Remove use of current->cgns bpf_cgroup_from_id
Subsystems (5)
Kumar Kartikeya Dwivedi
maintainercover Remove use of current->cgns bpf_cgroup_from_id
bpf_cgroup_from_id currently ends up doing a check on whether the cgroup being looked up is a descendant of the root cgroup of the current task's cgroup namespace. This leads to unreliable results since this kfunc can be invoked from any arbitrary context, for any arbitrary value of current. Fix this by removing namespace-awarness in the kfunc, and include a test that detects such a case and fails without the fix.
Kumar Kartikeya Dwivedi (2):
bpf: Do not limit bpf_cgroup_from_id to current's namespace selftests/bpf: Add a test for bpf_cgroup_from_id lookup in non-root
cgns
include/linux/cgroup.h | 2 +- kernel/bpf/cgroup_iter.c | 2 +- kernel/bpf/helpers.c | 2 +- kernel/cgroup/cgroup.c | 7 ++- .../selftests/bpf/prog_tests/cgrp_kfunc.c | 48 +++++++++++++++++++ .../selftests/bpf/progs/cgrp_kfunc_success.c | 12 +++++ 6 files changed, 69 insertions(+), 4 deletions(-)
base-commit: fa479132845e94b60068fad01c2a9979b3efe2dc
Signature · 1 line
2.47.3
Kumar Kartikeya Dwivedi
maintainerPATCH 1/2 bpf: Do not limit bpf_cgroup_from_id to current's namespace
The bpf_cgroup_from_id kfunc relies on cgroup_get_from_id to obtain the cgroup corresponding to a given cgroup ID. This helper can be called in a lot of contexts where the current thread can be random. A recent example was its use in sched_ext's ops.tick(), to obtain the root cgroup pointer. Since the current task can be whatever random user space task preempted by the timer tick, this makes the behavior of the helper unreliable.
Resolve this by refactoring cgroup_get_from_id to take a parameter to elide the cgroup_is_descendant check when root_cgns parameter is set to true.
There is no compatibility breakage here, since changing the namespace against which the lookup is being done to the root cgroup namespace only permits a wider set of lookups to succeed now. The cgroup IDs across namespaces are globally unique, and thus don't need to be retranslated.
--- include/linux/cgroup.h | 2 +- kernel/bpf/cgroup_iter.c | 2 +- kernel/bpf/helpers.c | 2 +- kernel/cgroup/cgroup.c | 7 ++++++- 4 files changed, 9 insertions(+), 4 deletions(-)
include/linux/cgroup.h +1 -1
@@ -650,7 +650,7 @@ static inline void cgroup_kthread_ready(void) }void cgroup_path_from_kernfs_id (u64 id,char *buf,size_t buflen); -struct cgroup *cgroup_get_from_id (u64 id); +struct cgroup *cgroup_get_from_id (u64 id,bool root_cgns);#else /* !CONFIG_CGROUPS */ struct cgroup_subsys_state ;
kernel/bpf/cgroup_iter.c +1 -1
@@ -212,7 +212,7 @@ static int bpf_iter_attach_cgroup(struct bpf_prog *prog,if (fd) cgrp =cgroup_v1v2_get_from_fd (fd);else if (id) - cgrp =cgroup_get_from_id (id); + cgrp =cgroup_get_from_id (id, false);else /* walk the entire hierarchy by default. */ cgrp =cgroup_get_from_path ("/" );
kernel/bpf/helpers.c +1 -1
@@ -2537,7 +2537,7 @@ __bpf_kfunc struct cgroup *bpf_cgroup_from_id(u64 cgid) {struct cgroup *cgrp; - cgrp =cgroup_get_from_id (cgid); + cgrp =cgroup_get_from_id (cgid, true);if (IS_ERR (cgrp))return NULL ;return cgrp;
kernel/cgroup/cgroup.c +6 -1
@@ -6345,10 +6345,11 @@ void cgroup_path_from_kernfs_id(u64 id, char *buf, size_t buflen)/* * cgroup_get_from_id : get the cgroup associated with cgroup id * @id: cgroup id +* @root_cgns: Select root cgroup namespace instead of current's. * On success return the cgrp or ERR_PTR on failure * Only cgroups within current task's cgroup NS are valid. */ -struct cgroup *cgroup_get_from_id (u64 id) +struct cgroup *cgroup_get_from_id (u64 id,bool root_cgns) {struct kernfs_node *kn;struct cgroup *cgrp, *root_cgrp;
@@ -6374,6 +6375,10 @@ struct cgroup *cgroup_get_from_id(u64 id)if (!cgrp)return ERR_PTR (-ENOENT ); +/* We don't need to namespace this operation against current. */ +if (root_cgns) +return cgrp; + root_cgrp =current_cgns_cgroup_dfl ();if (!cgroup_is_descendant (cgrp, root_cgrp)) {cgroup_put (cgrp);
Signature · 1 line
2.47.3
Re: [PATCH bpf-next v1 1/2] bpf: Do not limit bpf_cgroup_from_id to current's namespace
Quoted 12 lines · Kumar Kartikeya Dwivedi
The
bpf_cgroup_from_idkfunc relies oncgroup_get_from_idto obtain the cgroup corresponding to a given cgroup ID. This helper can be called in a lot of contexts where the current thread can be random. A recent example was its use in sched_ext'sops.tick(), to obtain the root cgroup pointer. Since the current task can be whatever random user space task preempted by the timer tick, this makes the behavior of the helper unreliable.Resolve this by refactoring
cgroup_get_from_idto take a parameter to elide thecgroup_is_descendantcheck whenroot_cgnsparameter is set to true.
Quoted 7 lines
There is no compatibility breakage here, since changing the namespace against which the lookup is being done to the root cgroup namespace only permits a wider set of lookups to succeed now. The cgroup IDs across namespaces are globally unique, and thus don't need to be retranslated.
Reported-by Dan Schatzberg <[email protected]>Signed-off-by Kumar Kartikeya Dwivedi <[email protected]>
Thanks.
Signature · 1 line
tejun
Kumar Kartikeya Dwivedi
maintainerPATCH 2/2 selftests/bpf: Add a test for bpf_cgroup_from_id lookup in non-root cgns
In reply to Kumar Kartikeya Dwivedi--- .../selftests/bpf/prog_tests/cgrp_kfunc.c | 48 +++++++++++++++++++ .../selftests/bpf/progs/cgrp_kfunc_success.c | 12 +++++ 2 files changed, 60 insertions(+)
tools/testing/selftests/bpf/prog_tests/cgrp_kfunc.c +48 -0
@@ -4,6 +4,7 @@#define _GNU_SOURCE#include <cgroup_helpers.h> #include <test_progs.h> +#include <sched.h> #include "cgrp_kfunc_failure.skel.h" #include "cgrp_kfunc_success.skel.h"
@@ -87,6 +88,50 @@ static const char * const success_tests[] = { "test_cgrp_from_id", }; +static void test_cgrp_from_id_ns (void ) +{ +LIBBPF_OPTS (bpf_test_run_opts, opts); +struct cgrp_kfunc_success *skel; +struct bpf_program *prog; +int fd, ret; + + skel =open_load_cgrp_kfunc_skel (); +if (!ASSERT_OK_PTR (skel,"open_load_skel" )) +return ; + +if (!ASSERT_OK (skel->bss->err,"pre_mkdir_err" )) + goto cleanup; + + prog =bpf_object__find_program_by_name (skel->obj,"test_cgrp_from_id_ns" ); +if (!ASSERT_OK_PTR (prog,"bpf_object__find_program_by_name" )) + goto cleanup; + + fd =create_and_get_cgroup ("cgrp_from_id_ns" ); +if (!ASSERT_GE (fd,0 ,"cgrp_fd" )) + goto cleanup; + +if (!ASSERT_OK (join_cgroup ("cgrp_from_id_ns" ),"join cgrp" )) + goto fd_cleanup; + +if (!ASSERT_OK (unshare (CLONE_NEWCGROUP ),"unshare cgns" )) + goto fd_cleanup; + + ret =bpf_prog_test_run_opts (bpf_program__fd (prog), &opts); +if (!ASSERT_OK (ret,"test run ret" )) + goto fd_cleanup; + +if (!ASSERT_OK (opts.retval,"test run retval" )) + goto fd_cleanup; + +remove_cgroup ("cgrp_from_id_ns" ); + +fd_cleanup: +close (fd); +cleanup: +cgrp_kfunc_success__destroy (skel); + +} +void test_cgrp_kfunc (void ) {int i, err;
@@ -102,6 +147,9 @@ void test_cgrp_kfunc(void)run_success_test (success_tests[i]); } +if (test__start_subtest ("test_cgrp_from_id_ns" )) +test_cgrp_from_id_ns (); +RUN_TESTS (cgrp_kfunc_failure); cleanup:
tools/testing/selftests/bpf/progs/cgrp_kfunc_success.c +12 -0
@@ -221,3 +221,15 @@ int BPF_PROG(test_cgrp_from_id, struct cgroup *cgrp, const char *path)return 0 ; } + +SEC("syscall") +int test_cgrp_from_id_ns (void *ctx) +{ +struct cgroup *cg; + + cg =bpf_cgroup_from_id (1 ); +if (!cg) +return 42 ; +bpf_cgroup_release (cg); +return 0 ; +}
Signature · 1 line
2.47.3
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v1 2/2] selftests/bpf: Add a test for bpf_cgroup_from_id lookup in non-root cgns
Quoted 1 line · Kumar Kartikeya Dwivedi
Signed-off-by Kumar Kartikeya Dwivedi <[email protected]>
Please add a commit message in the respin. Thanks.
Kumar Kartikeya Dwivedi
maintainerQuoted 8 lines · Kumar Kartikeya Dwivedi
bpf_cgroup_from_idcurrently ends up doing a check on whether the cgroup being looked up is a descendant of the root cgroup of the current task's cgroup namespace. This leads to unreliable results since this kfunc can be invoked from any arbitrary context, for any arbitrary value of current. Fix this by removing namespace-awarness in the kfunc, and include a test that detects such a case and fails without the fix.
The CI failure is probably because of the ugly unshare(2), I will fix and respin.
Quoted 18 lines
Kumar Kartikeya Dwivedi (2):
bpf: Do not limitbpf_cgroup_from_idto current's namespace selftests/bpf: Add a test forbpf_cgroup_from_idlookup in non-root
cgnsinclude/linux/cgroup.h | 2 +- kernel/bpf/cgroup_iter.c | 2 +- kernel/bpf/helpers.c | 2 +- kernel/cgroup/cgroup.c | 7 ++- .../selftests/bpf/prog_tests/cgrp_kfunc.c | 48 +++++++++++++++++++ .../selftests/bpf/progs/cgrp_kfunc_success.c | 12 +++++ 6 files changed, 69 insertions(+), 4 deletions(-)base-commit:
fa479132845e94b60068fad01c2a9979b3efe2dc--
2.47.3