This kfunc can be used in LSM hooks with dentry, such as:
security_inode_listxattr
security_inode_permission
and many more.
Acked-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20240806230904.71194-3-song@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Matt Bobrowski says:
====================
bpf: introduce new VFS based BPF kfuncs
G'day!
A respin based off v3, which can be found here [0]. Original
motivations for introducing this suite of BPF kfuncs can be found here
[1].
The primary difference in this version of the patch series is that the
suite of VFS related BPF kfuncs added can be used from both sleepable
and non-sleepable BPF LSM program types. IOW, the KF_SLEEPABLE
annotation has been removed from all of them.
Changes sinve v3:
* KF_SLEEPABLE annotation has been dropped from all newly introduced
VFS related BPF kfuncs. This includes bpf_get_task_exe_file(),
bpf_put_file(), and bpf_path_d_path(). Both negative and positive
selftests backing these new BPF kfuncs have also been updated
accordingly.
* buf__sz conditional in bpf_path_d_path() has been updated from
buf__sz <= 0, to !buf__sz.
* Syntax issues as reported so here [2] have been corrected.
[0] https://lore.kernel.org/bpf/20240726085604.2369469-1-mattbobrowski@google.com/
[1] https://lore.kernel.org/bpf/cover.1708377880.git.mattbobrowski@google.com/#t
[2] https://netdev.bots.linux.dev/static/nipa/874023/13742510/checkpatch/stdout
====================
Link: https://lore.kernel.org/r/20240731110833.1834742-1-mattbobrowski@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a bunch of positive selftests which extensively cover the various
contexts and parameters in which the new VFS based BPF kfuncs may be
used from.
Again, the following VFS based BPF kfuncs are thoroughly tested within
this new selftest:
* struct file *bpf_get_task_exe_file(struct task_struct *);
* void bpf_put_file(struct file *);
* int bpf_path_d_path(struct path *, char *, size_t);
Acked-by: Christian Brauner <brauner@kernel.org>
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Matt Bobrowski <mattbobrowski@google.com>
Link: https://lore.kernel.org/r/20240731110833.1834742-4-mattbobrowski@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a bunch of negative selftests responsible for asserting that the
BPF verifier successfully rejects a BPF program load when the
underlying BPF program misuses one of the newly introduced VFS based
BPF kfuncs.
The following VFS based BPF kfuncs are extensively tested within this
new selftest:
* struct file *bpf_get_task_exe_file(struct task_struct *);
* void bpf_put_file(struct file *);
* int bpf_path_d_path(struct path *, char *, size_t);
Acked-by: Christian Brauner <brauner@kernel.org>
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Matt Bobrowski <mattbobrowski@google.com>
Link: https://lore.kernel.org/r/20240731110833.1834742-3-mattbobrowski@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a new variant of bpf_d_path() named bpf_path_d_path() which takes
the form of a BPF kfunc and enforces KF_TRUSTED_ARGS semantics onto
its arguments.
This new d_path() based BPF kfunc variant is intended to address the
legacy bpf_d_path() BPF helper's susceptability to memory corruption
issues [0, 1, 2] by ensuring to only operate on supplied arguments
which are deemed trusted by the BPF verifier. Typically, this means
that only pointers to a struct path which have been referenced counted
may be supplied.
In addition to the new bpf_path_d_path() BPF kfunc, we also add a
KF_ACQUIRE based BPF kfunc bpf_get_task_exe_file() and KF_RELEASE
counterpart BPF kfunc bpf_put_file(). This is so that the new
bpf_path_d_path() BPF kfunc can be used more flexibily from within the
context of a BPF LSM program. It's rather common to ascertain the
backing executable file for the calling process by performing the
following walk current->mm->exe_file while instrumenting a given
operation from the context of the BPF LSM program. However, walking
current->mm->exe_file directly is never deemed to be OK, and doing so
from both inside and outside of BPF LSM program context should be
considered as a bug. Using bpf_get_task_exe_file() and in turn
bpf_put_file() will allow BPF LSM programs to reliably get and put
references to current->mm->exe_file.
As of now, all the newly introduced BPF kfuncs within this patch are
limited to BPF LSM program types. These can be either sleepable or
non-sleepable variants of BPF LSM program types.
[0] https://lore.kernel.org/bpf/CAG48ez0ppjcT=QxU-jtCUfb5xQb3mLr=5FcwddF_VKfEBPs_Dg@mail.gmail.com/
[1] https://lore.kernel.org/bpf/20230606181714.532998-1-jolsa@kernel.org/
[2] https://lore.kernel.org/bpf/20220219113744.1852259-1-memxor@gmail.com/
Acked-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Matt Bobrowski <mattbobrowski@google.com>
Acked-by: Song Liu <song@kernel.org>
Link: https://lore.kernel.org/r/20240731110833.1834742-2-mattbobrowski@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Jakub reported bpf selftest "btf_dump" failure after forwarding to
v6.11-rc1 with netdev.
Error: #33 btf_dump
Error: #33/15 btf_dump/btf_dump: var_data
btf_dump_data:FAIL:find type id unexpected find type id: actual -2 < expected 0
The reason for the failure is due to
commit 94ede2a3e9 ("profiling: remove stale percpu flip buffer variables")
where percpu static variable "cpu_profile_flip" is removed.
Let us replace "cpu_profile_flip" with a variable in bpf subsystem
so whenever that variable gets deleted or renamed, we can detect the
failure immediately. In this case, I picked a static percpu variable
"bpf_cgrp_storage_busy" which is defined in kernel/bpf/bpf_cgrp_storage.c.
Reported-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20240802185434.1749056-1-yonghong.song@linux.dev
Alexis Lothoré (eBPF Foundation) says:
====================
Hello,
this small series aims to integrate test_dev_cgroup in test_progs so it
could be run automatically in CI. The new version brings a few differences
with the current one:
- test now uses directly syscalls instead of wrapping commandline tools
into system() calls
- test_progs manipulates /dev/null (eg: redirecting test logs into it), so
disabling access to it in the bpf program confuses the tests. To fix this,
the first commit modifies the bpf program to allow access to char devices
1:3 (/dev/null), and disable access to char devices 1:5 (/dev/zero)
- once test is converted, add a small subtest to also check for device type
interpretation (char or block)
- paths used in mknod tests are now in /dev instead of /tmp: due to the CI
runner organisation and mountpoints manipulations, trying to create nodes
in /tmp leads to errors unrelated to the test (ie, mknod calls refused by
kernel, not the bpf program). I don't understand exactly the root cause
at the deepest point (all I see in CI is an -ENXIO error on mknod when trying to
create the node in tmp, and I can not make sense out of it neither
replicate it locally), so I would gladly take inputs from anyone more
educated than me about this.
The new test_progs part has been tested in a local qemu environment as well
as in upstream CI:
./test_progs -a cgroup_dev
47/1 cgroup_dev/allow-mknod:OK
47/2 cgroup_dev/allow-read:OK
47/3 cgroup_dev/allow-write:OK
47/4 cgroup_dev/deny-mknod:OK
47/5 cgroup_dev/deny-read:OK
47/6 cgroup_dev/deny-write:OK
47/7 cgroup_dev/deny-mknod-wrong-type:OK
47 cgroup_dev:OK
Summary: 1/7 PASSED, 0 SKIPPED, 0 FAILED
---
Changes in v4:
- Fix mixup between ret and errno by testing both
- Properly apply ack tag from Stanislas
- Link to v3: https://lore.kernel.org/r/20240730-convert_dev_cgroup-v3-0-93e573b74357@bootlin.com
Changes in v3:
- delete mknod file only if it has been created
- use bpf_program__attach_cgroup() instead of bpf_prog_attach
- reorganize subtests order
- collect review/ack tags from Alan and Stanislas
- Link to v2: https://lore.kernel.org/r/20240729-convert_dev_cgroup-v2-0-4c1fc0520545@bootlin.com
Changes in v2:
- directly pass expected ret code to subtests instead of boolean pass/not
pass
- fix faulty fd check in subtest expected to fail on open
- fix wrong subtest name
- pass test buffer and corresponding size to read/write subtests
- use correct series prefix
- Link to v1: https://lore.kernel.org/r/20240725-convert_dev_cgroup-v1-0-2c8cbd487c44@bootlin.com
====================
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Current cgroup_dev test mostly tests that device operation is accepted or
refused base on passed major/minor (and so, any operation performed during
test involves only char device)
Add a small subtest ensuring that the device type passed to bpf program
allows it to take decisions as well.
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Signed-off-by: Alexis Lothoré (eBPF Foundation) <alexis.lothore@bootlin.com>
Link: https://lore.kernel.org/r/20240731-convert_dev_cgroup-v4-3-849425d90de6@bootlin.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
test_dev_cgroup is defined as a standalone test program, and so is not
executed in CI.
Convert it to test_progs framework so it is tested automatically in CI, and
remove the old test. In order to be able to run it in test_progs, /dev/null
must remain usable, so change the new test to test operations on devices
1:3 as valid, and operations on devices 1:5 (/dev/zero) as invalid.
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Signed-off-by: Alexis Lothoré (eBPF Foundation) <alexis.lothore@bootlin.com>
Link: https://lore.kernel.org/r/20240731-convert_dev_cgroup-v4-2-849425d90de6@bootlin.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
test_dev_cgroup currently loads a small bpf program allowing any access on
urandom and zero devices, disabling access to any other device. It makes
migrating this test to test_progs impossible, since this one manipulates
extensively /dev/null.
Allow /dev/null manipulation in dev_cgroup program to make its usage in
test_progs framework possible. Update test_dev_cgroup.c as well to match
this change while it has not been removed.
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Signed-off-by: Alexis Lothoré (eBPF Foundation) <alexis.lothore@bootlin.com>
Link: https://lore.kernel.org/r/20240731-convert_dev_cgroup-v4-1-849425d90de6@bootlin.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
When building selftests with RELEASE=1 using GCC compiler, it complaints
about uninitialized err. Fix the problem.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20240730231805.1933923-1-andrii@kernel.org
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
We recently found out that extending xsk_umem_reg might be a bit
complicated due to not enforcing padding to be zero [0]. Add
a couple of things to make it less error-prone:
1. Remove xdp_umem_reg_v2 since its sizeof is the same as xdp_umem_reg
2. Add BUILD_BUG_ON that checks that the size of xdp_umem_reg_v1 is less
than xdp_umem_reg; presumably, when we get to v2, there is gonna
be a similar line to enforce that sizeof(v2) > sizeof(v1)
3. Add BUILD_BUG_ON to make sure the last field plus its size matches
the overall struct size. The intent is to demonstrate that we don't
have any lingering padding.
0: https://lore.kernel.org/bpf/ZqI29QE+5JnkdPmE@boxer/T/#me03113f7c2458fd08f3c4114a7a9472ac3646c98
Reported-by: Julian Schindel <mail@arctic-alpaca.de>
Cc: Magnus Karlsson <magnus.karlsson@gmail.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://lore.kernel.org/r/20240726222048.1397869-1-sdf@fomichev.me
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
After the commit 66665ad2f1 ("tracing/kprobe: bpf: Compare instruction
pointer with original one"), "bpf_kprobe_override" is not used anywhere
anymore, and we can remove it now.
Fixes: 66665ad2f1 ("tracing/kprobe: bpf: Compare instruction pointer with original one")
Signed-off-by: Menglong Dong <dongml2@chinatelecom.cn>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20240730053733.885785-1-dongml2@chinatelecom.cn
Linux 5.1 implemented 64-bit time types and related syscalls to address the
Y2038 problem generally across archs. Userspace handling of Y2038 varies
with the libc however. While musl libc uses 64-bit time across all 32-bit
and 64-bit platforms, GNU glibc uses 64-bit time on 64-bit platforms but
defaults to 32-bit time on 32-bit platforms unless they "opt-in" to 64-bit
time or explicitly use 64-bit syscalls and time structures.
One specific area is the standard setsockopt() call, SO_TIMESTAMPNS option
used for timestamping, and the related output 'struct timespec'. GNU glibc
defaults as above, also exposing the SO_TIMESTAMPNS_NEW flag to explicitly
use a 64-bit call and 'struct __kernel_timespec'. Since these are not
exposed or needed with musl libc, their use in tc_redirect.c leads to
compile errors building for mips64el/musl:
tc_redirect.c: In function 'rcv_tstamp':
tc_redirect.c:425:32: error: 'SO_TIMESTAMPNS_NEW' undeclared (first use in this function); did you mean 'SO_TIMESTAMPNS'?
425 | cmsg->cmsg_type == SO_TIMESTAMPNS_NEW)
| ^~~~~~~~~~~~~~~~~~
| SO_TIMESTAMPNS
tc_redirect.c:425:32: note: each undeclared identifier is reported only once for each function it appears in
tc_redirect.c: In function 'test_inet_dtime':
tc_redirect.c:491:49: error: 'SO_TIMESTAMPNS_NEW' undeclared (first use in this function); did you mean 'SO_TIMESTAMPNS'?
491 | err = setsockopt(listen_fd, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
| ^~~~~~~~~~~~~~~~~~
| SO_TIMESTAMPNS
However, using SO_TIMESTAMPNS_NEW isn't strictly needed, nor is Y2038 being
explicitly tested. The timestamp checks in tc_redirect.c are simple: the
packet receive timestamp is non-zero and processed/handled in less than 5
seconds.
Switch to using the standard setsockopt() call and SO_TIMESTAMPNS option to
ensure compatibility across glibc and musl libc. In the worst-case, there
is a 5-second window 14 years from now where tc_redirect tests may fail on
32-bit systems. However, we should reasonably expect glibc to adopt a
64-bit mandate rather than the current "opt-in" policy before the Y2038
roll-over.
Fixes: ce6f6cffae ("selftests/bpf: Wait for the netstamp_needed_key static key to be turned on")
Fixes: c803475fd8 ("bpf: selftests: test skb->tstamp in redirect_neigh")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/031d656c058b4e55ceae56ef49c4e1729b5090f3.1722244708.git.tony.ambardar@gmail.com
Typically stdin, stdout, stderr are treated as reserved identifiers under
ISO/ANSI C and libc implementations further define these as macros, both in
glibc and musl <stdio.h>.
However, while glibc defines:
...
/* Standard streams. */
extern FILE *stdin; /* Standard input stream. */
extern FILE *stdout; /* Standard output stream. */
extern FILE *stderr; /* Standard error output stream. */
/* C89/C99 say they're macros. Make them happy. */
#define stdin stdin
#define stdout stdout
#define stderr stderr
...
musl instead uses (legally):
...
extern FILE *const stdin;
extern FILE *const stdout;
extern FILE *const stderr;
#define stdin (stdin)
#define stdout (stdout)
#define stderr (stderr)
...
The latter results in compile errors when the names are reused as fields of
'struct test_env' and elsewhere in test_progs.[ch] and reg_bounds.c.
Rename the fields to stdout_saved and stderr_saved to avoid many errors
seen building against musl, e.g.:
In file included from test_progs.h:6,
from test_progs.c:5:
test_progs.c: In function 'print_test_result':
test_progs.c:237:21: error: expected identifier before '(' token
237 | fprintf(env.stdout, "#%-*d %s:", TEST_NUM_WIDTH, test->test_num, test->test_name);
| ^~~~~~
test_progs.c:237:9: error: too few arguments to function 'fprintf'
237 | fprintf(env.stdout, "#%-*d %s:", TEST_NUM_WIDTH, test->test_num, test->test_name);
| ^~~~~~~
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/ZqR2DuHdBXPX%2Fyx8@kodidev-ubuntu/
Link: https://lore.kernel.org/bpf/684ea17548e237f39dfb3f7a3d33450069015b21.1722244708.git.tony.ambardar@gmail.com
Include GNU <execinfo.h> header only with glibc and provide weak, stubbed
backtrace functions as a fallback in test_progs.c. This allows for non-GNU
replacements while avoiding compile errors (e.g. with musl libc) like:
test_progs.c:13:10: fatal error: execinfo.h: No such file or directory
13 | #include <execinfo.h> /* backtrace */
| ^~~~~~~~~~~~
test_progs.c: In function 'crash_handler':
test_progs.c:1034:14: error: implicit declaration of function 'backtrace' [-Werror=implicit-function-declaration]
1034 | sz = backtrace(bt, ARRAY_SIZE(bt));
| ^~~~~~~~~
test_progs.c:1045:9: error: implicit declaration of function 'backtrace_symbols_fd' [-Werror=implicit-function-declaration]
1045 | backtrace_symbols_fd(bt, sz, STDERR_FILENO);
| ^~~~~~~~~~~~~~~~~~~~
Fixes: 9fb156bb82 ("selftests/bpf: Print backtrace on SIGSEGV in test_progs")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/aa6dc8e23710cb457b278039d0081de7e7b4847d.1722244708.git.tony.ambardar@gmail.com
Compiling lwt_reroute.c with GCC 12.3 for mips64el/musl-libc yields errors:
In file included from .../include/arpa/inet.h:9,
from ./test_progs.h:18,
from tools/testing/selftests/bpf/prog_tests/lwt_helpers.h:11,
from tools/testing/selftests/bpf/prog_tests/lwt_reroute.c:52:
.../include/netinet/in.h:23:8: error: redefinition of 'struct in6_addr'
23 | struct in6_addr {
| ^~~~~~~~
In file included from .../include/linux/icmp.h:24,
from tools/testing/selftests/bpf/prog_tests/lwt_helpers.h:9:
.../include/linux/in6.h:33:8: note: originally defined here
33 | struct in6_addr {
| ^~~~~~~~
.../include/netinet/in.h:34:8: error: redefinition of 'struct sockaddr_in6'
34 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../include/linux/in6.h:50:8: note: originally defined here
50 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../include/netinet/in.h:42:8: error: redefinition of 'struct ipv6_mreq'
42 | struct ipv6_mreq {
| ^~~~~~~~~
.../include/linux/in6.h:60:8: note: originally defined here
60 | struct ipv6_mreq {
| ^~~~~~~~~
These errors occur because <linux/in6.h> is included before <netinet/in.h>,
bypassing the Linux uapi/libc compat mechanism's partial musl support. As
described in [1] and [2], fix these errors by including <netinet/in.h> in
lwt_reroute.c before any uapi headers.
[1]: commit c0bace7984 ("uapi libc compat: add fallback for unsupported libcs")
[2]: https://git.musl-libc.org/cgit/musl/commit/?id=04983f227238
Fixes: 6c77997bc6 ("selftests/bpf: Add lwt_xmit tests for BPF_REROUTE")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/bd2908aec0755ba8b75f5dc41848b00585f5c73e.1722244708.git.tony.ambardar@gmail.com
While building, bpftool makes a skeleton from test_core_extern.c, which
itself includes <stdbool.h> and uses the 'bool' type. However, the skeleton
test_core_extern.skel.h generated *does not* include <stdbool.h> or use the
'bool' type, instead using the C-only '_Bool' type. Compiling test_cpp.cpp
with g++ 12.3 for mips64el/musl-libc then fails with error:
In file included from test_cpp.cpp:9:
test_core_extern.skel.h:45:17: error: '_Bool' does not name a type
45 | _Bool CONFIG_BOOL;
| ^~~~~
This was likely missed previously because glibc uses a GNU extension for
<stdbool.h> with C++ (#define _Bool bool), not supported by musl libc.
Normally, a C fragment would include <stdbool.h> and use the 'bool' type,
and thus cleanly work after import by C++. The ideal fix would be for
'bpftool gen skeleton' to output the correct type/include supporting C++,
but in the meantime add a conditional define as above.
Fixes: 7c8dce4b16 ("bpftool: Make skeleton C code compilable with C++ compiler")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/6fc1dd28b8bda49e51e4f610bdc9d22f4455632d.1722244708.git.tony.ambardar@gmail.com
Although the post-increment in macro 'CPU_SET(next++, &cpuset)' seems safe,
the sequencing can raise compile errors, so move the increment outside the
macro. This avoids an error seen using gcc 12.3.0 for mips64el/musl-libc:
In file included from test_lru_map.c:11:
test_lru_map.c: In function 'sched_next_online':
test_lru_map.c:129:29: error: operation on 'next' may be undefined [-Werror=sequence-point]
129 | CPU_SET(next++, &cpuset);
| ^
cc1: all warnings being treated as errors
Fixes: 3fbfadce60 ("bpf: Fix test_lru_sanity5() in test_lru_map.c")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/22993dfb11ccf27925a626b32672fd3324cb76c4.1722244708.git.tony.ambardar@gmail.com
Current code parses arguments with strtok_r() using a construct like
char *state = NULL;
while ((next = strtok_r(state ? NULL : input, ",", &state))) {
...
}
where logic assumes the 'state' var can distinguish between first and
subsequent strtok_r() calls, and adjusts parameters accordingly. However,
'state' is strictly internal context for strtok_r() and no such assumptions
are supported in the man page. Moreover, the exact behaviour of 'state'
depends on the libc implementation, making the above code fragile.
Indeed, invoking "./test_progs -t <test_name>" on mips64el/musl will hang,
with the above code in an infinite loop.
Similarly, we see strange behaviour running 'veristat' on mips64el/musl:
$ ./veristat -e file,prog,verdict,insns -C two-ok add-failure
Can't specify more than 9 stats
Rewrite code using a counter to distinguish between strtok_r() calls.
Fixes: 61ddff373f ("selftests/bpf: Improve by-name subtest selection logic in prog_tests")
Fixes: 394169b079 ("selftests/bpf: add comparison mode to veristat")
Fixes: c8bc5e0509 ("selftests/bpf: Add veristat tool for mass-verifying BPF object files")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/392d8bf5559f85fa37926c1494e62312ef252c3d.1722244708.git.tony.ambardar@gmail.com
Use the POSIX version of basename() to allow compilation against non-gnu
libc (e.g. musl). Include <libgen.h> ahead of <string.h> to enable using
functions from the latter while preferring POSIX over GNU basename().
In veristat.c, rely on strdupa() to avoid basename() altering the passed
"const char" argument. This is not needed in xskxceiver.c since the arg
is mutable and the program exits immediately after usage.
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/0fd3c9f3c605e6cba33504213c9df287817ade04.1722244708.git.tony.ambardar@gmail.com
In prog_tests/test_global_maps_resize.c, we test various use cases for
resizing global maps. Commit 7244100e0389 ("libbpf: Don't take direct
pointers into BTF data from st_ops") updated libbpf to not store pointers
to volatile BTF data, which for some users, was causing a UAF when resizing
a datasec array.
Let's ensure we have coverage for resizing datasec arrays with struct_ops
progs by also including a struct_ops map and struct_ops prog in the
test_global_map_resize skeleton. The map is automatically loaded, so we
don't need to do anything other than add it to the BPF prog being tested
to get the coverage.
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20240725032214.50676-1-void@manifault.com
test_xdp_veth.sh tests that XDP return codes work as expected, by bringing
up multiple veth pairs isolated in different namespaces, attaching specific
xdp programs to each interface, and ensuring that the whole chain allows to
ping one end interface from the first one. The test runs well but is
currently not integrated in test_progs, which prevents it from being run
automatically in the CI infrastructure.
Rewrite it as a C test relying on libbpf to allow running it in the CI
infrastructure. The new code brings up the same network infrastructure and
reuses the same eBPF programs as test_xdp_veth.sh, for which skeletons are
already generated by the bpf tests makefile.
Signed-off-by: Alexis Lothoré (eBPF Foundation) <alexis.lothore@bootlin.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://lore.kernel.org/bpf/20240721-convert_test_xdp_veth-v4-2-23bdba21b2f9@bootlin.com
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
xdp_redirect_map.c is a bpf program used by test_xdp_veth.sh, which is not
handled by the generic test runner (test_progs). To allow converting this
test to test_progs, the corresponding program must be updated to allow
handling it through skeletons generated by bpftool and libbpf.
Update programs section names to allow to manipulate those with libbpf.
Signed-off-by: Alexis Lothoré (eBPF Foundation) <alexis.lothore@bootlin.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://lore.kernel.org/bpf/20240721-convert_test_xdp_veth-v4-1-23bdba21b2f9@bootlin.com
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
In struct bpf_struct_ops, we have take a pointer to a BTF type name, and
a struct btf_type. This was presumably done for convenience, but can
actually result in subtle and confusing bugs given that BTF data can be
invalidated before a program is loaded. For example, in sched_ext, we
may sometimes resize a data section after a skeleton has been opened,
but before the struct_ops scheduler map has been loaded. This may cause
the BTF data to be realloc'd, which can then cause a UAF when loading
the program because the struct_ops map has pointers directly into the
BTF data.
We're already storing the BTF type_id in struct bpf_struct_ops. Because
type_id is stable, we can therefore just update the places where we were
looking at those pointers to instead do the lookups we need from the
type_id.
Fixes: 590a008882 ("bpf: libbpf: Add STRUCT_OPS support")
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20240724171459.281234-1-void@manifault.com
Tony Ambardar says:
====================
selftests/bpf: Improve libc portability / musl support (part 1)
Hello all,
This series includes the bulk of libc-related compile fixes accumulated to
support systems using musl, with smaller numbers to follow. These patches
are simple and straightforward, and the series has been tested with the
kernel-patches/bpf CI and locally using mips64el-gcc/musl-libc and QEMU
with an OpenWrt rootfs.
The patches address a few general categories of libc portability issues:
- missing, redundant or incorrect include headers
- disabled GNU header extensions (i.e. missing #define _GNU_SOURCE)
- issues with types and casting
Feedback and suggestions for improvement are welcome!
Thanks,
Tony
====================
Link: https://lore.kernel.org/r/cover.1721713597.git.tony.ambardar@gmail.com
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Remove a redundant include of '<asm/types.h>', whose needed definitions are
already included (via '<linux/types.h>') in cg_storage_multi_egress_only.c,
cg_storage_multi_isolated.c, and cg_storage_multi_shared.c. This avoids
redefinition errors seen compiling for mips64el/musl-libc like:
In file included from progs/cg_storage_multi_egress_only.c:13:
In file included from progs/cg_storage_multi.h:6:
In file included from /usr/mips64el-linux-gnuabi64/include/asm/types.h:23:
/usr/include/asm-generic/int-l64.h:29:25: error: typedef redefinition with different types ('long' vs 'long long')
29 | typedef __signed__ long __s64;
| ^
/usr/include/asm-generic/int-ll64.h:30:44: note: previous definition is here
30 | __extension__ typedef __signed__ long long __s64;
| ^
Fixes: 9e5bd1f763 ("selftests/bpf: Test CGROUP_STORAGE map can't be used by multiple progs")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/4f4702e9f6115b7f84fea01b2326ca24c6df7ba8.1721713597.git.tony.ambardar@gmail.com
Remove a redundant include of '<linux/in6.h>', whose needed definitions are
already provided by 'test_progs.h'. This avoids errors seen compiling for
mips64el/musl-libc:
In file included from .../arpa/inet.h:9,
from ./test_progs.h:17,
from prog_tests/crypto_sanity.c:10:
.../netinet/in.h:23:8: error: redefinition of 'struct in6_addr'
23 | struct in6_addr {
| ^~~~~~~~
In file included from crypto_sanity.c:7:
.../linux/in6.h:33:8: note: originally defined here
33 | struct in6_addr {
| ^~~~~~~~
.../netinet/in.h:34:8: error: redefinition of 'struct sockaddr_in6'
34 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../linux/in6.h:50:8: note: originally defined here
50 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../netinet/in.h:42:8: error: redefinition of 'struct ipv6_mreq'
42 | struct ipv6_mreq {
| ^~~~~~~~~
.../linux/in6.h:60:8: note: originally defined here
60 | struct ipv6_mreq {
| ^~~~~~~~~
Fixes: 91541ab192 ("selftests: bpf: crypto skcipher algo selftests")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://lore.kernel.org/bpf/911293968f424ad7b462d8805aeb3baee8f4985b.1721713597.git.tony.ambardar@gmail.com
Remove a redundant include of '<linux/in6.h>', whose needed definitions are
already provided by 'test_progs.h'. This avoids errors seen compiling for
mips64el/musl-libc:
In file included from .../arpa/inet.h:9,
from ./test_progs.h:17,
from prog_tests/decap_sanity.c:9:
.../netinet/in.h:23:8: error: redefinition of 'struct in6_addr'
23 | struct in6_addr {
| ^~~~~~~~
In file included from decap_sanity.c:7:
.../linux/in6.h:33:8: note: originally defined here
33 | struct in6_addr {
| ^~~~~~~~
.../netinet/in.h:34:8: error: redefinition of 'struct sockaddr_in6'
34 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../linux/in6.h:50:8: note: originally defined here
50 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../netinet/in.h:42:8: error: redefinition of 'struct ipv6_mreq'
42 | struct ipv6_mreq {
| ^~~~~~~~~
.../linux/in6.h:60:8: note: originally defined here
60 | struct ipv6_mreq {
| ^~~~~~~~~
Fixes: 70a00e2f1d ("selftests/bpf: Test bpf_skb_adjust_room on CHECKSUM_PARTIAL")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/e986ba2d7edccd254b54f7cd049b98f10bafa8c3.1721713597.git.tony.ambardar@gmail.com
Remove a redundant include of '<linux/icmp.h>' which is already provided in
'lwt_helpers.h'. This avoids errors seen compiling for mips64el/musl-libc:
In file included from .../arpa/inet.h:9,
from lwt_redirect.c:51:
.../netinet/in.h:23:8: error: redefinition of 'struct in6_addr'
23 | struct in6_addr {
| ^~~~~~~~
In file included from .../linux/icmp.h:24,
from lwt_redirect.c:50:
.../linux/in6.h:33:8: note: originally defined here
33 | struct in6_addr {
| ^~~~~~~~
.../netinet/in.h:34:8: error: redefinition of 'struct sockaddr_in6'
34 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../linux/in6.h:50:8: note: originally defined here
50 | struct sockaddr_in6 {
| ^~~~~~~~~~~~
.../netinet/in.h:42:8: error: redefinition of 'struct ipv6_mreq'
42 | struct ipv6_mreq {
| ^~~~~~~~~
.../linux/in6.h:60:8: note: originally defined here
60 | struct ipv6_mreq {
| ^~~~~~~~~
Fixes: 43a7c3ef8a ("selftests/bpf: Add lwt_xmit tests for BPF_REDIRECT")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/3869dda876d5206d2f8d4dd67331c739ceb0c7f8.1721713597.git.tony.ambardar@gmail.com
The type 'loff_t' is a GNU extension and not exposed by the musl 'fcntl.h'
header unless _GNU_SOURCE is defined. Add this definition to fix errors
seen compiling for mips64el/musl-libc:
In file included from tools/testing/selftests/bpf/prog_tests/core_reloc.c:4:
./bpf_testmod/bpf_testmod.h:10:9: error: unknown type name 'loff_t'
10 | loff_t off;
| ^~~~~~
./bpf_testmod/bpf_testmod.h:16:9: error: unknown type name 'loff_t'
16 | loff_t off;
| ^~~~~~
Fixes: 6bcd39d366 ("selftests/bpf: Add CO-RE relocs selftest relying on kernel module BTF")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/11c3af75a7eb6bcb7ad9acfae6a6f470c572eb82.1721713597.git.tony.ambardar@gmail.com
The GNU version of 'struct tcp_info' in 'netinet/tcp.h' is not exposed by
musl headers unless _GNU_SOURCE is defined.
Add this definition to fix errors seen compiling for mips64el/musl-libc:
tcp_rtt.c: In function 'wait_for_ack':
tcp_rtt.c:24:25: error: storage size of 'info' isn't known
24 | struct tcp_info info;
| ^~~~
tcp_rtt.c:24:25: error: unused variable 'info' [-Werror=unused-variable]
cc1: all warnings being treated as errors
Fixes: 1f4f80fed2 ("selftests/bpf: test_progs: convert test_tcp_rtt")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/f2329767b15df206f08a5776d35a47c37da855ae.1721713597.git.tony.ambardar@gmail.com
The GNU version of 'struct tcphdr' has members 'doff', 'source' and 'dest',
which are not exposed by musl libc headers unless _GNU_SOURCE is defined.
Add this definition to fix errors seen compiling for mips64el/musl-libc:
flow_dissector.c:118:30: error: 'struct tcphdr' has no member named 'doff'
118 | .tcp.doff = 5,
| ^~~~
flow_dissector.c:119:30: error: 'struct tcphdr' has no member named 'source'
119 | .tcp.source = 80,
| ^~~~~~
flow_dissector.c:120:30: error: 'struct tcphdr' has no member named 'dest'
120 | .tcp.dest = 8080,
| ^~~~
Fixes: ae173a9157 ("selftests/bpf: support BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/8f7ab21a73f678f9cebd32b26c444a686e57414d.1721713597.git.tony.ambardar@gmail.com
The GNU version of 'struct tcphdr' with member 'doff' is not exposed by
musl headers unless _GNU_SOURCE is defined. Add this definition to fix
errors seen compiling for mips64el/musl-libc:
In file included from kfree_skb.c:2:
kfree_skb.c: In function 'on_sample':
kfree_skb.c:45:30: error: 'struct tcphdr' has no member named 'doff'
45 | if (CHECK(pkt_v6->tcp.doff != 5, "check_tcp",
| ^
Fixes: 580d656d80 ("selftests/bpf: Add kfree_skb raw_tp test")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/e2d8cedc790959c10d6822a51f01a7a3616bea1b.1721713597.git.tony.ambardar@gmail.com
The GNU version of 'struct tcphdr', with members 'doff' and 'urg_ptr', is
not exposed by musl headers unless _GNU_SOURCE is defined.
Add this definition to fix errors seen compiling for mips64el/musl-libc:
parse_tcp_hdr_opt.c:18:21: error: 'struct tcphdr' has no member named 'urg_ptr'
18 | .pk6_v6.tcp.urg_ptr = 123,
| ^~~~~~~
parse_tcp_hdr_opt.c:19:21: error: 'struct tcphdr' has no member named 'doff'
19 | .pk6_v6.tcp.doff = 9, /* 16 bytes of options */
| ^~~~
Fixes: cfa7b01189 ("selftests/bpf: tests for using dynptrs to parse skb and xdp buffers")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/ac5440213c242c62cb4e0d9e0a9cd5058b6a31f6.1721713597.git.tony.ambardar@gmail.com
Update ns_current_pid_tgid.c to use '#include <fcntl.h>' and avoid compile
error against mips64el/musl libc:
In file included from .../prog_tests/ns_current_pid_tgid.c:14:
.../include/sys/fcntl.h:1:2: error: #warning redirecting incorrect #include <sys/fcntl.h> to <fcntl.h> [-Werror=cpp]
1 | #warning redirecting incorrect #include <sys/fcntl.h> to <fcntl.h>
| ^~~~~~~
cc1: all warnings being treated as errors
Fixes: 09c02d553c ("bpf, selftests: Fold test_current_pid_tgid_new_ns into test_progs.")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/8bdc869749177b575025bf69600a4ce591822609.1721713597.git.tony.ambardar@gmail.com
Explicitly include '<linux/build_bug.h>' to fix errors seen compiling with
gcc targeting mips64el/musl-libc:
user_ringbuf.c: In function 'test_user_ringbuf_loop':
user_ringbuf.c:426:9: error: implicit declaration of function 'BUILD_BUG_ON' [-Werror=implicit-function-declaration]
426 | BUILD_BUG_ON(total_samples <= c_max_entries);
| ^~~~~~~~~~~~
cc1: all warnings being treated as errors
Fixes: e5a9df51c7 ("selftests/bpf: Add selftests validating the user ringbuf")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/b28575f9221ec54871c46a2e87612bb4bbf46ccd.1721713597.git.tony.ambardar@gmail.com
Include <limits.h> in 'bench.h' to provide a UINT_MAX definition and avoid
multiple compile errors against mips64el/musl-libc like:
benchs/bench_local_storage.c: In function 'parse_arg':
benchs/bench_local_storage.c:40:38: error: 'UINT_MAX' undeclared (first use in this function)
40 | if (ret < 1 || ret > UINT_MAX) {
| ^~~~~~~~
benchs/bench_local_storage.c:11:1: note: 'UINT_MAX' is defined in header '<limits.h>'; did you forget to '#include <limits.h>'?
10 | #include <test_btf.h>
+++ |+#include <limits.h>
11 |
seen with bench_local_storage.c, bench_local_storage_rcu_tasks_trace.c, and
bench_bpf_hashmap_lookup.c.
Fixes: 7308748925 ("selftests/bpf: Add benchmark for local_storage get")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/8f64a9d9fcff40a7fca090a65a68a9b62a468e16.1721713597.git.tony.ambardar@gmail.com
Add a "bpf_util.h" include to avoid the following error seen compiling for
mips64el with musl libc:
bench.c: In function 'find_benchmark':
bench.c:590:25: error: implicit declaration of function 'ARRAY_SIZE' [-Werror=implicit-function-declaration]
590 | for (i = 0; i < ARRAY_SIZE(benchs); i++) {
| ^~~~~~~~~~
cc1: all warnings being treated as errors
Fixes: 8e7c2a023a ("selftests/bpf: Add benchmark runner infrastructure")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/bc4dde77dfcd17a825d8f28f72f3292341966810.1721713597.git.tony.ambardar@gmail.com
The addition of general support for unprivileged tests in test_loader.c
breaks building test_verifier on non-glibc (e.g. musl) systems, due to the
inclusion of glibc extension '<error.h>' in 'unpriv_helpers.c'. However,
the header is actually not needed, so remove it to restore building.
Similarly for sk_lookup.c and flow_dissector.c, error.h is not necessary
and causes problems, so drop them.
Fixes: 1d56ade032 ("selftests/bpf: Unprivileged tests for test_loader.c")
Fixes: 0ab5539f85 ("selftests/bpf: Tests for BPF_SK_LOOKUP attach point")
Fixes: 0905beec9f ("selftests/bpf: run flow dissector tests in skb-less mode")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/5664367edf5fea4f3f4b4aec3b182bcfc6edff9c.1721713597.git.tony.ambardar@gmail.com
Existing code calls getsockname() with a 'struct sockaddr_in6 *' argument
where a 'struct sockaddr *' argument is declared, yielding compile errors
when building for mips64el/musl-libc:
bpf_iter_setsockopt.c: In function 'get_local_port':
bpf_iter_setsockopt.c:98:30: error: passing argument 2 of 'getsockname' from incompatible pointer type [-Werror=incompatible-pointer-types]
98 | if (!getsockname(fd, &addr, &addrlen))
| ^~~~~
| |
| struct sockaddr_in6 *
In file included from .../netinet/in.h:10,
from .../arpa/inet.h:9,
from ./test_progs.h:17,
from bpf_iter_setsockopt.c:5:
.../sys/socket.h:391:23: note: expected 'struct sockaddr * restrict' but argument is of type 'struct sockaddr_in6 *'
391 | int getsockname (int, struct sockaddr *__restrict, socklen_t *__restrict);
| ^
cc1: all warnings being treated as errors
This compiled under glibc only because the argument is declared to be a
"funky" transparent union which includes both types above. Explicitly cast
the argument to allow compiling for both musl and glibc.
Fixes: eed92afdd1 ("bpf: selftest: Test batching and bpf_(get|set)sockopt in bpf tcp iter")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Geliang Tang <geliang@kernel.org>
Link: https://lore.kernel.org/bpf/f41def0f17b27a23b1709080e4e3f37f4cc11ca9.1721713597.git.tony.ambardar@gmail.com
Cast 'rlim_t' argument to match expected type of printf() format and avoid
compile errors seen building for mips64el/musl-libc:
In file included from map_tests/sk_storage_map.c:20:
map_tests/sk_storage_map.c: In function 'test_sk_storage_map_stress_free':
map_tests/sk_storage_map.c:414:56: error: format '%lu' expects argument of type 'long unsigned int', but argument 2 has type 'rlim_t' {aka 'long long unsigned int'} [-Werror=format=]
414 | CHECK(err, "setrlimit(RLIMIT_NOFILE)", "rlim_new:%lu errno:%d",
| ^~~~~~~~~~~~~~~~~~~~~~~
415 | rlim_new.rlim_cur, errno);
| ~~~~~~~~~~~~~~~~~
| |
| rlim_t {aka long long unsigned int}
./test_maps.h:12:24: note: in definition of macro 'CHECK'
12 | printf(format); \
| ^~~~~~
map_tests/sk_storage_map.c:414:68: note: format string is defined here
414 | CHECK(err, "setrlimit(RLIMIT_NOFILE)", "rlim_new:%lu errno:%d",
| ~~^
| |
| long unsigned int
| %llu
cc1: all warnings being treated as errors
Fixes: 51a0e301a5 ("bpf: Add BPF_MAP_TYPE_SK_STORAGE test to test_maps")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/1e00a1fa7acf91b4ca135c4102dc796d518bad86.1721713597.git.tony.ambardar@gmail.com
Use pid_t rather than __pid_t when allocating memory for 'worker_pids' in
'struct test_env', as this is its declared type and also avoids compile
errors seen building against musl libc on mipsel64:
test_progs.c:1738:49: error: '__pid_t' undeclared (first use in this function); did you mean 'pid_t'?
1738 | env.worker_pids = calloc(sizeof(__pid_t), env.workers);
| ^~~~~~~
| pid_t
test_progs.c:1738:49: note: each undeclared identifier is reported only once for each function it appears in
Fixes: 91b2c0afd0 ("selftests/bpf: Add parallelism to test_progs")
Signed-off-by: Tony Ambardar <tony.ambardar@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Geliang Tang <geliang@kernel.org>
Link: https://lore.kernel.org/bpf/c6447da51a94babc1931711a43e2ceecb135c93d.1721713597.git.tony.ambardar@gmail.com
Eduard Zingerman says:
====================
no_caller_saved_registers attribute for helper calls
This patch-set seeks to allow using no_caller_saved_registers gcc/clang
attribute with some BPF helper functions (and kfuncs in the future).
As documented in [1], this attribute means that function scratches
only some of the caller saved registers defined by ABI.
For BPF the set of such registers could be defined as follows:
- R0 is scratched only if function is non-void;
- R1-R5 are scratched only if corresponding parameter type is defined
in the function prototype.
The goal of the patch-set is to implement no_caller_saved_registers
(nocsr for short) in a backwards compatible manner:
- for kernels that support the feature, gain some performance boost
from better register allocation;
- for kernels that don't support the feature, allow programs execution
with minor performance losses.
To achieve this, use a scheme suggested by Alexei Starovoitov:
- for nocsr calls clang allocates registers as-if relevant r0-r5
registers are not scratched by the call;
- as a post-processing step, clang visits each nocsr call and adds
spill/fill for every live r0-r5;
- stack offsets used for spills/fills are allocated as lowest
stack offsets in whole function and are not used for any other
purpose;
- when kernel loads a program, it looks for such patterns
(nocsr function surrounded by spills/fills) and checks if
spill/fill stack offsets are used exclusively in nocsr patterns;
- if so, and if current JIT inlines the call to the nocsr function
(e.g. a helper call), kernel removes unnecessary spill/fill pairs;
- when old kernel loads a program, presence of spill/fill pairs
keeps BPF program valid, albeit slightly less efficient.
Corresponding clang/llvm changes are available in [2].
The patch-set uses bpf_get_smp_processor_id() function as a canary,
making it the first helper with nocsr attribute.
For example, consider the following program:
#define __no_csr __attribute__((no_caller_saved_registers))
#define SEC(name) __attribute__((section(name), used))
#define bpf_printk(fmt, ...) bpf_trace_printk((fmt), sizeof(fmt), __VA_ARGS__)
typedef unsigned int __u32;
static long (* const bpf_trace_printk)(const char *fmt, __u32 fmt_size, ...) = (void *) 6;
static __u32 (*const bpf_get_smp_processor_id)(void) __no_csr = (void *)8;
SEC("raw_tp")
int test(void *ctx)
{
__u32 task = bpf_get_smp_processor_id();
bpf_printk("ctx=%p, smp=%d", ctx, task);
return 0;
}
char _license[] SEC("license") = "GPL";
Compiled (using [2]) as follows:
$ clang --target=bpf -O2 -g -c -o nocsr.bpf.o nocsr.bpf.c
$ llvm-objdump --no-show-raw-insn -Sd nocsr.bpf.o
...
3rd parameter for printk call removable spill/fill pair
.--- 0: r3 = r1 |
; | __u32 task = bpf_get_smp_processor_id(); |
| 1: *(u64 *)(r10 - 0x8) = r3 <----------|
| 2: call 0x8 |
| 3: r3 = *(u64 *)(r10 - 0x8) <----------'
; | bpf_printk("ctx=%p, smp=%d", ctx, task);
| 4: r1 = 0x0 ll
| 6: r2 = 0xf
| 7: r4 = r0
'--> 8: call 0x6
; return 0;
9: r0 = 0x0
10: exit
Here is how the program looks after verifier processing:
# bpftool prog load ./nocsr.bpf.o /sys/fs/bpf/nocsr-test
# bpftool prog dump xlated pinned /sys/fs/bpf/nocsr-test
int test(void * ctx):
0: (bf) r3 = r1 <--- 3rd printk parameter
; __u32 task = bpf_get_smp_processor_id();
1: (b4) w0 = 197324 <--. inlined helper call,
2: (bf) r0 = &(void __percpu *)(r0) <--- spill/fill
3: (61) r0 = *(u32 *)(r0 +0) <--' pair removed
; bpf_printk("ctx=%p, smp=%d", ctx, task);
4: (18) r1 = map[id:5][0]+0
6: (b7) r2 = 15
7: (bf) r4 = r0
8: (85) call bpf_trace_printk#-125920
; return 0;
9: (b7) r0 = 0
10: (95) exit
[1] https://clang.llvm.org/docs/AttributeReference.html#no-caller-saved-registers
[2] https://github.com/eddyz87/llvm-project/tree/bpf-no-caller-saved-registers
Change list:
- v3 -> v4:
- When nocsr spills/fills are removed in the subprogram, allow these
spills/fills to reside in [-MAX_BPF_STACK-48..MAX_BPF_STACK) range
(suggested by Alexei);
- Dropped patches with special handling for bpf_probe_read_kernel()
(requested by Alexei);
- Reset aux .nocsr_pattern and .nocsr_spills_num fields in
check_nocsr_stack_contract() (requested by Andrii).
Andrii, I have not added an additional flag to
struct bpf_subprog_info, it currently does not have holes
and I really don't like adding a bool field there just as an
alternative indicator that nocsr is disabled.
Indicator at the moment:
- nocsr_stack_off >= S16_MIN means that nocsr rewrite is enabled;
- nocsr_stack_off == S16_MIN means that nocsr rewrite is disabled.
- v2 -> v3:
- As suggested by Andrii, 'nocsr_stack_off' is no longer checked at
rewrite time, instead mark_nocsr_patterns() now does two passes
over BPF program:
- on a first pass it computes the lowest stack spill offset for
the subprogram;
- on a second pass this offset is used to recognize nocsr pattern.
- As suggested by Alexei, a new mechanic is added to work around a
situation mentioned by Andrii, when more helper functions are
marked as nocsr at compile time than current kernel supports:
- all {spill*,helper call,fill*} patterns are now marked as
insn_aux_data[*].nocsr_pattern, thus relaxing failure condition
for check_nocsr_stack_contract();
- spill/fill pairs are not removed for patterns where helper can't
be inlined;
- see mark_nocsr_pattern_for_call() for details an example.
- As suggested by Alexei, subprogram stack depth is now adjusted
if all spill/fill pairs could be removed. This adjustment has
to take place before optimize_bpf_loop(), hence the rewrite
is moved from do_misc_fixups() to remove_nocsr_spills_fills()
(again).
- As suggested by Andrii, special measures are taken to work around
bpf_probe_read_kernel() access to BPF stack, see patches 11, 12.
Patch #11 is very simplistic, a more comprehensive solution would
be to change the type of the third parameter of the
bpf_probe_read_kernel() from ARG_ANYTHING to something else and
not only check nocsr contract, but also propagate stack slot
liveness information. However, such change would require update in
struct bpf_call_arg_meta processing, which currently implies that
every memory parameter is followed by a size parameter.
I can work on these changes, please comment.
- Stylistic changes suggested by Andrii.
- Added acks from Andrii.
- Dropped RFC tag.
- v1 -> v2:
- assume that functions inlined by either jit or verifier
conform to no_caller_saved_registers contract (Andrii, Puranjay);
- allow nocsr rewrite for bpf_get_smp_processor_id()
on arm64 and riscv64 architectures (Puranjay);
- __arch_{x86_64,arm64,riscv64} macro for test_loader;
- moved remove_nocsr_spills_fills() inside do_misc_fixups() (Andrii);
- moved nocsr pattern detection from check_cfg() to a separate pass
(Andrii);
- various stylistic/correctness changes according to Andrii's
comments.
Revisions:
- v1 https://lore.kernel.org/bpf/20240629094733.3863850-1-eddyz87@gmail.com/
- v2 https://lore.kernel.org/bpf/20240704102402.1644916-1-eddyz87@gmail.com/
- v3 https://lore.kernel.org/bpf/20240715230201.3901423-1-eddyz87@gmail.com/
====================
Link: https://lore.kernel.org/r/20240722233844.1406874-1-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Tests for no_caller_saved_registers processing logic
(see verifier.c:match_and_mark_nocsr_pattern()):
- a canary positive test case;
- a canary test case for arm64 and riscv64;
- various tests with broken patterns;
- tests with read/write fixed/varying stack access that violate nocsr
stack access contract;
- tests with multiple subprograms;
- tests using nocsr in combination with may_goto/bpf_loop,
as all of these features affect stack depth;
- tests for nocsr stack spills below max stack depth.
Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20240722233844.1406874-11-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Add annotations __arch_x86_64, __arch_arm64, __arch_riscv64
to specify on which architecture the test case should be tested.
Several __arch_* annotations could be specified at once.
When test case is not run on current arch it is marked as skipped.
For example, the following would be tested only on arm64 and riscv64:
SEC("raw_tp")
__arch_arm64
__arch_riscv64
__xlated("1: *(u64 *)(r10 - 16) = r1")
__xlated("2: call")
__xlated("3: r1 = *(u64 *)(r10 - 16);")
__success
__naked void canary_arm64_riscv64(void)
{
asm volatile (
"r1 = 1;"
"*(u64 *)(r10 - 16) = r1;"
"call %[bpf_get_smp_processor_id];"
"r1 = *(u64 *)(r10 - 16);"
"exit;"
:
: __imm(bpf_get_smp_processor_id)
: __clobber_all);
}
On x86 it would be skipped:
#467/2 verifier_nocsr/canary_arm64_riscv64:SKIP
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20240722233844.1406874-10-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Add a macro __xlated("...") for use with test_loader tests.
When such annotations are present for the test case:
- bpf_prog_get_info_by_fd() is used to get BPF program after all
rewrites are applied by verifier.
- the program is disassembled and patterns specified in __xlated are
searched for in the disassembly text.
__xlated matching follows the same mechanics as __msg:
each subsequent pattern is matched from the point where
previous pattern ended.
This allows to write tests like below, where the goal is to verify the
behavior of one of the of the transformations applied by verifier:
SEC("raw_tp")
__xlated("1: w0 = ")
__xlated("2: r0 = &(void __percpu *)(r0)")
__xlated("3: r0 = *(u32 *)(r0 +0)")
__xlated("4: exit")
__success __naked void simple(void)
{
asm volatile (
"call %[bpf_get_smp_processor_id];"
"exit;"
:
: __imm(bpf_get_smp_processor_id)
: __clobber_all);
}
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20240722233844.1406874-9-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
