_xmit call,
further optimizations are possible, like bulk DMA-mapping for TX.
Testing without CONFIG_RETPOLINE show the same performance for
physical NIC drivers.
The virtual NIC driver tun sees a huge performance boost, as it can
avoid doing per frame producer locking, but instead amortize the
locking cost over the bulk.
V2: Fix compile errors reported by kbuild test robot <lkp@intel.com>
V4: Isolated ndo, driver changes and callers.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In commit 6870de435b ("bpf: make virtio compatible w/
bpf_xdp_adjust_tail") i didn't account for vi->hdr_len during new
packet's length calculation after bpf_prog_run in receive_mergeable.
because of this all packets, if they were passed to the kernel,
were truncated by 12 bytes.
Fixes:6870de435b90 ("bpf: make virtio compatible w/ bpf_xdp_adjust_tail")
Reported-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Nikita V. Shirokov <tehnerd@tehnerd.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
offloads is a buffer in virtio format, should use
the __virtio64 tag.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Programming vids (adding or removing them) still passes
guest-endian values in the DMA buffer. That's wrong
if guest is big-endian and when virtio 1 is enabled.
Note: this is on top of a previous patch:
virtio_net: split out ctrl buffer
Fixes: 9465a7a6f ("virtio_net: enable v1.0 support")
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When sending control commands, virtio net sets up several buffers for
DMA. The buffers are all part of the net device which means it's
actually allocated by kvmalloc so it's in theory (on extreme memory
pressure) possible to get a vmalloc'ed buffer which on some platforms
means we can't DMA there.
Fix up by moving the DMA buffers into a separate structure.
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
w/ bpf_xdp_adjust_tail helper xdp's data_end pointer could be changed as
well (only "decrease" of pointer's location is going to be supported).
changing of this pointer will change packet's size.
for virtio driver we need to adjust XDP_PASS handling by recalculating
length of the packet if it was passed to the TCP/IP stack
Reviewed-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Nikita V. Shirokov <tehnerd@tehnerd.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Changing API ndo_xdp_xmit to take a struct xdp_frame instead of struct
xdp_buff. This brings xdp_return_frame and ndp_xdp_xmit in sync.
This builds towards changing the API further to become a bulk API,
because xdp_buff is not a queue-able object while xdp_frame is.
V4: Adjust for commit 59655a5b6c ("tuntap: XDP_TX can use native XDP")
V7: Adjust for commit d9314c474d ("i40e: add support for XDP_REDIRECT")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Changing API xdp_return_frame() to take struct xdp_frame as argument,
seems like a natural choice. But there are some subtle performance
details here that needs extra care, which is a deliberate choice.
When de-referencing xdp_frame on a remote CPU during DMA-TX
completion, result in the cache-line is change to "Shared"
state. Later when the page is reused for RX, then this xdp_frame
cache-line is written, which change the state to "Modified".
This situation already happens (naturally) for, virtio_net, tun and
cpumap as the xdp_frame pointer is the queued object. In tun and
cpumap, the ptr_ring is used for efficiently transferring cache-lines
(with pointers) between CPUs. Thus, the only option is to
de-referencing xdp_frame.
It is only the ixgbe driver that had an optimization, in which it can
avoid doing the de-reference of xdp_frame. The driver already have
TX-ring queue, which (in case of remote DMA-TX completion) have to be
transferred between CPUs anyhow. In this data area, we stored a
struct xdp_mem_info and a data pointer, which allowed us to avoid
de-referencing xdp_frame.
To compensate for this, a prefetchw is used for telling the cache
coherency protocol about our access pattern. My benchmarks show that
this prefetchw is enough to compensate the ixgbe driver.
V7: Adjust for commit d9314c474d ("i40e: add support for XDP_REDIRECT")
V8: Adjust for commit bd658dda42 ("net/mlx5e: Separate dma base address
and offset in dma_sync call")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the IDA infrastructure for getting a cyclic increasing ID number,
that is used for keeping track of each registered allocator per
RX-queue xdp_rxq_info. Instead of using the IDR infrastructure, which
uses a radix tree, use a dynamic rhashtable, for creating ID to
pointer lookup table, because this is faster.
The problem that is being solved here is that, the xdp_rxq_info
pointer (stored in xdp_buff) cannot be used directly, as the
guaranteed lifetime is too short. The info is needed on a
(potentially) remote CPU during DMA-TX completion time . In an
xdp_frame the xdp_mem_info is stored, when it got converted from an
xdp_buff, which is sufficient for the simple page refcnt based recycle
schemes.
For more advanced allocators there is a need to store a pointer to the
registered allocator. Thus, there is a need to guard the lifetime or
validity of the allocator pointer, which is done through this
rhashtable ID map to pointer. The removal and validity of of the
allocator and helper struct xdp_mem_allocator is guarded by RCU. The
allocator will be created by the driver, and registered with
xdp_rxq_info_reg_mem_model().
It is up-to debate who is responsible for freeing the allocator
pointer or invoking the allocator destructor function. In any case,
this must happen via RCU freeing.
Use the IDA infrastructure for getting a cyclic increasing ID number,
that is used for keeping track of each registered allocator per
RX-queue xdp_rxq_info.
V4: Per req of Jason Wang
- Use xdp_rxq_info_reg_mem_model() in all drivers implementing
XDP_REDIRECT, even-though it's not strictly necessary when
allocator==NULL for type MEM_TYPE_PAGE_SHARED (given it's zero).
V6: Per req of Alex Duyck
- Introduce rhashtable_lookup() call in later patch
V8: Address sparse should be static warnings (from kbuild test robot)
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The virtio_net driver assumes XDP frames are always released based on
page refcnt (via put_page). Thus, is only queues the XDP data pointer
address and uses virt_to_head_page() to retrieve struct page.
Use the XDP return API to get away from such assumptions. Instead
queue an xdp_frame, which allow us to use the xdp_return_frame API,
when releasing the frame.
V8: Avoid endianness issues (found by kbuild test robot)
V9: Change __virtnet_xdp_xmit from bool to int return value (found by Dan Carpenter)
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We tends to batch submitting packets during XDP_TX. This requires to
kick virtqueue after a batch, we tried to do it through
xdp_do_flush_map() which only makes sense for devmap not XDP_TX. So
explicitly kick the virtqueue in this case.
Reported-by: Kimitoshi Takahashi <ktaka@nii.ac.jp>
Tested-by: Kimitoshi Takahashi <ktaka@nii.ac.jp>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Fixes: 186b3c998c ("virtio-net: support XDP_REDIRECT")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The operstate update logic will leave an interface in the
default UNKNOWN operstate if the interface carrier state never changes
from the default carrier up state set at creation. This includes the
case of an explicit call to netif_carrier_on, as the carrier on to on
transition has no effect on operstate.
This affects virtio-net for the case that the virtio peer does
not support VIRTIO_NET_F_STATUS (the feature that provides carrier state
updates). Without this feature, the virtio specification states that
"the link should be assumed active," so, logically, the operstate should
be UP instead of UNKNOWN. This has impact on user space applications
that use the operstate to make availability decisions for the interface.
Resolve this by changing the virtio probe logic slightly to call
netif_carrier_off for both the "with" and "without" VIRTIO_NET_F_STATUS
cases, and then the existing call to netif_carrier_on for the "without"
case will cause an operstate transition.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
XDP_REDIRECT support for mergeable buffer was removed since commit
7324f5399b ("virtio_net: disable XDP_REDIRECT in receive_mergeable()
case"). This is because we don't reserve enough tailroom for struct
skb_shared_info which breaks XDP assumption. So this patch fixes this
by reserving enough tailroom and using fixed size of rx buffer.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We try to disable NAPI to prevent a single XDP TX queue being used by
multiple cpus. But we don't check if device is up (NAPI is enabled),
this could result stall because of infinite wait in
napi_disable(). Fixing this by checking device state through
netif_running() before.
Fixes: 4941d472bf ("virtio-net: do not reset during XDP set")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a driver implements the ndo_xdp_xmit() function, there is
(currently) no generic way to determine whether it is safe to call.
It is e.g. unsafe to call the drivers ndo_xdp_xmit, if it have not
allocated the needed XDP TX queues yet. This is the case for
virtio_net, which first allocates the XDP TX queues once an XDP/bpf
prog is attached (in virtnet_xdp_set()).
Thus, a crash will occur for virtio_net when redirecting to another
virtio_net device's ndo_xdp_xmit, which have not attached a XDP prog.
The sample xdp_redirect_map tries to attach a dummy XDP prog to take
this into account, but it can also easily fail if the virtio_net (or
actually underlying vhost driver) have not allocated enough extra
queues for the device.
Allocating more queue this is currently a manual config.
Hint for libvirt XML add:
<driver name='vhost' queues='16'>
<host mrg_rxbuf='off'/>
<guest tso4='off' tso6='off' ecn='off' ufo='off'/>
</driver>
The solution in this patch is to check that the device have loaded an
XDP/bpf prog before proceeding. This is similar to the check
performed in driver ixgbe.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
XDP_REDIRECT calling xdp_do_redirect() can fail for multiple reasons
(which can be inspected by tracepoints). The current semantics is that
on failure the driver calling xdp_do_redirect() must handle freeing or
recycling the page associated with this frame. This can be seen as an
optimization, as drivers usually have an optimized XDP_DROP code path
for frame recycling in place already.
The virtio_net driver didn't handle when xdp_do_redirect() failed.
This caused a memory leak as the page refcnt wasn't decremented on
failures.
The function __virtnet_xdp_xmit() did handle one type of failure,
when the xmit queue virtqueue_add_outbuf() is full, which "hides"
releasing a refcnt on the page. Instead the function __virtnet_xdp_xmit()
must follow API of xdp_do_redirect(), which on errors leave it up to
the caller to free the page, of the failed send operation.
Fixes: 186b3c998c ("virtio-net: support XDP_REDIRECT")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When configuring virtio_net to use the code path 'receive_small()',
in-order to get correct XDP_REDIRECT support, I discovered TCP packets
would get silently dropped when loading an XDP program action XDP_PASS.
The bug seems to be that receive_small() when XDP is loaded check that
hdr->hdr.flags is zero, which seems wrong as hdr.flags contains the
flags VIRTIO_NET_HDR_F_* :
#define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /* Use csum_start, csum_offset */
#define VIRTIO_NET_HDR_F_DATA_VALID 2 /* Csum is valid */
TCP got dropped as it had the VIRTIO_NET_HDR_F_DATA_VALID flag set.
The flags that are relevant here are the VIRTIO_NET_HDR_GSO_* flags
stored in hdr->hdr.gso_type. Thus, the fix is just check that none of
the gso_type flags have been set.
Fixes: bb91accf27 ("virtio-net: XDP support for small buffers")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The virtio_net code have three different RX code-paths in receive_buf().
Two of these code paths can handle XDP, but one of them is broken for
at least XDP_REDIRECT.
Function(1): receive_big() does not support XDP.
Function(2): receive_small() support XDP fully and uses build_skb().
Function(3): receive_mergeable() broken XDP_REDIRECT uses napi_alloc_skb().
The simple explanation is that receive_mergeable() is broken because
it uses napi_alloc_skb(), which violates XDP given XDP assumes packet
header+data in single page and enough tail room for skb_shared_info.
The longer explaination is that receive_mergeable() tries to
work-around and satisfy these XDP requiresments e.g. by having a
function xdp_linearize_page() that allocates and memcpy RX buffers
around (in case packet is scattered across multiple rx buffers). This
does currently satisfy XDP_PASS, XDP_DROP and XDP_TX (but only because
we have not implemented bpf_xdp_adjust_tail yet).
The XDP_REDIRECT action combined with cpumap is broken, and cause hard
to debug crashes. The main issue is that the RX packet does not have
the needed tail-room (SKB_DATA_ALIGN(skb_shared_info)), causing
skb_shared_info to overlap the next packets head-room (in which cpumap
stores info).
Reproducing depend on the packet payload length and if RX-buffer size
happened to have tail-room for skb_shared_info or not. But to make
this even harder to troubleshoot, the RX-buffer size is runtime
dynamically change based on an Exponentially Weighted Moving Average
(EWMA) over the packet length, when refilling RX rings.
This patch only disable XDP_REDIRECT support in receive_mergeable()
case, because it can cause a real crash.
IMHO we should consider NOT supporting XDP in receive_mergeable() at
all, because the principles behind XDP are to gain speed by (1) code
simplicity, (2) sacrificing memory and (3) where possible moving
runtime checks to setup time. These principles are clearly being
violated in receive_mergeable(), that e.g. runtime track average
buffer size to save memory consumption.
In the longer run, we should consider introducing a separate receive
function when attaching an XDP program, and also change the memory
model to be compatible with XDP when attaching an XDP prog.
Fixes: 186b3c998c ("virtio-net: support XDP_REDIRECT")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The main purpose of this patch is adding a way of checking per-queue stats.
It's useful to debug performance problems on multiqueue environment.
$ ethtool -S ens10
NIC statistics:
rx_queue_0_packets: 2090408
rx_queue_0_bytes: 3164825094
rx_queue_1_packets: 2082531
rx_queue_1_bytes: 3152932314
tx_queue_0_packets: 2770841
tx_queue_0_bytes: 4194955474
tx_queue_1_packets: 3084697
tx_queue_1_bytes: 4670196372
This change converts existing per-cpu stats structure into per-queue one.
This should not impact on performance since each queue counter is not
updated concurrently by multiple cpus.
Performance numbers:
- Guest has 2 vcpus and 2 queues
- Guest runs netserver
- Host runs 100-flow super_netperf
Before After Diff
UDP_STREAM 18byte 86.22 87.00 +0.90%
UDP_STREAM 1472byte 4055.27 4042.18 -0.32%
TCP_STREAM 16956.32 16890.63 -0.39%
UDP_RR 178667.11 185862.70 +4.03%
TCP_RR 128473.04 124985.81 -2.71%
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ability to set speed and duplex for virtio_net is useful in various
scenarios as described here:
16032be virtio_net: add ethtool support for set and get of settings
However, it would be nice to be able to set this from the hypervisor,
such that virtio_net doesn't require custom guest ethtool commands.
Introduce a new feature flag, VIRTIO_NET_F_SPEED_DUPLEX, which allows
the hypervisor to export a linkspeed and duplex setting. The user can
subsequently overwrite it later if desired via: 'ethtool -s'.
Note that VIRTIO_NET_F_SPEED_DUPLEX is defined as bit 63, the intention
is that device feature bits are to grow down from bit 63, since the
transports are starting from bit 24 and growing up.
Signed-off-by: Jason Baron <jbaron@akamai.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: virtio-dev@lists.oasis-open.org
Signed-off-by: David S. Miller <davem@davemloft.net>
The virtio_net driver doesn't dynamically change the RX-ring queue
layout and backing pages, but instead reject XDP setup if all the
conditions for XDP is not meet. Thus, the xdp_rxq_info also remains
fairly static. This allow us to simply add the reg/unreg to
net_device open/close functions.
Driver hook points for xdp_rxq_info:
* reg : virtnet_open
* unreg: virtnet_close
V3:
- bugfix, also setup xdp.rxq in receive_mergeable()
- Tested bpf-sample prog inside guest on a virtio_net device
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: virtualization@lists.linux-foundation.org
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A couple of minor bugfixes.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJaKW26AAoJECgfDbjSjVRpDwIH/1tq1vcjd5i00nh+gQD7jWm2
qWrpcKbsS0ANTvhm9UNFY24BVoPuNpNrApQpuBxmgxE/XGqx7A+8xXhwPAM5lLiG
uDSPB2nsfjvEUOde7bgeR+t6ay+Ki2UWKzY46lSoCTcIN7BSVFWQZonsGu2xLDzz
kKpCtlobXRnXzeWm+fh1oOZu/cn/TuAF0mbb+6TUQqSsHAt6PQ3Hwsly2EmQV5xR
of6Si/TnxFOkhZUKpezbuTU/g/20ZkHUSzgfrOuZLGonaIvZIE6BUm8m21/IgCbw
WnwSqe66WsWF2vd0RDLFq4L3qaRyGn+W7iYo3KPbwIPEEotSRFMUgixBWBB3IgU=
=Pf4w
-----END PGP SIGNATURE-----
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio bugfixes from Michael Tsirkin:
"A couple of minor bugfixes"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
virtio_net: fix return value check in receive_mergeable()
virtio_mmio: add cleanup for virtio_mmio_remove
virtio_mmio: add cleanup for virtio_mmio_probe
Since commit 39e6c8208d ("net: solve a NAPI race") napi has been able
to be rescheduled within napi_complete_done() even in non-busypoll case,
but virtnet_poll() always enabled interrupts before complete, and when
napi was rescheduled within napi_complete_done() it did not disable
interrupts.
This caused more interrupts when event idx is disabled.
According to commit cbdadbbf0c ("virtio_net: fix race in RX VQ
processing") we cannot place virtqueue_enable_cb_prepare() after
NAPI_STATE_SCHED is cleared, so disable interrupts again if
napi_complete_done() returned false.
Tested with vhost-user of OVS 2.7 on host, which does not have the event
idx feature.
* Before patch:
$ netperf -t UDP_STREAM -H 192.168.150.253 -l 60 -- -m 1472
MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.150.253 () port 0 AF_INET
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
212992 1472 60.00 32763206 0 6430.32
212992 60.00 23384299 4589.56
Interrupts on guest: 9872369
Packets/interrupt: 2.37
* After patch
$ netperf -t UDP_STREAM -H 192.168.150.253 -l 60 -- -m 1472
MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.150.253 () port 0 AF_INET
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
212992 1472 60.00 32794646 0 6436.49
212992 60.00 32793501 6436.27
Interrupts on guest: 4941299
Packets/interrupt: 6.64
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function virtqueue_get_buf_ctx() could return NULL, the return
value 'buf' need to be checked with NULL, not value 'ctx'.
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Merge updates from Andrew Morton:
- a few misc bits
- ocfs2 updates
- almost all of MM
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (131 commits)
memory hotplug: fix comments when adding section
mm: make alloc_node_mem_map a void call if we don't have CONFIG_FLAT_NODE_MEM_MAP
mm: simplify nodemask printing
mm,oom_reaper: remove pointless kthread_run() error check
mm/page_ext.c: check if page_ext is not prepared
writeback: remove unused function parameter
mm: do not rely on preempt_count in print_vma_addr
mm, sparse: do not swamp log with huge vmemmap allocation failures
mm/hmm: remove redundant variable align_end
mm/list_lru.c: mark expected switch fall-through
mm/shmem.c: mark expected switch fall-through
mm/page_alloc.c: broken deferred calculation
mm: don't warn about allocations which stall for too long
fs: fuse: account fuse_inode slab memory as reclaimable
mm, page_alloc: fix potential false positive in __zone_watermark_ok
mm: mlock: remove lru_add_drain_all()
mm, sysctl: make NUMA stats configurable
shmem: convert shmem_init_inodecache() to void
Unify migrate_pages and move_pages access checks
mm, pagevec: rename pagevec drained field
...
As the page free path makes no distinction between cache hot and cold
pages, there is no real useful ordering of pages in the free list that
allocation requests can take advantage of. Juding from the users of
__GFP_COLD, it is likely that a number of them are the result of copying
other sites instead of actually measuring the impact. Remove the
__GFP_COLD parameter which simplifies a number of paths in the page
allocator.
This is potentially controversial but bear in mind that the size of the
per-cpu pagelists versus modern cache sizes means that the whole per-cpu
list can often fit in the L3 cache. Hence, there is only a potential
benefit for microbenchmarks that alloc/free pages in a tight loop. It's
even worse when THP is taken into account which has little or no chance
of getting a cache-hot page as the per-cpu list is bypassed and the
zeroing of multiple pages will thrash the cache anyway.
The truncate microbenchmarks are not shown as this patch affects the
allocation path and not the free path. A page fault microbenchmark was
tested but it showed no sigificant difference which is not surprising
given that the __GFP_COLD branches are a miniscule percentage of the
fault path.
Link: http://lkml.kernel.org/r/20171018075952.10627-9-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ndo_xdp is a control path callback for setting up XDP in the
driver. We can reuse it for other forms of communication
between the eBPF stack and the drivers. Rename the callback
and associated structures and definitions.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This work enables generic transfer of metadata from XDP into skb. The
basic idea is that we can make use of the fact that the resulting skb
must be linear and already comes with a larger headroom for supporting
bpf_xdp_adjust_head(), which mangles xdp->data. Here, we base our work
on a similar principle and introduce a small helper bpf_xdp_adjust_meta()
for adjusting a new pointer called xdp->data_meta. Thus, the packet has
a flexible and programmable room for meta data, followed by the actual
packet data. struct xdp_buff is therefore laid out that we first point
to data_hard_start, then data_meta directly prepended to data followed
by data_end marking the end of packet. bpf_xdp_adjust_head() takes into
account whether we have meta data already prepended and if so, memmove()s
this along with the given offset provided there's enough room.
xdp->data_meta is optional and programs are not required to use it. The
rationale is that when we process the packet in XDP (e.g. as DoS filter),
we can push further meta data along with it for the XDP_PASS case, and
give the guarantee that a clsact ingress BPF program on the same device
can pick this up for further post-processing. Since we work with skb
there, we can also set skb->mark, skb->priority or other skb meta data
out of BPF, thus having this scratch space generic and programmable
allows for more flexibility than defining a direct 1:1 transfer of
potentially new XDP members into skb (it's also more efficient as we
don't need to initialize/handle each of such new members). The facility
also works together with GRO aggregation. The scratch space at the head
of the packet can be multiple of 4 byte up to 32 byte large. Drivers not
yet supporting xdp->data_meta can simply be set up with xdp->data_meta
as xdp->data + 1 as bpf_xdp_adjust_meta() will detect this and bail out,
such that the subsequent match against xdp->data for later access is
guaranteed to fail.
The verifier treats xdp->data_meta/xdp->data the same way as we treat
xdp->data/xdp->data_end pointer comparisons. The requirement for doing
the compare against xdp->data is that it hasn't been modified from it's
original address we got from ctx access. It may have a range marking
already from prior successful xdp->data/xdp->data_end pointer comparisons
though.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We should set xdp_xmit only when xdp_do_redirect() succeed.
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch tries to add XDP_REDIRECT for virtio-net. The changes are
not complex as we could use exist XDP_TX helpers for most of the
work. The rest is passing the XDP_TX to NAPI handler for implementing
batching.
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's no need to add packet len average in the case of XDP_PASS
since it will be done soon after skb is created.
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change is needed to not fool drop monitor.
(perf record ... -e skb:kfree_skb )
Packets were properly sent and are consumed after TX completion.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The show and store functions don't need/use the attribute.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Kernel coding style is to put paren around operand of sizeof.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The array guest_offloads is local to the source and does not need to
be in global scope, so make it static. Also tweak formatting.
Cleans up sparse warnings:
symbol 'guest_offloads' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Two minor conflicts in virtio_net driver (bug fix overlapping addition
of a helper) and MAINTAINERS (new driver edit overlapping revamp of
PHY entry).
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
1) Handle notifier registry failures properly in tun/tap driver, from
Tonghao Zhang.
2) Fix bpf verifier handling of subtraction bounds and add a testcase
for this, from Edward Cree.
3) Increase reset timeout in ftgmac100 driver, from Ben Herrenschmidt.
4) Fix use after free in prd_retire_rx_blk_timer_exired() in AF_PACKET,
from Cong Wang.
5) Fix SElinux regression due to recent UDP optimizations, from Paolo
Abeni.
6) We accidently increment IPSTATS_MIB_FRAGFAILS in the ipv6 code
paths, fix from Stefano Brivio.
7) Fix some mem leaks in dccp, from Xin Long.
8) Adjust MDIO_BUS kconfig deps to avoid build errors, from Arnd
Bergmann.
9) Mac address length check and buffer size fixes from Cong Wang.
10) Don't leak sockets in ipv6 udp early demux, from Paolo Abeni.
11) Fix return value when copy_from_user() fails in
bpf_prog_get_info_by_fd(), from Daniel Borkmann.
12) Handle PHY_HALTED properly in phy library state machine, from
Florian Fainelli.
13) Fix OOPS in fib_sync_down_dev(), from Ido Schimmel.
14) Fix truesize calculation in virtio_net which led to performance
regressions, from Michael S Tsirkin.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (76 commits)
samples/bpf: fix bpf tunnel cleanup
udp6: fix jumbogram reception
ppp: Fix a scheduling-while-atomic bug in del_chan
Revert "net: bcmgenet: Remove init parameter from bcmgenet_mii_config"
virtio_net: fix truesize for mergeable buffers
mv643xx_eth: fix of_irq_to_resource() error check
MAINTAINERS: Add more files to the PHY LIBRARY section
ipv4: fib: Fix NULL pointer deref during fib_sync_down_dev()
net: phy: Correctly process PHY_HALTED in phy_stop_machine()
sunhme: fix up GREG_STAT and GREG_IMASK register offsets
bpf: fix bpf_prog_get_info_by_fd to dump correct xlated_prog_len
tcp: avoid bogus gcc-7 array-bounds warning
net: tc35815: fix spelling mistake: "Intterrupt" -> "Interrupt"
bpf: don't indicate success when copy_from_user fails
udp6: fix socket leak on early demux
net: thunderx: Fix BGX transmit stall due to underflow
Revert "vhost: cache used event for better performance"
team: use a larger struct for mac address
net: check dev->addr_len for dev_set_mac_address()
phy: bcm-ns-usb3: fix MDIO_BUS dependency
...
Seth Forshee noticed a performance degradation with some workloads.
This turns out to be due to packet drops. Euan Kemp noticed that this
is because we drop all packets where length exceeds the truesize, but
for some packets we add in extra memory without updating the truesize.
This in turn was kept around unchanged from ab7db91705 ("virtio-net:
auto-tune mergeable rx buffer size for improved performance"). That
commit had an internal reason not to account for the extra space: not
enough bits to do it. No longer true so let's account for the allocated
length exactly.
Many thanks to Seth Forshee for the report and bisecting and Euan Kemp
for debugging the issue.
Fixes: 680557cf79 ("virtio_net: rework mergeable buffer handling")
Reported-by: Euan Kemp <euan.kemp@coreos.com>
Tested-by: Euan Kemp <euan.kemp@coreos.com>
Reported-by: Seth Forshee <seth.forshee@canonical.com>
Tested-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After removing the reset function, the freeze and restore functions
are now unused when CONFIG_PM_SLEEP is disabled:
drivers/net/virtio_net.c:1881:12: error: 'virtnet_restore_up' defined but not used [-Werror=unused-function]
static int virtnet_restore_up(struct virtio_device *vdev)
drivers/net/virtio_net.c:1859:13: error: 'virtnet_freeze_down' defined but not used [-Werror=unused-function]
static void virtnet_freeze_down(struct virtio_device *vdev)
A more robust way to do this is to remove the #ifdef around the callers
and instead mark them as __maybe_unused. The compiler will now just
silently drop the unused code.
Fixes: 4941d472bf ("virtio-net: do not reset during XDP set")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unregister the driver before removing multi-instance hotplug
callbacks. This order avoids the warning issued from
__cpuhp_remove_state_cpuslocked when the number of remaining
instances isn't yet zero.
Fixes: 8017c27919 ("net/virtio-net: Convert to hotplug state machine")
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Current XDP implementation wants guest offloads feature to be disabled
on device. This is inconvenient and means guest can't benefit from
offloads if XDP is not used. This patch tries to address this
limitation by disabling the offloads on demand through control guest
offloads. Guest offloads will be disabled and enabled on demand on XDP
set.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently reset the device during XDP set, the main reason is
that we allocate more headroom with XDP (for header adjustment).
This works but causes network downtime for users.
Previous patches encoded the headroom in the buffer context,
this makes it possible to detect the case where a buffer
with headroom insufficient for XDP is added to the queue and
XDP is enabled afterwards.
Upon detection, we handle this case by copying the packet
(slow, but it's a temporary condition).
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use ctx API to store headroom for small buffers.
Following patches will retrieve this info and use it for XDP.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pack headroom into ctx - this way when we get a buffer we can figure out
the actual headroom that was allocated for the buffer. Will be helpful
to optimize switching between XDP and non-XDP modes which have different
headroom requirements.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes: commit d45b897b11 ("virtio_net: allow specifying context for rx")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
