Relax tracing prog recursive attach rules
Subsystems (4)
v5 cover Relax tracing prog recursive attach rules
Currently, it's not allowed to attach an fentry/fexit prog to another fentry/fexit. At the same time it's not uncommon to see a tracing program with lots of logic in use, and the attachment limitation prevents usage of fentry/fexit for performance analysis (e.g. with "bpftool prog profile" command) in this case. An example could be falcosecurity libs project that uses tp_btf tracing programs for offloading certain part of logic into tail-called programs, but the use-case is still generic enough -- a tracing program could be complicated and heavy enough to warrant its profiling, yet frustratingly it's not possible to do so use best tooling for that.
Following the corresponding discussion [1], the reason for that is to avoid tracing progs call cycles without introducing more complex solutions. But currently it seems impossible to load and attach tracing programs in a way that will form such a cycle. Replace "no same type" requirement with verification that no more than one level of attachment nesting is allowed. In this way only one fentry/fexit program could be attached to another fentry/fexit to cover profiling use case, and still no cycle could be formed.
The series contains a test for recursive attachment, as well as a fix + test for an issue in re-attachment branch of bpf_tracing_prog_attach. When preparing the test for the main change set, I've stumbled upon the possibility to construct a sequence of events when attach_btf would be NULL while computing a trampoline key. It doesn't look like this issue is triggered by the main change, because the reproduces doesn't actually need to have an fentry attachment chain.
[1]: https://lore.kernel.org/bpf/[email protected]/
Dmitrii Dolgov (3):
bpf: Relax tracing prog recursive attach rules selftests/bpf: Add test for recursive attachment of tracing progs selftests/bpf: Test re-attachment fix for bpf_tracing_prog_attach
Jiri Olsa (1):
bpf: Fix re-attachment branch in bpf_tracing_prog_attach
include/linux/bpf.h | 1 + include/uapi/linux/bpf.h | 1 + kernel/bpf/syscall.c | 16 +++ kernel/bpf/verifier.c | 33 ++--- tools/include/uapi/linux/bpf.h | 1 + .../bpf/prog_tests/recursive_attach.c | 117 ++++++++++++++++++ .../selftests/bpf/progs/fentry_recursive.c | 19 +++ .../bpf/progs/fentry_recursive_target.c | 31 +++++ 8 files changed, 205 insertions(+), 14 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/recursive_attach.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive_target.c
base-commit: 40d0eb0259ae77ace3e81d7454d1068c38bc95c2
Signature · 1 line
2.41.0
PATCH v5 1/4 bpf: Relax tracing prog recursive attach rules
Currently, it's not allowed to attach an fentry/fexit prog to another one fentry/fexit. At the same time it's not uncommon to see a tracing program with lots of logic in use, and the attachment limitation prevents usage of fentry/fexit for performance analysis (e.g. with "bpftool prog profile" command) in this case. An example could be falcosecurity libs project that uses tp_btf tracing programs.
Following the corresponding discussion [1], the reason for that is to avoid tracing progs call cycles without introducing more complex solutions. But currently it seems impossible to load and attach tracing programs in a way that will form such a cycle. The limitation is coming from the fact that attach_prog_fd is specified at the prog load (thus making it impossible to attach to a program loaded after it in this way), as well as tracing progs not implementing link_detach.
Replace "no same type" requirement with verification that no more than one level of attachment nesting is allowed. In this way only one fentry/fexit program could be attached to another fentry/fexit to cover profiling use case, and still no cycle could be formed. To implement, add a new field into bpf_prog_aux to track the depth of attachment.
[1]: https://lore.kernel.org/bpf/[email protected]/
---
Previous discussion: https://lore.kernel.org/bpf/[email protected]/
Changes in v5:
- Remove follower_cnt and drop unreachable cycle prevention condition
- Allow only one level of attachment nesting
- Do not display attach_depth in bpftool, as it doesn't make sense anymore
Changes in v3:
- Fix incorrect decreasing of attach_depth, setting to 0 instead
- Place bookkeeping later, to not miss a cleanup if needed
- Display attach_depth in bpftool only if the value is not 0
Changes in v2:
- Verify tgt_prog is not null
- Replace boolean followed with number of followers, to handle multiple progs attaching/detaching
include/linux/bpf.h | 1 + include/uapi/linux/bpf.h | 1 + kernel/bpf/syscall.c | 7 +++++++ kernel/bpf/verifier.c | 33 +++++++++++++++++++-------------- tools/include/uapi/linux/bpf.h | 1 + 5 files changed, 29 insertions(+), 14 deletions(-)
include/linux/bpf.h +1 -0
@@ -1400,6 +1400,7 @@ struct bpf_prog_aux {u32 real_func_cnt;/* includes hidden progs, only used for JIT and freeing progs */ u32 func_idx;/* 0 for non-func prog, the index in func array for func prog */ u32 attach_btf_id;/* in-kernel BTF type id to attach to */ +u32 attach_depth;/* position of the prog in the attachment chain */ u32 ctx_arg_info_size;u32 max_rdonly_access;u32 max_rdwr_access;
include/uapi/linux/bpf.h +1 -0
@@ -6468,6 +6468,7 @@ struct bpf_prog_info {__u32 verified_insns;__u32 attach_btf_obj_id;__u32 attach_btf_id; +__u32 attach_depth; } __attribute__((aligned(8)));struct bpf_map_info {
kernel/bpf/syscall.c +7 -0
@@ -3039,6 +3039,7 @@ static void bpf_tracing_link_release(struct bpf_link *link)bpf_trampoline_put (tr_link->trampoline); + link->prog->aux->attach_depth =0 ;/* tgt_prog is NULL if target is a kernel function */ if (tr_link->tgt_prog)bpf_prog_put (tr_link->tgt_prog);
@@ -3243,6 +3244,11 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog, goto out_unlock; } +if (tgt_prog) { +/* Bookkeeping for managing the prog attachment chain. */ + prog->aux->attach_depth = tgt_prog->aux->attach_depth +1 ; + } + link->tgt_prog = tgt_prog; link->trampoline = tr;
@@ -4510,6 +4516,7 @@ static int bpf_prog_get_info_by_fd(struct file *file,if (prog->aux->btf) info.btf_id =btf_obj_id (prog->aux->btf); info.attach_btf_id = prog->aux->attach_btf_id; + info.attach_depth = prog->aux->attach_depth;if (attach_btf) info.attach_btf_obj_id =btf_obj_id (attach_btf);
kernel/bpf/verifier.c +19 -14
@@ -20109,6 +20109,11 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,if (tgt_prog) {struct bpf_prog_aux *aux = tgt_prog->aux; +if (aux->attach_depth >=1 ) { +bpf_log (log,"Cannot attach with more than one level of nesting\n" ); +return -EINVAL ; + } +if (bpf_prog_is_dev_bound (prog->aux) && !bpf_prog_dev_bound_match (prog, tgt_prog)) {bpf_log (log,"Target program bound device mismatch" );
@@ -20147,10 +20152,11 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,bpf_log (log,"Can attach to only JITed progs\n" );return -EINVAL ; } -if (tgt_prog->type == prog->type) { -/* Cannot fentry/fexit another fentry/fexit program. -* Cannot attach program extension to another extension. -* It's ok to attach fentry/fexit to extension program. +if (tgt_prog->type == prog->type && prog_extension) { +/* +* To avoid potential call chain cycles, prevent attaching of a +* program extension to another extension. It's ok to attach +* fentry/fexit to extension program. */ bpf_log (log,"Cannot recursively attach\n" );return -EINVAL ;
@@ -20163,16 +20169,15 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,* except fentry/fexit. The reason is the following. * The fentry/fexit programs are used for performance * analysis, stats and can be attached to any program -* type except themselves. When extension program is -* replacing XDP function it is necessary to allow -* performance analysis of all functions. Both original -* XDP program and its program extension. Hence -* attaching fentry/fexit to BPF_PROG_TYPE_EXT is -* allowed. If extending of fentry/fexit was allowed it -* would be possible to create long call chain -* fentry->extension->fentry->extension beyond -* reasonable stack size. Hence extending fentry is not -* allowed. +* type. When extension program is replacing XDP function +* it is necessary to allow performance analysis of all +* functions. Both original XDP program and its program +* extension. Hence attaching fentry/fexit to +* BPF_PROG_TYPE_EXT is allowed. If extending of +* fentry/fexit was allowed it would be possible to create +* long call chain fentry->extension->fentry->extension +* beyond reasonable stack size. Hence extending fentry +* is not allowed. */ bpf_log (log,"Cannot extend fentry/fexit\n" );return -EINVAL ;
tools/include/uapi/linux/bpf.h +1 -0
@@ -6468,6 +6468,7 @@ struct bpf_prog_info {__u32 verified_insns;__u32 attach_btf_obj_id;__u32 attach_btf_id; +__u32 attach_depth; } __attribute__((aligned(8)));struct bpf_map_info {
Signature · 1 line
2.41.0
PATCH v5 2/4 selftests/bpf: Add test for recursive attachment of tracing progs
In reply to Dmitrii DolgovVerify the fact that only one fentry prog could be attached to another fentry, building up an attachment chain of limited size. Use existing bpf_testmod as a start of the chain.
---
Changes in v5:
- Test only one level of attachment
.../bpf/prog_tests/recursive_attach.c | 69 +++++++++++++++++++ .../selftests/bpf/progs/fentry_recursive.c | 19 +++++ .../bpf/progs/fentry_recursive_target.c | 20 ++++++ 3 files changed, 108 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/recursive_attach.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive_target.c
tools/testing/selftests/bpf/prog_tests/recursive_attach.c new file +69 -0
@@ -0,0 +1,69 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2023 Red Hat, Inc. */ +#include <test_progs.h> +#include "fentry_recursive.skel.h" +#include "fentry_recursive_target.skel.h" +#include <bpf/btf.h> +#include "bpf/libbpf_internal.h" + +/* +* Test following scenarios: +* - attach one fentry progs to another one +* - more than one nesting levels are not allowed +*/ +void test_recursive_fentry_attach (void ) +{ +struct fentry_recursive_target *target_skel =NULL ; +struct fentry_recursive *tracing_chain[2 ] = {}; +struct bpf_program *prog; +int prev_fd, err; + + target_skel =fentry_recursive_target__open_and_load (); +if (!ASSERT_OK_PTR (target_skel,"fentry_recursive_target__open_and_load" )) + goto close_prog; + +/* Create an attachment chain with two fentry progs */ +for (int i =0 ; i <2 ; i++) { + tracing_chain[i] =fentry_recursive__open (); +if (!ASSERT_OK_PTR (tracing_chain[i],"fentry_recursive__open" )) + goto close_prog; + +/* +* The first prog in the chain is going to be attached to the target +* fentry program, the second one to the previous in the chain. +*/ +if (i ==0 ) { + prog = tracing_chain[0 ]->progs.recursive_attach; + prev_fd =bpf_program__fd (target_skel->progs.test1); + err =bpf_program__set_attach_target (prog, prev_fd,"test1" ); + }else { + prog = tracing_chain[i]->progs.recursive_attach; + prev_fd =bpf_program__fd (tracing_chain[i-1 ]->progs.recursive_attach); + err =bpf_program__set_attach_target (prog, prev_fd,"recursive_attach" ); + } + +if (!ASSERT_OK (err,"bpf_program__set_attach_target" )) + goto close_prog; + + err =fentry_recursive__load (tracing_chain[i]); +/* The first attach should succeed, the second fail */ +if (i ==0 ) { +if (!ASSERT_OK (err,"fentry_recursive__load" )) + goto close_prog; + + err =fentry_recursive__attach (tracing_chain[i]); +if (!ASSERT_OK (err,"fentry_recursive__attach" )) + goto close_prog; + }else { +if (!ASSERT_ERR (err,"fentry_recursive__load" )) + goto close_prog; + } + } + +close_prog: +fentry_recursive_target__destroy (target_skel); +for (int i =0 ; i <2 ; i++) { +if (tracing_chain[i]) +fentry_recursive__destroy (tracing_chain[i]); + } +}
tools/testing/selftests/bpf/progs/fentry_recursive.c new file +19 -0
@@ -0,0 +1,19 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2023 Red Hat, Inc. */ +#include <linux/bpf.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +char _license[] SEC("license") = "GPL"; + +__u64 test1_result =0 ; + +/* +* Dummy fentry bpf prog for testing fentry attachment chains +*/ +SEC("fentry/XXX") +int BPF_PROG (recursive_attach ,int a) +{ + test1_result = a ==1 ; +return 0 ; +}
tools/testing/selftests/bpf/progs/fentry_recursive_target.c new file +20 -0
@@ -0,0 +1,20 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2023 Red Hat, Inc. */ +#include <linux/bpf.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +char _license[] SEC("license") = "GPL"; + +__u64 test1_result =0 ; + +/* +* Dummy fentry bpf prog for testing fentry attachment chains. It's going to be +* a start of the chain. +*/ +SEC("fentry/bpf_testmod_fentry_test1") +int BPF_PROG (test1 ,int a) +{ + test1_result = a ==1 ; +return 0 ; +}
Signature · 1 line
2.41.0
PATCH v5 3/4 bpf: Fix re-attachment branch in bpf_tracing_prog_attach
In reply to Dmitrii DolgovFrom: Jiri Olsa <[email protected]>
The following case can cause a crash due to missing attach_btf:
1) load rawtp program
2) load fentry program with rawtp as target_fd
3) create tracing link for fentry program with target_fd = 0
4) repeat 3
In the end we have:
- prog->aux->dst_trampoline == NULL
- tgt_prog == NULL (because we did not provide target_fd to link_create)
- prog->aux->attach_btf == NULL (the program was loaded with attach_prog_fd=X)
- the program was loaded for tgt_prog but we have no way to find out which one
BUG: kernel NULL pointer dereference, address: 0000000000000058
Call Trace:
<TASK>
? __die+0x20/0x70
? page_fault_oops+0x15b/0x430
? fixup_exception+0x22/0x330
? exc_page_fault+0x6f/0x170
? asm_exc_page_fault+0x22/0x30
? bpf_tracing_prog_attach+0x279/0x560
? btf_obj_id+0x5/0x10
bpf_tracing_prog_attach+0x439/0x560
__sys_bpf+0x1cf4/0x2de0
__x64_sys_bpf+0x1c/0x30
do_syscall_64+0x41/0xf0
entry_SYSCALL_64_after_hwframe+0x6e/0x76Return -EINVAL in this situation.
--- kernel/bpf/syscall.c | 9 +++++++++ 1 file changed, 9 insertions(+)
kernel/bpf/syscall.c +9 -0
@@ -3181,6 +3181,10 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog,* * - if prog->aux->dst_trampoline and tgt_prog is NULL, the program * was detached and is going for re-attachment. +* +* - if prog->aux->dst_trampoline is NULL and tgt_prog and prog->aux->attach_btf +* are NULL, then program was already attached and user did not provide +* tgt_prog_fd so we have no way to find out or create trampoline */ if (!prog->aux->dst_trampoline && !tgt_prog) {/*
@@ -3194,6 +3198,11 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog, err = -EINVAL ; goto out_unlock; } +/* We can allow re-attach only if we have valid attach_btf. */ +if (!prog->aux->attach_btf) { + err = -EINVAL ; + goto out_unlock; + } btf_id = prog->aux->attach_btf_id; key =bpf_trampoline_compute_key (NULL , prog->aux->attach_btf, btf_id); }
Signature · 1 line
2.41.0
PATCH v5 4/4 selftests/bpf: Test re-attachment fix for bpf_tracing_prog_attach
In reply to Dmitrii DolgovAdd a test case to verify the fix for "prog->aux->dst_trampoline and tgt_prog is NULL" branch in bpf_tracing_prog_attach. The sequence of events:
1. load rawtp program
2. load fentry program with rawtp as target_fd
3. create tracing link for fentry program with target_fd = 0
4. repeat 3
--- .../bpf/prog_tests/recursive_attach.c | 48 +++++++++++++++++++ .../bpf/progs/fentry_recursive_target.c | 11 +++++ 2 files changed, 59 insertions(+)
tools/testing/selftests/bpf/prog_tests/recursive_attach.c +48 -0
@@ -67,3 +67,51 @@ void test_recursive_fentry_attach(void)fentry_recursive__destroy (tracing_chain[i]); } } + +/* +* Test that a tracing prog reattachment (when we land in +* "prog->aux->dst_trampoline and tgt_prog is NULL" branch in +* bpf_tracing_prog_attach) does not lead to a crash due to missing attach_btf +*/ +void test_fentry_attach_btf_presence (void ) +{ +struct fentry_recursive_target *target_skel =NULL ; +struct fentry_recursive *tracing_skel =NULL ; +struct bpf_program *prog; +int err, link_fd, tgt_prog_fd; + + target_skel =fentry_recursive_target__open_and_load (); +if (!ASSERT_OK_PTR (target_skel,"fentry_recursive_target__open_and_load" )) + goto close_prog; + + tracing_skel =fentry_recursive__open (); +if (!ASSERT_OK_PTR (tracing_skel,"fentry_recursive__open" )) + goto close_prog; + + prog = tracing_skel->progs.recursive_attach; + tgt_prog_fd =bpf_program__fd (target_skel->progs.fentry_target); + err =bpf_program__set_attach_target (prog, tgt_prog_fd,"fentry_target" ); +if (!ASSERT_OK (err,"bpf_program__set_attach_target" )) + goto close_prog; + + err =fentry_recursive__load (tracing_skel); +if (!ASSERT_OK (err,"fentry_recursive__load" )) + goto close_prog; + +LIBBPF_OPTS (bpf_link_create_opts, link_opts); + + link_fd =bpf_link_create (bpf_program__fd (tracing_skel->progs.recursive_attach), +0 ,BPF_TRACE_FENTRY , &link_opts); +if (!ASSERT_GE (link_fd,0 ,"link_fd" )) + goto close_prog; + +fentry_recursive__detach (tracing_skel); + + err =fentry_recursive__attach (tracing_skel); +if (!ASSERT_ERR (err,"fentry_recursive__attach" )) + goto close_prog; + +close_prog: +fentry_recursive_target__destroy (target_skel); +fentry_recursive__destroy (tracing_skel); +}
tools/testing/selftests/bpf/progs/fentry_recursive_target.c +11 -0
@@ -18,3 +18,14 @@ int BPF_PROG(test1, int a) test1_result = a ==1 ;return 0 ; } + +/* +* Dummy bpf prog for testing attach_btf presence when attaching an fentry +* program. +*/ +SEC("raw_tp/sys_enter") +int BPF_PROG (fentry_target ,struct pt_regs *regs,long id) +{ + test1_result = id ==1 ; +return 0 ; +}
Signature · 1 line
2.41.0
Song Liu
reviewerQuoted 21 lines · Dmitrii Dolgov
Currently, it's not allowed to attach an fentry/fexit prog to another fentry/fexit. At the same time it's not uncommon to see a tracing program with lots of logic in use, and the attachment limitation prevents usage of fentry/fexit for performance analysis (e.g. with "bpftool prog profile" command) in this case. An example could be falcosecurity libs project that uses
tp_btftracing programs for offloading certain part of logic into tail-called programs, but the use-case is still generic enough -- a tracing program could be complicated and heavy enough to warrant its profiling, yet frustratingly it's not possible to do so use best tooling for that.Following the corresponding discussion [1], the reason for that is to avoid tracing progs call cycles without introducing more complex solutions. But currently it seems impossible to load and attach tracing programs in a way that will form such a cycle. Replace "no same type" requirement with verification that no more than one level of attachment nesting is allowed. In this way only one fentry/fexit program could be attached to another fentry/fexit to cover profiling use case, and still no cycle could be formed.
Quoted 7 lines
The series contains a test for recursive attachment, as well as a fix + test for an issue in re-attachment branch of
bpf_tracing_prog_attach. When preparing the test for the main change set, I've stumbled upon the possibility to construct a sequence of events whenattach_btfwould be NULL while computing a trampoline key. It doesn't look like this issue is triggered by the main change, because the reproduces doesn't actually need to have an fentry attachment chain.
It appears this set breaks test_progs/trace_ext:
https://github.com/kernel-patches/bpf/actions/runs/7062243664/job/19225827450
Thanks,
Song
Quoted 29 lines
[1]: https://lore.kernel.org/bpf/[email protected]/
Dmitrii Dolgov (3):
bpf: Relax tracing prog recursive attach rules selftests/bpf: Add test for recursive attachment of tracing progs selftests/bpf: Test re-attachment fix forbpf_tracing_prog_attachJiri Olsa (1):
bpf: Fix re-attachment branch inbpf_tracing_prog_attachinclude/linux/bpf.h | 1 + include/uapi/linux/bpf.h | 1 + kernel/bpf/syscall.c | 16 +++ kernel/bpf/verifier.c | 33 ++--- tools/include/uapi/linux/bpf.h | 1 + .../bpf/prog_tests/recursive_attach.c | 117 ++++++++++++++++++ .../selftests/bpf/progs/fentry_recursive.c | 19 +++ .../bpf/progs/fentry_recursive_target.c | 31 +++++ 8 files changed, 205 insertions(+), 14 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/recursive_attach.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive.c create mode 100644 tools/testing/selftests/bpf/progs/fentry_recursive_target.cbase-commit:
40d0eb0259ae77ace3e81d7454d1068c38bc95c2--
2.41.0
Quoted 4 lines · Song Liu
It appears this set breaks test_progs/trace_ext:
https://github.com/kernel-patches/bpf/actions/runs/7062243664/job/19225827450
Yeah, my bad. There was an issue in the original patch, the attach depth check was applied not only to "fentry->fentry" chains, but also to "fentry->extension". Reducing nesting level to one revealed the problem. I'm going to modify the attach function to maintain attach_depth only for fentry progs, this will resolve the problem. Thanks!