Previously, the xsk code did not record which umem was bound to a
specific queue id. This was not required if all drivers were zero-copy
enabled as this had to be recorded in the driver anyway. So if a user
tried to bind two umems to the same queue, the driver would say
no. But if copy-mode was first enabled and then zero-copy mode (or the
reverse order), we mistakenly enabled both of them on the same umem
leading to buggy behavior. The main culprit for this is that we did
not store the association of umem to queue id in the copy case and
only relied on the driver reporting this. As this relation was not
stored in the driver for copy mode (it does not rely on the AF_XDP
NDOs), this obviously could not work.
This patch fixes the problem by always recording the umem to queue id
relationship in the netdev_queue and netdev_rx_queue structs. This way
we always know what kind of umem has been bound to a queue id and can
act appropriately at bind time.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit gets rid of the structure xdp_umem_props. It was there to
be able to break a dependency at one point, but this is no longer
needed. The values in the struct are instead stored directly in the
xdp_umem structure. This simplifies the xsk code as well as af_xdp
zero-copy drivers and as a bonus gets rid of one internal header file.
The i40e driver is also adapted to the new interface in this commit.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, the AF_XDP (XDP_DRV/XDP_SKB copy-mode) ingress logic did
not include XDP meta data in the data buffers copied out to the user
application.
In this commit, we check if meta data is available, and if so, it is
prepended to the frame.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
xdp_return_buff() is used when frame has been successfully
handled (transmitted) or if an error occurred during delayed
processing and there is no way to report it back to
xdp_do_redirect().
In case of __xsk_rcv_zc() error is propagated all the way
back to the driver, so there is no need to call
xdp_return_buff(). Driver will recycle the frame anyway
after seeing that error happened.
Fixes: 173d3adb6f ("xsk: add zero-copy support for Rx")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EMSGSIZE from sendmsg in copy mode when the
size of the packet is larger than the MTU. Just send it to the device
so that it will drop it as in zero-copy mode. This makes the error
reporting consistent between copy mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch makes sure ENOBUFS is always returned from sendmsg if there
is no TX queue configured. This was not the case for zero-copy
mode. With this patch this error reporting is consistent between copy
mode and zero-copy mode.
Fixes: ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EAGAIN in TX copy mode when the completion
queue is full as zero-copy does not do this. Instead this situation
can be detected by comparing the head and tail pointers of the
completion queue in both modes. In any case, EAGAIN was not the
correct error code here since no amount of calling sendmsg will solve
the problem. Only consuming one or more messages on the completion
queue will fix this.
With this patch, the error reporting becomes consistent between copy
mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch removes the ENXIO return code from TX copy-mode when
someone has forcefully changed the number of queues on the device so
that the queue bound to the socket is no longer available. Just
silently stop sending anything as in zero-copy mode so the error
reporting gets consistent between the two modes.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
There is a potential race in the TX completion code for the SKB
case. One process enters the sendmsg code of an AF_XDP socket in order
to send a frame. The execution eventually trickles down to the driver
that is told to send the packet. However, it decides to drop the
packet due to some error condition (e.g., rings full) and frees the
SKB. This will trigger the SKB destructor and a completion will be
sent to the AF_XDP user space through its
single-producer/single-consumer queues.
At the same time a TX interrupt has fired on another core and it
dispatches the TX completion code in the driver. It does its HW
specific things and ends up freeing the SKB associated with the
transmitted packet. This will trigger the SKB destructor and a
completion will be sent to the AF_XDP user space through its
single-producer/single-consumer queues. With a pseudo call stack, it
would look like this:
Core 1:
sendmsg() being called in the application
netdev_start_xmit()
Driver entered through ndo_start_xmit
Driver decides to free the SKB for some reason (e.g., rings full)
Destructor of SKB called
xskq_produce_addr() is called to signal completion to user space
Core 2:
TX completion irq
NAPI loop
Driver irq handler for TX completions
Frees the SKB
Destructor of SKB called
xskq_produce_addr() is called to signal completion to user space
We now have a violation of the single-producer/single-consumer
principle for our queues as there are two threads trying to produce at
the same time on the same queue.
Fixed by introducing a spin_lock in the destructor. In regards to the
performance, I get around 1.74 Mpps for txonly before and after the
introduction of the spinlock. There is of course some impact due to
the spin lock but it is in the less significant digits that are too
noisy for me to measure. But let us say that the version without the
spin lock got 1.745 Mpps in the best case and the version with 1.735
Mpps in the worst case, then that would mean a maximum drop in
performance of 0.5%.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Fixed a bug in which a frame could be completed more than once
when an error was returned from dev_direct_xmit(). The code
erroneously retried sending the message leading to multiple
calls to the SKB destructor and therefore multiple completions
of the same buffer to user space.
The error code in this case has been changed from EAGAIN to EBUSY
in order to tell user space that the sending of the packet failed
and the buffer has been return to user space through the completion
queue.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Reported-by: Pavel Odintsov <pavel@fastnetmon.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The poll() changes were not well thought out, and completely
unexplained. They also caused a huge performance regression, because
"->poll()" was no longer a trivial file operation that just called down
to the underlying file operations, but instead did at least two indirect
calls.
Indirect calls are sadly slow now with the Spectre mitigation, but the
performance problem could at least be largely mitigated by changing the
"->get_poll_head()" operation to just have a per-file-descriptor pointer
to the poll head instead. That gets rid of one of the new indirections.
But that doesn't fix the new complexity that is completely unwarranted
for the regular case. The (undocumented) reason for the poll() changes
was some alleged AIO poll race fixing, but we don't make the common case
slower and more complex for some uncommon special case, so this all
really needs way more explanations and most likely a fundamental
redesign.
[ This revert is a revert of about 30 different commits, not reverted
individually because that would just be unnecessarily messy - Linus ]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit 173d3adb6f ("xsk: add zero-copy support for Rx") introduced a
regression on the XDP_SKB receive path, when the queue id checks were
removed. Now, they are back again.
Fixes: 173d3adb6f ("xsk: add zero-copy support for Rx")
Reported-by: Qi Zhang <qi.z.zhang@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
With gcc-4.1.2 on 32-bit:
net/xdp/xsk.c:663: warning: integer constant is too large for ‘long’ type
net/xdp/xsk.c:665: warning: integer constant is too large for ‘long’ type
Add the missing "ULL" suffixes to the large XDP_UMEM_PGOFF_*_RING values
to fix this.
net/xdp/xsk.c:663: warning: comparison is always false due to limited range of data type
net/xdp/xsk.c:665: warning: comparison is always false due to limited range of data type
"unsigned long" is 32-bit on 32-bit systems, hence the offset is
truncated, and can never be equal to any of the XDP_UMEM_PGOFF_*_RING
values. Use loff_t (and the required cast) to fix this.
Fixes: 423f38329d ("xsk: add umem fill queue support and mmap")
Fixes: fe2308328c ("xsk: add umem completion queue support and mmap")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Pull networking updates from David Miller:
1) Add Maglev hashing scheduler to IPVS, from Inju Song.
2) Lots of new TC subsystem tests from Roman Mashak.
3) Add TCP zero copy receive and fix delayed acks and autotuning with
SO_RCVLOWAT, from Eric Dumazet.
4) Add XDP_REDIRECT support to mlx5 driver, from Jesper Dangaard
Brouer.
5) Add ttl inherit support to vxlan, from Hangbin Liu.
6) Properly separate ipv6 routes into their logically independant
components. fib6_info for the routing table, and fib6_nh for sets of
nexthops, which thus can be shared. From David Ahern.
7) Add bpf_xdp_adjust_tail helper, which can be used to generate ICMP
messages from XDP programs. From Nikita V. Shirokov.
8) Lots of long overdue cleanups to the r8169 driver, from Heiner
Kallweit.
9) Add BTF ("BPF Type Format"), from Martin KaFai Lau.
10) Add traffic condition monitoring to iwlwifi, from Luca Coelho.
11) Plumb extack down into fib_rules, from Roopa Prabhu.
12) Add Flower classifier offload support to igb, from Vinicius Costa
Gomes.
13) Add UDP GSO support, from Willem de Bruijn.
14) Add documentation for eBPF helpers, from Quentin Monnet.
15) Add TLS tx offload to mlx5, from Ilya Lesokhin.
16) Allow applications to be given the number of bytes available to read
on a socket via a control message returned from recvmsg(), from
Soheil Hassas Yeganeh.
17) Add x86_32 eBPF JIT compiler, from Wang YanQing.
18) Add AF_XDP sockets, with zerocopy support infrastructure as well.
From Björn Töpel.
19) Remove indirect load support from all of the BPF JITs and handle
these operations in the verifier by translating them into native BPF
instead. From Daniel Borkmann.
20) Add GRO support to ipv6 gre tunnels, from Eran Ben Elisha.
21) Allow XDP programs to do lookups in the main kernel routing tables
for forwarding. From David Ahern.
22) Allow drivers to store hardware state into an ELF section of kernel
dump vmcore files, and use it in cxgb4. From Rahul Lakkireddy.
23) Various RACK and loss detection improvements in TCP, from Yuchung
Cheng.
24) Add TCP SACK compression, from Eric Dumazet.
25) Add User Mode Helper support and basic bpfilter infrastructure, from
Alexei Starovoitov.
26) Support ports and protocol values in RTM_GETROUTE, from Roopa
Prabhu.
27) Support bulking in ->ndo_xdp_xmit() API, from Jesper Dangaard
Brouer.
28) Add lots of forwarding selftests, from Petr Machata.
29) Add generic network device failover driver, from Sridhar Samudrala.
* ra.kernel.org:/pub/scm/linux/kernel/git/davem/net-next: (1959 commits)
strparser: Add __strp_unpause and use it in ktls.
rxrpc: Fix terminal retransmission connection ID to include the channel
net: hns3: Optimize PF CMDQ interrupt switching process
net: hns3: Fix for VF mailbox receiving unknown message
net: hns3: Fix for VF mailbox cannot receiving PF response
bnx2x: use the right constant
Revert "net: sched: cls: Fix offloading when ingress dev is vxlan"
net: dsa: b53: Fix for brcm tag issue in Cygnus SoC
enic: fix UDP rss bits
netdev-FAQ: clarify DaveM's position for stable backports
rtnetlink: validate attributes in do_setlink()
mlxsw: Add extack messages for port_{un, }split failures
netdevsim: Add extack error message for devlink reload
devlink: Add extack to reload and port_{un, }split operations
net: metrics: add proper netlink validation
ipmr: fix error path when ipmr_new_table fails
ip6mr: only set ip6mr_table from setsockopt when ip6mr_new_table succeeds
net: hns3: remove unused hclgevf_cfg_func_mta_filter
netfilter: provide udp*_lib_lookup for nf_tproxy
qed*: Utilize FW 8.37.2.0
...
Here we add the functionality required to support zero-copy Tx, and
also exposes various zero-copy related functions for the netdevs.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Extend the xsk_rcv to support the new MEM_TYPE_ZERO_COPY memory, and
wireup ndo_bpf call in bind.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, AF_XDP only supports a fixed frame-size memory scheme where
each frame is referenced via an index (idx). A user passes the frame
index to the kernel, and the kernel acts upon the data. Some NICs,
however, do not have a fixed frame-size model, instead they have a
model where a memory window is passed to the hardware and multiple
frames are filled into that window (referred to as the "type-writer"
model).
By changing the descriptor format from the current frame index
addressing scheme, AF_XDP can in the future be extended to support
these kinds of NICs.
In the index-based model, an idx refers to a frame of size
frame_size. Addressing a frame in the UMEM is done by offseting the
UMEM starting address by a global offset, idx * frame_size + offset.
Communicating via the fill- and completion-rings are done by means of
idx.
In this commit, the idx is removed in favor of an address (addr),
which is a relative address ranging over the UMEM. To convert an
idx-based address to the new addr is simply: addr = idx * frame_size +
offset.
We also stop referring to the UMEM "frame" as a frame. Instead it is
simply called a chunk.
To transfer ownership of a chunk to the kernel, the addr of the chunk
is passed in the fill-ring. Note, that the kernel will mask addr to
make it chunk aligned, so there is no need for userspace to do
that. E.g., for a chunk size of 2k, passing an addr of 2048, 2050 or
3000 to the fill-ring will refer to the same chunk.
On the completion-ring, the addr will match that of the Tx descriptor,
passed to the kernel.
Changing the descriptor format to use chunks/addr will allow for
future changes to move to a type-writer based model, where multiple
frames can reside in one chunk. In this model passing one single chunk
into the fill-ring, would potentially result in multiple Rx
descriptors.
This commit changes the uapi of AF_XDP sockets, and updates the
documentation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, rx_dropped could be updated incorrectly, e.g. if the XDP
program redirected the frame to a socket bound to a different queue
than where the XDP program was executing.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously the fill queue descriptor was not copied to kernel space
prior validating it, making it possible for userland to change the
descriptor post-kernel-validation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
As suggested by Daniel Borkmann, the umem setup code was a too
defensive and complex. Here, we reduce the number of checks. Also, the
memory pinning is now folded into the umem creation, and we do correct
locking.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Here, we add a missing write-barrier, and use READ_ONCE for the
data-dependency barrier.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In this commit we remove the explicit ring structure from the the
uapi. It is tricky for an uapi to depend on a certain L1 cache line
size, since it can differ for variants of the same architecture. Now,
we let the user application determine the offsets of the producer,
consumer and descriptors by asking the socket via getsockopt.
A typical flow would be (Rx ring):
struct xdp_mmap_offsets off;
struct xdp_desc *ring;
u32 *prod, *cons;
void *map;
...
getsockopt(fd, SOL_XDP, XDP_MMAP_OFFSETS, &off, &optlen);
map = mmap(NULL, off.rx.desc +
NUM_DESCS * sizeof(struct xdp_desc),
PROT_READ | PROT_WRITE,
MAP_SHARED | MAP_POPULATE, sfd,
XDP_PGOFF_RX_RING);
prod = map + off.rx.producer;
cons = map + off.rx.consumer;
ring = map + off.rx.desc;
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Validate the queue id against both Rx and Tx on the netdev. Also, make
sure that the queue exists at xmit time.
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Tested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Supporting rebind, i.e. after a successful bind the process can call
bind again without closing the socket, makes the AF_XDP setup state
machine more complex. Constrain the state space, by not supporting
rebind.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Clean up SPDX-License-Identifier and removing licensing leftovers.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In this commit, a new getsockopt is added: XDP_STATISTICS. This is
used to obtain stats from the sockets.
v2: getsockopt now returns size of stats structure.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Here, Tx support is added. The user fills the Tx queue with frames to
be sent by the kernel, and let's the kernel know using the sendmsg
syscall.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Another setsockopt (XDP_TX_QUEUE) is added to let the process allocate
a queue, where the user process can pass frames to be transmitted by
the kernel.
The mmapping of the queue is done using the XDP_PGOFF_TX_QUEUE offset.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Here, we add another setsockopt for registered user memory (umem)
called XDP_UMEM_COMPLETION_QUEUE. Using this socket option, the
process can ask the kernel to allocate a queue (ring buffer) and also
mmap it (XDP_UMEM_PGOFF_COMPLETION_QUEUE) into the process.
The queue is used to explicitly pass ownership of umem frames from the
kernel to user process. This will be used by the TX path to tell user
space that a certain frame has been transmitted and user space can use
it for something else, if it wishes.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The xskmap is yet another BPF map, very much inspired by
dev/cpu/sockmap, and is a holder of AF_XDP sockets. A user application
adds AF_XDP sockets into the map, and by using the bpf_redirect_map
helper, an XDP program can redirect XDP frames to an AF_XDP socket.
Note that a socket that is bound to certain ifindex/queue index will
*only* accept XDP frames from that netdev/queue index. If an XDP
program tries to redirect from a netdev/queue index other than what
the socket is bound to, the frame will not be received on the socket.
A socket can reside in multiple maps.
v3: Fixed race and simplified code.
v2: Removed one indirection in map lookup.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Here the actual receive functions of AF_XDP are implemented, that in a
later commit, will be called from the XDP layers.
There's one set of functions for the XDP_DRV side and another for
XDP_SKB (generic).
A new XDP API, xdp_return_buff, is also introduced.
Adding xdp_return_buff, which is analogous to xdp_return_frame, but
acts upon an struct xdp_buff. The API will be used by AF_XDP in future
commits.
Support for the poll syscall is also implemented.
v2: xskq_validate_id did not update cons_tail.
The entries variable was calculated twice in xskq_nb_avail.
Squashed xdp_return_buff commit.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Here, the bind syscall is added. Binding an AF_XDP socket, means
associating the socket to an umem, a netdev and a queue index. This
can be done in two ways.
The first way, creating a "socket from scratch". Create the umem using
the XDP_UMEM_REG setsockopt and an associated fill queue with
XDP_UMEM_FILL_QUEUE. Create the Rx queue using the XDP_RX_QUEUE
setsockopt. Call bind passing ifindex and queue index ("channel" in
ethtool speak).
The second way to bind a socket, is simply skipping the
umem/netdev/queue index, and passing another already setup AF_XDP
socket. The new socket will then have the same umem/netdev/queue index
as the parent so it will share the same umem. You must also set the
flags field in the socket address to XDP_SHARED_UMEM.
v2: Use PTR_ERR instead of passing error variable explicitly.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Another setsockopt (XDP_RX_QUEUE) is added to let the process allocate
a queue, where the kernel can pass completed Rx frames from the kernel
to user process.
The mmapping of the queue is done using the XDP_PGOFF_RX_QUEUE offset.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Here, we add another setsockopt for registered user memory (umem)
called XDP_UMEM_FILL_QUEUE. Using this socket option, the process can
ask the kernel to allocate a queue (ring buffer) and also mmap it
(XDP_UMEM_PGOFF_FILL_QUEUE) into the process.
The queue is used to explicitly pass ownership of umem frames from the
user process to the kernel. These frames will in a later patch be
filled in with Rx packet data by the kernel.
v2: Fixed potential crash in xsk_mmap.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In this commit the base structure of the AF_XDP address family is set
up. Further, we introduce the abilty register a window of user memory
to the kernel via the XDP_UMEM_REG setsockopt syscall. The memory
window is viewed by an AF_XDP socket as a set of equally large
frames. After a user memory registration all frames are "owned" by the
user application, and not the kernel.
v2: More robust checks on umem creation and unaccount on error.
Call set_page_dirty_lock on cleanup.
Simplified xdp_umem_reg.
Co-authored-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Magnus Karlsson