Compare commits

..

9 Commits

Author SHA1 Message Date
Andrii Nakryiko
5579664205 libbpf: Fix build with latest gcc/binutils with LTO
After updating to binutils 2.35, the build began to fail with an
assembler error. A bug was opened on the Red Hat Bugzilla a few days
later for the same issue.

Work around the problem by using the new `symver` attribute (introduced
in GCC 10) as needed instead of assembler directives.

This addresses Red Hat ([0]) and OpenSUSE ([1]) bug reports, as well as libbpf
issue ([2]).

  [0]: https://bugzilla.redhat.com/show_bug.cgi?id=1863059
  [1]: https://bugzilla.opensuse.org/show_bug.cgi?id=1188749
  [2]: Closes: https://github.com/libbpf/libbpf/issues/338

Co-developed-by: Patrick McCarty <patrick.mccarty@intel.com>
Co-developed-by: Michal Suchanek <msuchanek@suse.de>
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Signed-off-by: Michal Suchanek <msuchanek@suse.de>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210907221023.2660953-1-andrii@kernel.org
2021-09-08 16:01:51 -07:00
Matt Smith
860b201cd0 libbpf: Change bpf_object_skeleton data field to const pointer
This change was necessary to enforce the implied contract
that bpf_object_skeleton->data should not be mutated.  The data
will be cast to `void *` during assignment to handle the case
where a user is compiling with older libbpf headers to avoid
a compiler warning of `const void *` data being cast to `void *`

Signed-off-by: Matt Smith <alastorze@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210901194439.3853238-2-alastorze@fb.com
2021-09-08 16:01:51 -07:00
Toke Høiland-Jørgensen
50e13993a1 libbpf: Don't crash on object files with no symbol tables
If libbpf encounters an ELF file that has been stripped of its symbol
table, it will crash in bpf_object__add_programs() when trying to
dereference the obj->efile.symbols pointer.

Fix this by erroring out of bpf_object__elf_collect() if it is not able
able to find the symbol table.

v2:
  - Move check into bpf_object__elf_collect() and add nice error message

Fixes: 6245947c1b3c ("libbpf: Allow gaps in BPF program sections to support overriden weak functions")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210901114812.204720-1-toke@redhat.com
2021-09-08 16:01:51 -07:00
Andrii Nakryiko
72fd44da53 ci: blacklist few new selftests for 5.5
Blacklist netns_cookie, sockopt_qos_to_cc, task_pt_regs for 5.5.

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2021-08-31 14:21:44 -07:00
Andrii Nakryiko
d6e9681b0d sync: latest libbpf changes from kernel
Syncing latest libbpf commits from kernel repository.
Baseline bpf-next commit:   d20b41115ad53293201cc07ee429a38740cb056b
Checkpoint bpf-next commit: 47bb27a20d6ea22cd092c1fc2bb4fcecac374838
Baseline bpf commit:        3776f3517ed94d40ff0e3851d7ce2ce17b63099f
Checkpoint bpf commit:      3776f3517ed94d40ff0e3851d7ce2ce17b63099f

Daniel Xu (1):
  bpf: Add bpf_task_pt_regs() helper

Dave Marchevsky (1):
  bpf: Migrate cgroup_bpf to internal cgroup_bpf_attach_type enum

 include/uapi/linux/bpf.h | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

--
2.30.2
2021-08-31 14:21:44 -07:00
Andrii Nakryiko
517762deca sync: auto-generate latest BPF helpers
Latest changes to BPF helper definitions.
2021-08-31 14:21:44 -07:00
Daniel Xu
1670e6100b bpf: Add bpf_task_pt_regs() helper
The motivation behind this helper is to access userspace pt_regs in a
kprobe handler.

uprobe's ctx is the userspace pt_regs. kprobe's ctx is the kernelspace
pt_regs. bpf_task_pt_regs() allows accessing userspace pt_regs in a
kprobe handler. The final case (kernelspace pt_regs in uprobe) is
pretty rare (usermode helper) so I think that can be solved later if
necessary.

More concretely, this helper is useful in doing BPF-based DWARF stack
unwinding. Currently the kernel can only do framepointer based stack
unwinds for userspace code. This is because the DWARF state machines are
too fragile to be computed in kernelspace [0]. The idea behind
DWARF-based stack unwinds w/ BPF is to copy a chunk of the userspace
stack (while in prog context) and send it up to userspace for unwinding
(probably with libunwind) [1]. This would effectively enable profiling
applications with -fomit-frame-pointer using kprobes and uprobes.

[0]: https://lkml.org/lkml/2012/2/10/356
[1]: https://github.com/danobi/bpf-dwarf-walk

Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/e2718ced2d51ef4268590ab8562962438ab82815.1629772842.git.dxu@dxuuu.xyz
2021-08-31 14:21:44 -07:00
Dave Marchevsky
ed529685db bpf: Migrate cgroup_bpf to internal cgroup_bpf_attach_type enum
Add an enum (cgroup_bpf_attach_type) containing only valid cgroup_bpf
attach types and a function to map bpf_attach_type values to the new
enum. Inspired by netns_bpf_attach_type.

Then, migrate cgroup_bpf to use cgroup_bpf_attach_type wherever
possible.  Functionality is unchanged as attach_type_to_prog_type
switches in bpf/syscall.c were preventing non-cgroup programs from
making use of the invalid cgroup_bpf array slots.

As a result struct cgroup_bpf uses 504 fewer bytes relative to when its
arrays were sized using MAX_BPF_ATTACH_TYPE.

bpf_cgroup_storage is notably not migrated as struct
bpf_cgroup_storage_key is part of uapi and contains a bpf_attach_type
member which is not meant to be opaque. Similarly, bpf_cgroup_link
continues to report its bpf_attach_type member to userspace via fdinfo
and bpf_link_info.

To ease disambiguation, bpf_attach_type variables are renamed from
'type' to 'atype' when changed to cgroup_bpf_attach_type.

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210819092420.1984861-2-davemarchevsky@fb.com
2021-08-31 14:21:44 -07:00
Andrii Nakryiko
c92a5d043e ci: don't hang on kernel crashes in qemu
Backport of kernel-patches PR ([0]), done by @thefallentree.

  [0] https://github.com/kernel-patches/vmtest/pull/29

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2021-08-30 12:52:51 -07:00
9 changed files with 52 additions and 13 deletions

View File

@@ -1 +1 @@
d20b41115ad53293201cc07ee429a38740cb056b
47bb27a20d6ea22cd092c1fc2bb4fcecac374838

View File

@@ -84,7 +84,7 @@ struct bpf_lpm_trie_key {
struct bpf_cgroup_storage_key {
__u64 cgroup_inode_id; /* cgroup inode id */
__u32 attach_type; /* program attach type */
__u32 attach_type; /* program attach type (enum bpf_attach_type) */
};
union bpf_iter_link_info {
@@ -4871,6 +4871,12 @@ union bpf_attr {
* Return
* Value specified by user at BPF link creation/attachment time
* or 0, if it was not specified.
*
* long bpf_task_pt_regs(struct task_struct *task)
* Description
* Get the struct pt_regs associated with **task**.
* Return
* A pointer to struct pt_regs.
*/
#define __BPF_FUNC_MAPPER(FN) \
FN(unspec), \
@@ -5048,6 +5054,7 @@ union bpf_attr {
FN(timer_cancel), \
FN(get_func_ip), \
FN(get_attach_cookie), \
FN(task_pt_regs), \
/* */
/* integer value in 'imm' field of BPF_CALL instruction selects which helper

View File

@@ -4033,4 +4033,14 @@ static __u64 (*bpf_get_func_ip)(void *ctx) = (void *) 173;
*/
static __u64 (*bpf_get_attach_cookie)(void *ctx) = (void *) 174;
/*
* bpf_task_pt_regs
*
* Get the struct pt_regs associated with **task**.
*
* Returns
* A pointer to struct pt_regs.
*/
static long (*bpf_task_pt_regs)(struct task_struct *task) = (void *) 175;

View File

@@ -2993,6 +2993,12 @@ static int bpf_object__elf_collect(struct bpf_object *obj)
}
}
if (!obj->efile.symbols) {
pr_warn("elf: couldn't find symbol table in %s, stripped object file?\n",
obj->path);
return -ENOENT;
}
scn = NULL;
while ((scn = elf_nextscn(elf, scn)) != NULL) {
idx++;

View File

@@ -854,7 +854,7 @@ struct bpf_object_skeleton {
size_t sz; /* size of this struct, for forward/backward compatibility */
const char *name;
void *data;
const void *data;
size_t data_sz;
struct bpf_object **obj;

View File

@@ -90,17 +90,30 @@
/* Symbol versioning is different between static and shared library.
* Properly versioned symbols are needed for shared library, but
* only the symbol of the new version is needed for static library.
* Starting with GNU C 10, use symver attribute instead of .symver assembler
* directive, which works better with GCC LTO builds.
*/
#ifdef SHARED
# define COMPAT_VERSION(internal_name, api_name, version) \
#if defined(SHARED) && defined(__GNUC__) && __GNUC__ >= 10
#define DEFAULT_VERSION(internal_name, api_name, version) \
__attribute__((symver(#api_name "@@" #version)))
#define COMPAT_VERSION(internal_name, api_name, version) \
__attribute__((symver(#api_name "@" #version)))
#elif defined(SHARED)
#define COMPAT_VERSION(internal_name, api_name, version) \
asm(".symver " #internal_name "," #api_name "@" #version);
# define DEFAULT_VERSION(internal_name, api_name, version) \
#define DEFAULT_VERSION(internal_name, api_name, version) \
asm(".symver " #internal_name "," #api_name "@@" #version);
#else
# define COMPAT_VERSION(internal_name, api_name, version)
# define DEFAULT_VERSION(internal_name, api_name, version) \
#else /* !SHARED */
#define COMPAT_VERSION(internal_name, api_name, version)
#define DEFAULT_VERSION(internal_name, api_name, version) \
extern typeof(internal_name) api_name \
__attribute__((alias(#internal_name)));
#endif
extern void libbpf_print(enum libbpf_print_level level,

View File

@@ -281,6 +281,7 @@ out_mmap:
return err;
}
DEFAULT_VERSION(xsk_umem__create_v0_0_4, xsk_umem__create, LIBBPF_0.0.4)
int xsk_umem__create_v0_0_4(struct xsk_umem **umem_ptr, void *umem_area,
__u64 size, struct xsk_ring_prod *fill,
struct xsk_ring_cons *comp,
@@ -345,6 +346,7 @@ struct xsk_umem_config_v1 {
__u32 frame_headroom;
};
COMPAT_VERSION(xsk_umem__create_v0_0_2, xsk_umem__create, LIBBPF_0.0.2)
int xsk_umem__create_v0_0_2(struct xsk_umem **umem_ptr, void *umem_area,
__u64 size, struct xsk_ring_prod *fill,
struct xsk_ring_cons *comp,
@@ -358,8 +360,6 @@ int xsk_umem__create_v0_0_2(struct xsk_umem **umem_ptr, void *umem_area,
return xsk_umem__create_v0_0_4(umem_ptr, umem_area, size, fill, comp,
&config);
}
COMPAT_VERSION(xsk_umem__create_v0_0_2, xsk_umem__create, LIBBPF_0.0.2)
DEFAULT_VERSION(xsk_umem__create_v0_0_4, xsk_umem__create, LIBBPF_0.0.4)
static enum xsk_prog get_xsk_prog(void)
{

View File

@@ -45,6 +45,7 @@ mmap # 5.5 kernel is too permissive with re-mmaping
modify_return # fmod_ret support is missing
module_attach # module BTF support missing (v5.11+)
netcnt
netns_cookie # v5.15+
ns_current_pid_tgid # bpf_get_ns_current_pid_tgid() helper is missing
pe_preserve_elems # v5.10+
perf_branches # bpf_read_branch_records() helper is missing
@@ -74,10 +75,12 @@ socket_cookie # v5.12+
sockmap_basic # uses new socket fields, 5.8+
sockmap_listen # no listen socket supportin SOCKMAP
sockopt_sk
sockopt_qos_to_cc # v5.15+
stacktrace_build_id # v5.9+
stack_var_off # v5.12+
syscall # v5.14+
task_local_storage # v5.12+
task_pt_regs # v5.15+
tcp_hdr_options # v5.10+, new TCP header options feature in BPF
tcpbpf_user # LINK_CREATE is missing
tc_redirect # v5.14+

View File

@@ -459,10 +459,10 @@ if kvm-ok ; then
else
accel="-cpu qemu64 -machine accel=tcg"
fi
qemu-system-x86_64 -nodefaults -display none -serial mon:stdio \
qemu-system-x86_64 -nodefaults -display none -serial mon:stdio -no-reboot \
${accel} -smp "$(nproc)" -m 4G \
-drive file="$IMG",format=raw,index=1,media=disk,if=virtio,cache=none \
-kernel "$vmlinuz" -append "root=/dev/vda rw console=ttyS0,115200$APPEND"
-kernel "$vmlinuz" -append "root=/dev/vda rw console=ttyS0,115200 kernel.panic=-1 $APPEND"
sudo mount -o loop "$IMG" "$mnt"
if exitstatus="$(cat "$mnt/exitstatus" 2>/dev/null)"; then
printf '\nTests exit status: %s\n' "$exitstatus" >&2