IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Write data to fd by calling "vdprintf", in most implementations
of the standard library, the data is finally written by the writev syscall.
But "uprobe_events/kprobe_events" does not allow segmented writes,
so switch the "append_to_file" function to explicit write() call.
Signed-off-by: Liu Pan <patteliu@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230320030720.650-1-patteliu@gmail.com
Unlike normal libbpf the light skeleton 'loader' program is doing
btf_find_by_name_kind() call at run-time to find ksym in the kernel and
populate its {btf_id, btf_obj_fd} pair in ld_imm64 insn. To avoid doing the
search multiple times for the same ksym it remembers the first patched ld_imm64
insn and copies {btf_id, btf_obj_fd} from it into subsequent ld_imm64 insn.
Fix a bug in copying logic, since it may incorrectly clear BPF_PSEUDO_BTF_ID flag.
Also replace always true if (btf_obj_fd >= 0) check with unconditional JMP_JA
to clarify the code.
Fixes: d995816b77 ("libbpf: Avoid reload of imm for weak, unresolved, repeating ksym")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230319203014.55866-1-alexei.starovoitov@gmail.com
This patch documents overview of libbpf, including its features for
developing BPF programs.
Signed-off-by: Sreevani Sreejith <ssreevani@meta.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/bpf/20230315195405.2051559-1-ssreevani@meta.com
Currently, test_progs outputs all stdout/stderr as it runs, and when it
is done, prints a summary.
It is non-trivial for tooling to parse that output and extract meaningful
information from it.
This change adds a new option, `--json-summary`/`-J` that let the caller
specify a file where `test_progs{,-no_alu32}` can write a summary of the
run in a json format that can later be parsed by tooling.
Currently, it creates a summary section with successes/skipped/failures
followed by a list of failed tests and subtests.
A test contains the following fields:
- name: the name of the test
- number: the number of the test
- message: the log message that was printed by the test.
- failed: A boolean indicating whether the test failed or not. Currently
we only output failed tests, but in the future, successful tests could
be added.
- subtests: A list of subtests associated with this test.
A subtest contains the following fields:
- name: same as above
- number: sanme as above
- message: the log message that was printed by the subtest.
- failed: same as above but for the subtest
An example run and json content below:
```
$ sudo ./test_progs -a $(grep -v '^#' ./DENYLIST.aarch64 | awk '{print
$1","}' | tr -d '\n') -j -J /tmp/test_progs.json
$ jq < /tmp/test_progs.json | head -n 30
{
"success": 29,
"success_subtest": 23,
"skipped": 3,
"failed": 28,
"results": [
{
"name": "bpf_cookie",
"number": 10,
"message": "test_bpf_cookie:PASS:skel_open 0 nsec\n",
"failed": true,
"subtests": [
{
"name": "multi_kprobe_link_api",
"number": 2,
"message": "kprobe_multi_link_api_subtest:PASS:load_kallsyms 0 nsec\nlibbpf: extern 'bpf_testmod_fentry_test1' (strong): not resolved\nlibbpf: failed to load object 'kprobe_multi'\nlibbpf: failed to load BPF skeleton 'kprobe_multi': -3\nkprobe_multi_link_api_subtest:FAIL:fentry_raw_skel_load unexpected error: -3\n",
"failed": true
},
{
"name": "multi_kprobe_attach_api",
"number": 3,
"message": "libbpf: extern 'bpf_testmod_fentry_test1' (strong): not resolved\nlibbpf: failed to load object 'kprobe_multi'\nlibbpf: failed to load BPF skeleton 'kprobe_multi': -3\nkprobe_multi_attach_api_subtest:FAIL:fentry_raw_skel_load unexpected error: -3\n",
"failed": true
},
{
"name": "lsm",
"number": 8,
"message": "lsm_subtest:PASS:lsm.link_create 0 nsec\nlsm_subtest:FAIL:stack_mprotect unexpected stack_mprotect: actual 0 != expected -1\n",
"failed": true
}
```
The file can then be used to print a summary of the test run and list of
failing tests/subtests:
```
$ jq -r < /tmp/test_progs.json '"Success: \(.success)/\(.success_subtest), Skipped: \(.skipped), Failed: \(.failed)"'
Success: 29/23, Skipped: 3, Failed: 28
$ jq -r < /tmp/test_progs.json '.results | map([
if .failed then "#\(.number) \(.name)" else empty end,
(
. as {name: $tname, number: $tnum} | .subtests | map(
if .failed then "#\($tnum)/\(.number) \($tname)/\(.name)" else empty end
)
)
]) | flatten | .[]' | head -n 20
#10 bpf_cookie
#10/2 bpf_cookie/multi_kprobe_link_api
#10/3 bpf_cookie/multi_kprobe_attach_api
#10/8 bpf_cookie/lsm
#15 bpf_mod_race
#15/1 bpf_mod_race/ksym (used_btfs UAF)
#15/2 bpf_mod_race/kfunc (kfunc_btf_tab UAF)
#36 cgroup_hierarchical_stats
#61 deny_namespace
#61/1 deny_namespace/unpriv_userns_create_no_bpf
#73 fexit_stress
#83 get_func_ip_test
#99 kfunc_dynptr_param
#99/1 kfunc_dynptr_param/dynptr_data_null
#99/4 kfunc_dynptr_param/dynptr_data_null
#100 kprobe_multi_bench_attach
#100/1 kprobe_multi_bench_attach/kernel
#100/2 kprobe_multi_bench_attach/modules
#101 kprobe_multi_test
#101/1 kprobe_multi_test/skel_api
```
Signed-off-by: Manu Bretelle <chantr4@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230317163256.3809328-1-chantr4@gmail.com
Alexei Starovoitov says:
====================
From: Alexei Starovoitov <ast@kernel.org>
Allow BPF programs detect at load time whether particular kfunc exists.
Patch 1: Allow ld_imm64 to point to kfunc in the kernel.
Patch 2: Fix relocation of kfunc in ld_imm64 insn when kfunc is in kernel module.
Patch 3: Introduce bpf_ksym_exists() macro.
Patch 4: selftest.
NOTE: detection of kfuncs from light skeleton is not supported yet.
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Add load and run time test for bpf_ksym_exists() and check that the verifier
performs dead code elimination for non-existing kfunc.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Martin KaFai Lau <martin.lau@kernel.org>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20230317201920.62030-5-alexei.starovoitov@gmail.com
Introduce bpf_ksym_exists() macro that can be used by BPF programs
to detect at load time whether particular ksym (either variable or kfunc)
is present in the kernel.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230317201920.62030-4-alexei.starovoitov@gmail.com
void *p = kfunc; -> generates ld_imm64 insn.
kfunc() -> generates bpf_call insn.
libbpf patches bpf_call insn correctly while only btf_id part of ld_imm64 is
set in the former case. Which means that pointers to kfuncs in modules are not
patched correctly and the verifier rejects load of such programs due to btf_id
being out of range. Fix libbpf to patch ld_imm64 for kfunc.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230317201920.62030-3-alexei.starovoitov@gmail.com
Allow ld_imm64 insn with BPF_PSEUDO_BTF_ID to hold the address of kfunc. The
ld_imm64 pointing to a valid kfunc will be seen as non-null PTR_TO_MEM by
is_branch_taken() logic of the verifier, while libbpf will resolve address to
unknown kfunc as ld_imm64 reg, 0 which will also be recognized by
is_branch_taken() and the verifier will proceed dead code elimination. BPF
programs can use this logic to detect at load time whether kfunc is present in
the kernel with bpf_ksym_exists() macro that is introduced in the next patches.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Martin KaFai Lau <martin.lau@kernel.org>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20230317201920.62030-2-alexei.starovoitov@gmail.com
Commit d56b0c461d ("bpf, docs: Fix link to netdev-FAQ target")
attempts to fix linking problem to undefined "netdev-FAQ" label
introduced in 287f4fa99a ("docs: Update references to netdev-FAQ")
by changing internal cross reference to netdev subsystem documentation
(Documentation/process/maintainer-netdev.rst) to external one at
docs.kernel.org. However, the linking problem is still not
resolved, as the generated link points to non-existent netdev-FAQ
section of the external doc, which when clicked, will instead going
to the top of the doc.
Revert back to internal linking by simply mention the doc path while
massaging the leading text to the link, since the netdev subsystem
doc contains no FAQs but rather general information about the subsystem.
Fixes: d56b0c461d ("bpf, docs: Fix link to netdev-FAQ target")
Fixes: 287f4fa99a ("docs: Update references to netdev-FAQ")
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20230314074449.23620-1-bagasdotme@gmail.com
Moving find_kallsyms_symbol_value from kernel/module/internal.h to
include/linux/module.h. The reason is that internal.h is not prepared to
be included when CONFIG_MODULES=n. find_kallsyms_symbol_value is used by
kernel/bpf/verifier.c and including internal.h from it (without modules)
leads into a compilation error:
In file included from ../include/linux/container_of.h:5,
from ../include/linux/list.h:5,
from ../include/linux/timer.h:5,
from ../include/linux/workqueue.h:9,
from ../include/linux/bpf.h:10,
from ../include/linux/bpf-cgroup.h:5,
from ../kernel/bpf/verifier.c:7:
../kernel/bpf/../module/internal.h: In function 'mod_find':
../include/linux/container_of.h:20:54: error: invalid use of undefined type 'struct module'
20 | static_assert(__same_type(*(ptr), ((type *)0)->member) || \
| ^~
[...]
This patch fixes the above error.
Fixes: 31bf1dbccf ("bpf: Fix attaching fentry/fexit/fmod_ret/lsm to modules")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Viktor Malik <vmalik@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/oe-kbuild-all/202303161404.OrmfCy09-lkp@intel.com/
Link: https://lore.kernel.org/bpf/20230317095601.386738-1-vmalik@redhat.com
Alexander Lobakin says:
====================
Enabling skb PP recycling revealed a couple issues in the bpf_test_run
code. Recycling broke the assumption that the headroom won't ever be
touched during the test_run execution: xdp_scrub_frame() invalidates the
XDP frame at the headroom start, while neigh xmit code overwrites 2 bytes
to the left of the Ethernet header. The first makes the kernel panic in
certain cases, while the second breaks xdp_do_redirect selftest on BE.
test_run is a limited-scope entity, so let's hope no more corner cases
will happen here or at least they will be as easy and pleasant to fix
as those two.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Alexei noticed xdp_do_redirect test on BPF CI started failing on
BE systems after skb PP recycling was enabled:
test_xdp_do_redirect:PASS:prog_run 0 nsec
test_xdp_do_redirect:PASS:pkt_count_xdp 0 nsec
test_xdp_do_redirect:PASS:pkt_count_zero 0 nsec
test_xdp_do_redirect:FAIL:pkt_count_tc unexpected pkt_count_tc: actual
220 != expected 9998
test_max_pkt_size:PASS:prog_run_max_size 0 nsec
test_max_pkt_size:PASS:prog_run_too_big 0 nsec
close_netns:PASS:setns 0 nsec
#289 xdp_do_redirect:FAIL
Summary: 270/1674 PASSED, 30 SKIPPED, 1 FAILED
and it doesn't happen on LE systems.
Ilya then hunted it down to:
#0 0x0000000000aaeee6 in neigh_hh_output (hh=0x83258df0,
skb=0x88142200) at linux/include/net/neighbour.h:503
#1 0x0000000000ab2cda in neigh_output (skip_cache=false,
skb=0x88142200, n=<optimized out>) at linux/include/net/neighbour.h:544
#2 ip6_finish_output2 (net=net@entry=0x88edba00, sk=sk@entry=0x0,
skb=skb@entry=0x88142200) at linux/net/ipv6/ip6_output.c:134
#3 0x0000000000ab4cbc in __ip6_finish_output (skb=0x88142200, sk=0x0,
net=0x88edba00) at linux/net/ipv6/ip6_output.c:195
#4 ip6_finish_output (net=0x88edba00, sk=0x0, skb=0x88142200) at
linux/net/ipv6/ip6_output.c:206
xdp_do_redirect test places a u32 marker (0x42) right before the Ethernet
header to check it then in the XDP program and return %XDP_ABORTED if it's
not there. Neigh xmit code likes to round up hard header length to speed
up copying the header, so it overwrites two bytes in front of the Eth
header. On LE systems, 0x42 is one byte at `data - 4`, while on BE it's
`data - 1`, what explains why it happens only there.
It didn't happen previously due to that %XDP_PASS meant the page will be
discarded and replaced by a new one, but now it can be recycled as well,
while bpf_test_run code doesn't reinitialize the content of recycled
pages. This mark is limited to this particular test and its setup though,
so there's no need to predict 1000 different possible cases. Just move
it 4 bytes to the left, still keeping it 32 bit to match on more bytes.
Fixes: 9c94bbf9a8 ("xdp: recycle Page Pool backed skbs built from XDP frames")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/CAADnVQ+B_JOU+EpP=DKhbY9yXdN6GiRPnpTTXfEZ9sNkUeb-yQ@mail.gmail.com
Reported-by: Ilya Leoshkevich <iii@linux.ibm.com> # + debugging
Link: https://lore.kernel.org/bpf/8341c1d9f935f410438e79d3bd8a9cc50aefe105.camel@linux.ibm.com
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Tested-by: Ilya Leoshkevich <iii@linux.ibm.com>
Link: https://lore.kernel.org/r/20230316175051.922550-3-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
syzbot and Ilya faced the splats when %XDP_PASS happens for bpf_test_run
after skb PP recycling was enabled for {__,}xdp_build_skb_from_frame():
BUG: kernel NULL pointer dereference, address: 0000000000000d28
RIP: 0010:memset_erms+0xd/0x20 arch/x86/lib/memset_64.S:66
[...]
Call Trace:
<TASK>
__finalize_skb_around net/core/skbuff.c:321 [inline]
__build_skb_around+0x232/0x3a0 net/core/skbuff.c:379
build_skb_around+0x32/0x290 net/core/skbuff.c:444
__xdp_build_skb_from_frame+0x121/0x760 net/core/xdp.c:622
xdp_recv_frames net/bpf/test_run.c:248 [inline]
xdp_test_run_batch net/bpf/test_run.c:334 [inline]
bpf_test_run_xdp_live+0x1289/0x1930 net/bpf/test_run.c:362
bpf_prog_test_run_xdp+0xa05/0x14e0 net/bpf/test_run.c:1418
[...]
This happens due to that it calls xdp_scrub_frame(), which nullifies
xdpf->data. bpf_test_run code doesn't reinit the frame when the XDP
program doesn't adjust head or tail. Previously, %XDP_PASS meant the
page will be released from the pool and returned to the MM layer, but
now it does return to the Pool with the nullified xdpf->data, which
doesn't get reinitialized then.
So, in addition to checking whether the head and/or tail have been
adjusted, check also for a potential XDP frame corruption. xdpf->data
is 100% affected and also xdpf->flags is the field closest to the
metadata / frame start. Checking for these two should be enough for
non-extreme cases.
Fixes: 9c94bbf9a8 ("xdp: recycle Page Pool backed skbs built from XDP frames")
Reported-by: syzbot+e1d1b65f7c32f2a86a9f@syzkaller.appspotmail.com
Link: https://lore.kernel.org/bpf/000000000000f1985705f6ef2243@google.com
Reported-by: Ilya Leoshkevich <iii@linux.ibm.com>
Link: https://lore.kernel.org/bpf/e07dd94022ad5731705891b9487cc9ed66328b94.camel@linux.ibm.com
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Tested-by: Ilya Leoshkevich <iii@linux.ibm.com>
Link: https://lore.kernel.org/r/20230316175051.922550-2-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
For every BPF_ADD/SUB involving a pointer, adjust_ptr_min_max_vals()
ensures that the resulting pointer has a constant offset if
bypass_spec_v1 is false. This is ensured by calling sanitize_check_bounds()
which in turn calls check_stack_access_for_ptr_arithmetic(). There,
-EACCESS is returned if the register's offset is not constant, thereby
rejecting the program.
In summary, an unprivileged user must never be able to create stack
pointers with a variable offset. That is also the case, because a
respective check in check_stack_write() is missing. If they were able
to create a variable-offset pointer, users could still use it in a
stack-write operation to trigger unsafe speculative behavior [1].
Because unprivileged users must already be prevented from creating
variable-offset stack pointers, viable options are to either remove
this check (replacing it with a clarifying comment), or to turn it
into a "verifier BUG"-message, also adding a similar check in
check_stack_write() (for consistency, as a second-level defense).
This patch implements the first option to reduce verifier bloat.
This check was introduced by commit 01f810ace9 ("bpf: Allow
variable-offset stack access") which correctly notes that
"variable-offset reads and writes are disallowed (they were already
disallowed for the indirect access case) because the speculative
execution checking code doesn't support them". However, it does not
further discuss why the check in check_stack_read() is necessary.
The code which made this check obsolete was also introduced in this
commit.
I have compiled ~650 programs from the Linux selftests, Linux samples,
Cilium, and libbpf/examples projects and confirmed that none of these
trigger the check in check_stack_read() [2]. Instead, all of these
programs are, as expected, already rejected when constructing the
variable-offset pointers. Note that the check in
check_stack_access_for_ptr_arithmetic() also prints "off=%d" while the
code removed by this patch does not (the error removed does not appear
in the "verification_error" values). For reproducibility, the
repository linked includes the raw data and scripts used to create
the plot.
[1] https://arxiv.org/pdf/1807.03757.pdf
[2] 53dc19fcf4/data/plots/23-02-26_23-56_bpftool/bpftool/0004-errors.pdf
Fixes: 01f810ace9 ("bpf: Allow variable-offset stack access")
Signed-off-by: Luis Gerhorst <gerhorst@cs.fau.de>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20230315165358.23701-1-gerhorst@cs.fau.de
David Vernet says:
====================
The struct bpf_cpumask type is currently not RCU safe. It uses the
bpf_mem_cache_{alloc,free}() APIs to allocate and release cpumasks, and
those allocations may be reused before an RCU grace period has elapsed.
We want to be able to enable using this pattern in BPF programs:
private(MASK) static struct bpf_cpumask __kptr *global;
int BPF_PROG(prog, ...)
{
struct bpf_cpumask *cpumask;
bpf_rcu_read_lock();
cpumask = global;
if (!cpumask) {
bpf_rcu_read_unlock();
return -1;
}
bpf_cpumask_setall(cpumask);
...
bpf_rcu_read_unlock();
}
In other words, to be able to pass a kptr to KF_RCU bpf_cpumask kfuncs
without requiring the acquisition and release of refcounts using
bpf_cpumask_kptr_get(). This patchset enables this by making the struct
bpf_cpumask type RCU safe, and removing the bpf_cpumask_kptr_get()
function.
---
v1: https://lore.kernel.org/all/20230316014122.678082-2-void@manifault.com/
Changelog:
----------
v1 -> v2:
- Add doxygen comment for new @rcu field in struct bpf_cpumask.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Now that the kfunc no longer exists, we can remove it and instead
describe how RCU can be used to get a struct bpf_cpumask from a map
value. This patch updates the BPF documentation accordingly.
Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230316054028.88924-6-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Now that struct bpf_cpumask is RCU safe, there's no need for this kfunc.
Rather than doing the following:
private(MASK) static struct bpf_cpumask __kptr *global;
int BPF_PROG(prog, s32 cpu, ...)
{
struct bpf_cpumask *cpumask;
bpf_rcu_read_lock();
cpumask = bpf_cpumask_kptr_get(&global);
if (!cpumask) {
bpf_rcu_read_unlock();
return -1;
}
bpf_cpumask_setall(cpumask);
...
bpf_cpumask_release(cpumask);
bpf_rcu_read_unlock();
}
Programs can instead simply do (assume same global cpumask):
int BPF_PROG(prog, ...)
{
struct bpf_cpumask *cpumask;
bpf_rcu_read_lock();
cpumask = global;
if (!cpumask) {
bpf_rcu_read_unlock();
return -1;
}
bpf_cpumask_setall(cpumask);
...
bpf_rcu_read_unlock();
}
In other words, no extra atomic acquire / release, and less boilerplate
code.
This patch removes both the kfunc, as well as its selftests and
documentation.
Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230316054028.88924-5-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Now that struct bpf_cpumask * is considered an RCU-safe type according
to the verifier, we should add tests that validate its common usages.
This patch adds those tests to the cpumask test suite. A subsequent
changes will remove bpf_cpumask_kptr_get(), and will adjust the selftest
and BPF documentation accordingly.
Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230316054028.88924-4-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
struct bpf_cpumask is a BPF-wrapper around the struct cpumask type which
can be instantiated by a BPF program, and then queried as a cpumask in
similar fashion to normal kernel code. The previous patch in this series
makes the type fully RCU safe, so the type can be included in the
rcu_protected_type BTF ID list.
A subsequent patch will remove bpf_cpumask_kptr_get(), as it's no longer
useful now that we can just treat the type as RCU safe by default and do
our own if check.
Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230316054028.88924-3-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The struct bpf_cpumask type uses the bpf_mem_cache_{alloc,free}() APIs
to allocate and free its cpumasks. The bpf_mem allocator may currently
immediately reuse some memory when its freed, without waiting for an RCU
read cycle to elapse. We want to be able to treat struct bpf_cpumask
objects as completely RCU safe.
This is necessary for two reasons:
1. bpf_cpumask_kptr_get() currently does an RCU-protected
refcnt_inc_not_zero(). This of course assumes that the underlying
memory is not reused, and is therefore unsafe in its current form.
2. We want to be able to get rid of bpf_cpumask_kptr_get() entirely, and
intead use the superior kptr RCU semantics now afforded by the
verifier.
This patch fixes (1), and enables (2), by making struct bpf_cpumask RCU
safe. A subsequent patch will update the verifier to allow struct
bpf_cpumask * pointers to be passed to KF_RCU kfuncs, and then a latter
patch will remove bpf_cpumask_kptr_get().
Fixes: 516f4d3397 ("bpf: Enable cpumasks to be queried and used as kptrs")
Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230316054028.88924-2-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Some consumers of libbpf compile the code base with different warnings
enabled. In a report for perf, for example, -Wpacked was set which
caused warnings about "inefficient alignment" to be emitted on a subset
of supported architectures.
With this change we silence specifically those warnings, as we intentionally
worked with packed structs.
This is a similar resolution as in b2f10cd4e8 ("perf cpumap: Fix alignment
for masks in event encoding").
Fixes: 1eebcb6063 ("libbpf: Implement basic zip archive parsing support")
Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/bpf/CA+G9fYtBnwxAWXi2+GyNByApxnf_DtP1-6+_zOKAdJKnJBexjg@mail.gmail.com/
Link: https://lore.kernel.org/bpf/20230315171550.1551603-1-deso@posteo.net
In __start_server, it leaks a fd when setsockopt(SO_REUSEPORT) fails.
This patch fixes it.
Fixes: eed92afdd1 ("bpf: selftest: Test batching and bpf_(get|set)sockopt in bpf tcp iter")
Reported-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20230316000726.1016773-2-martin.lau@linux.dev
In tcp_hdr_options test, it ensures the received tcp hdr option
and the sk local storage have the expected values. It uses memcmp
to check that. Testing the memcmp result with ASSERT_OK is confusing
because ASSERT_OK will print out the errno which is not set.
This patch uses ASSERT_EQ to check for 0 instead.
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20230316000726.1016773-1-martin.lau@linux.dev
Viktor Malik says:
====================
I noticed that the verifier behaves incorrectly when attaching to fentry
of multiple functions of the same name located in different modules (or
in vmlinux). The reason for this is that if the target program is not
specified, the verifier will search kallsyms for the trampoline address
to attach to. The entire kallsyms is always searched, not respecting the
module in which the function to attach to is located.
As Yonghong correctly pointed out, there is yet another issue - the
trampoline acquires the module reference in register_fentry which means
that if the module is unloaded between the place where the address is
found in the verifier and register_fentry, it is possible that another
module is loaded to the same address in the meantime, which may lead to
errors.
This patch fixes the above issues by extracting the module name from the
BTF of the attachment target (which must be specified) and by doing the
search in kallsyms of the correct module. At the same time, the module
reference is acquired right after the address is found and only released
right before the program itself is unloaded.
---
Changes in v10:
- added the new test to DENYLIST.aarch64 (suggested by Andrii)
- renamed the test source file to match the test name
Changes in v9:
- two small changes suggested by Jiri Olsa and Jiri's ack
Changes in v8:
- added module_put to error paths in bpf_check_attach_target after the
module reference is acquired
Changes in v7:
- refactored the module reference manipulation (comments by Jiri Olsa)
- cleaned up the test (comments by Andrii Nakryiko)
Changes in v6:
- storing the module reference inside bpf_prog_aux instead of
bpf_trampoline and releasing it when the program is unloaded
(suggested by Jiri Olsa)
Changes in v5:
- fixed acquiring and releasing of module references by trampolines to
prevent modules being unloaded between address lookup and trampoline
allocation
Changes in v4:
- reworked module kallsyms lookup approach using existing functions,
verifier now calls btf_try_get_module to retrieve the module and
find_kallsyms_symbol_value to get the symbol address (suggested by
Alexei)
- included Jiri Olsa's comments
- improved description of the new test and added it as a comment into
the test source
Changes in v3:
- added trivial implementation for kallsyms_lookup_name_in_module() for
!CONFIG_MODULES (noticed by test robot, fix suggested by Hao Luo)
Changes in v2:
- introduced and used more space-efficient kallsyms lookup function,
suggested by Jiri Olsa
- included Hao Luo's comments
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Adds a new test that tries to attach a program to fentry of two
functions of the same name, one located in vmlinux and the other in
bpf_testmod.
To avoid conflicts with existing tests, a new function
"bpf_fentry_shadow_test" was created both in vmlinux and in bpf_testmod.
The previous commit fixed a bug which caused this test to fail. The
verifier would always use the vmlinux function's address as the target
trampoline address, hence trying to create two trampolines for a single
address, which is forbidden.
The test (similarly to other fentry/fexit tests) is not working on arm64
at the moment.
Signed-off-by: Viktor Malik <vmalik@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/5fe2f364190b6f79b085066ed7c5989c5bc475fa.1678432753.git.vmalik@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This resolves two problems with attachment of fentry/fexit/fmod_ret/lsm
to functions located in modules:
1. The verifier tries to find the address to attach to in kallsyms. This
is always done by searching the entire kallsyms, not respecting the
module in which the function is located. Such approach causes an
incorrect attachment address to be computed if the function to attach
to is shadowed by a function of the same name located earlier in
kallsyms.
2. If the address to attach to is located in a module, the module
reference is only acquired in register_fentry. If the module is
unloaded between the place where the address is found
(bpf_check_attach_target in the verifier) and register_fentry, it is
possible that another module is loaded to the same address which may
lead to potential errors.
Since the attachment must contain the BTF of the program to attach to,
we extract the module from it and search for the function address in the
correct module (resolving problem no. 1). Then, the module reference is
taken directly in bpf_check_attach_target and stored in the bpf program
(in bpf_prog_aux). The reference is only released when the program is
unloaded (resolving problem no. 2).
Signed-off-by: Viktor Malik <vmalik@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/3f6a9d8ae850532b5ef864ef16327b0f7a669063.1678432753.git.vmalik@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The commit 332ea1f697 ("bpf: Add bpf_cgroup_from_id() kfunc") added
bpf_cgroup_from_id() which calls current_cgns_cgroup_dfl() through
cgroup_get_from_id(). However, BPF programs may be attached to a point where
current->nsproxy has already been cleared to NULL by exit_task_namespace()
and calling bpf_cgroup_from_id() would cause an oops.
Just return the system-wide root if nsproxy has been cleared. This allows
all cgroups to be looked up after the task passed through
exit_task_namespace(), which semantically makes sense. Given that the only
way to get this behavior is through BPF programs, it seems safe but let's
see what others think.
Fixes: 332ea1f697 ("bpf: Add bpf_cgroup_from_id() kfunc")
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/ZBDuVWiFj2jiz3i8@slm.duckdns.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
LLVM commit https://reviews.llvm.org/D143726 introduced hoistMinMax optimization
that transformed
(i < VIRTIO_MAX_SGS) && (i < out_sgs)
into
i < MIN(VIRTIO_MAX_SGS, out_sgs)
and caused the verifier to stop recognizing such loop as bounded.
Which resulted in the following test failure:
libbpf: prog 'trace_virtqueue_add_sgs': BPF program load failed: Bad address
libbpf: prog 'trace_virtqueue_add_sgs': -- BEGIN PROG LOAD LOG --
The sequence of 8193 jumps is too complex.
verification time 789206 usec
stack depth 56
processed 156446 insns (limit 1000000) max_states_per_insn 7 total_states 1746 peak_states 1701 mark_read 12
-- END PROG LOAD LOG --
libbpf: prog 'trace_virtqueue_add_sgs': failed to load: -14
libbpf: failed to load object 'loop6.bpf.o'
Workaround the verifier limitation for now with inline asm that
prevents this particular optimization.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Alexander Lobakin says:
====================
Yeah, I still remember that "Who needs cpumap nowadays" (c), but anyway.
__xdp_build_skb_from_frame() missed the moment when the networking stack
became able to recycle skb pages backed by a page_pool. This was making
e.g. cpumap redirect even less effective than simple %XDP_PASS. veth was
also affected in some scenarios.
A lot of drivers use skb_mark_for_recycle() already, it's been almost
two years and seems like there are no issues in using it in the generic
code too. {__,}xdp_release_frame() can be then removed as it losts its
last user.
Page Pool becomes then zero-alloc (or almost) in the abovementioned
cases, too. Other memory type models (who needs them at this point)
have no changes.
Some numbers on 1 Xeon Platinum core bombed with 27 Mpps of 64-byte
IPv6 UDP, iavf w/XDP[0] (CONFIG_PAGE_POOL_STATS is enabled):
Plain %XDP_PASS on baseline, Page Pool driver:
src cpu Rx drops dst cpu Rx
2.1 Mpps N/A 2.1 Mpps
cpumap redirect (cross-core, w/o leaving its NUMA node) on baseline:
6.8 Mpps 5.0 Mpps 1.8 Mpps
cpumap redirect with skb PP recycling:
7.9 Mpps 5.7 Mpps 2.2 Mpps
+22% (from cpumap redir on baseline)
[0] https://github.com/alobakin/linux/commits/iavf-xdp
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
__xdp_build_skb_from_frame() was the last user of
{__,}xdp_release_frame(), which detaches pages from the page_pool.
All the consumers now recycle Page Pool skbs and page, except mlx5,
stmmac and tsnep drivers, which use page_pool_release_page() directly
(might change one day). It's safe to assume this functionality is not
needed anymore and can be removed (in favor of recycling).
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20230313215553.1045175-5-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
__xdp_build_skb_from_frame() state(d):
/* Until page_pool get SKB return path, release DMA here */
Page Pool got skb pages recycling in April 2021, but missed this
function.
xdp_release_frame() is relevant only for Page Pool backed frames and it
detaches the page from the corresponding page_pool in order to make it
freeable via page_frag_free(). It can instead just mark the output skb
as eligible for recycling if the frame is backed by a pp. No change for
other memory model types (the same condition check as before).
cpumap redirect and veth on Page Pool drivers now become zero-alloc (or
almost).
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20230313215553.1045175-4-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
skb_mark_for_recycle() is guarded with CONFIG_PAGE_POOL, this creates
unneeded complication when using it in the generic code. For now, it's
only used in the drivers always selecting Page Pool, so this works.
Move the guards so that preprocessor will cut out only the operation
itself and the function will still be a noop on !PAGE_POOL systems,
but available there as well.
No functional changes.
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/oe-kbuild-all/202303020342.Wi2PRFFH-lkp@intel.com
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20230313215553.1045175-3-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, the test relies on that only dropped ("xmitted") frames will
be recycled and if a frame became an skb, it will be freed later by the
stack and never come back to its page_pool.
So, it easily gets broken by trying to recycle skbs[0]:
test_xdp_do_redirect:PASS:pkt_count_xdp 0 nsec
test_xdp_do_redirect:FAIL:pkt_count_zero unexpected pkt_count_zero:
actual 9936 != expected 2
test_xdp_do_redirect:PASS:pkt_count_tc 0 nsec
That huge mismatch happened because after the TC ingress hook zeroes the
magic, the page gets recycled when skb is freed, not returned to the MM
layer. "Live frames" mode initializes only new pages and keeps the
recycled ones as is by design, so they appear with zeroed magic on the
Rx path again.
Expand the possible magic values from two: 0 (was "xmitted"/dropped or
did hit the TC hook) and 0x42 (hit the input XDP prog) to three: the new
one will mark frames hit the TC hook, so that they will elide both
@pkt_count_zero and @pkt_count_xdp. They can then be recycled to their
page_pool or returned to the page allocator, this won't affect the
counters anyhow. Just make sure to mark them as "input" (0x42) when they
appear on the Rx path again.
Also make an enum from those magics, so that they will be always visible
and can be changed in just one place anytime. This also eases adding any
new marks later on.
Link: https://github.com/kernel-patches/bpf/actions/runs/4386538411/jobs/7681081789
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20230313215553.1045175-2-aleksander.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The verifier rejects the code:
bpf_strncmp(task->comm, 16, "my_task");
with the message:
16: (85) call bpf_strncmp#182
R1 type=trusted_ptr_ expected=fp, pkt, pkt_meta, map_key, map_value, mem, ringbuf_mem, buf
Teach the verifier that such access pattern is safe.
Do not allow untrusted and legacy ptr_to_btf_id to be passed into helpers.
Reported-by: David Vernet <void@manifault.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230313235845.61029-3-alexei.starovoitov@gmail.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
bpf_strncmp() doesn't write into its first argument.
Make sure that the verifier knows about it.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230313235845.61029-2-alexei.starovoitov@gmail.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Improve clarity by adding an example of a signed comparison instruction
Signed-off-by: Dave Thaler <dthaler@microsoft.com>
Acked-by: David Vernet <void@manifault.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/r/20230310233814.4641-1-dthaler1968@googlemail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The canonical location for the tracefs filesystem is at
/sys/kernel/tracing.
But, from Documentation/trace/ftrace.rst:
Before 4.1, all ftrace tracing control files were within the debugfs
file system, which is typically located at /sys/kernel/debug/tracing.
For backward compatibility, when mounting the debugfs file system,
the tracefs file system will be automatically mounted at:
/sys/kernel/debug/tracing
Many tests in the bpf selftest code still refer to this older debugfs
path, so let's update them to avoid confusion.
Signed-off-by: Ross Zwisler <zwisler@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20230313205628.1058720-3-zwisler@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The canonical location for the tracefs filesystem is at /sys/kernel/tracing.
But, from Documentation/trace/ftrace.rst:
Before 4.1, all ftrace tracing control files were within the debugfs
file system, which is typically located at /sys/kernel/debug/tracing.
For backward compatibility, when mounting the debugfs file system,
the tracefs file system will be automatically mounted at:
/sys/kernel/debug/tracing
Many comments and samples in the bpf code still refer to this older
debugfs path, so let's update them to avoid confusion. There are a few
spots where the bpf code explicitly checks both tracefs and debugfs
(tools/bpf/bpftool/tracelog.c and tools/lib/api/fs/fs.c) and I've left
those alone so that the tools can continue to work with both paths.
Signed-off-by: Ross Zwisler <zwisler@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20230313205628.1058720-2-zwisler@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When a local kptr is stashed in a map and freed when the map goes away,
currently an error like the below appears:
[ 39.195695] BUG: using smp_processor_id() in preemptible [00000000] code: kworker/u32:15/2875
[ 39.196549] caller is bpf_mem_free+0x56/0xc0
[ 39.196958] CPU: 15 PID: 2875 Comm: kworker/u32:15 Tainted: G O 6.2.0-13016-g22df776a9a86 #4477
[ 39.197897] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[ 39.198949] Workqueue: events_unbound bpf_map_free_deferred
[ 39.199470] Call Trace:
[ 39.199703] <TASK>
[ 39.199911] dump_stack_lvl+0x60/0x70
[ 39.200267] check_preemption_disabled+0xbf/0xe0
[ 39.200704] bpf_mem_free+0x56/0xc0
[ 39.201032] ? bpf_obj_new_impl+0xa0/0xa0
[ 39.201430] bpf_obj_free_fields+0x1cd/0x200
[ 39.201838] array_map_free+0xad/0x220
[ 39.202193] ? finish_task_switch+0xe5/0x3c0
[ 39.202614] bpf_map_free_deferred+0xea/0x210
[ 39.203006] ? lockdep_hardirqs_on_prepare+0xe/0x220
[ 39.203460] process_one_work+0x64f/0xbe0
[ 39.203822] ? pwq_dec_nr_in_flight+0x110/0x110
[ 39.204264] ? do_raw_spin_lock+0x107/0x1c0
[ 39.204662] ? lockdep_hardirqs_on_prepare+0xe/0x220
[ 39.205107] worker_thread+0x74/0x7a0
[ 39.205451] ? process_one_work+0xbe0/0xbe0
[ 39.205818] kthread+0x171/0x1a0
[ 39.206111] ? kthread_complete_and_exit+0x20/0x20
[ 39.206552] ret_from_fork+0x1f/0x30
[ 39.206886] </TASK>
This happens because the call to __bpf_obj_drop_impl I added in the patch
adding support for stashing local kptrs doesn't disable migration. Prior
to that patch, __bpf_obj_drop_impl logic only ran when called by a BPF
progarm, whereas now it can be called from map free path, so it's
necessary to explicitly disable migration.
Also, refactor a bit to just call __bpf_obj_drop_impl directly instead
of bothering w/ dtor union and setting pointer-to-obj_drop.
Fixes: c8e1875409 ("bpf: Support __kptr to local kptrs")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230313214641.3731908-1-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In commit 3fbd7ee285 ("tasks: Add a count of task RCU users"), a
count on the number of RCU users was added to struct task_struct. This
was done so as to enable the removal of task_rcu_dereference(), and
allow tasks to be protected by RCU even after exiting and being removed
from the runqueue. In this commit, the 'refcount_t rcu_users' field that
keeps track of this refcount was put into a union co-located with
'struct rcu_head rcu', so as to avoid taking up any extra space in
task_struct. This was possible to do safely, because the field was only
ever decremented by a static set of specific callers, and then never
incremented again.
While this restriction of there only being a small, static set of users
of this field has worked fine, it prevents us from leveraging the field
to use RCU to protect tasks in other contexts.
During tracing, for example, it would be useful to be able to collect
some tasks that performed a certain operation, put them in a map, and
then periodically summarize who they are, which cgroup they're in, how
much CPU time they've utilized, etc. While this can currently be done
with 'usage', it becomes tricky when a task is already in a map, or if a
reference should only be taken if a task is valid and will not soon be
reaped. Ideally, we could do something like pass a reference to a map
value, and then try to acquire a reference to the task in an RCU read
region by using refcount_inc_not_zero().
Similarly, in sched_ext, schedulers are using integer pids to remember
tasks, and then looking them up with find_task_by_pid_ns(). This is
slow, error prone, and adds complexity. It would be more convenient and
performant if BPF schedulers could instead store tasks directly in maps,
and then leverage RCU to ensure they can be safely accessed with low
overhead.
Finally, overloading fields like this is error prone. Someone that wants
to use 'rcu_users' could easily overlook the fact that once the rcu
callback is scheduled, the refcount will go back to being nonzero, thus
precluding the use of refcount_inc_not_zero(). Furthermore, as described
below, it's possible to extract the fields of the union without changing
the size of task_struct.
There are several possible ways to enable this:
1. The lightest touch approach is likely the one proposed in this patch,
which is to simply extract 'rcu_users' and 'rcu' from the union, so
that scheduling the 'rcu' callback doesn't overwrite the 'rcu_users'
refcount. If we have a trusted task pointer, this would allow us to
use refcnt_inc_not_zero() inside of an RCU region to determine if we
can safely acquire a reference to the task and store it in a map. As
mentioned below, this can be done without changing the size of
task_struct, by moving the location of the union to another location
that has padding gaps we can fill in.
2. Removing 'refcount_t rcu_users', and instead having the entire task
be freed in an rcu callback. This is likely the most sound overall
design, though it changes the behavioral semantics exposed to
callers, who currently expect that a task that's successfully looked
up in e.g. the pid_list with find_task_by_pid_ns(), can always have a
'usage' reference acquired on them, as it's guaranteed to be >
0 until after the next gp. In order for this approach to work, we'd
have to audit all callers. This approach also slightly changes
behavior observed by user space by not invoking
trace_sched_process_free() until the whole task_struct is actually being
freed, rather than just after it's exited. It also may change
timings, as memory will be freed in an RCU callback rather than
immediately when the final 'usage' refcount drops to 0. This also is
arguably a benefit, as it provides more predictable performance to
callers who are refcounting tasks.
3. There may be other solutions as well that don't require changing the
layout of task_struct. For example, we could possibly do something
complex from the BPF side, such as listen for task exit and remove a
task from a map when the task is exiting. This would likely require
significant custom handling for task_struct in the verifier, so a
more generalizable solution is likely warranted.
As mentioned above, this patch proposes the lightest-touch approach
which allows callers elsewhere in the kernel to use 'rcu_users' to
ensure the lifetime of a task, by extracting 'rcu_users' and 'rcu' from
the union. There is no size change in task_struct with this patch.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Signed-off-by: David Vernet <void@manifault.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lore.kernel.org/r/20230215233033.889644-1-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Fix wrong order of frame index vs register/slot index in precision
propagation verbose (level 2) output. It's wrong and very confusing as is.
Fixes: 529409ea92 ("bpf: propagate precision across all frames, not just the last one")
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20230313184017.4083374-1-andrii@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Dave Marchevsky says:
====================
Local kptrs are kptrs allocated via bpf_obj_new with a type specified in program
BTF. A BPF program which creates a local kptr has exclusive control of the
lifetime of the kptr, and, prior to terminating, must:
* free the kptr via bpf_obj_drop
* If the kptr is a {list,rbtree} node, add the node to a {list, rbtree},
thereby passing control of the lifetime to the collection
This series adds a third option:
* stash the kptr in a map value using bpf_kptr_xchg
As indicated by the use of "stash" to describe this behavior, the intended use
of this feature is temporary storage of local kptrs. For example, a sched_ext
([0]) scheduler may want to create an rbtree node for each new cgroup on cgroup
init, but to add that node to the rbtree as part of a separate program which
runs on enqueue. Stashing the node in a map_value allows its lifetime to outlive
the execution of the cgroup_init program.
Behavior:
There is no semantic difference between adding a kptr to a graph collection and
"stashing" it in a map. In both cases exclusive ownership of the kptr's lifetime
is passed to some containing data structure, which is responsible for
bpf_obj_drop'ing it when the container goes away.
Since graph collections also expect exclusive ownership of the nodes they
contain, graph nodes cannot be both stashed in a map_value and contained by
their corresponding collection.
Implementation:
Two observations simplify the verifier changes for this feature. First, kptrs
("referenced kptrs" until a recent renaming) require registration of a
dtor function as part of their acquire/release semantics, so that a referenced
kptr which is placed in a map_value is properly released when the map goes away.
We want this exact behavior for local kptrs, but with bpf_obj_drop as the dtor
instead of a per-btf_id dtor.
The second observation is that, in terms of identification, "referenced kptr"
and "local kptr" already don't interfere with one another. Consider the
following example:
struct node_data {
long key;
long data;
struct bpf_rb_node node;
};
struct map_value {
struct node_data __kptr *node;
};
struct {
__uint(type, BPF_MAP_TYPE_ARRAY);
__type(key, int);
__type(value, struct map_value);
__uint(max_entries, 1);
} some_nodes SEC(".maps");
struct map_value *mapval;
struct node_data *res;
int key = 0;
res = bpf_obj_new(typeof(*res));
if (!res) { /* err handling */ }
mapval = bpf_map_lookup_elem(&some_nodes, &key);
if (!mapval) { /* err handling */ }
res = bpf_kptr_xchg(&mapval->node, res);
if (res)
bpf_obj_drop(res);
The __kptr tag identifies map_value's node as a referenced kptr, while the
PTR_TO_BTF_ID which bpf_obj_new returns - a type in some non-vmlinux,
non-module BTF - identifies res as a local kptr. Type tag on the pointer
indicates referenced kptr, while the type of the pointee indicates local kptr.
So using existing facilities we can tell the verifier about a "referenced kptr"
pointer to a "local kptr" pointee.
When kptr_xchg'ing a kptr into a map_value, the verifier can recognize local
kptr types and treat them like referenced kptrs with a properly-typed
bpf_obj_drop as a dtor.
Other implementation notes:
* We don't need to do anything special to enforce "graph nodes cannot be
both stashed in a map_value and contained by their corresponding collection"
* bpf_kptr_xchg both returns and takes as input a (possibly-null) owning
reference. It does not accept non-owning references as input by virtue
of requiring a ref_obj_id. By definition, if a program has an owning
ref to a node, the node isn't in a collection, so it's safe to pass
ownership via bpf_kptr_xchg.
Summary of patches:
* Patch 1 modifies BTF plumbing to support using bpf_obj_drop as a dtor
* Patch 2 adds verifier plumbing to support MEM_ALLOC-flagged param for
bpf_kptr_xchg
* Patch 3 adds selftests exercising the new behavior
Changelog:
v1 -> v2: https://lore.kernel.org/bpf/20230309180111.1618459-1-davemarchevsky@fb.com/
Patch #s used below refer to the patch's position in v1 unless otherwise
specified.
Patches 1-3 were applied and are not included in v2.
Rebase onto latest bpf-next: "libbpf: Revert poisoning of strlcpy"
Patch 4: "bpf: Support __kptr to local kptrs"
* Remove !btf_is_kernel(btf) check, WARN_ON_ONCE instead (Alexei)
Patch 6: "selftests/bpf: Add local kptr stashing test"
* Add test which stashes 2 nodes and later unstashes one of them using a
separate BPF program (Alexei)
* Fix incorrect runner subtest name for original test (was
"rbtree_add_nodes")
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a new selftest, local_kptr_stash, which uses bpf_kptr_xchg to stash
a bpf_obj_new-allocated object in a map. Test the following scenarios:
* Stash two rb_nodes in an arraymap, don't unstash them, rely on map
free to destruct them
* Stash two rb_nodes in an arraymap, unstash the second one in a
separate program, rely on map free to destruct first
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230310230743.2320707-4-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The previous patch added necessary plumbing for verifier and runtime to
know what to do with non-kernel PTR_TO_BTF_IDs in map values, but didn't
provide any way to get such local kptrs into a map value. This patch
modifies verifier handling of bpf_kptr_xchg to allow MEM_ALLOC kptr
types.
check_reg_type is modified accept MEM_ALLOC-flagged input to
bpf_kptr_xchg despite such types not being in btf_ptr_types. This could
have been done with a MAYBE_MEM_ALLOC equivalent to MAYBE_NULL, but
bpf_kptr_xchg is the only helper that I can forsee using
MAYBE_MEM_ALLOC, so keep it special-cased for now.
The verifier tags bpf_kptr_xchg retval MEM_ALLOC if and only if the BTF
associated with the retval is not kernel BTF.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230310230743.2320707-3-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
If a PTR_TO_BTF_ID type comes from program BTF - not vmlinux or module
BTF - it must have been allocated by bpf_obj_new and therefore must be
free'd with bpf_obj_drop. Such a PTR_TO_BTF_ID is considered a "local
kptr" and is tagged with MEM_ALLOC type tag by bpf_obj_new.
This patch adds support for treating __kptr-tagged pointers to "local
kptrs" as having an implicit bpf_obj_drop destructor for referenced kptr
acquire / release semantics. Consider the following example:
struct node_data {
long key;
long data;
struct bpf_rb_node node;
};
struct map_value {
struct node_data __kptr *node;
};
struct {
__uint(type, BPF_MAP_TYPE_ARRAY);
__type(key, int);
__type(value, struct map_value);
__uint(max_entries, 1);
} some_nodes SEC(".maps");
If struct node_data had a matching definition in kernel BTF, the verifier would
expect a destructor for the type to be registered. Since struct node_data does
not match any type in kernel BTF, the verifier knows that there is no kfunc
that provides a PTR_TO_BTF_ID to this type, and that such a PTR_TO_BTF_ID can
only come from bpf_obj_new. So instead of searching for a registered dtor,
a bpf_obj_drop dtor can be assumed.
This allows the runtime to properly destruct such kptrs in
bpf_obj_free_fields, which enables maps to clean up map_vals w/ such
kptrs when going away.
Implementation notes:
* "kernel_btf" variable is renamed to "kptr_btf" in btf_parse_kptr.
Before this patch, the variable would only ever point to vmlinux or
module BTFs, but now it can point to some program BTF for local kptr
type. It's later used to populate the (btf, btf_id) pair in kptr btf
field.
* It's necessary to btf_get the program BTF when populating btf_field
for local kptr. btf_record_free later does a btf_put.
* Behavior for non-local referenced kptrs is not modified, as
bpf_find_btf_id helper only searches vmlinux and module BTFs for
matching BTF type. If such a type is found, btf_field_kptr's btf will
pass btf_is_kernel check, and the associated release function is
some one-argument dtor. If btf_is_kernel check fails, associated
release function is two-arg bpf_obj_drop_impl. Before this patch
only btf_field_kptr's w/ kernel or module BTFs were created.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230310230743.2320707-2-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add brief text about existence of helper functions, with details to go in
separate psABI text.
Note that text about runtime functions (kfuncs) is part of a separate patch,
not this one.
Signed-off-by: Dave Thaler <dthaler@microsoft.com>
Link: https://lore.kernel.org/r/20230308205303.1308-1-dthaler1968@googlemail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
