I've seen a situation, where a process that's under pprof constantly
generates SIGPROF which prevents program loading indefinitely.
The right thing to do probably is to disable signals in the upper
layers while loading, but it still would be nice to get some error from
libbpf instead of an endless loop.
Let's add some small retry limit to the program loading:
try loading the program 5 (arbitrary) times and give up.
v2:
* 10 -> 5 retires (Andrii Nakryiko)
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20201202231332.3923644-1-sdf@google.com
When we added sanitising of map names before loading programs to libbpf, we
still allowed periods in the name. While the kernel will accept these for
the map names themselves, they are not allowed in file names when pinning
maps. This means that bpf_object__pin_maps() will fail if called on an
object that contains internal maps (such as sections .rodata).
Fix this by replacing periods with underscores when constructing map pin
paths. This only affects the paths generated by libbpf when
bpf_object__pin_maps() is called with a path argument. Any pin paths set
by bpf_map__set_pin_path() are unaffected, and it will still be up to the
caller to avoid invalid characters in those.
Fixes: 113e6b7e15e2 ("libbpf: Sanitise internal map names so they are not rejected by the kernel")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20201203093306.107676-1-toke@redhat.com
Before this patch, a program with unspecified type
(BPF_PROG_TYPE_UNSPEC) would be passed to the BPF syscall, only to have
the kernel reject it with an opaque invalid argument error. This patch
makes libbpf reject such programs with a nicer error message - in
particular libbpf now tries to diagnose bad ELF section names at both
open time and load time.
Signed-off-by: Andrei Matei <andreimatei1@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20201203043410.59699-1-andreimatei1@gmail.com
Add support for separation of eBPF program load and xsk socket
creation.
This is needed for use-case when you want to privide as little
privileges as possible to the data plane application that will
handle xsk socket creation and incoming traffic.
With this patch the data entity container can be run with only
CAP_NET_RAW capability to fulfill its purpose of creating xsk
socket and handling packages. In case your umem is larger or
equal process limit for MEMLOCK you need either increase the
limit or CAP_IPC_LOCK capability.
To resolve privileges issue two APIs are introduced:
- xsk_setup_xdp_prog - loads the built in XDP program. It can
also return xsks_map_fd which is needed by unprivileged process
to update xsks_map with AF_XDP socket "fd"
- xsk_socket__update_xskmap - inserts an AF_XDP socket into an xskmap
for a particular xsk_socket
Signed-off-by: Mariusz Dudek <mariuszx.dudek@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20201203090546.11976-2-mariuszx.dudek@intel.com
Replace size_t with __u32 in the xsk interfaces that contain this.
There is no reason to have size_t since the internal variable that
is manipulated is a __u32. The following APIs are affected:
__u32 xsk_ring_prod__reserve(struct xsk_ring_prod *prod, __u32 nb, __u32 *idx)
void xsk_ring_prod__submit(struct xsk_ring_prod *prod, __u32 nb)
__u32 xsk_ring_cons__peek(struct xsk_ring_cons *cons, __u32 nb, __u32 *idx)
void xsk_ring_cons__cancel(struct xsk_ring_cons *cons, __u32 nb)
void xsk_ring_cons__release(struct xsk_ring_cons *cons, __u32 nb)
The "nb" variable and the return values have been changed from size_t
to __u32.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1606383455-8243-1-git-send-email-magnus.karlsson@gmail.com
Provide a wrapper function to get the IMA hash of an inode. This helper
is useful in fingerprinting files (e.g executables on execution) and
using these fingerprints in detections like an executable unlinking
itself.
Since the ima_inode_hash can sleep, it's only allowed for sleepable
LSM hooks.
Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20201124151210.1081188-3-kpsingh@chromium.org
Add a new function for returning descriptors the user received
after an xsk_ring_cons__peek call. After the application has
gotten a number of descriptors from a ring, it might not be able
to or want to process them all for various reasons. Therefore,
it would be useful to have an interface for returning or
cancelling a number of them so that they are returned to the ring.
This patch adds a new function called xsk_ring_cons__cancel that
performs this operation on nb descriptors counted from the end of
the batch of descriptors that was received through the peek call.
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
[ Magnus Karlsson: rewrote changelog ]
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/1606202474-8119-1-git-send-email-lirongqing@baidu.com
The helper uses CLOCK_MONOTONIC_COARSE source of time that is less
accurate but more performant.
We have a BPF CGROUP_SKB firewall that supports event logging through
bpf_perf_event_output(). Each event has a timestamp and currently we use
bpf_ktime_get_ns() for it. Use of bpf_ktime_get_coarse_ns() saves ~15-20
ns in time required for event logging.
bpf_ktime_get_ns():
EgressLogByRemoteEndpoint 113.82ns 8.79M
bpf_ktime_get_coarse_ns():
EgressLogByRemoteEndpoint 95.40ns 10.48M
Signed-off-by: Dmitrii Banshchikov <me@ubique.spb.ru>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20201117184549.257280-1-me@ubique.spb.ru
The helper allows modification of certain bits on the linux_binprm
struct starting with the secureexec bit which can be updated using the
BPF_F_BPRM_SECUREEXEC flag.
secureexec can be set by the LSM for privilege gaining executions to set
the AT_SECURE auxv for glibc. When set, the dynamic linker disables the
use of certain environment variables (like LD_PRELOAD).
Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20201117232929.2156341-1-kpsingh@chromium.org
When operating on split BTF, btf__find_by_name[_kind] will not
iterate over all types since they use btf->nr_types to show
the number of types to iterate over. For split BTF this is
the number of types _on top of base BTF_, so it will
underestimate the number of types to iterate over, especially
for vmlinux + module BTF, where the latter is much smaller.
Use btf__get_nr_types() instead.
Fixes: ba451366bf44 ("libbpf: Implement basic split BTF support")
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/1605437195-2175-1-git-send-email-alan.maguire@oracle.com
Background:
Broadcast and multicast packages are enqueued for later processing.
This queue was previously hardcoded to 1000.
This proved insufficient for handling very high packet rates.
This resulted in packet drops for multicast.
While at the same time unicast worked fine.
The change:
This patch make the queue length adjustable to accommodate
for environments with very high multicast packet rate.
But still keeps the default value of 1000 unless specified.
The queue length is specified as a request per macvlan
using the IFLA_MACVLAN_BC_QUEUE_LEN parameter.
The actual used queue length will then be the maximum of
any macvlan connected to the same port. The actual used
queue length for the port can be retrieved (read only)
by the IFLA_MACVLAN_BC_QUEUE_LEN_USED parameter for verification.
This will be followed up by a patch to iproute2
in order to adjust the parameter from userspace.
Signed-off-by: Thomas Karlsson <thomas.karlsson@paneda.se>
Link: https://lore.kernel.org/r/dd4673b2-7eab-edda-6815-85c67ce87f63@paneda.se
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If BPF code contains unused BPF subprogram and there are no other subprogram
calls (which can realistically happen in real-world applications given
sufficiently smart Clang code optimizations), libbpf will erroneously assume
that subprograms are entry-point programs and will attempt to load them with
UNSPEC program type.
Fix by not relying on subcall instructions and rather detect it based on the
structure of BPF object's sections.
Fixes: 9a94f277c4fb ("tools: libbpf: restore the ability to load programs from .text section")
Reported-by: Dmitrii Banshchikov <dbanschikov@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20201107000251.256821-1-andrii@kernel.org
Allocate ID for vmlinux BTF. This makes it visible when iterating over all BTF
objects in the system. To allow distinguishing vmlinux BTF (and later kernel
module BTF) from user-provided BTFs, expose extra kernel_btf flag, as well as
BTF name ("vmlinux" for vmlinux BTF, will equal to module's name for module
BTF). We might want to later allow specifying BTF name for user-provided BTFs
as well, if that makes sense. But currently this is reserved only for
in-kernel BTFs.
Having in-kernel BTFs exposed IDs will allow to extend BPF APIs that require
in-kernel BTF type with ability to specify BTF types from kernel modules, not
just vmlinux BTF. This will be implemented in a follow up patch set for
fentry/fexit/fmod_ret/lsm/etc.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201110011932.3201430-3-andrii@kernel.org
The currently available bpf_get_current_task returns an unsigned integer
which can be used along with BPF_CORE_READ to read data from
the task_struct but still cannot be used as an input argument to a
helper that accepts an ARG_PTR_TO_BTF_ID of type task_struct.
In order to implement this helper a new return type, RET_PTR_TO_BTF_ID,
is added. This is similar to RET_PTR_TO_BTF_ID_OR_NULL but does not
require checking the nullness of returned pointer.
Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20201106103747.2780972-6-kpsingh@chromium.org
Similar to bpf_local_storage for sockets and inodes add local storage
for task_struct.
The life-cycle of storage is managed with the life-cycle of the
task_struct. i.e. the storage is destroyed along with the owning task
with a callback to the bpf_task_storage_free from the task_free LSM
hook.
The BPF LSM allocates an __rcu pointer to the bpf_local_storage in
the security blob which are now stackable and can co-exist with other
LSMs.
The userspace map operations can be done by using a pid fd as a key
passed to the lookup, update and delete operations.
Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20201106103747.2780972-3-kpsingh@chromium.org
test_progs's IMA selftests requires extra Kconfig values, so update
latest.config to accommodate those.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Move gory details about libbpf mirror and sync into a
separate section at the bottom of README.
Also add references to libbpf-bootstrap and blog about it,
as well as libbpf-tools reference.
tcpbpf_user uses cgroup bpf_link, not available in 5.5. hash_large_key is
testing a more permissive verifier check, implemented in 5.11. So blacklist
both.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
If bits is 0, the case when the map is empty, then the >> is the size of
the register which is undefined behavior - on x86 it is the same as a
shift by 0.
Fix by handling the 0 case explicitly and guarding calls to hash_bits for
empty maps in hashmap__for_each_key_entry and hashmap__for_each_entry_safe.
Fixes: e3b924224028 ("libbpf: add resizable non-thread safe internal hashmap")
Suggested-by: Andrii Nakryiko <andriin@fb.com>,
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201029223707.494059-1-irogers@google.com
In some cases compiler seems to generate distinct DWARF types for identical
arrays within the same CU. That seems like a bug, but it's already out there
and breaks type graph equivalence checks, so accommodate it anyway by checking
for identical arrays, regardless of their type ID.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201105043402.2530976-10-andrii@kernel.org
Add support for deduplication split BTFs. When deduplicating split BTF, base
BTF is considered to be immutable and can't be modified or adjusted. 99% of
BTF deduplication logic is left intact (module some type numbering adjustments).
There are only two differences.
First, each type in base BTF gets hashed (expect VAR and DATASEC, of course,
those are always considered to be self-canonical instances) and added into
a table of canonical table candidates. Hashing is a shallow, fast operation,
so mostly eliminates the overhead of having entire base BTF to be a part of
BTF dedup.
Second difference is very critical and subtle. While deduplicating split BTF
types, it is possible to discover that one of immutable base BTF BTF_KIND_FWD
types can and should be resolved to a full STRUCT/UNION type from the split
BTF part. This is, obviously, can't happen because we can't modify the base
BTF types anymore. So because of that, any type in split BTF that directly or
indirectly references that newly-to-be-resolved FWD type can't be considered
to be equivalent to the corresponding canonical types in base BTF, because
that would result in a loss of type resolution information. So in such case,
split BTF types will be deduplicated separately and will cause some
duplication of type information, which is unavoidable.
With those two changes, the rest of the algorithm manages to deduplicate split
BTF correctly, pointing all the duplicates to their canonical counter-parts in
base BTF, but also is deduplicating whatever unique types are present in split
BTF on their own.
Also, theoretically, split BTF after deduplication could end up with either
empty type section or empty string section. This is handled by libbpf
correctly in one of previous patches in the series.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201105043402.2530976-9-andrii@kernel.org
Make data section layout checks stricter, disallowing overlap of types and
strings data.
Additionally, allow BTFs with no type data. There is nothing inherently wrong
with having BTF with no types (put potentially with some strings). This could
be a situation with kernel module BTFs, if module doesn't introduce any new
type information.
Also fix invalid offset alignment check for btf->hdr->type_off.
Fixes: 8a138aed4a80 ("bpf: btf: Add BTF support to libbpf")
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20201105043402.2530976-8-andrii@kernel.org
Support split BTF operation, in which one BTF (base BTF) provides basic set of
types and strings, while another one (split BTF) builds on top of base's types
and strings and adds its own new types and strings. From API standpoint, the
fact that the split BTF is built on top of the base BTF is transparent.
Type numeration is transparent. If the base BTF had last type ID #N, then all
types in the split BTF start at type ID N+1. Any type in split BTF can
reference base BTF types, but not vice versa. Programmatically construction of
a split BTF on top of a base BTF is supported: one can create an empty split
BTF with btf__new_empty_split() and pass base BTF as an input, or pass raw
binary data to btf__new_split(), or use btf__parse_xxx_split() variants to get
initial set of split types/strings from the ELF file with .BTF section.
String offsets are similarly transparent and are a logical continuation of
base BTF's strings. When building BTF programmatically and adding a new string
(explicitly with btf__add_str() or implicitly through appending new
types/members), string-to-be-added would first be looked up from the base
BTF's string section and re-used if it's there. If not, it will be looked up
and/or added to the split BTF string section. Similarly to type IDs, types in
split BTF can refer to strings from base BTF absolutely transparently (but not
vice versa, of course, because base BTF doesn't "know" about existence of
split BTF).
Internal type index is slightly adjusted to be zero-indexed, ignoring a fake
[0] VOID type. This allows to handle split/base BTF type lookups transparently
by using btf->start_id type ID offset, which is always 1 for base/non-split
BTF and equals btf__get_nr_types(base_btf) + 1 for the split BTF.
BTF deduplication is not yet supported for split BTF and support for it will
be added in separate patch.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201105043402.2530976-5-andrii@kernel.org
Revamp BTF dedup's string deduplication to match the approach of writable BTF
string management. This allows to transfer deduplicated strings index back to
BTF object after deduplication without expensive extra memory copying and hash
map re-construction. It also simplifies the code and speeds it up, because
hashmap-based string deduplication is faster than sort + unique approach.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20201105043402.2530976-4-andrii@kernel.org
This avoids compilation warning if `struct bpf_redir_neigh` is not provided by
other kernel headers.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Yaniv reported a compilation error after pulling latest libbpf:
[...]
../libbpf/src/root/usr/include/bpf/bpf_helpers.h:99:10: error:
unknown register name 'r0' in asm
: "r0", "r1", "r2", "r3", "r4", "r5");
[...]
The issue got triggered given Yaniv was compiling tracing programs with native
target (e.g. x86) instead of BPF target, hence no BTF generated vmlinux.h nor
CO-RE used, and later llc with -march=bpf was invoked to compile from LLVM IR
to BPF object file. Given that clang was expecting x86 inline asm and not BPF
one the error complained that these regs don't exist on the former.
Guard bpf_tail_call_static() with defined(__bpf__) where BPF inline asm is valid
to use. BPF tracing programs on more modern kernels use BPF target anyway and
thus the bpf_tail_call_static() function will be available for them. BPF inline
asm is supported since clang 7 (clang <= 6 otherwise throws same above error),
and __bpf_unreachable() since clang 8, therefore include the latter condition
in order to prevent compilation errors for older clang versions. Given even an
old Ubuntu 18.04 LTS has official LLVM packages all the way up to llvm-10, I did
not bother to special case the __bpf_unreachable() inside bpf_tail_call_static()
further.
Also, undo the sockex3_kern's use of bpf_tail_call_static() sample given they
still have the old hacky way to even compile networking progs with native instead
of BPF target so bpf_tail_call_static() won't be defined there anymore.
Fixes: 0e9f6841f664 ("bpf, libbpf: Add bpf_tail_call_static helper for bpf programs")
Reported-by: Yaniv Agman <yanivagman@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Tested-by: Yaniv Agman <yanivagman@gmail.com>
Link: https://lore.kernel.org/bpf/CAMy7=ZUk08w5Gc2Z-EKi4JFtuUCaZYmE4yzhJjrExXpYKR4L8w@mail.gmail.com
Link: https://lore.kernel.org/bpf/20201021203257.26223-1-daniel@iogearbox.net
Recent work in f4d05259213f ("bpf: Add map_meta_equal map ops") and 134fede4eecf
("bpf: Relax max_entries check for most of the inner map types") added support
for dynamic inner max elements for most map-in-map types. Exceptions were maps
like array or prog array where the map_gen_lookup() callback uses the maps'
max_entries field as a constant when emitting instructions.
We recently implemented Maglev consistent hashing into Cilium's load balancer
which uses map-in-map with an outer map being hash and inner being array holding
the Maglev backend table for each service. This has been designed this way in
order to reduce overall memory consumption given the outer hash map allows to
avoid preallocating a large, flat memory area for all services. Also, the
number of service mappings is not always known a-priori.
The use case for dynamic inner array map entries is to further reduce memory
overhead, for example, some services might just have a small number of back
ends while others could have a large number. Right now the Maglev backend table
for small and large number of backends would need to have the same inner array
map entries which adds a lot of unneeded overhead.
Dynamic inner array map entries can be realized by avoiding the inlined code
generation for their lookup. The lookup will still be efficient since it will
be calling into array_map_lookup_elem() directly and thus avoiding retpoline.
The patch adds a BPF_F_INNER_MAP flag to map creation which therefore skips
inline code generation and relaxes array_map_meta_equal() check to ignore both
maps' max_entries. This also still allows to have faster lookups for map-in-map
when BPF_F_INNER_MAP is not specified and hence dynamic max_entries not needed.
Example code generation where inner map is dynamic sized array:
# bpftool p d x i 125
int handle__sys_enter(void * ctx):
; int handle__sys_enter(void *ctx)
0: (b4) w1 = 0
; int key = 0;
1: (63) *(u32 *)(r10 -4) = r1
2: (bf) r2 = r10
;
3: (07) r2 += -4
; inner_map = bpf_map_lookup_elem(&outer_arr_dyn, &key);
4: (18) r1 = map[id:468]
6: (07) r1 += 272
7: (61) r0 = *(u32 *)(r2 +0)
8: (35) if r0 >= 0x3 goto pc+5
9: (67) r0 <<= 3
10: (0f) r0 += r1
11: (79) r0 = *(u64 *)(r0 +0)
12: (15) if r0 == 0x0 goto pc+1
13: (05) goto pc+1
14: (b7) r0 = 0
15: (b4) w6 = -1
; if (!inner_map)
16: (15) if r0 == 0x0 goto pc+6
17: (bf) r2 = r10
;
18: (07) r2 += -4
; val = bpf_map_lookup_elem(inner_map, &key);
19: (bf) r1 = r0 | No inlining but instead
20: (85) call array_map_lookup_elem#149280 | call to array_map_lookup_elem()
; return val ? *val : -1; | for inner array lookup.
21: (15) if r0 == 0x0 goto pc+1
; return val ? *val : -1;
22: (61) r6 = *(u32 *)(r0 +0)
; }
23: (bc) w0 = w6
24: (95) exit
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20201010234006.7075-4-daniel@iogearbox.net
Add an efficient ingress to ingress netns switch that can be used out of tc BPF
programs in order to redirect traffic from host ns ingress into a container
veth device ingress without having to go via CPU backlog queue [0]. For local
containers this can also be utilized and path via CPU backlog queue only needs
to be taken once, not twice. On a high level this borrows from ipvlan which does
similar switch in __netif_receive_skb_core() and then iterates via another_round.
This helps to reduce latency for mentioned use cases.
Pod to remote pod with redirect(), TCP_RR [1]:
# percpu_netperf 10.217.1.33
RT_LATENCY: 122.450 (per CPU: 122.666 122.401 122.333 122.401 )
MEAN_LATENCY: 121.210 (per CPU: 121.100 121.260 121.320 121.160 )
STDDEV_LATENCY: 120.040 (per CPU: 119.420 119.910 125.460 115.370 )
MIN_LATENCY: 46.500 (per CPU: 47.000 47.000 47.000 45.000 )
P50_LATENCY: 118.500 (per CPU: 118.000 119.000 118.000 119.000 )
P90_LATENCY: 127.500 (per CPU: 127.000 128.000 127.000 128.000 )
P99_LATENCY: 130.750 (per CPU: 131.000 131.000 129.000 132.000 )
TRANSACTION_RATE: 32666.400 (per CPU: 8152.200 8169.842 8174.439 8169.897 )
Pod to remote pod with redirect_peer(), TCP_RR:
# percpu_netperf 10.217.1.33
RT_LATENCY: 44.449 (per CPU: 43.767 43.127 45.279 45.622 )
MEAN_LATENCY: 45.065 (per CPU: 44.030 45.530 45.190 45.510 )
STDDEV_LATENCY: 84.823 (per CPU: 66.770 97.290 84.380 90.850 )
MIN_LATENCY: 33.500 (per CPU: 33.000 33.000 34.000 34.000 )
P50_LATENCY: 43.250 (per CPU: 43.000 43.000 43.000 44.000 )
P90_LATENCY: 46.750 (per CPU: 46.000 47.000 47.000 47.000 )
P99_LATENCY: 52.750 (per CPU: 51.000 54.000 53.000 53.000 )
TRANSACTION_RATE: 90039.500 (per CPU: 22848.186 23187.089 22085.077 21919.130 )
[0] https://linuxplumbersconf.org/event/7/contributions/674/attachments/568/1002/plumbers_2020_cilium_load_balancer.pdf
[1] https://github.com/borkmann/netperf_scripts/blob/master/percpu_netperf
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20201010234006.7075-3-daniel@iogearbox.net
Add support for patching instructions of the following form:
- rX = *(T *)(rY + <off>);
- *(T *)(rX + <off>) = rY;
- *(T *)(rX + <off>) = <imm>, where T is one of {u8, u16, u32, u64}.
For such instructions, if the actual kernel field recorded in CO-RE relocation
has a different size than the one recorded locally (e.g., from vmlinux.h),
then libbpf will adjust T to an appropriate 1-, 2-, 4-, or 8-byte loads.
In general, such transformation is not always correct and could lead to
invalid final value being loaded or stored. But two classes of cases are
always safe:
- if both local and target (kernel) types are unsigned integers, but of
different sizes, then it's OK to adjust load/store instruction according to
the necessary memory size. Zero-extending nature of such instructions and
unsignedness make sure that the final value is always correct;
- pointer size mismatch between BPF target architecture (which is always
64-bit) and 32-bit host kernel architecture can be similarly resolved
automatically, because pointer is essentially an unsigned integer. Loading
32-bit pointer into 64-bit BPF register with zero extension will leave
correct pointer in the register.
Both cases are necessary to support CO-RE on 32-bit kernels, as `unsigned
long` in vmlinux.h generated from 32-bit kernel is 32-bit, but when compiled
with BPF program for BPF target it will be treated by compiler as 64-bit
integer. Similarly, pointers in vmlinux.h are 32-bit for kernel, but treated
as 64-bit values by compiler for BPF target. Both problems are now resolved by
libbpf for direct memory reads.
But similar transformations are useful in general when kernel fields are
"resized" from, e.g., unsigned int to unsigned long (or vice versa).
Now, similar transformations for signed integers are not safe to perform as
they will result in incorrect sign extension of the value. If such situation
is detected, libbpf will emit helpful message and will poison the instruction.
Not failing immediately means that it's possible to guard the instruction
based on kernel version (or other conditions) and make sure it's not
reachable.
If there is a need to read signed integers that change sizes between different
kernels, it's possible to use BPF_CORE_READ_BITFIELD() macro, which works both
with bitfields and non-bitfield integers of any signedness and handles
sign-extension properly. Also, bpf_core_read() with proper size and/or use of
bpf_core_field_size() relocation could allow to deal with such complicated
situations explicitly, if not so conventiently as direct memory reads.
Selftests added in a separate patch in progs/test_core_autosize.c demonstrate
both direct memory and probed use cases.
BPF_CORE_READ() is not changed and it won't deal with such situations as
automatically as direct memory reads due to the signedness integer
limitations, which are much harder to detect and control with compiler macro
magic. So it's encouraged to utilize direct memory reads as much as possible.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20201008001025.292064-3-andrii@kernel.org
Bypass CO-RE relocations step for BPF programs that are not going to be
loaded. This allows to have BPF programs compiled in and disabled dynamically
if kernel is not supposed to provide enough relocation information. In such
case, there won't be unnecessary warnings about failed relocations.
Fixes: d929758101fc ("libbpf: Support disabling auto-loading BPF programs")
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20201008001025.292064-2-andrii@kernel.org
Fix a compatibility problem when the old XDP_SHARED_UMEM mode is used
together with the xsk_socket__create() call. In the old XDP_SHARED_UMEM
mode, only sharing of the same device and queue id was allowed, and
in this mode, the fill ring and completion ring were shared between
the AF_XDP sockets.
Therefore, it was perfectly fine to call the xsk_socket__create() API
for each socket and not use the new xsk_socket__create_shared() API.
This behavior was ruined by the commit introducing XDP_SHARED_UMEM
support between different devices and/or queue ids. This patch restores
the ability to use xsk_socket__create in these circumstances so that
backward compatibility is not broken.
Fixes: 2f6324a3937f ("libbpf: Support shared umems between queues and devices")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1602070946-11154-1-git-send-email-magnus.karlsson@gmail.com
