Magnus Karlsson [Mon, 21 Oct 2019 08:57:04 +0000 (10:57 +0200)]
xsk: Improve documentation for AF_XDP
Added sections on all the bind flags, libbpf, all the setsockopts and
all the getsockopts. Also updated the document to reflect the latest
features and to correct some spelling errors.
v1 -> v2:
* Updated XDP program with latest BTF map format
* Added one more FAQ entry
* Some minor edits and corrections
v2 -> v3:
* Simplified XDP_SHARED_UMEM example XDP program
Björn Töpel [Tue, 22 Oct 2019 07:22:06 +0000 (09:22 +0200)]
libbpf: Use implicit XSKMAP lookup from AF_XDP XDP program
In commit 43e74c0267a3 ("bpf_xdp_redirect_map: Perform map lookup in
eBPF helper") the bpf_redirect_map() helper learned to do map lookup,
which means that the explicit lookup in the XDP program for AF_XDP is
not needed for post-5.3 kernels.
This commit adds the implicit map lookup with default action, which
improves the performance for the "rx_drop" [1] scenario with ~4%.
For pre-5.3 kernels, the bpf_redirect_map() returns XDP_ABORTED, and a
fallback path for backward compatibility is entered, where explicit
lookup is still performed. This means a slight regression for older
kernels (an additional bpf_redirect_map() call), but I consider that a
fair punishment for users not upgrading their kernels. ;-)
v1->v2: Backward compatibility (Toke) [2]
v2->v3: Avoid masking/zero-extension by using JMP32 [3]
xdp: Fix type of string pointer in __XDP_ACT_SYM_TAB
The table entry in __XDP_ACT_SYM_TAB for the last item is set
to { -1, 0 } where it should be { -1, NULL } as the second item
is a pointer to a string.
Fixes the following sparse warnings:
./include/trace/events/xdp.h:28:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:53:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:82:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:140:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:155:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:190:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:225:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:260:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:318:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:356:1: warning: Using plain integer as NULL pointer
./include/trace/events/xdp.h:390:1: warning: Using plain integer as NULL pointer
Andrii Nakryiko [Tue, 22 Oct 2019 17:21:00 +0000 (10:21 -0700)]
libbpf: Make DECLARE_LIBBPF_OPTS macro strictly a variable declaration
LIBBPF_OPTS is implemented as a mix of field declaration and memset
+ assignment. This makes it neither variable declaration nor purely
statements, which is a problem, because you can't mix it with either
other variable declarations nor other function statements, because C90
compiler mode emits warning on mixing all that together.
This patch changes LIBBPF_OPTS into a strictly declaration of variable
and solves this problem, as can be seen in case of bpftool, which
previously would emit compiler warning, if done this way (LIBBPF_OPTS as
part of function variables declaration block).
This patch also renames LIBBPF_OPTS into DECLARE_LIBBPF_OPTS to follow
kernel convention for similar macros more closely.
v1->v2:
- rename LIBBPF_OPTS into DECLARE_LIBBPF_OPTS (Jakub Sitnicki).
Experiments showed that in general performance is better with alu32
enabled:
https://lwn.net/Articles/775316/
This patch turns on alu32 with no-flavor test_progs which is tested
most often. The flavor test at no_alu32/test_progs can be used to
test without alu32 enabled. The Makefile check for whether LLVM
supports '-mattr=+alu32 -mcpu=v3' is removed as LLVM7 should be
available for recent distributions and also latest LLVM is preferred
to run BPF selftests.
Note that jmp32 is checked by -mcpu=probe and will be enabled if the
host kernel supports it.
Daniel Borkmann [Mon, 21 Oct 2019 12:49:12 +0000 (14:49 +0200)]
Merge branch 'bpf-libbpf-cleanups'
Andrii Nakryiko says:
====================
This patch set's main goal is to teach bpf_object__open() (and its variants)
to automatically derive BPF program type/expected attach type from section
names, similarly to how bpf_prog_load() was doing it. This significantly
improves user experience by eliminating yet another
obvious-only-in-the-hindsight surprise, when using libbpf APIs.
There are a bunch of auxiliary clean-ups and improvements. E.g.,
bpf_program__get_type() and bpf_program__get_expected_attach_type() are added
for completeness and symmetry with corresponding setter APIs. Some clean up
and fixes in selftests/bpf are done as well.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Andrii Nakryiko [Mon, 21 Oct 2019 03:39:01 +0000 (20:39 -0700)]
selftests/bpf: Make reference_tracking test use subtests
reference_tracking is actually a set of 9 sub-tests. Make it explicitly so.
Also, add explicit "classifier/" prefix to BPF program section names to
let libbpf correctly guess program type. Thus, also remove explicit
bpf_prog__set_type() call.
Andrii Nakryiko [Mon, 21 Oct 2019 03:39:00 +0000 (20:39 -0700)]
selftests/bpf: Make a copy of subtest name
test_progs never created a copy of subtest name, rather just stored
pointer to whatever string test provided. This is bad as that string
might be freed or modified by the end of subtest. Fix this by creating
a copy of given subtest name when subtest starts.
Andrii Nakryiko [Mon, 21 Oct 2019 03:38:59 +0000 (20:38 -0700)]
libbpf: Teach bpf_object__open to guess program types
Teach bpf_object__open how to guess program type and expected attach
type from section names, similar to what bpf_prog_load() does. This
seems like a really useful features and an oversight to not have this
done during bpf_object_open(). To preserver backwards compatible
behavior of bpf_prog_load(), its attr->prog_type is treated as an
override of bpf_object__open() decisions, if attr->prog_type is not
UNSPECIFIED.
There is a slight difference in behavior for bpf_prog_load().
Previously, if bpf_prog_load() was loading BPF object with more than one
program, first program's guessed program type and expected attach type
would determine corresponding attributes of all the subsequent program
types, even if their sections names suggest otherwise. That seems like
a rather dubious behavior and with this change it will behave more
sanely: each program's type is determined individually, unless they are
forced to uniformity through attr->prog_type.
Kefeng Wang [Mon, 21 Oct 2019 05:55:32 +0000 (13:55 +0800)]
tools, bpf: Rename pr_warning to pr_warn to align with kernel logging
For kernel logging macros, pr_warning() is completely removed and
replaced by pr_warn(). By using pr_warn() in tools/lib/bpf/ for
symmetry to kernel logging macros, we could eventually drop the
use of pr_warning() in the whole kernel tree.
Jakub Sitnicki [Sun, 20 Oct 2019 11:23:44 +0000 (13:23 +0200)]
scripts/bpf: Print an error when known types list needs updating
Don't generate a broken bpf_helper_defs.h header if the helper script needs
updating because it doesn't recognize a newly added type. Instead print an
error that explains why the build is failing, clean up the partially
generated header and stop.
v1->v2:
- Switched from temporary file to .DELETE_ON_ERROR.
Fixes: 456a513bb5d4 ("scripts/bpf: Emit an #error directive known types list needs updating") Suggested-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Yonghong Song <yhs@fb.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/20191020112344.19395-1-jakub@cloudflare.com
John Fastabend [Fri, 18 Oct 2019 14:41:26 +0000 (07:41 -0700)]
bpf, libbpf: Add kernel version section parsing back
With commit "libbpf: stop enforcing kern_version,..." we removed the
kernel version section parsing in favor of querying for the kernel
using uname() and populating the version using the result of the
query. After this any version sections were simply ignored.
Unfortunately, the world of kernels is not so friendly. I've found some
customized kernels where uname() does not match the in kernel version.
To fix this so programs can load in this environment this patch adds
back parsing the section and if it exists uses the user specified
kernel version to override the uname() result. However, keep most the
kernel uname() discovery bits so users are not required to insert the
version except in these odd cases.
====================
This patch set extensively revamps selftests/bpf's Makefile to generalize test
runner concept and apply it uniformly to test_maps and test_progs test
runners, along with test_progs' few build "flavors", exercising various ways
to build BPF programs.
As we do that, we fix dependencies between various phases of test runners, and
simplify some one-off rules and dependencies currently present in Makefile.
test_progs' flavors are now built into root $(OUTPUT) directory and can be run
without any extra steps right from there. E.g., test_progs-alu32 is built and
is supposed to be run from $(OUTPUT). It will cd into alu32/ subdirectory to
load correct set of BPF object files (which are different from the ones built
for test_progs).
Outline:
- patch #1 teaches test_progs about flavor sub-directories;
- patch #2 fixes one of CO-RE tests to not depend strictly on process name;
- patch #3 changes test_maps's usage of map_tests/tests.h to be the same as
test_progs' one;
- patch #4 adds convenient short `make test_progs`-like targets to build only
individual tests, if necessary;
- patch #5 is a main patch in the series; it uses a bunch of make magic
(mainly $(call) and $(eval)) to define test runner "skeleton" and apply it
to 4 different test runners, lots more details in corresponding commit
description;
- patch #6 does a bit of post-clean up for test_queue_map and test_stack_map
BPF programs;
- patch #7 cleans up test_libbpf.sh/test_libbpf_open superseded by test_progs.
v3->v4:
- remove accidentally checked in binaries;
v2->v3:
- drop test_xdp.o mixed compilation mode, remove test_libbpf.sh (Alexei);
v1->v2:
- drop test_progs-native causing compilation failures due to
__builtin_preserve_field_access, add back test_xdp.o override, which will
now emit rule re-definition warning.
====================
Andrii Nakryiko [Wed, 16 Oct 2019 06:00:50 +0000 (23:00 -0700)]
selftests/bpf: Move test_queue_stack_map.h into progs/ where it belongs
test_queue_stack_map.h is used only from BPF programs. Thus it should be
part of progs/ subdir. An added benefit of moving it there is that new
TEST_RUNNER_DEFINE_RULES macro-rule will properly capture dependency on
this header for all BPF objects and trigger re-build, if it changes.
Andrii Nakryiko [Wed, 16 Oct 2019 06:00:49 +0000 (23:00 -0700)]
selftests/bpf: Replace test_progs and test_maps w/ general rule
Define test runner generation meta-rule that codifies dependencies
between test runner, its tests, and its dependent BPF programs. Use that
for defining test_progs and test_maps test-runners. Also additionally define
2 flavors of test_progs:
- alu32, which builds BPF programs with 32-bit registers codegen;
- bpf_gcc, which build BPF programs using GCC, if it supports BPF target.
Overall, this is accomplished through $(eval)'ing a set of generic
rules, which defines Makefile targets dynamically at runtime. See
comments explaining the need for 2 $(evals), though.
For each test runner we have (test_maps and test_progs, currently), and,
optionally, their flavors, the logic of build process is modeled as
follows (using test_progs as an example):
- all BPF objects are in progs/:
- BPF object's .o file is built into output directory from
corresponding progs/.c file;
- all BPF objects in progs/*.c depend on all progs/*.h headers;
- all BPF objects depend on bpf_*.h helpers from libbpf (but not
libbpf archive). There is an extra rule to trigger bpf_helper_defs.h
(re-)build, if it's not present/outdated);
- build recipe for BPF object can be re-defined per test runner/flavor;
- test files are built from prog_tests/*.c:
- all such test file objects are built on individual file basis;
- currently, every single test file depends on all BPF object files;
this might be improved in follow up patches to do 1-to-1 dependency,
but allowing to customize this per each individual test;
- each test runner definition can specify a list of extra .c and .h
files to be built along test files and test runner binary; all such
headers are becoming automatic dependency of each test .c file;
- due to test files sometimes embedding (using .incbin assembly
directive) contents of some BPF objects at compilation time, which are
expected to be in CWD of compiler, compilation for test file object does
cd into test runner's output directory; to support this mode all the
include paths are turned into absolute paths using $(abspath) make
function;
- prog_tests/test.h is automatically (re-)generated with an entry for
each .c file in prog_tests/;
- final test runner binary is linked together from test object files and
extra object files, linking together libbpf's archive as well;
- it's possible to specify extra "resource" files/targets, which will be
copied into test runner output directory, if it differes from
Makefile-wide $(OUTPUT). This is used to ensure btf_dump test cases and
urandom_read binary is put into a test runner's CWD for tests to find
them in runtime.
For flavored test runners, their output directory is a subdirectory of
common Makefile-wide $(OUTPUT) directory with flavor name used as
subdirectory name.
BPF objects targets might be reused between different test runners, so
extra checks are employed to not double-define them. Similarly, we have
redefinition guards for output directories and test headers.
test_verifier follows slightly different patterns and is simple enough
to not justify generalizing TEST_RUNNER_DEFINE/TEST_RUNNER_DEFINE_RULES
further to accomodate these differences. Instead, rules for
test_verifier are minimized and simplified, while preserving correctness
of dependencies.
Andrii Nakryiko [Wed, 16 Oct 2019 06:00:48 +0000 (23:00 -0700)]
selftests/bpf: Add simple per-test targets to Makefile
Currently it's impossible to do `make test_progs` and have only
test_progs be built, because all the binary targets are defined in terms
of $(OUTPUT)/<binary>, and $(OUTPUT) is absolute path to current
directory (or whatever gets overridden to by user).
This patch adds simple re-directing targets for all test targets making
it possible to do simple and nice `make test_progs` (and any other
target).
Andrii Nakryiko [Wed, 16 Oct 2019 06:00:46 +0000 (23:00 -0700)]
selftests/bpf: Make CO-RE reloc test impartial to test_progs flavor
test_core_reloc_kernel test captures its own process name and validates
it as part of the test. Given extra "flavors" of test_progs, this break
for anything by default test_progs binary. Fix the test to cut out
flavor part of the process name.
Andrii Nakryiko [Wed, 16 Oct 2019 06:00:45 +0000 (23:00 -0700)]
selftests/bpf: Teach test_progs to cd into subdir
We are building a bunch of "flavors" of test_progs, e.g., w/ alu32 flag
for Clang when building BPF object. test_progs setup is relying on
having all the BPF object files and extra resources to be available in
current working directory, though. But we actually build all these files
into a separate sub-directory. Next set of patches establishes
convention of naming "flavored" test_progs (and test runner binaries in
general) as test_progs-flavor (e.g., test_progs-alu32), for each such
extra flavor. This patch teaches test_progs binary to automatically
detect its own extra flavor based on its argv[0], and if present, to
change current directory to a flavor-specific subdirectory.
Jakub Sitnicki [Thu, 17 Oct 2019 08:37:52 +0000 (10:37 +0200)]
selftests/bpf: Restore the netns after flow dissector reattach test
flow_dissector_reattach test changes the netns we run in but does not
restore it to the one we started in when finished. This interferes with
tests that run after it. Fix it by restoring the netns when done.
Fixes: f97eea1756f3 ("selftests/bpf: Check that flow dissector can be re-attached") Reported-by: Alexei Starovoitov <ast@kernel.org> Reported-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191017083752.30999-1-jakub@cloudflare.com
Daniel Borkmann [Thu, 17 Oct 2019 14:44:36 +0000 (16:44 +0200)]
Merge branch 'bpf-btf-trace'
Alexei Starovoitov says:
====================
v2->v3:
- while trying to adopt btf-based tracing in production service realized
that disabling bpf_probe_read() was premature. The real tracing program
needs to see much more than this type safe tracking can provide.
With these patches the verifier will be able to see that skb->data
is a pointer to 'u8 *', but it cannot possibly know how many bytes
of it is readable. Hence bpf_probe_read() is necessary to do basic
packet reading from tracing program. Some helper can be introduced to
solve this particular problem, but there are other similar structures.
Another issue is bitfield reading. The support for bitfields
is coming to llvm. libbpf will be supporting it eventually as well,
but there will be corner cases where bpf_probe_read() is necessary.
The long term goal is still the same: get rid of probe_read eventually.
- fixed build issue with clang reported by Nathan Chancellor.
- addressed a ton of comments from Andrii.
bitfields and arrays are explicitly unsupported in btf-based tracking.
This will be improved in the future.
Right now the verifier is more strict than necessary.
In some cases it can fall back to 'scalar' instead of rejecting
the program, but rejection today allows to make better decisions
in the future.
- adjusted testcase to demo bitfield and skb->data reading.
v1->v2:
- addressed feedback from Andrii and Eric. Thanks a lot for review!
- added missing check at raw_tp attach time.
- Andrii noticed that expected_attach_type cannot be reused.
Had to introduce new field to bpf_attr.
- cleaned up logging nicely by introducing bpf_log() helper.
- rebased.
Revolutionize bpf tracing and bpf C programming.
C language allows any pointer to be typecasted to any other pointer
or convert integer to a pointer.
Though bpf verifier is operating at assembly level it has strict type
checking for fixed number of types.
Known types are defined in 'enum bpf_reg_type'.
For example:
PTR_TO_FLOW_KEYS is a pointer to 'struct bpf_flow_keys'
PTR_TO_SOCKET is a pointer to 'struct bpf_sock',
and so on.
When it comes to bpf tracing there are no types to track.
bpf+kprobe receives 'struct pt_regs' as input.
bpf+raw_tracepoint receives raw kernel arguments as an array of u64 values.
It was up to bpf program to interpret these integers.
Typical tracing program looks like:
int bpf_prog(struct pt_regs *ctx)
{
struct net_device *dev;
struct sk_buff *skb;
int ifindex;
skb = (struct sk_buff *) ctx->di;
bpf_probe_read(&dev, sizeof(dev), &skb->dev);
bpf_probe_read(&ifindex, sizeof(ifindex), &dev->ifindex);
}
Addressing mistakes will not be caught by C compiler or by the verifier.
The program above could have typecasted ctx->si to skb and page faulted
on every bpf_probe_read().
bpf_probe_read() allows reading any address and suppresses page faults.
Typical program has hundreds of bpf_probe_read() calls to walk
kernel data structures.
Not only tracing program would be slow, but there was always a risk
that bpf_probe_read() would read mmio region of memory and cause
unpredictable hw behavior.
With introduction of Compile Once Run Everywhere technology in libbpf
and in LLVM and BPF Type Format (BTF) the verifier is finally ready
for the next step in program verification.
Now it can use in-kernel BTF to type check bpf assembly code.
Equivalent program will look like:
struct trace_kfree_skb {
struct sk_buff *skb;
void *location;
};
SEC("raw_tracepoint/kfree_skb")
int trace_kfree_skb(struct trace_kfree_skb* ctx)
{
struct sk_buff *skb = ctx->skb;
struct net_device *dev;
int ifindex;
__builtin_preserve_access_index(({
dev = skb->dev;
ifindex = dev->ifindex;
}));
}
These patches teach bpf verifier to recognize kfree_skb's first argument
as 'struct sk_buff *' because this is what kernel C code is doing.
The bpf program cannot 'cheat' and say that the first argument
to kfree_skb raw_tracepoint is some other type.
The verifier will catch such type mismatch between bpf program
assumption of kernel code and the actual type in the kernel.
Furthermore skb->dev access is type tracked as well.
The verifier can see which field of skb is being read
in bpf assembly. It will match offset to type.
If bpf program has code:
struct net_device *dev = (void *)skb->len;
C compiler will not complain and generate bpf assembly code,
but the verifier will recognize that integer 'len' field
is being accessed at offsetof(struct sk_buff, len) and will reject
further dereference of 'dev' variable because it contains
integer value instead of a pointer.
Such sophisticated type tracking allows calling networking
bpf helpers from tracing programs.
This patchset allows calling bpf_skb_event_output() that dumps
skb data into perf ring buffer.
It greatly improves observability.
Now users can not only see packet lenth of the skb
about to be freed in kfree_skb() kernel function, but can
dump it to user space via perf ring buffer using bpf helper
that was previously available only to TC and socket filters.
See patch 10 for full example.
The end result is safer and faster bpf tracing.
Safer - because type safe direct load can be used most of the time
instead of bpf_probe_read().
Faster - because direct loads are used to walk kernel data structures
instead of bpf_probe_read() calls.
Note that such loads can page fault and are supported by
hidden bpf_probe_read() in interpreter and via exception table
if program is JITed.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Load basic cls_bpf program.
Load raw_tracepoint program and attach to kfree_skb raw tracepoint.
Trigger cls_bpf via prog_test_run.
At the end of test_run kernel will call kfree_skb
which will trigger trace_kfree_skb tracepoint.
Which will call our raw_tracepoint program.
Which will take that skb and will dump it into perf ring buffer.
Check that user space received correct packet.
Introduce new helper that reuses existing skb perf_event output
implementation, but can be called from raw_tracepoint programs
that receive 'struct sk_buff *' as tracepoint argument or
can walk other kernel data structures to skb pointer.
In order to do that teach verifier to resolve true C types
of bpf helpers into in-kernel BTF ids.
The type of kernel pointer passed by raw tracepoint into bpf
program will be tracked by the verifier all the way until
it's passed into helper function.
For example:
kfree_skb() kernel function calls trace_kfree_skb(skb, loc);
bpf programs receives that skb pointer and may eventually
pass it into bpf_skb_output() bpf helper which in-kernel is
implemented via bpf_skb_event_output() kernel function.
Its first argument in the kernel is 'struct sk_buff *'.
The verifier makes sure that types match all the way.
Pointer to BTF object is a pointer to kernel object or NULL.
Such pointers can only be used by BPF_LDX instructions.
The verifier changed their opcode from LDX|MEM|size
to LDX|PROBE_MEM|size to make JITing easier.
The number of entries in extable is the number of BPF_LDX insns
that access kernel memory via "pointer to BTF type".
Only these load instructions can fault.
Since x86 extable is relative it has to be allocated in the same
memory region as JITed code.
Allocate it prior to last pass of JITing and let the last pass populate it.
Pointer to extable in bpf_prog_aux is necessary to make page fault
handling fast.
Page fault handling is done in two steps:
1. bpf_prog_kallsyms_find() finds BPF program that page faulted.
It's done by walking rb tree.
2. then extable for given bpf program is binary searched.
This process is similar to how page faulting is done for kernel modules.
The exception handler skips over faulting x86 instruction and
initializes destination register with zero. This mimics exact
behavior of bpf_probe_read (when probe_kernel_read faults dest is zeroed).
JITs for other architectures can add support in similar way.
Until then they will reject unknown opcode and fallback to interpreter.
Since extable should be aligned and placed near JITed code
make bpf_jit_binary_alloc() return 4 byte aligned image offset,
so that extable aligning formula in bpf_int_jit_compile() doesn't need
to rely on internal implementation of bpf_jit_binary_alloc().
On x86 gcc defaults to 16-byte alignment for regular kernel functions
due to better performance. JITed code may be aligned to 16 in the future,
but it will use 4 in the meantime.
Pointer to BTF object is a pointer to kernel object or NULL.
The memory access in the interpreter has to be done via probe_kernel_read
to avoid page faults.
BTF type id specified at program load time has all
necessary information to attach that program to raw tracepoint.
Use kernel type name to find raw tracepoint.
bpf: Implement accurate raw_tp context access via BTF
libbpf analyzes bpf C program, searches in-kernel BTF for given type name
and stores it into expected_attach_type.
The kernel verifier expects this btf_id to point to something like:
typedef void (*btf_trace_kfree_skb)(void *, struct sk_buff *skb, void *loc);
which represents signature of raw_tracepoint "kfree_skb".
Then btf_ctx_access() matches ctx+0 access in bpf program with 'skb'
and 'ctx+8' access with 'loc' arguments of "kfree_skb" tracepoint.
In first case it passes btf_id of 'struct sk_buff *' back to the verifier core
and 'void *' in second case.
Then the verifier tracks PTR_TO_BTF_ID as any other pointer type.
Like PTR_TO_SOCKET points to 'struct bpf_sock',
PTR_TO_TCP_SOCK points to 'struct bpf_tcp_sock', and so on.
PTR_TO_BTF_ID points to in-kernel structs.
If 1234 is btf_id of 'struct sk_buff' in vmlinux's BTF
then PTR_TO_BTF_ID#1234 points to one of in kernel skbs.
When PTR_TO_BTF_ID#1234 is dereferenced (like r2 = *(u64 *)r1 + 32)
the btf_struct_access() checks which field of 'struct sk_buff' is
at offset 32. Checks that size of access matches type definition
of the field and continues to track the dereferenced type.
If that field was a pointer to 'struct net_device' the r2's type
will be PTR_TO_BTF_ID#456. Where 456 is btf_id of 'struct net_device'
in vmlinux's BTF.
Such verifier analysis prevents "cheating" in BPF C program.
The program cannot cast arbitrary pointer to 'struct sk_buff *'
and access it. C compiler would allow type cast, of course,
but the verifier will notice type mismatch based on BPF assembly
and in-kernel BTF.
libbpf: Auto-detect btf_id of BTF-based raw_tracepoints
It's a responsiblity of bpf program author to annotate the program
with SEC("tp_btf/name") where "name" is a valid raw tracepoint.
The libbpf will try to find "name" in vmlinux BTF and error out
in case vmlinux BTF is not available or "name" is not found.
If "name" is indeed a valid raw tracepoint then in-kernel BTF
will have "btf_trace_##name" typedef that points to function
prototype of that raw tracepoint. BTF description captures
exact argument the kernel C code is passing into raw tracepoint.
The kernel verifier will check the types while loading bpf program.
libbpf keeps BTF type id in expected_attach_type, but since
kernel ignores this attribute for tracing programs copy it
into attach_btf_id attribute before loading.
Later the kernel will use prog->attach_btf_id to select raw tracepoint
during bpf_raw_tracepoint_open syscall command.
Add attach_btf_id attribute to prog_load command.
It's similar to existing expected_attach_type attribute which is
used in several cgroup based program types.
Unfortunately expected_attach_type is ignored for
tracing programs and cannot be reused for new purpose.
Hence introduce attach_btf_id to verify bpf programs against
given in-kernel BTF type id at load time.
It is strictly checked to be valid for raw_tp programs only.
In a later patches it will become:
btf_id == 0 semantics of existing raw_tp progs.
btd_id > 0 raw_tp with BTF and additional type safety.
If in-kernel BTF exists parse it and prepare 'struct btf *btf_vmlinux'
for further use by the verifier.
In-kernel BTF is trusted just like kallsyms and other build artifacts
embedded into vmlinux.
Yet run this BTF image through BTF verifier to make sure
that it is valid and it wasn't mangled during the build.
If in-kernel BTF is incorrect it means either gcc or pahole or kernel
are buggy. In such case disallow loading BPF programs.
bpf: Add typecast to bpf helpers to help BTF generation
When pahole converts dwarf to btf it emits only used types.
Wrap existing bpf helper functions into typedef and use it in
typecast to make gcc emits this type into dwarf.
Then pahole will convert it to btf.
The "btf_#name_of_helper" types will be used to figure out
types of arguments of bpf helpers.
The generated code before and after is the same.
Only dwarf and btf sections are different.
Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-3-ast@kernel.org
bpf: Add typecast to raw_tracepoints to help BTF generation
When pahole converts dwarf to btf it emits only used types.
Wrap existing __bpf_trace_##template() function into
btf_trace_##template typedef and use it in type cast to
make gcc emits this type into dwarf. Then pahole will convert it to btf.
The "btf_trace_" prefix will be used to identify BTF enabled raw tracepoints.
Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Andrii Nakryiko <andriin@fb.com> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Link: https://lore.kernel.org/bpf/20191016032505.2089704-2-ast@kernel.org
This can be reproduced by:
1. Start a multi-thread program that does parallel mmap() and malloc();
2. taskset the program to 2 CPUs;
3. Attach bpf program to trace_sched_switch and gather stackmap with
build-id, e.g. with trace.py from bcc tools:
trace.py -U -p <pid> -s <some-bin,some-lib> t:sched:sched_switch
A sample reproducer is attached at the end.
This could also trigger deadlock with other locks that are nested with
rq_lock.
Fix this by checking whether irqs are disabled. Since rq_lock and all
other nested locks are irq safe, it is safe to do up_read() when irqs are
not disable. If the irqs are disabled, postpone up_read() in irq_work.
Fixes: 615755a77b24 ("bpf: extend stackmap to save binary_build_id+offset instead of address") Signed-off-by: Song Liu <songliubraving@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/bpf/20191014171223.357174-1-songliubraving@fb.com
Reproducer:
============================ 8< ============================
char *filename;
void *worker(void *p)
{
void *ptr;
int fd;
char *pptr;
Jakub Sitnicki [Wed, 16 Oct 2019 08:58:11 +0000 (10:58 +0200)]
scripts/bpf: Emit an #error directive known types list needs updating
Make the compiler report a clear error when bpf_helpers_doc.py needs
updating rather than rely on the fact that Clang fails to compile
English:
../../../lib/bpf/bpf_helper_defs.h:2707:1: error: unknown type name 'Unrecognized'
Unrecognized type 'struct bpf_inet_lookup', please add it to known types!
It's useful for implementing EDT related tests (set tstamp, run the
test, see how the tstamp is changed or observe some other parameter).
Note that bpf_ktime_get_ns() helper is using monotonic clock, so for
the BPF programs that compare tstamp against it, tstamp should be
derived from clock_gettime(CLOCK_MONOTONIC, ...).
====================
This patch set generalizes libbpf's CO-RE relocation support. In addition to
existing field's byte offset relocation, libbpf now supports field existence
relocations, which are emitted by Clang when using
__builtin_preserve_field_info(<field>, BPF_FIELD_EXISTS). A convenience
bpf_core_field_exists() macro is added to bpf_core_read.h BPF-side header,
along the bpf_field_info_kind enum containing currently supported types of
field information libbpf supports. This list will grow as libbpf gains support
for other relo kinds.
This patch set upgrades the format of .BTF.ext's relocation record to match
latest Clang's format (12 -> 16 bytes). This is not a breaking change, as the
previous format hasn't been released yet as part of official Clang version
release.
v1->v2:
- unify bpf_field_info_kind enum and naming changes (Alexei);
- added bpf_core_field_exists() to bpf_core_read.h.
====================
Andrii Nakryiko [Tue, 15 Oct 2019 18:28:49 +0000 (11:28 -0700)]
selftests/bpf: Add field existence CO-RE relocs tests
Add a bunch of tests validating CO-RE is handling field existence
relocation. Relaxed CO-RE relocation mode is activated for these new
tests to prevent libbpf from rejecting BPF object for no-match
relocation, even though test BPF program is not going to use that
relocation, if field is missing.
Andrii Nakryiko [Tue, 15 Oct 2019 18:28:48 +0000 (11:28 -0700)]
libbpf: Add BPF-side definitions of supported field relocation kinds
Add enum definition for Clang's __builtin_preserve_field_info()
second argument (info_kind). Currently only byte offset and existence
are supported. Corresponding Clang changes introducing this built-in can
be found at [0]
Andrii Nakryiko [Tue, 15 Oct 2019 18:28:47 +0000 (11:28 -0700)]
libbpf: Add support for field existance CO-RE relocation
Add support for BPF_FRK_EXISTS relocation kind to detect existence of
captured field in a destination BTF, allowing conditional logic to
handle incompatible differences between kernels.
Also introduce opt-in relaxed CO-RE relocation handling option, which
makes libbpf emit warning for failed relocations, but proceed with other
relocations. Instruction, for which relocation failed, is patched with
(u32)-1 value.
Andrii Nakryiko [Tue, 15 Oct 2019 18:28:46 +0000 (11:28 -0700)]
libbpf: Refactor bpf_object__open APIs to use common opts
Refactor all the various bpf_object__open variations to ultimately
specify common bpf_object_open_opts struct. This makes it easy to keep
extending this common struct w/ extra parameters without having to
update all the legacy APIs.
Andrii Nakryiko [Tue, 15 Oct 2019 18:28:45 +0000 (11:28 -0700)]
libbpf: Update BTF reloc support to latest Clang format
BTF offset reloc was generalized in recent Clang into field relocation,
capturing extra u32 field, specifying what aspect of captured field
needs to be relocated. This changes .BTF.ext's record size for this
relocation from 12 bytes to 16 bytes. Given these format changes
happened in Clang before official released version, it's ok to not
support outdated 12-byte record size w/o breaking ABI.
Vladimir Oltean [Fri, 11 Oct 2019 22:31:15 +0000 (01:31 +0300)]
net: dsa: sja1105: Switch to scatter/gather API for SPI
This reworks the SPI transfer implementation to make use of more of the
SPI core features. The main benefit is to avoid the memcpy in
sja1105_xfer_buf().
The memcpy was only needed because the function was transferring a
single buffer at a time. So it needed to copy the caller-provided buffer
at buf + 4, to store the SPI message header in the "headroom" area.
But the SPI core supports scatter-gather messages, comprised of multiple
transfers. We can actually use those to break apart every SPI message
into 2 transfers: one for the header and one for the actual payload.
To keep the behavior the same regarding the chip select signal, it is
necessary to tell the SPI core to de-assert the chip select after each
chunk. This was not needed before, because each spi_message contained
only 1 single transfer.
The meaning of the per-transfer cs_change=1 is:
- If the transfer is the last one of the message, keep CS asserted
- Otherwise, deassert CS
We need to deassert CS in the "otherwise" case, which was implicit
before.
Avoiding the memcpy creates yet another opportunity. The device can't
process more than 256 bytes of SPI payload at a time, so the
sja1105_xfer_long_buf() function used to exist, to split the larger
caller buffer into chunks.
But these chunks couldn't be used as scatter/gather buffers for
spi_message until now, because of that memcpy (we would have needed more
memory for each chunk). So we can now remove the sja1105_xfer_long_buf()
function and have a single implementation for long and short buffers.
Another benefit is lower usage of stack memory. Previously we had to
store 2 SPI buffers for each chunk. Due to the elimination of the
memcpy, we can now send pointers to the actual chunks from the
caller-supplied buffer to the SPI core.
Since the patch merges two functions into a rewritten implementation,
the function prototype was also changed, mainly for cosmetic consistency
with the structures used within it.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Fri, 11 Oct 2019 17:22:32 +0000 (18:22 +0100)]
net: b44: remove redundant assignment to variable reg
The variable reg is being assigned a value that is never read
and is being re-assigned in the following for-loop. The
assignment is redundant and hence can be removed.
Addresses-Coverity: ("Unused value") Signed-off-by: Colin Ian King <colin.king@canonical.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
PTP driver refactoring for SJA1105 DSA
This series creates a better separation between the driver core and the
PTP portion. Therefore, users who are not interested in PTP can get a
simpler and smaller driver by compiling it out.
This is in preparation for further patches: SPI transfer timestamping,
synchronizing the hardware clock (as opposed to keeping it
free-running), PPS input/output, etc.
====================
Acked-by: Richard Cochran <richardcochran@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Fri, 11 Oct 2019 23:18:16 +0000 (02:18 +0300)]
net: dsa: sja1105: Change the PTP command access pattern
The PTP command register contains enable bits for:
- Putting the 64-bit PTPCLKVAL register in add/subtract or write mode
- Taking timestamps off of the corrected vs free-running clock
- Starting/stopping the TTEthernet scheduling
- Starting/stopping PPS output
- Resetting the switch
When a command needs to be issued (e.g. "change the PTPCLKVAL from write
mode to add/subtract mode"), one cannot simply write to the command
register setting the PTPCLKADD bit to 1, because that would zeroize the
other settings. One also cannot do a read-modify-write (that would be
too easy for this hardware) because not all bits of the command register
are readable over SPI.
So this leaves us with the only option of keeping the value of the PTP
command register in the driver, and operating on that.
Actually there are 2 types of PTP operations now:
- Operations that modify the cached PTP command. These operate on
ptp_data->cmd as a pointer.
- Operations that apply all previously cached PTP settings, but don't
otherwise cache what they did themselves. The sja1105_ptp_reset
function is such an example. It copies the ptp_data->cmd on stack
before modifying and writing it to SPI.
This practically means that struct sja1105_ptp_cmd is no longer an
implementation detail, since it needs to be stored in full into struct
sja1105_ptp_data, and hence in struct sja1105_private. So the (*ptp_cmd)
function prototype can change and take struct sja1105_ptp_cmd as second
argument now.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Fri, 11 Oct 2019 23:18:15 +0000 (02:18 +0300)]
net: dsa: sja1105: Move PTP data to its own private structure
This is a non-functional change with 2 goals (both for the case when
CONFIG_NET_DSA_SJA1105_PTP is not enabled):
- Reduce the size of the sja1105_private structure.
- Make the PTP code more self-contained.
Leaving priv->ptp_data.lock to be initialized in sja1105_main.c is not a
leftover: it will be used in a future patch "net: dsa: sja1105: Restore
PTP time after switch reset".
Signed-off-by: Vladimir Oltean <olteanv@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean [Fri, 11 Oct 2019 23:18:14 +0000 (02:18 +0300)]
net: dsa: sja1105: Make all public PTP functions take dsa_switch as argument
The new rule (as already started for sja1105_tas.h) is for functions of
optional driver components (ones which may be disabled via Kconfig - PTP
and TAS) to take struct dsa_switch *ds instead of struct sja1105_private
*priv as first argument.
This is so that forward-declarations of struct sja1105_private can be
avoided.
So make sja1105_ptp.h the second user of this rule.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Sun, 13 Oct 2019 18:29:07 +0000 (11:29 -0700)]
Merge tag 'mac80211-next-for-net-next-2019-10-11' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next
Johannes Berg says:
====================
A few more small things, nothing really stands out:
* minstrel improvements from Felix
* a TX aggregation simplification
* some additional capabilities for hwsim
* minor cleanups & docs updates
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Michal Kubecek [Fri, 11 Oct 2019 07:40:09 +0000 (09:40 +0200)]
genetlink: do not parse attributes for families with zero maxattr
Commit c10e6cf85e7d ("net: genetlink: push attrbuf allocation and parsing
to a separate function") moved attribute buffer allocation and attribute
parsing from genl_family_rcv_msg_doit() into a separate function
genl_family_rcv_msg_attrs_parse() which, unlike the previous code, calls
__nlmsg_parse() even if family->maxattr is 0 (i.e. the family does its own
parsing). The parser error is ignored and does not propagate out of
genl_family_rcv_msg_attrs_parse() but an error message ("Unknown attribute
type") is set in extack and if further processing generates no error or
warning, it stays there and is interpreted as a warning by userspace.
Dumpit requests are not affected as genl_family_rcv_msg_dumpit() bypasses
the call of genl_family_rcv_msg_attrs_parse() if family->maxattr is zero.
Move this logic inside genl_family_rcv_msg_attrs_parse() so that we don't
have to handle it in each caller.
v3: put the check inside genl_family_rcv_msg_attrs_parse()
v2: adjust also argument of genl_family_rcv_msg_attrs_free()
Fixes: c10e6cf85e7d ("net: genetlink: push attrbuf allocation and parsing to a separate function") Signed-off-by: Michal Kubecek <mkubecek@suse.cz> Acked-by: Jiri Pirko <jiri@mellanox.com> Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by: David S. Miller <davem@davemloft.net>
tcp: improve recv_skip_hint for tcp_zerocopy_receive
tcp_zerocopy_receive() rounds down the zc->length a multiple of
PAGE_SIZE. This results in two issues:
- tcp_zerocopy_receive sets recv_skip_hint to the length of the
receive queue if the zc->length input is smaller than the
PAGE_SIZE, even though the data in receive queue could be
zerocopied.
- tcp_zerocopy_receive would set recv_skip_hint of 0, in cases
where we have a little bit of data after the perfectly-sized
packets.
To fix these issues, do not store the rounded down value in
zc->length. Round down the length passed to zap_page_range(),
and return min(inq, zc->length) when the zap_range is 0.
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
Patch #1 enforces libbpf build to have bpf_helper_defs.h ready before test BPF
programs are built.
Patch #2 drops obsolete BTF/pahole detection logic from Makefile.
v1->v2:
- drop CPU and PROBE (Martin).
====================
Acked-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Andrii Nakryiko [Fri, 11 Oct 2019 22:01:46 +0000 (15:01 -0700)]
selftests/bpf: Remove obsolete pahole/BTF support detection
Given lots of selftests won't work without recent enough Clang/LLVM that
fully supports BTF, there is no point in maintaining outdated BTF
support detection and fall-back to pahole logic. Just assume we have
everything we need.
Andrii Nakryiko [Fri, 11 Oct 2019 22:01:45 +0000 (15:01 -0700)]
selftests/bpf: Enforce libbpf build before BPF programs are built
Given BPF programs rely on libbpf's bpf_helper_defs.h, which is
auto-generated during libbpf build, libbpf build has to happen before
we attempt progs/*.c build. Enforce it as order-only dependency.
====================
This series contains mainly fixes/improvements for cross-compilation
but not only, tested for arm, arm64, and intended for any arch.
Also verified on native build (not cross compilation) for x86_64
and arm, arm64.
Besides the patches given here, the RFC also contains couple patches
related to llvm clang
arm: include: asm: swab: mask rev16 instruction for clang
arm: include: asm: unified: mask .syntax unified for clang
They are necessarily to verify arm 32 build.
Also, couple more fixes were added but are not merged in bpf-next yet,
they can be needed for verification/configuration steps, if not in
your tree the fixes can be taken here:
https://www.spinics.net/lists/netdev/msg601716.html
https://www.spinics.net/lists/netdev/msg601714.html
https://www.spinics.net/lists/linux-kbuild/msg23468.html
Now, to build samples, SAMPLE_BPF should be enabled in config.
The change touches not only cross-compilation and can have impact on
other archs and build environments, so might be good idea to verify
it in order to add appropriate changes, some warn options could be
tuned also.
All is tested on x86-64 with clang installed (has to be built containing
targets for arm, arm64..., see llc --version, usually it's present already)
Instructions to test native on x86_64
=================================================
Native build on x86_64 is done in usual way and shouldn't have difference
except HOSTCC is now printed as CC wile building the samples.
Instructions to test cross compilation on arm64
=================================================
gcc version 8.3.0
(GNU Toolchain for the A-profile Architecture 8.3-2019.03 (arm-rel-8.36))
I've used sdk for TI am65x got here:
http://downloads.ti.com/processor-sdk-linux/esd/AM65X/latest/exports/\
ti-processor-sdk-linux-am65xx-evm-06.00.00.07-Linux-x86-Install.bin
make ARCH=arm64 -C tools/ clean
make ARCH=arm64 -C samples/bpf clean
make ARCH=arm64 clean
make ARCH=arm64 defconfig
make ARCH=arm64 headers_install
make ARCH=arm64 INSTALL_HDR_PATH=/../sdk/\
ti-processor-sdk-linux-am65xx-evm-06.00.00.07/linux-devkit/sysroots/\
aarch64-linux/usr headers_install
make samples/bpf/ ARCH=arm64 CROSS_COMPILE="aarch64-linux-gnu-"\
SYSROOT="/../sdk/ti-processor-sdk-linux-am65xx-evm-06.00.00.07/\
linux-devkit/sysroots/aarch64-linux"
Instructions to test cross compilation on arm
=================================================
arm-linux-gnueabihf-gcc (Linaro GCC 7.2-2017.11) 7.2.1 20171011
or
arm-linux-gnueabihf-gcc
(GNU Toolchain for the A-profile Architecture 8.3-2019.03 \
(arm-rel-8.36)) 8.3.0
make ARCH=arm -C tools/ clean
make ARCH=arm -C samples/bpf clean
make ARCH=arm clean
make ARCH=arm omap2plus_defconfig
make ARCH=arm headers_install
make ARCH=arm INSTALL_HDR_PATH=/../sdk/\
ti-processor-sdk-linux-am57xx-evm-05.03.00.07/linux-devkit/sysroots/\
armv7ahf-neon-linux-gnueabi/usr headers_install
make samples/bpf/ ARCH=arm CROSS_COMPILE="arm-linux-gnueabihf-"\
SYSROOT="/../sdk/ti-processor-sdk-linux-am57xx-evm-05.03\
.00.07/linux-devkit/sysroots/armv7ahf-neon-linux-gnueabi"
Based on bpf-next/master
v5..v4:
- any changes, only missed SOBs are added
v4..v3:
- renamed CLANG_EXTRA_CFLAGS on BPF_EXTRA_CFLAGS
- used filter for ARCH_ARM_SELECTOR
- omit "-fomit-frame-pointer" and use same flags for native and "cross"
- used sample/bpf prefixes
- use C instead of C++ compiler for test_libbpf target
v3..v2:
- renamed makefile.progs to makeifle.target, as more appropriate
- left only __LINUX_ARM_ARCH__ for D options for arm
- for host build - left options from KBUILD_HOST for compatibility reasons
- split patch adding c/cxx/ld flags to libbpf by modules
- moved readme change to separate patch
- added patch setting options for cross-compile
- fixed issue with option error for syscall_nrs.S,
avoiding overlap for ccflags-y.
v2..v1:
- restructured patches order
- split "samples: bpf: Makefile: base progs build on Makefile.progs"
to make change more readable. It added couple nice extra patches.
- removed redundant patch:
"samples: bpf: Makefile: remove target for native build"
- added fix:
"samples: bpf: makefile: fix cookie_uid_helper_example obj build"
- limited -D option filter only for arm
- improved comments
- added couple instructions to verify cross compilation for arm and
arm64 arches based on TI am57xx and am65xx sdks.
- corrected include a little order
====================
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:07 +0000 (03:28 +0300)]
samples/bpf: Add sysroot support
Basically it only enables that was added by previous couple fixes.
Sysroot contains correct libs installed and its headers. Useful when
working with NFC or virtual machine.
Usage example:
clean (on demand)
make ARCH=arm -C samples/bpf clean
make ARCH=arm -C tools clean
make ARCH=arm clean
configure and install headers:
make ARCH=arm defconfig
make ARCH=arm headers_install
build samples/bpf:
make ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- samples/bpf/ \
SYSROOT="path/to/sysroot"
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:05 +0000 (03:28 +0300)]
libbpf: Add C/LDFLAGS to libbpf.so and test_libpf targets
In case of C/LDFLAGS there is no way to pass them correctly to build
command, for instance when --sysroot is used or external libraries
are used, like -lelf, wich can be absent in toolchain. This can be
used for samples/bpf cross-compiling allowing to get elf lib from
sysroot.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:04 +0000 (03:28 +0300)]
libbpf: Don't use cxx to test_libpf target
No need to use C++ for test_libbpf target when libbpf is on C and it
can be tested with C, after this change the CXXFLAGS in makefiles can
be avoided, at least in bpf samples, when sysroot is used, passing
same C/LDFLAGS as for lib.
Add "return 0" in test_libbpf to avoid warn, but also remove spaces at
start of the lines to keep same style and avoid warns while apply.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:02 +0000 (03:28 +0300)]
samples/bpf: Use own flags but not HOSTCFLAGS
While compiling natively, the host's cflags and ldflags are equal to
ones used from HOSTCFLAGS and HOSTLDFLAGS. When cross compiling it
should have own, used for target arch. While verification, for arm,
arm64 and x86_64 the following flags were used always:
So, add them as they were verified and used before adding
Makefile.target and lets omit "-fomit-frame-pointer" as were proposed
while review, as no sense in such optimization for samples.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:01 +0000 (03:28 +0300)]
samples/bpf: Base target programs rules on Makefile.target
The main reason for that - HOSTCC and CC have different aims.
HOSTCC is used to build programs running on host, that can
cross-comple target programs with CC. It was tested for arm and arm64
cross compilation, based on linaro toolchain, but should work for
others.
So, in order to split cross compilation (CC) with host build (HOSTCC),
lets base samples on Makefile.target. It allows to cross-compile
samples/bpf programs with CC while auxialry tools running on host
built with HOSTCC.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:28:00 +0000 (03:28 +0300)]
samples/bpf: Add makefile.target for separate CC target build
The Makefile.target is added only and will be used in
sample/bpf/Makefile later in order to switch cross-compiling to CC
from HOSTCC environment.
The HOSTCC is supposed to build binaries and tools running on the host
afterwards, in order to simplify build or so, like "fixdep" or else.
In case of cross compiling "fixdep" is executed on host when the rest
samples should run on target arch. In order to build binaries for
target arch with CC and tools running on host with HOSTCC, lets add
Makefile.target for simplicity, having definition and routines similar
to ones, used in script/Makefile.host. This allows later add
cross-compilation to samples/bpf with minimum changes.
The tprog stands for target programs built with CC.
Makefile.target contains only stuff needed for samples/bpf, potentially
can be reused later and now needed only for unblocking tricky
samples/bpf cross compilation.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:27:59 +0000 (03:27 +0300)]
samples/bpf: Drop unnecessarily inclusion for bpf_load
Drop inclusion for bpf_load -I$(objtree)/usr/include as it is
included for all objects anyway, with above line:
KBUILD_HOSTCFLAGS += -I$(objtree)/usr/include
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:27:58 +0000 (03:27 +0300)]
samples/bpf: Use __LINUX_ARM_ARCH__ selector for arm
For arm, -D__LINUX_ARM_ARCH__=X is min version used as instruction
set selector and is absolutely required while parsing some parts of
headers. It's present in KBUILD_CFLAGS but not in autoconf.h, so let's
retrieve it from and add to programs cflags. In another case errors
like "SMP is not supported" for armv7 and bunch of other errors are
issued resulting to incorrect final object.
Ivan Khoronzhuk [Fri, 11 Oct 2019 00:27:56 +0000 (03:27 +0300)]
samples/bpf: Use --target from cross-compile
For cross compiling the target triple can be inherited from
cross-compile prefix as it's done in CLANG_FLAGS from kernel makefile.
So copy-paste this decision from kernel Makefile.
Don't list userspace "cookie_uid_helper_example" object in list for
bpf objects.
'always' target is used for listing bpf programs, but
'cookie_uid_helper_example.o' is a user space ELF file, and covered
by rule `per_socket_stats_example`, so shouldn't be in 'always'.
Let us remove `always += cookie_uid_helper_example.o`, which avoids
breaking cross compilation due to mismatched includes.
Jiri Pirko [Thu, 10 Oct 2019 13:18:50 +0000 (15:18 +0200)]
netdevsim: implement couple of testing devlink health reporters
Implement "empty" and "dummy" reporters. The first one is really simple
and does nothing. The other one has debugfs files to trigger breakage
and it is able to do recovery. The ops also implement dummy fmsg
content.
Signed-off-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Peter Fink [Thu, 10 Oct 2019 13:00:22 +0000 (15:00 +0200)]
net: usb: ax88179_178a: write mac to hardware in get_mac_addr
When the MAC address is supplied via device tree or a random
MAC is generated it has to be written to the asix chip in
order to receive any data.
Previously in 9fb137aef34e ("net: usb: ax88179_178a: allow
optionally getting mac address from device tree") this line was
omitted because it seemed to work perfectly fine without it.
But it was simply not detected because the chip keeps the mac
stored even beyond a reset and it was tested on a hardware
with an integrated UPS where the asix chip was permanently
powered on even throughout power cycles.
Fixes: 9fb137aef34e ("net: usb: ax88179_178a: allow optionally getting mac address from device tree") Signed-off-by: Peter Fink <pfink@christ-es.de> Signed-off-by: David S. Miller <davem@davemloft.net>
Vito Caputo [Thu, 10 Oct 2019 03:43:47 +0000 (20:43 -0700)]
af_unix: __unix_find_socket_byname() cleanup
Remove pointless return variable dance.
Appears vestigial from when the function did locking as seen in
unix_find_socket_byinode(), but locking is handled in
unix_find_socket_byname() for __unix_find_socket_byname().
Signed-off-by: Vito Caputo <vcaputo@pengaru.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
net: ftgmac100: Ungate RCLK for RMII on ASPEED MACs
This series slightly extends the devicetree binding and driver for the
FTGMAC100 to describe an optional RMII RCLK gate in the clocks property.
Currently it's necessary for the kernel to ungate RCLK on the AST2600 in NCSI
configurations as u-boot does not yet support NCSI (which uses the
R(educed)MII).
v2:
* Clear up Reduced vs Reversed MII in the cover letter
* Mitigate anxiety in the commit message for 1/3
* Clarify that AST2500 is also affected in the clocks property description in
2/3
* Rework the error paths and update some comments in 3/3
v1 can be found here: https://lore.kernel.org/netdev/20191008115143.14149-1-andrew@aj.id.au/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Critically, the AST2600 requires ungating the RMII RCLK if e.g. NCSI is
in use.
Signed-off-by: Andrew Jeffery <andrew@aj.id.au> Acked-by: Joel Stanley <joel@jms.id.au> Acked-by: Rob Herring <robh@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
The AST2600 contains an FTGMAC100-compatible MAC, although the MDIO
controller previously embedded in the MAC has been moved out to a
dedicated MDIO block.
Signed-off-by: Andrew Jeffery <andrew@aj.id.au> Acked-by: Joel Stanley <joel@jms.id.au> Acked-by: Rob Herring <robh@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Andrii Nakryiko [Fri, 11 Oct 2019 03:29:01 +0000 (20:29 -0700)]
libbpf: Handle invalid typedef emitted by old GCC
Old GCC versions are producing invalid typedef for __gnuc_va_list
pointing to void. Special-case this and emit valid:
typedef __builtin_va_list __gnuc_va_list;
Reported-by: John Fastabend <john.fastabend@gmail.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Martin KaFai Lau <kafai@fb.com> Acked-by: John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/bpf/20191011032901.452042-1-andriin@fb.com
Andrii Nakryiko [Fri, 11 Oct 2019 02:38:47 +0000 (19:38 -0700)]
libbpf: Generate more efficient BPF_CORE_READ code
Existing BPF_CORE_READ() macro generates slightly suboptimal code. If
there are intermediate pointers to be read, initial source pointer is
going to be assigned into a temporary variable and then temporary
variable is going to be uniformly used as a "source" pointer for all
intermediate pointer reads. Schematically (ignoring all the type casts),
BPF_CORE_READ(s, a, b, c) is expanded into:
({
const void *__t = src;
bpf_probe_read(&__t, sizeof(*__t), &__t->a);
bpf_probe_read(&__t, sizeof(*__t), &__t->b);
This initial `__t = src` makes calls more uniform, but causes slightly
less optimal register usage sometimes when compiled with Clang. This can
cascase into, e.g., more register spills.
This patch fixes this issue by generating more optimal sequence:
({
const void *__t;
bpf_probe_read(&__t, sizeof(*__t), &src->a); /* <-- src here */
bpf_probe_read(&__t, sizeof(*__t), &__t->b);
projects / mirror_ubuntu-kernels.git / log