When listing maps, read their "frozen" status from procfs, and tell if
maps are frozen.
As commit log for map freezing command mentions that the feature might
be extended with flags (e.g. for write-only instead of read-only) in the
future, use an integer and not a boolean for JSON output.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
I opened /sys/kernel/tracing/trace once and kept reading from it.
bpf_trace_printk somehow did not seem to work, no entries were appended
to that trace file. It turns out that tracing is disabled when that file
is open. Save the next person some time and document this.
The trace file is described in Documentation/trace/ftrace.rst, however
the implication "tracing is disabled" did not immediate translate to
"bpf_trace_printk silently discards entries".
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
There is no 'struct pt_reg'.
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
PERF_EVENT_IOC_SET_BPF supports uprobes since v4.3, and tracepoints
since v4.7 via commit 04a22fae4c ("tracing, perf: Implement BPF
programs attached to uprobes"), and commit 98b5c2c65c ("perf, bpf:
allow bpf programs attach to tracepoints") respectively.
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Building s390 kernel with CONFIG_DEBUG_INFO_BTF fails, because
CONFIG_OUTPUT_FORMAT is not defined. As a matter of fact, this variable
appears to be x86-only, so other arches might be affected as well.
Fix by obtaining this value from objdump output, just like it's already
done for bin_arch. The exact objdump invocation is "inspired" by
arch/powerpc/boot/wrapper.
Also, use LANG=C for the existing bin_arch objdump invocation to avoid
potential build issues on systems with non-English locale.
Fixes: 341dfcf8d7 ("btf: expose BTF info through sysfs")
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For arm32 xdp sockets mmap2 is preferred, so use it if it's defined.
Declaration of __NR_mmap can be skipped and it breaks build.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For 64-bit there is no reason to use vmap/vunmap, so use page_address
as it was initially. For 32 bits, in some apps, like in samples
xdpsock_user.c when number of pgs in use is quite big, the kmap
memory can be not enough, despite on this, kmap looks like is
deprecated in such cases as it can block and should be used rather
for dynamic mm.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Drop __NR_mmap2 fork in flavor of LFS, that is _FILE_OFFSET_BITS=64
(glibc & bionic) / LARGEFILE64_SOURCE (for musl) decision. It allows
mmap() to use 64bit offset that is passed to mmap2 syscall. As result
pgoff is not truncated and no need to use direct access to mmap2 for
32 bits systems.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Quentin Monnet says:
====================
This set adds a new command BPF_BTF_GET_NEXT_ID to the bpf() system call,
adds the relevant API function in libbpf, and uses it in bpftool to list
all BTF objects loaded on the system (and to dump the ids of maps and
programs associated with them, if any).
The main motivation of listing BTF objects is introspection and debugging
purposes. By getting BPF program and map information, it should already be
possible to list all BTF objects associated to at least one map or one
program. But there may be unattached BTF objects, held by a file descriptor
from a user space process only, and we may want to list them too.
As a side note, it also turned useful for examining the BTF objects
attached to offloaded programs, which would not show in program information
because the BTF id is not copied when retrieving such info. A fix is in
progress on that side.
v2:
- Rebase patch with new libbpf function on top of Andrii's changes
regarding libbpf versioning.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a "btf list" (alias: "btf show") subcommand to bpftool in order to
dump all BTF objects loaded on a system.
When running the command, hash tables are built in bpftool to retrieve
all the associations between BTF objects and BPF maps and programs. This
allows for printing all such associations when listing the BTF objects.
The command is added at the top of the subcommands for "bpftool btf", so
that typing only "bpftool btf" also comes down to listing the programs.
We could not have this with the previous command ("dump"), which
required a BTF object id, so it should not break any previous behaviour.
This also makes the "btf" command behaviour consistent with "prog" or
"map".
Bash completion is updated to use "bpftool btf" instead of "bpftool
prog" to list the BTF ids, as it looks more consistent.
Example output (plain):
# bpftool btf show
9: size 2989B prog_ids 21 map_ids 15
17: size 2847B prog_ids 36 map_ids 30,29,28
26: size 2847B
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add an API function taking a BTF object id and providing the id of the
next BTF object in the kernel. This can be used to list all BTF objects
loaded on the system.
v2:
- Rebase on top of Andrii's changes regarding libbpf versioning.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In preparation for the introduction of a similar function for retrieving
the id of the next BTF object, consolidate the code from
bpf_prog_get_next_id() and bpf_map_get_next_id() in libbpf.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Synchronise the bpf.h header under tools, to report the addition of the
new BPF_BTF_GET_NEXT_ID syscall command for bpf().
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a new command for the bpf() system call: BPF_BTF_GET_NEXT_ID is used
to cycle through all BTF objects loaded on the system.
The motivation is to be able to inspect (list) all BTF objects presents
on the system.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
r369217 in clang added a new warning about potential misuse of the xor
operator as an exponentiation operator:
../lib/test_bpf.c:870:13: warning: result of '10 ^ 300' is 294; did you
mean '1e300'? [-Wxor-used-as-pow]
{ { 4, 10 ^ 300 }, { 20, 10 ^ 300 } },
~~~^~~~~
1e300
../lib/test_bpf.c:870:13: note: replace expression with '0xA ^ 300' to
silence this warning
../lib/test_bpf.c:870:31: warning: result of '10 ^ 300' is 294; did you
mean '1e300'? [-Wxor-used-as-pow]
{ { 4, 10 ^ 300 }, { 20, 10 ^ 300 } },
~~~^~~~~
1e300
../lib/test_bpf.c:870:31: note: replace expression with '0xA ^ 300' to
silence this warning
The commit link for this new warning has some good logic behind wanting
to add it but this instance appears to be a false positive. Adopt its
suggestion to silence the warning but not change the code. According to
the differential review link in the clang commit, GCC may eventually
adopt this warning as well.
Link: https://github.com/ClangBuiltLinux/linux/issues/643
Link: 920890e268
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add a header include guard just in case.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Implement the show_fdinfo hook for BTF FDs file operations, and make it
print the id of the BTF object. This allows for a quick retrieval of the
BTF id from its FD; or it can help understanding what type of object
(BTF) the file descriptor points to.
v2:
- Do not expose data_size, only btf_id, in FD info.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Use PTR_ERR_OR_ZERO rather than if(IS_ERR(...)) + PTR_ERR.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Björn Töpel says:
====================
This series (v5 and counting) add two improvements for the XSKMAP,
used by AF_XDP sockets.
1. Automatic cleanup when an AF_XDP socket goes out of scope/is
released. Instead of require that the user manually clears the
"released" state socket from the map, this is done
automatically. Each socket tracks which maps it resides in, and
remove itself from those maps at relase. A notable implementation
change, is that the sockets references the map, instead of the map
referencing the sockets. Which implies that when the XSKMAP is
freed, it is by definition cleared of sockets.
2. The XSKMAP did not honor the BPF_EXIST/BPF_NOEXIST flag on insert,
which this patch addresses.
v1->v2: Fixed deadlock and broken cleanup. (Daniel)
v2->v3: Rebased onto bpf-next
v3->v4: {READ, WRITE}_ONCE consistency. (Daniel)
Socket release/map update race. (Daniel)
v4->v5: Avoid use-after-free on XSKMAP self-assignment [1]. (Daniel)
Removed redundant assignment in xsk_map_update_elem().
Variable name consistency; Use map_entry everywhere.
[1] https://lore.kernel.org/bpf/20190802081154.30962-1-bjorn.topel@gmail.com/T/#mc68439e97bc07fa301dad9fc4850ed5aa392f385
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The XSKMAP did not honor the BPF_EXIST/BPF_NOEXIST flags when updating
an entry. This patch addresses that.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When an AF_XDP socket is released/closed the XSKMAP still holds a
reference to the socket in a "released" state. The socket will still
use the netdev queue resource, and block newly created sockets from
attaching to that queue, but no user application can access the
fill/complete/rx/tx queues. This results in that all applications need
to explicitly clear the map entry from the old "zombie state"
socket. This should be done automatically.
In this patch, the sockets tracks, and have a reference to, which maps
it resides in. When the socket is released, it will remove itself from
all maps.
Suggested-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Stanislav Fomichev says:
====================
Currently there is no way to propagate sk storage from the listener
socket to a newly accepted one. Consider the following use case:
fd = socket();
setsockopt(fd, SOL_IP, IP_TOS,...);
/* ^^^ setsockopt BPF program triggers here and saves something
* into sk storage of the listener.
*/
listen(fd, ...);
while (client = accept(fd)) {
/* At this point all association between listener
* socket and newly accepted one is gone. New
* socket will not have any sk storage attached.
*/
}
Let's add new BPF_F_CLONE flag that can be specified when creating
a socket storage map. This new flag indicates that map contents
should be cloned when the socket is cloned.
v4:
* drop 'goto err' in bpf_sk_storage_clone (Yonghong Song)
* add comment about race with bpf_sk_storage_map_free to the
bpf_sk_storage_clone side as well (Daniel Borkmann)
v3:
* make sure BPF_F_NO_PREALLOC is always present when creating
a map (Martin KaFai Lau)
* don't call bpf_sk_storage_free explicitly, rely on
sk_free_unlock_clone to do the cleanup (Martin KaFai Lau)
v2:
* remove spinlocks around selem_link_map/sk (Martin KaFai Lau)
* BPF_F_CLONE on a map, not selem (Martin KaFai Lau)
* hold a map while cloning (Martin KaFai Lau)
* use BTF maps in selftests (Yonghong Song)
* do proper cleanup selftests; don't call close(-1) (Yonghong Song)
* export bpf_map_inc_not_zero
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add a test that calls setsockopt on the listener socket which triggers
BPF program. This BPF program writes to the sk storage and sets
clone flag. Make sure that sk storage is cloned for a newly
accepted connection.
We have two cloned maps in the tests to make sure we hit both cases
in bpf_sk_storage_clone: first element (sk_storage_alloc) and
non-first element(s) (selem_link_map).
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Sync new sk storage clone flag.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add new helper bpf_sk_storage_clone which optionally clones sk storage
and call it from sk_clone_lock.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Rename existing bpf_map_inc_not_zero to __bpf_map_inc_not_zero to
indicate that it's caller's responsibility to do proper locking.
Create and export bpf_map_inc_not_zero wrapper that properly
locks map_idr_lock. Will be used in the next commit to
hold a map while cloning a socket.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
There is a race in this test between receiving the ACK for the
single-byte packet sent in the test, and reading the values from the
map.
This patch fixes this by having the client wait until there are no more
unacknowledged packets.
Before:
for i in {1..1000}; do ../net/in_netns.sh ./test_tcp_rtt; \
done | grep -c PASSED
< trimmed error messages >
993
After:
for i in {1..10000}; do ../net/in_netns.sh ./test_tcp_rtt; \
done | grep -c PASSED
10000
Fixes: b55873984d ("selftests/bpf: test BPF_SOCK_OPS_RTT_CB")
Signed-off-by: Petar Penkov <ppenkov@google.com>
Reviewed-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
bpf_helpers.h and bpf_endian.h contain useful macros and BPF helper
definitions essential to almost every BPF program. Which makes them
useful not just for selftests. To be able to expose them as part of
libbpf, though, we need them to be dual-licensed as LGPL-2.1 OR
BSD-2-Clause. This patch updates licensing of those two files.
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Hechao Li <hechaol@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Adam Barth <arb@fb.com>
Acked-by: Roman Gushchin <guro@fb.com>
Acked-by: Josef Bacik <jbacik@fb.com>
Acked-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: David Ahern <dsahern@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>
Acked-by: Lorenz Bauer <lmb@cloudflare.com>
Acked-by: Adrian Ratiu <adrian.ratiu@collabora.com>
Acked-by: Nikita V. Shirokov <tehnerd@tehnerd.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Petar Penkov <ppenkov@google.com>
Acked-by: Teng Qin <palmtenor@gmail.com>
Cc: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Michal Rostecki <mrostecki@opensuse.org>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Sargun Dhillon <sargun@sargun.me>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Don't uninstall an XDP program when none is installed, and don't install
an XDP program that has the same ID as the one already installed.
dev_change_xdp_fd doesn't perform any checks in case it uninstalls an
XDP program. It means that the driver's ndo_bpf can be called with
XDP_SETUP_PROG asking to set it to NULL even if it's already NULL. This
case happens if the user runs `ip link set eth0 xdp off` when there is
no XDP program attached.
The symmetrical case is possible when the user tries to set the program
that is already set.
The drivers typically perform some heavy operations on XDP_SETUP_PROG,
so they all have to handle these cases internally to return early if
they happen. This patch puts this check into the kernel code, so that
all drivers will benefit from it.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Magnus Karlsson says:
====================
This patch set adds support for a new flag called need_wakeup in the
AF_XDP Tx and fill rings. When this flag is set by the driver, it
means that the application has to explicitly wake up the kernel Rx
(for the bit in the fill ring) or kernel Tx (for bit in the Tx ring)
processing by issuing a syscall. Poll() can wake up both and sendto()
will wake up Tx processing only.
The main reason for introducing this new flag is to be able to
efficiently support the case when application and driver is executing
on the same core. Previously, the driver was just busy-spinning on the
fill ring if it ran out of buffers in the HW and there were none to
get from the fill ring. This approach works when the application and
driver is running on different cores as the application can replenish
the fill ring while the driver is busy-spinning. Though, this is a
lousy approach if both of them are running on the same core as the
probability of the fill ring getting more entries when the driver is
busy-spinning is zero. With this new feature the driver now sets the
need_wakeup flag and returns to the application. The application can
then replenish the fill queue and then explicitly wake up the Rx
processing in the kernel using the syscall poll(). For Tx, the flag is
only set to one if the driver has no outstanding Tx completion
interrupts. If it has some, the flag is zero as it will be woken up by
a completion interrupt anyway. This flag can also be used in other
situations where the driver needs to be woken up explicitly.
As a nice side effect, this new flag also improves the Tx performance
of the case where application and driver are running on two different
cores as it reduces the number of syscalls to the kernel. The kernel
tells user space if it needs to be woken up by a syscall, and this
eliminates many of the syscalls. The Rx performance of the 2-core case
is on the other hand slightly worse, since there is a need to use a
syscall now to wake up the driver, instead of the driver
busy-spinning. It does waste less CPU cycles though, which might lead
to better overall system performance.
This new flag needs some simple driver support. If the driver does not
support it, the Rx flag is always zero and the Tx flag is always
one. This makes any application relying on this feature default to the
old behavior of not requiring any syscalls in the Rx path and always
having to call sendto() in the Tx path.
For backwards compatibility reasons, this feature has to be explicitly
turned on using a new bind flag (XDP_USE_NEED_WAKEUP). I recommend
that you always turn it on as it has a large positive performance
impact for the one core case and does not degrade 2 core performance
and actually improves it for Tx heavy workloads.
Here are some performance numbers measured on my local,
non-performance optimized development system. That is why you are
seeing numbers lower than the ones from Björn and Jesper. 64 byte
packets at 40Gbit/s line rate. All results in Mpps. Cores == 1 means
that both application and driver is executing on the same core. Cores
== 2 that they are on different cores.
Applications
need_wakeup cores txpush rxdrop l2fwd
---------------------------------------------------------------
n 1 0.07 0.06 0.03
y 1 21.6 8.2 6.5
n 2 32.3 11.7 8.7
y 2 33.1 11.7 8.7
Overall, the need_wakeup flag provides the same or better performance
in all the micro-benchmarks. The reduction of sendto() calls in txpush
is large. Only a few per second is needed. For l2fwd, the drop is 50%
for the 1 core case and more than 99.9% for the 2 core case. Do not
know why I am not seeing the same drop for the 1 core case yet.
The name and inspiration of the flag has been taken from io_uring by
Jens Axboe. Details about this feature in io_uring can be found in
http://kernel.dk/io_uring.pdf, section 8.3. It also addresses most of
the denial of service and sendto() concerns raised by Maxim
Mikityanskiy in https://www.spinics.net/lists/netdev/msg554657.html.
The typical Tx part of an application will have to change from:
ret = sendto(fd,....)
to:
if (xsk_ring_prod__needs_wakeup(&xsk->tx))
ret = sendto(fd,....)
and th Rx part from:
rcvd = xsk_ring_cons__peek(&xsk->rx, BATCH_SIZE, &idx_rx);
if (!rcvd)
return;
to:
rcvd = xsk_ring_cons__peek(&xsk->rx, BATCH_SIZE, &idx_rx);
if (!rcvd) {
if (xsk_ring_prod__needs_wakeup(&xsk->umem->fq))
ret = poll(fd,.....);
return;
}
v3 -> v4:
* Maxim found a possible race in the Tx part of the driver. The
setting of the flag needs to happen before the sending, otherwise it
might trigger this race. Fixed in ixgbe and i40e driver.
* Mellanox support contributed by Maxim
* Removed the XSK_DRV_CAN_SLEEP flag as it was not used
anymore. Thanks to Sridhar for discovering this.
* For consistency the feature is now always called need_wakeup. There
were some places where it was referred to as might_sleep, but they
have been removed. Thanks to Sridhar for spotting.
* Fixed some typos in the commit messages
v2 -> v3:
* Converted the Mellanox driver to the new ndo in patch 1 as pointed
out by Maxim
* Fixed the compatibility code of XDP_MMAP_OFFSETS so it now works.
v1 -> v2:
* Fixed bisectability problem pointed out by Jakub
* Added missing initiliztion of the Tx need_wakeup flag to 1
This patch has been applied against commit b753c5a7f9 ("Merge branch 'r8152-RX-improve'")
Structure of the patch set:
Patch 1: Replaces the ndo_xsk_async_xmit with ndo_xsk_wakeup to
support waking up both Rx and Tx processing
Patch 2: Implements the need_wakeup functionality in common code
Patch 3-4: Add need_wakeup support to the i40e and ixgbe drivers
Patch 5: Add need_wakeup support to libbpf
Patch 6: Add need_wakeup support to the xdpsock sample application
Patch 7-8: Add need_wakeup support to the Mellanox mlx5 driver
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds support for the new need_wakeup feature of AF_XDP. The
applications can opt-in by using the XDP_USE_NEED_WAKEUP bind() flag.
When this feature is enabled, some behavior changes:
RX side: If the Fill Ring is empty, instead of busy-polling, set the
flag to tell the application to kick the driver when it refills the Fill
Ring.
TX side: If there are pending completions or packets queued for
transmission, set the flag to tell the application that it can skip the
sendto() syscall and save time.
The performance testing was performed on a machine with the following
configuration:
- 24 cores of Intel Xeon E5-2620 v3 @ 2.40 GHz
- Mellanox ConnectX-5 Ex with 100 Gbit/s link
The results with retpoline disabled:
| without need_wakeup | with need_wakeup |
|----------------------|----------------------|
| one core | two cores | one core | two cores |
-------|----------|-----------|----------|-----------|
txonly | 20.1 | 33.5 | 29.0 | 34.2 |
rxdrop | 0.065 | 14.1 | 12.0 | 14.1 |
l2fwd | 0.032 | 7.3 | 6.6 | 7.2 |
"One core" means the application and NAPI run on the same core. "Two
cores" means they are pinned to different cores.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Two XSK tasks are performed during NAPI polling, that are not bound to
hardware interrupts: TXing packets and polling for frames in the Fill
Ring. They are special in a way that the hardware doesn't know about
these tasks, so it doesn't trigger interrupts if there is still some
work to be done, it's our driver's responsibility to ensure NAPI will be
rescheduled if needed.
Create a new function to handle these tasks and move the corresponding
code from mlx5e_napi_poll to the new function to improve modularity and
prepare for the changes in the following patch.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds using the need_wakeup flag to the xdpsock sample
application. It is turned on by default as we think it is a feature
that seems to always produce a performance benefit, if the application
has been written taking advantage of it. It can be turned off in the
sample app by using the '-m' command line option.
The txpush and l2fwd sub applications have also been updated to
support poll() with multiple sockets.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds support for the new need_wakeup flag in AF_XDP. The
xsk_socket__create function is updated to handle this and a new
function is introduced called xsk_ring_prod__needs_wakeup(). This
function can be used by the application to check if Rx and/or Tx
processing needs to be explicitly woken up.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch adds support for the need_wakeup feature of AF_XDP. If the
application has told the kernel that it might sleep using the new bind
flag XDP_USE_NEED_WAKEUP, the driver will then set this flag if it has
no more buffers on the NIC Rx ring and yield to the application. For
Tx, it will set the flag if it has no outstanding Tx completion
interrupts and return to the application.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch adds support for the need_wakeup feature of AF_XDP. If the
application has told the kernel that it might sleep using the new bind
flag XDP_USE_NEED_WAKEUP, the driver will then set this flag if it has
no more buffers on the NIC Rx ring and yield to the application. For
Tx, it will set the flag if it has no outstanding Tx completion
interrupts and return to the application.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds support for a new flag called need_wakeup in the
AF_XDP Tx and fill rings. When this flag is set, it means that the
application has to explicitly wake up the kernel Rx (for the bit in
the fill ring) or kernel Tx (for bit in the Tx ring) processing by
issuing a syscall. Poll() can wake up both depending on the flags
submitted and sendto() will wake up tx processing only.
The main reason for introducing this new flag is to be able to
efficiently support the case when application and driver is executing
on the same core. Previously, the driver was just busy-spinning on the
fill ring if it ran out of buffers in the HW and there were none on
the fill ring. This approach works when the application is running on
another core as it can replenish the fill ring while the driver is
busy-spinning. Though, this is a lousy approach if both of them are
running on the same core as the probability of the fill ring getting
more entries when the driver is busy-spinning is zero. With this new
feature the driver now sets the need_wakeup flag and returns to the
application. The application can then replenish the fill queue and
then explicitly wake up the Rx processing in the kernel using the
syscall poll(). For Tx, the flag is only set to one if the driver has
no outstanding Tx completion interrupts. If it has some, the flag is
zero as it will be woken up by a completion interrupt anyway.
As a nice side effect, this new flag also improves the performance of
the case where application and driver are running on two different
cores as it reduces the number of syscalls to the kernel. The kernel
tells user space if it needs to be woken up by a syscall, and this
eliminates many of the syscalls.
This flag needs some simple driver support. If the driver does not
support this, the Rx flag is always zero and the Tx flag is always
one. This makes any application relying on this feature default to the
old behaviour of not requiring any syscalls in the Rx path and always
having to call sendto() in the Tx path.
For backwards compatibility reasons, this feature has to be explicitly
turned on using a new bind flag (XDP_USE_NEED_WAKEUP). I recommend
that you always turn it on as it so far always have had a positive
performance impact.
The name and inspiration of the flag has been taken from io_uring by
Jens Axboe. Details about this feature in io_uring can be found in
http://kernel.dk/io_uring.pdf, section 8.3.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit replaces ndo_xsk_async_xmit with ndo_xsk_wakeup. This new
ndo provides the same functionality as before but with the addition of
a new flags field that is used to specifiy if Rx, Tx or both should be
woken up. The previous ndo only woke up Tx, as implied by the
name. The i40e and ixgbe drivers (which are all the supported ones)
are updated with this new interface.
This new ndo will be used by the new need_wakeup functionality of XDP
sockets that need to be able to wake up both Rx and Tx driver
processing.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In case of error, the function kobject_create_and_add() returns NULL
pointer not ERR_PTR(). The IS_ERR() test in the return value check
should be replaced with NULL test.
Fixes: 341dfcf8d7 ("btf: expose BTF info through sysfs")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Quentin Monnet says:
====================
Because the "__printf()" attributes were used only where the functions are
implemented, and not in header files, the checks have not been enforced on
all the calls to printf()-like functions, and a number of errors slipped in
bpftool over time.
This set cleans up such errors, and then moves the "__printf()" attributes
to header files, so that the checks are performed at all locations.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Some functions in bpftool have a "__printf()" format attributes to tell
the compiler they should expect printf()-like arguments. But because
these attributes are not used for the function prototypes in the header
files, the compiler does not run the checks everywhere the functions are
used, and some mistakes on format string and corresponding arguments
slipped in over time.
Let's move the __printf() attributes to the correct places.
Note: We add guards around the definition of GCC_VERSION in
tools/include/linux/compiler-gcc.h to prevent a conflict in jit_disasm.c
on GCC_VERSION from headers pulled via libbfd.
Fixes: c101189bc9 ("tools: bpftool: fix -Wmissing declaration warnings")
Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
There is one call to the p_err() function in detect_common_prefix()
where the message to print is passed directly as the first argument,
without using a format string. This is harmless, but may trigger
warnings if the "__printf()" attribute is used correctly for the p_err()
function. Let's fix it by using a "%s" format string.
Fixes: ba95c74524 ("tools: bpftool: add "prog run" subcommand to test-run programs")
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The format string passed to one call to the p_err() function in
query_flow_dissector() does not match the value that should be printed,
resulting in some garbage integer being printed instead of
strerror(errno) if /proc/self/ns/net cannot be open. Let's fix the
format string.
Fixes: 7f0c57fec8 ("bpftool: show flow_dissector attachment status")
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The last argument passed to one call to the p_err() function is not
correct, it should be "*argv" instead of "**argv". This may lead to a
segmentation fault error if BTF id cannot be parsed correctly. Let's fix
this.
Fixes: c93cc69004dt ("bpftool: add ability to dump BTF types")
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
There are some mismatches between format strings and arguments passed to
jsonw_printf() in the BTF dumper for bpftool, which seems harmless but
may result in warnings if the "__printf()" attribute is used correctly
for jsonw_printf(). Let's fix relevant format strings and type cast.
Fixes: b12d6ec097 ("bpf: btf: add btf print functionality")
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The last argument passed to some calls to the p_err() functions is not
correct, it should be "*argv" instead of "**argv". This may lead to a
segmentation fault error if CPU IDs or indices from the command line
cannot be parsed correctly. Let's fix this.
Fixes: f412eed9df ("tools: bpftool: add simple perf event output reader")
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently libbpf version is specified in 2 places: libbpf.map and
Makefile. They easily get out of sync and it's very easy to update one,
but forget to update another one. In addition, Github projection of
libbpf has to maintain its own version which has to be remembered to be
kept in sync manually, which is very error-prone approach.
This patch makes libbpf.map a source of truth for libbpf version and
uses shell invocation to parse out correct full and major libbpf version
to use during build. Now we need to make sure that once new release
cycle starts, we need to add (initially) empty section to libbpf.map
with correct latest version.
This also will make it possible to keep Github projection consistent
with kernel sources version of libbpf by adopting similar parsing of
version from libbpf.map.
v2->v3:
- grep -o + sort -rV (Andrey);
v1->v2:
- eager version vars evaluation (Jakub);
- simplified version regex (Andrey);
Cc: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Daniel T. Lee says:
====================
Currently, bpftool net only supports dumping progs attached on the
interface. To attach XDP prog on interface, user must use other tool
(eg. iproute2). By this patch, with `bpftool net attach/detach`, user
can attach/detach XDP prog on interface.
# bpftool prog
16: xdp name xdp_prog1 tag 539ec6ce11b52f98 gpl
loaded_at 2019-08-07T08:30:17+0900 uid 0
...
20: xdp name xdp_fwd_prog tag b9cb69f121e4a274 gpl
loaded_at 2019-08-07T08:30:17+0900 uid 0
# bpftool net attach xdpdrv id 16 dev enp6s0np0
# bpftool net
xdp:
enp6s0np0(4) driver id 16
# bpftool net attach xdpdrv id 20 dev enp6s0np0 overwrite
# bpftool net
xdp:
enp6s0np0(4) driver id 20
# bpftool net detach xdpdrv dev enp6s0np0
# bpftool net
xdp:
While this patch only contains support for XDP, through `net
attach/detach`, bpftool can further support other prog attach types.
XDP attach/detach tested on Mellanox ConnectX-4 and Netronome Agilio.
---
Changes in v5:
- fix wrong error message, from errno to err with do_attach/detach
Changes in v4:
- rename variable, attach/detach error message enhancement
- bash-completion cleanup, doc update with brief description (attach
types)
Changes in v3:
- added 'overwrite' option for replacing previously attached XDP prog
- command argument order has been changed ('ATTACH_TYPE' comes first)
- add 'dev' keyword in front of <devname>
- added bash-completion and documentation
Changes in v2:
- command 'load/unload' changed to 'attach/detach' for the consistency
====================
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Since, new sub-command 'net attach/detach' has been added for
attaching XDP program on interface,
this commit documents usage and sample output of `net attach/detach`.
Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Quentin Monnet
Peter Wu
Ilya Leoshkevich
Ivan Khoronzhuk
Alexei Starovoitov
Nathan Chancellor
Masahiro Yamada
YueHaibing
Daniel Borkmann
Björn Töpel
Stanislav Fomichev
Petar Penkov
Andrii Nakryiko
Maxim Mikityanskiy
Magnus Karlsson
Wei Yongjun
Daniel T. Lee