bpf: Add bpf_vma_build_id_parse kfunc
Subsystems (8)
Jiri Olsa
reviewerv4 cover bpf: Add bpf_vma_build_id_parse kfunc
hi,
first version of this patchset added helper for this functionality, but based Alexei's feedback [1], changing it to kfunc.
With the current build_id_parse function as kfunc we can't effectively check buffer size provided by user. Therefore adding new function as bpf kfunc:
int bpf_vma_build_id_parse(struct vm_area_struct *vma,
unsigned char *build_id, size_t build_id__sz);
that triggers kfunc's verifier check for build_id/build_id__sz buffer size and calls build_id_parse.
v4 changes:
- vma object is now passed as trusted pointer argument [Alexei]
- marked bpf_vma_build_id_parse with KF_TRUSTED_ARGS flag so it can be called only with PTR_TRUSTED args
- marked vma objects from task_vma iter and find_vma callback as PTR_TRUSTED
- added test for task_vma iterator
v3 changes:
- restrict bpf_vma_build_id_parse to bpf_find_vma callback
- move bpf_vma_build_id_parse to kernel/trace/bpf_trace.c and add new tracing_kfunc_set
thanks,
jirka
[1] https://lore.kernel.org/bpf/CAADnVQKyT4Mm4EdTCYK8c070E-BwPZS_FOkWKLJC80riSGmLTg@mail.gmail.com/
---
Jiri Olsa (4):
bpf: Mark vma objects as trusted for task_vma iter and find_vma callback
bpf: Add bpf_vma_build_id_parse function and kfunc
selftests/bpf: Add bpf_vma_build_id_parse find_vma callback test selftests/bpf: Add bpf_vma_build_id_parse task vma iterator test
include/linux/bpf.h | 4 ++++ kernel/bpf/task_iter.c | 2 +- kernel/bpf/verifier.c | 2 +- kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/prog_tests/bpf_iter.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/prog_tests/bpf_vma_build_id_parse.c | 55 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/progs/bpf_iter_build_id.c | 41 +++++++++++++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c | 40 ++++++++++++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/trace_helpers.c | 40 ++++++++++++++++++++++++++++++++++++++++ tools/testing/selftests/bpf/trace_helpers.h | 1 + 10 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_vma_build_id_parse.c create mode 100644 tools/testing/selftests/bpf/progs/bpf_iter_build_id.c create mode 100644 tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c
Jiri Olsa
reviewerPATCH v4 1/4 bpf: Mark vma objects as trusted for task_vma iter and find_vma callback
Marking following vma objects as trusted so they can be used as arguments for kfunc function added in following changes:
- vma object argument in find_vma callback function
- vma object in context of task_vma iterator program
Both places lock vma object so it can't go away while running the bpf program.
--- kernel/bpf/task_iter.c | 2 +- kernel/bpf/verifier.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
kernel/bpf/task_iter.c +1 -1
@@ -755,7 +755,7 @@ static struct bpf_iter_reg task_vma_reg_info = { { offsetof(struct bpf_iter__task_vma , task), PTR_TO_BTF_ID_OR_NULL }, { offsetof(struct bpf_iter__task_vma, vma), - PTR_TO_BTF_ID_OR_NULL }, + PTR_TO_BTF_ID_OR_NULL | PTR_TRUSTED }, }, .seq_info = &task_vma_seq_info, .fill_link_info = bpf_iter_fill_link_info,
kernel/bpf/verifier.c +1 -1
@@ -7206,7 +7206,7 @@ static int set_find_vma_callback_state(struct bpf_verifier_env *env,*/ callee->regs[BPF_REG_1 ] = caller->regs[BPF_REG_1 ]; - callee->regs[BPF_REG_2 ].type =PTR_TO_BTF_ID ; + callee->regs[BPF_REG_2 ].type =PTR_TO_BTF_ID |PTR_TRUSTED ;__mark_reg_known_zero (&callee->regs[BPF_REG_2 ]); callee->regs[BPF_REG_2 ].btf = btf_vmlinux; callee->regs[BPF_REG_2].btf_id = btf_tracing_ids[BTF_TRACING_TYPE_VMA],
Signature · 1 line
2.38.1
Re: [PATCHv4 bpf-next 1/4] bpf: Mark vma objects as trusted for task_vma iter and find_vma callback
Quoted 21 lines · Jiri Olsa
Marking following vma objects as trusted so they can be used as arguments for kfunc function added in following changes:
- vma object argument in
find_vmacallback function
- vma object in context oftask_vmaiterator programBoth places lock vma object so it can't go away while running the bpf program.
Signed-off-by Jiri Olsa <[email protected]>--- kernel/bpf/task_iter.c | 2 +- kernel/bpf/verifier.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)kernel/bpf/task_iter.c +0
-0
kernel/bpf/task_iter.c static struct bpf_iter_reg task_vma_reg_info = { · line 758
Quoted 5 lines
kernel/bpf/task_iter.c +1
-1@@ -755,7 +755,7 @@ static struct bpf_iter_reg task_vma_reg_info = { { offsetof(struct bpf_iter__task_vma , task), PTR_TO_BTF_ID_OR_NULL }, { offsetof(struct bpf_iter__task_vma, vma), - PTR_TO_BTF_ID_OR_NULL }, + PTR_TO_BTF_ID_OR_NULL | PTR_TRUSTED },
Yonghong, Song,
Do you remember when task or vma is NULL here? Maybe we can do: if (!task || !vma) skip prog run in __task_vma_seq_show()
and make both pointers as PTR_TO_BTF_ID | PTR_TRUSTED?
Re: [PATCHv4 bpf-next 1/4] bpf: Mark vma objects as trusted for task_vma iter and find_vma callback
Quoted 30 lines · Alexei Starovoitov
Quoted 21 lines · Jiri Olsa
Marking following vma objects as trusted so they can be used as arguments for kfunc function added in following changes:
- vma object argument in
find_vmacallback function
- vma object in context oftask_vmaiterator programBoth places lock vma object so it can't go away while running the bpf program.
Signed-off-by Jiri Olsa <[email protected]>--- kernel/bpf/task_iter.c | 2 +- kernel/bpf/verifier.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)kernel/bpf/task_iter.c +0
-0kernel/bpf/task_iter.c static struct bpf_iter_reg task_vma_reg_info = { · line 758
Quoted 5 lines
kernel/bpf/task_iter.c +1
-1@@ -755,7 +755,7 @@ static struct bpf_iter_reg task_vma_reg_info = { { offsetof(struct bpf_iter__task_vma , task), PTR_TO_BTF_ID_OR_NULL }, { offsetof(struct bpf_iter__task_vma, vma), - PTR_TO_BTF_ID_OR_NULL }, + PTR_TO_BTF_ID_OR_NULL | PTR_TRUSTED },Yonghong, Song,
Quoted 4 lines
Do you remember when task or vma is NULL here? Maybe we can do: if (!task || !vma) skip prog run in
__task_vma_seq_show()
and make both pointers asPTR_TO_BTF_ID|PTR_TRUSTED?
The 'NULL' is to indicate the last bpf prog run before iteration ends. It is to provide an opportunity for bpf program to know all regular iterations are done and the bpf program can do end aggregation or print a footer if the prog link is cat'able.
Re: [PATCHv4 bpf-next 1/4] bpf: Mark vma objects as trusted for task_vma iter and find_vma callback
Quoted 39 lines · Yonghong Song
Quoted 30 lines · Alexei Starovoitov
Quoted 21 lines · Jiri Olsa
Marking following vma objects as trusted so they can be used as arguments for kfunc function added in following changes:
- vma object argument in
find_vmacallback function
- vma object in context oftask_vmaiterator programBoth places lock vma object so it can't go away while running the bpf program.
Signed-off-by Jiri Olsa <[email protected]>--- kernel/bpf/task_iter.c | 2 +- kernel/bpf/verifier.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)kernel/bpf/task_iter.c +0
-0kernel/bpf/task_iter.c static struct bpf_iter_reg task_vma_reg_info = { · line 758
Quoted 5 lines
kernel/bpf/task_iter.c +1
-1@@ -755,7 +755,7 @@ static struct bpf_iter_reg task_vma_reg_info = { { offsetof(struct bpf_iter__task_vma , task), PTR_TO_BTF_ID_OR_NULL }, { offsetof(struct bpf_iter__task_vma, vma), - PTR_TO_BTF_ID_OR_NULL }, + PTR_TO_BTF_ID_OR_NULL | PTR_TRUSTED },Yonghong, Song,
Quoted 4 lines
Do you remember when task or vma is NULL here? Maybe we can do: if (!task || !vma) skip prog run in
__task_vma_seq_show()
and make both pointers asPTR_TO_BTF_ID|PTR_TRUSTED?
Quoted 4 lines
The 'NULL' is to indicate the last bpf prog run before iteration ends. It is to provide an opportunity for bpf program to know all regular iterations are done and the bpf program can do end aggregation or print a footer if the prog link is cat'able.
Ahh. Right. Now I remember :)
I think we're fine with PTR_TRUSTED here.
The pointer still has to be checked for != NULL before being dereferenced or passed into kfunc.
Jiri Olsa
reviewerPATCH v4 2/4 bpf: Add bpf_vma_build_id_parse function and kfunc
In reply to Jiri OlsaAdding bpf_vma_build_id_parse function to retrieve build id from passed vma object and making it available as bpf kfunc.
We can't use build_id_parse directly as kfunc, because we would not have control over the build id buffer size provided by user.
Instead we are adding new bpf_vma_build_id_parse function with 'build_id__sz' argument that instructs verifier to check for the available space in build_id buffer.
This way we check that there's always available memory space behind build_id pointer. We also check that the build_id__sz is at least BUILD_ID_SIZE_MAX so we can place any buildid in.
The bpf_vma_build_id_parse kfunc is marked as KF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only by find_vma callback function and task_vma iterator program.
--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)
include/linux/bpf.h +4 -0
@@ -2839,4 +2839,8 @@ static inline bool type_is_alloc(u32 type) return type & MEM_ALLOC; } +int bpf_vma_build_id_parse (struct vm_area_struct *vma, +unsigned char *build_id, +size_t build_id__sz); +#endif /* _LINUX_BPF_H */
kernel/trace/bpf_trace.c +31 -0
@@ -23,6 +23,7 @@#include <linux/sort.h> #include <linux/key.h> #include <linux/verification.h> +#include <linux/buildid.h> #include <net/bpf_sk_storage.h>
@@ -1383,6 +1384,36 @@ static int __init bpf_key_sig_kfuncs_init(void)late_initcall (bpf_key_sig_kfuncs_init);#endif /* CONFIG_KEYS */ +int bpf_vma_build_id_parse (struct vm_area_struct *vma, +unsigned char *build_id, +size_t build_id__sz) +{ +__u32 size; +int err; + +if (build_id__sz <BUILD_ID_SIZE_MAX ) +return -EINVAL ; + + err =build_id_parse (vma, build_id, &size); +return err ?: (int ) size; +} + +BTF_SET8_START (tracing_btf_ids) +BTF_ID_FLAGS (func ,bpf_vma_build_id_parse ,KF_TRUSTED_ARGS ) +BTF_SET8_END (tracing_btf_ids) + +static const struct btf_kfunc_id_set tracing_kfunc_set = { + .owner =THIS_MODULE , + .set = &tracing_btf_ids, +}; + +static int __init kfunc_tracing_init (void ) +{ +return register_btf_kfunc_id_set (BPF_PROG_TYPE_TRACING , &tracing_kfunc_set); +} + +late_initcall (kfunc_tracing_init); +static const struct bpf_func_proto *bpf_tracing_func_proto (enum bpf_func_id func_id,const struct bpf_prog *prog) {
Signature · 1 line
2.38.1
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 57 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)include/linux/bpf.h +4
-0@@ -2839,4 +2839,8 @@ static inline bool type_is_alloc(u32 type) return type & MEM_ALLOC; } +int bpf_vma_build_id_parse (struct vm_area_struct *vma, +unsigned char *build_id, +size_t build_id__sz); +#endif /* _LINUX_BPF_H */ kernel/trace/bpf_trace.c +3
-0@@ -23,6 +23,7 @@#include <linux/sort.h> #include <linux/key.h> #include <linux/verification.h> +#include <linux/buildid.h> #include <net/bpf_sk_storage.h> @@ -1383,6 +1384,36 @@ static int __init bpf_key_sig_kfuncs_init(void)late_initcall (bpf_key_sig_kfuncs_init);#endif /* CONFIG_KEYS */ +int bpf_vma_build_id_parse(struct vm_area_struct *vma, + unsigned char *build_id,
kernel/trace/bpf_trace.c static int __init bpf_key_sig_kfuncs_init(void) · line 1398
Quoted 10 lines
kernel/trace/bpf_trace.c +10
-0+ size_t build_id__sz) +{ +__u32 size; +int err; + +if (build_id__sz <BUILD_ID_SIZE_MAX ) +return -EINVAL ; + + err =build_id_parse (vma, build_id, &size); +return err ?: (int ) size;
if err is positive the caller won't be able
to distinguish it vs size.
Quoted 7 lines
kernel/trace/bpf_trace.c +7
-0+} + +BTF_SET8_START (tracing_btf_ids) +BTF_ID_FLAGS (func ,bpf_vma_build_id_parse ,KF_TRUSTED_ARGS ) +BTF_SET8_END (tracing_btf_ids) + +static const struct btf_kfunc_id_set tracing_kfunc_set = {
kernel/trace/bpf_trace.c static int __init bpf_key_sig_kfuncs_init(void) · line 1415
Quoted 10 lines
kernel/trace/bpf_trace.c +10
-0+ .owner = THIS_MODULE, + .set = &tracing_btf_ids, +}; + +static int __init kfunc_tracing_init (void ) +{ +return register_btf_kfunc_id_set (BPF_PROG_TYPE_TRACING , &tracing_kfunc_set); +} + +late_initcall (kfunc_tracing_init);
Its own btf_id set and its own late_initcall just for one kfunc? Please reduce this boilerplate code.
Move it to kernel/bpf/helpers.c ?
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 56 lines · Alexei Starovoitov
Quoted 55 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)include/linux/bpf.h +4
-0@@ -2839,4 +2839,8 @@ static inline bool type_is_alloc(u32 type) return type & MEM_ALLOC; } +int bpf_vma_build_id_parse (struct vm_area_struct *vma, +unsigned char *build_id, +size_t build_id__sz); +#endif /* _LINUX_BPF_H */ kernel/trace/bpf_trace.c +1
-0@@ -23,6 +23,7 @@#include <linux/sort.h> #include <linux/key.h> #include <linux/verification.h> +#include <linux/buildid.h> #include <net/bpf_sk_storage.h> @@ -1383,6 +1384,36 @@ static int __init bpf_key_sig_kfuncs_init(void)late_initcall (bpf_key_sig_kfuncs_init);#endif /* CONFIG_KEYS */
Quoted 15 lines
kernel/trace/bpf_trace.c static int __init bpf_key_sig_kfuncs_init(void) · line 1398
Quoted 12 lines
kernel/trace/bpf_trace.c +12
-0+int bpf_vma_build_id_parse (struct vm_area_struct *vma, +unsigned char *build_id, +size_t build_id__sz) +{ +__u32 size; +int err; + +if (build_id__sz <BUILD_ID_SIZE_MAX ) +return -EINVAL ; + + err =build_id_parse (vma, build_id, &size); +return err ?: (int ) size;if err is positive the caller won't be able
to distinguish it vs size.
that should not happen, build_id_parse returns either < 0 for error, or 0 for success
Quoted 14 lines
Quoted 14 lines
kernel/trace/bpf_trace.c +13
-0+} + +BTF_SET8_START (tracing_btf_ids) +BTF_ID_FLAGS (func ,bpf_vma_build_id_parse ,KF_TRUSTED_ARGS ) +BTF_SET8_END (tracing_btf_ids) + +static const struct btf_kfunc_id_set tracing_kfunc_set = { + .owner =THIS_MODULE , + .set = &tracing_btf_ids, +}; + +static int __init kfunc_tracing_init (void ) +{
Quoted 8 lines
kernel/trace/bpf_trace.c static int __init bpf_key_sig_kfuncs_init(void) · line 1415
Quoted 4 lines
kernel/trace/bpf_trace.c +4
-0+return register_btf_kfunc_id_set (BPF_PROG_TYPE_TRACING , &tracing_kfunc_set); +} + +late_initcall (kfunc_tracing_init);Its own
btf_idset and its ownlate_initcalljust for one kfunc? Please reduce this boilerplate code.
Move it tokernel/bpf/helpers.c?
ok, will move it
thanks,
jirka
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
In reply to Jiri OlsaQuoted 3 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.
As a completely different way of solving this problem of retrieving build_id for tracing needs, can we teach kernel itself to parse and store build_id (probably gated behind Kconfig option) in struct file (ideally)? On exec() and when mmap()'ing with executable permissions, Linux kernel will try to fetch build_id from ELF file and if successful store it in struct file. Given build_id can be up to 20 bytes (currently) and not each struct file points to executable, we might want to only add a pointer field to struct file itself, which, if build_id is present, will point to
struct build_id {
char sz;
char data[];
};This way we don't increase struct file by much.
And then any BPF program would be able to easily probe_read_kernel such build_id from vma_area_struct->vm_file->build_id?
I'm sure I'm oversimplifying, but this seems like a good solution for all kinds of profiling BPF programs without the need to maintain all these allowlists and adding new helpers/kfuncs?
I know Hao was looking at the problem of getting build_id, I'm curious if something like this would work for their use cases as well?
Quoted 23 lines
We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)
[...]
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 26 lines · Andrii Nakryiko
Quoted 3 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.As a completely different way of solving this problem of retrieving
build_idfor tracing needs, can we teach kernel itself to parse and storebuild_id(probably gated behindKconfigoption) in struct file (ideally)? Onexec()and when mmap()'ing with executable permissions, Linux kernel will try to fetchbuild_idfrom ELF file and if successful store it in struct file. Givenbuild_idcan be up to 20 bytes (currently) and not each struct file points to executable, we might want to only add a pointer field tostruct fileitself, which, ifbuild_idis present, will point tostruct build_id { char sz; char data[]; };This way we don't increase
struct fileby much.And then any BPF program would be able to easily
probe_read_kernelsuchbuild_idfromvma_area_struct->vm_file->build_id?
Quoted 7 lines
I'm sure I'm oversimplifying, but this seems like a good solution for all kinds of profiling BPF programs without the need to maintain all these allowlists and adding new helpers/kfuncs?
I know Hao was looking at the problem of getting
build_id, I'm curious if something like this would work for their use cases as well?
This helps a little. We would like to get build_id reliably. There are two problems we encountered.
First, sometimes we need to get build_id from an atomic context. We fail to get build_id if the page that contains the build_id has been evicted from pagecache. Storing the build_id in struct file or struct inode is a good and natural solution. But, this problem can also be solved by using mlock to pin the page in memory. We are using mlock, it seems to be working well right now.
The other problem we encountered may be very specific to our own use case. Sometimes we execute code that is mapped in an anonymous page (not backed by file). In that case, we can't get build_id either. What we are doing right now is writing the build_id into the vm_area_struct->anon_name field and teach build_id_parse to try parsing from there, when seeing an anonymous page. I can send this with upstream if there are other users who have the same problem.
Quoted 26 lines
Quoted 24 lines
We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)[...]
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 40 lines · Hao Luo
Quoted 26 lines · Andrii Nakryiko
Quoted 3 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.As a completely different way of solving this problem of retrieving
build_idfor tracing needs, can we teach kernel itself to parse and storebuild_id(probably gated behindKconfigoption) in struct file (ideally)? Onexec()and when mmap()'ing with executable permissions, Linux kernel will try to fetchbuild_idfrom ELF file and if successful store it in struct file. Givenbuild_idcan be up to 20 bytes (currently) and not each struct file points to executable, we might want to only add a pointer field tostruct fileitself, which, ifbuild_idis present, will point tostruct build_id { char sz; char data[]; };This way we don't increase
struct fileby much.And then any BPF program would be able to easily
probe_read_kernelsuchbuild_idfromvma_area_struct->vm_file->build_id?Quoted 7 lines
I'm sure I'm oversimplifying, but this seems like a good solution for all kinds of profiling BPF programs without the need to maintain all these allowlists and adding new helpers/kfuncs?
I know Hao was looking at the problem of getting
build_id, I'm curious if something like this would work for their use cases as well?This helps a little. We would like to get
build_idreliably. There are two problems we encountered.
Quoted 6 lines
First, sometimes we need to get
build_idfrom an atomic context. We fail to getbuild_idif the page that contains thebuild_idhas been evicted from pagecache. Storing thebuild_idinstruct fileorstruct inodeis a good and natural solution. But, this problem can also be solved by using mlock to pin the page in memory. We are using mlock, it seems to be working well right now.
This is hardly a generic solution, as it requires instrumenting every application to do this, right? So what I'm proposing is exactly to avoid having each individual application do something special just to allow profiling tools to capture build_id.
Quoted 9 lines
The other problem we encountered may be very specific to our own use case. Sometimes we execute code that is mapped in an anonymous page (not backed by file). In that case, we can't get
build_ideither. What we are doing right now is writing thebuild_idinto thevm_area_struct->anon_namefield and teachbuild_id_parseto try parsing from there, when seeing an anonymous page. I can send this with upstream if there are other users who have the same problem.
Is this due to remapping some binary onto huge pages?
But regardless, your custom BPF applications can fetch this build_id from vm_area_struct->anon_name in pure BPF code, can't it? Why do you need to modify in-kernel build_id_parse implementation?
Quoted 26 lines
Quoted 26 lines
Quoted 24 lines
We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)[...]
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 4 lines · Andrii Nakryiko
This is hardly a generic solution, as it requires instrumenting every application to do this, right? So what I'm proposing is exactly to avoid having each individual application do something special just to allow profiling tools to capture
build_id.
I agree. Because the mlock approach is working, we didn't look further or try improving it. But an upstreamable and generic solution would be nice. I think Jiri has started looking at it, I am happy to help there.
Quoted 1 line
Is this due to remapping some binary onto huge pages?
I think so, but I'm not sure.
Quoted 3 lines
But regardless, your custom BPF applications can fetch this
build_idfromvm_area_struct->anon_namein pure BPF code, can't it? Why do you need to modify in-kernelbuild_id_parseimplementation?
The user is using bpf_get_stack() to collect stack traces. They don't implement walking the stack and fetching build_id from vma in their BPF code.
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
Quoted 8 lines · Hao Luo
Quoted 4 lines · Andrii Nakryiko
This is hardly a generic solution, as it requires instrumenting every application to do this, right? So what I'm proposing is exactly to avoid having each individual application do something special just to allow profiling tools to capture
build_id.
Quoted 5 lines
I agree. Because the mlock approach is working, we didn't look further or try improving it. But an upstreamable and generic solution would be nice. I think Jiri has started looking at it, I am happy to help there.
Ok, cool, it would be great to have this work reliably and not rely on user-space apps doing something special here.
Quoted 4 lines
Quoted 1 line
Is this due to remapping some binary onto huge pages?
I think so, but I'm not sure.
We used to have this problem, but then Song added some in-kernel support that we now preserve the original file information. Song, do you mind providing details?
Quoted 7 lines
Quoted 3 lines
But regardless, your custom BPF applications can fetch this
build_idfromvm_area_struct->anon_namein pure BPF code, can't it? Why do you need to modify in-kernelbuild_id_parseimplementation?The user is using
bpf_get_stack()to collect stack traces. They don't implement walking the stack and fetchingbuild_idfrom vma in their BPF code.
Ah, I see. Let's figure out why Song's approach doesn't work in your case, because this anon_name hack is just that -- hack.
Re: [PATCHv4 bpf-next 2/4] bpf: Add bpf_vma_build_id_parse function and kfunc
In reply to Andrii NakryikoQuoted 20 lines · Andrii Nakryiko
Quoted 3 lines · Jiri Olsa
Adding
bpf_vma_build_id_parsefunction to retrieve build id from passed vma object and making it available as bpf kfunc.As a completely different way of solving this problem of retrieving
build_idfor tracing needs, can we teach kernel itself to parse and storebuild_id(probably gated behindKconfigoption) in struct file (ideally)? Onexec()and when mmap()'ing with executable permissions, Linux kernel will try to fetchbuild_idfrom ELF file and if successful store it in struct file. Givenbuild_idcan be up to 20 bytes (currently) and not each struct file points to executable, we might want to only add a pointer field tostruct fileitself, which, ifbuild_idis present, will point tostruct build_id { char sz; char data[]; };
Quoted 8 lines
This way we don't increase
struct fileby much.And then any BPF program would be able to easily
probe_read_kernelsuchbuild_idfromvma_area_struct->vm_file->build_id?I'm sure I'm oversimplifying, but this seems like a good solution for all kinds of profiling BPF programs without the need to maintain all these allowlists and adding new helpers/kfuncs?
good idea, that would make it much easier, will check
jirka
Quoted 30 lines
I know Hao was looking at the problem of getting
build_id, I'm curious if something like this would work for their use cases as well?Quoted 23 lines
We can't use
build_id_parsedirectly as kfunc, because we would not have control over the build id buffer size provided by user.Instead we are adding new
bpf_vma_build_id_parsefunction with 'build_id__sz' argument that instructs verifier to check for the available space inbuild_idbuffer.This way we check that there's always available memory space behind
build_idpointer. We also check that thebuild_id__szis at leastBUILD_ID_SIZE_MAXso we can place any buildid in.The
bpf_vma_build_id_parsekfunc is marked asKF_TRUSTED_ARGS, so it can be only called with trusted vma objects. These are currently provided only byfind_vmacallback function andtask_vmaiterator program.Signed-off-by Jiri Olsa <[email protected]>--- include/linux/bpf.h | 4 ++++ kernel/trace/bpf_trace.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 35 insertions(+)[...]
Jiri Olsa
reviewerPATCH v4 3/4 selftests/bpf: Add bpf_vma_build_id_parse find_vma callback test
In reply to Jiri OlsaAdding tests for using new bpf_vma_build_id_parse kfunc in find_vma callback function.
On bpf side the test finds the vma of the test_progs text through the test function pointer and reads its build id with the new kfunc.
On user side the test uses readelf to get test_progs build id and compares it with the one from bpf side.
--- .../bpf/prog_tests/bpf_vma_build_id_parse.c | 55 +++++++++++++++++++ .../bpf/progs/bpf_vma_build_id_parse.c | 40 ++++++++++++++ tools/testing/selftests/bpf/trace_helpers.c | 40 ++++++++++++++ tools/testing/selftests/bpf/trace_helpers.h | 1 + 4 files changed, 136 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_vma_build_id_parse.c create mode 100644 tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c
tools/testing/selftests/bpf/prog_tests/bpf_vma_build_id_parse.c new file +55 -0
@@ -0,0 +1,55 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <test_progs.h> +#include <sys/types.h> +#include <unistd.h> +#include <stdlib.h> +#include "bpf_vma_build_id_parse.skel.h" +#include "trace_helpers.h" + +#define BUILDID_STR_SIZE (BPF_BUILD_ID_SIZE*2 + 1) + +void test_bpf_vma_build_id_parse (void ) +{ +char bpf_build_id[BUILDID_STR_SIZE ] = {}, *build_id; +LIBBPF_OPTS (bpf_test_run_opts, topts); +struct bpf_vma_build_id_parse *skel; +int i, err, prog_fd; + + skel =bpf_vma_build_id_parse__open_and_load (); +if (!ASSERT_OK_PTR (skel,"bpf_vma_build_id_parse__open_and_load" )) +return ; + + skel->bss->target_pid =getpid (); + skel->bss->addr = (__u64 )(uintptr_t )test_bpf_vma_build_id_parse; + + err =bpf_vma_build_id_parse__attach (skel); +if (!ASSERT_OK (err,"bpf_vma_build_id_parse__attach" )) + goto out; + + prog_fd =bpf_program__fd (skel->progs.test1); + err =bpf_prog_test_run_opts (prog_fd, &topts); +ASSERT_OK (err,"test_run_err" ); +ASSERT_EQ (topts.retval,0 ,"test_run_retval" ); + +ASSERT_EQ (skel->data->ret,0 ,"ret" ); + +ASSERT_GT (skel->data->size_pass,0 ,"size_pass" ); +ASSERT_EQ (skel->data->size_fail, -EINVAL ,"size_fail" ); + +/* Read build id via readelf to compare with build_id. */ +if (!ASSERT_OK (read_self_buildid (&build_id),"read_buildid" )) + goto out; + +ASSERT_EQ (skel->data->size_pass,strlen (build_id)/2 ,"build_id_size" ); + +/* Convert bpf build id to string, so we can compare it later. */ +for (i =0 ; i < skel->data->size_pass; i++) { +sprintf (bpf_build_id + i*2 ,"%02x" , + (unsigned char ) skel->bss->build_id[i]); + } +ASSERT_STREQ (bpf_build_id, build_id,"build_id_match" ); + +free (build_id); +out: +bpf_vma_build_id_parse__destroy (skel); +}
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +40 -0
@@ -0,0 +1,40 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "vmlinux.h" +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> + +char _license[] SEC("license") = "GPL"; + +#define BPF_BUILD_ID_SIZE 20 + +extern int bpf_vma_build_id_parse (struct vm_area_struct *vma,unsigned char *build_id, + size_t build_id__sz) __ksym; + +pid_t target_pid = 0; +__u64 addr =0 ; + +int ret =-1 ; +int size_pass =-1 ; +int size_fail =-1 ; + +unsigned char build_id[BPF_BUILD_ID_SIZE ]; + +static long check_vma (struct task_struct *task,struct vm_area_struct *vma, +void *data) +{ + size_fail =bpf_vma_build_id_parse (vma, build_id,sizeof (build_id)/2 ); + size_pass =bpf_vma_build_id_parse (vma, build_id,sizeof (build_id)); +return 0 ; +} + +SEC("fentry/bpf_fentry_test1") +int BPF_PROG (test1 ,int a) +{ +struct task_struct *task =bpf_get_current_task_btf (); + +if (task->pid != target_pid) +return 0 ; + + ret =bpf_find_vma (task, addr, check_vma,NULL ,0 ); +return 0 ; +}
tools/testing/selftests/bpf/trace_helpers.c +40 -0
@@ -9,6 +9,7 @@#include <poll.h> #include <unistd.h> #include <linux/perf_event.h> +#include <linux/limits.h> #include <sys/mman.h> #include "trace_helpers.h"
@@ -230,3 +231,42 @@ ssize_t get_rel_offset(uintptr_t addr)fclose (f);return -EINVAL ; } + +int read_self_buildid (char **build_id) +{ +char path[PATH_MAX ], buf[PATH_MAX +200 ]; +char tmp[] ="/tmp/dataXXXXXX" ; +int err, fd; +FILE *f; + + fd =mkstemp (tmp); +if (fd ==-1 ) +return -1 ; +close (fd); + + err =readlink ("/proc/self/exe" , path,sizeof (path)); +if (err ==-1 ) + goto out; + path[err] =0 ; + +snprintf (buf,sizeof (buf), +"readelf -n %s 2>/dev/null | grep 'Build ID' | awk '{print $3}' > %s" , + path, tmp); + + err =system (buf); +if (err) + goto out; + + f =fopen (tmp,"r" ); +if (f) { +if (fscanf (f,"%ms$*\n" , build_id) !=1 ) { + *build_id =NULL ; + err =-1 ; + } +fclose (f); + } + +out: +unlink (tmp); +return err; +}
tools/testing/selftests/bpf/trace_helpers.h +1 -0
@@ -23,4 +23,5 @@ void read_trace_pipe(void);ssize_t get_uprobe_offset (const void *addr);ssize_t get_rel_offset (uintptr_t addr); +int read_self_buildid (char **build_id);#endif
Signature · 1 line
2.38.1
Re: [PATCHv4 bpf-next 3/4] selftests/bpf: Add bpf_vma_build_id_parse find_vma callback test
Quoted 12 lines · Jiri Olsa
Adding tests for using new
bpf_vma_build_id_parsekfunc infind_vmacallback function.On bpf side the test finds the vma of the
test_progstext through the test function pointer and reads its build id with the new kfunc.On user side the test uses readelf to get
test_progsbuild id and compares it with the one from bpf side.Signed-off-by Jiri Olsa <[email protected]>---
<...>
Quoted 5 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +0
-0
<...>
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c line 36
Quoted 8 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +8
-0+ +SEC("fentry/bpf_fentry_test1") +int BPF_PROG (test1 ,int a) +{ +struct task_struct *task =bpf_get_current_task_btf (); + +if (task->pid != target_pid) +return 0 ;
I think here we should use task->tgid. IIRC, task->pid corresponds to thread id in the userspace. task->tgid is the process id.
Quoted 4 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +4
-0+ + ret =bpf_find_vma (task, addr, check_vma,NULL ,0 ); +return 0 ; +}
<...>
Quoted 3 lines
--
2.38.1
Re: [PATCHv4 bpf-next 3/4] selftests/bpf: Add bpf_vma_build_id_parse find_vma callback test
Quoted 20 lines · Hao Luo
Quoted 12 lines · Jiri Olsa
Adding tests for using new
bpf_vma_build_id_parsekfunc infind_vmacallback function.On bpf side the test finds the vma of the
test_progstext through the test function pointer and reads its build id with the new kfunc.On user side the test uses readelf to get
test_progsbuild id and compares it with the one from bpf side.Signed-off-by Jiri Olsa <[email protected]>---
<...>
Quoted 5 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +0
-0<...>
Quoted 11 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c line 36
Quoted 8 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +8
-0+ +SEC("fentry/bpf_fentry_test1") +int BPF_PROG (test1 ,int a) +{ +struct task_struct *task =bpf_get_current_task_btf (); + +if (task->pid != target_pid) +return 0 ;I think here we should use
task->tgid. IIRC,task->pidcorresponds to thread id in the userspace.task->tgidis the process id.
right, will change
thanks,
jirka
Quoted 9 lines
Quoted 5 lines
tools/testing/selftests/bpf/progs/bpf_vma_build_id_parse.c new file +4
-0+ + ret =bpf_find_vma (task, addr, check_vma,NULL ,0 ); +return 0 ; +}<...>
Quoted 3 lines
--
2.38.1
Jiri Olsa
reviewerPATCH v4 4/4 selftests/bpf: Add bpf_vma_build_id_parse task vma iterator test
In reply to Jiri OlsaAdding tests for using new bpf_vma_build_id_parse kfunc in task_vma iterator program.
On bpf program side the iterator filters test proccess and proper vma by provided function pointer and reads its build id with the new kfunc.
On user side the test uses readelf to get test_progs build id and compares it with the one read from iterator.
--- .../selftests/bpf/prog_tests/bpf_iter.c | 44 +++++++++++++++++++ .../selftests/bpf/progs/bpf_iter_build_id.c | 41 +++++++++++++++++ 2 files changed, 85 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/bpf_iter_build_id.c
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +44 -0
@@ -33,6 +33,9 @@#include "bpf_iter_bpf_link.skel.h" #include "bpf_iter_ksym.skel.h" #include "bpf_iter_sockmap.skel.h" +#include "bpf_iter_build_id.skel.h" + +#define BUILDID_STR_SIZE (BPF_BUILD_ID_SIZE*2 + 1)static int duration;
@@ -1560,6 +1563,45 @@ static void test_task_vma_offset(void)test_task_vma_offset_common (NULL , false); } +static void test_task_vma_build_id (void ) +{ +struct bpf_iter_build_id *skel; +char buf[BUILDID_STR_SIZE ] = {}; +int iter_fd, len; +char *build_id; + + skel =bpf_iter_build_id__open_and_load (); +if (!ASSERT_OK_PTR (skel,"bpf_iter_vma_offset__open_and_load" )) +return ; + + skel->bss->pid =getpid (); + skel->bss->address = (uintptr_t )trigger_func; + + skel->links.vma_build_id =bpf_program__attach_iter (skel->progs.vma_build_id,NULL ); +if (!ASSERT_OK_PTR (skel->links.vma_build_id,"attach_iter" )) + goto exit; + + iter_fd =bpf_iter_create (bpf_link__fd (skel->links.vma_build_id)); +if (!ASSERT_GT (iter_fd,0 ,"create_iter" )) + goto exit; + +while ((len =read (iter_fd, buf,sizeof (buf))) >0 ) + ; + buf[BUILDID_STR_SIZE ] =0 ; + +/* Read build_id via readelf to compare with iterator buf. */ +if (!ASSERT_OK (read_self_buildid (&build_id),"read_buildid" )) + goto exit; + +ASSERT_STREQ (buf, build_id,"build_id_match" ); +ASSERT_GT (skel->data->size,0 ,"size" ); + +free (build_id); +close (iter_fd); +exit: +bpf_iter_build_id__destroy (skel); +} +void test_bpf_iter (void ) {ASSERT_OK (pthread_mutex_init (&do_nothing_mutex,NULL ),"pthread_mutex_init" );
@@ -1640,4 +1682,6 @@ void test_bpf_iter(void)test_bpf_sockmap_map_iter_fd ();if (test__start_subtest ("vma_offset" ))test_task_vma_offset (); +if (test__start_subtest ("vma_build_id" )) +test_task_vma_build_id (); }
tools/testing/selftests/bpf/progs/bpf_iter_build_id.c new file +41 -0
41 lines not on this page · raw message
Signature · 1 line
2.38.1
Re: [PATCHv4 bpf-next 4/4] selftests/bpf: Add bpf_vma_build_id_parse task vma iterator test
Quoted 23 lines · Jiri Olsa
Adding tests for using new
bpf_vma_build_id_parsekfunc intask_vmaiterator program.On bpf program side the iterator filters test proccess and proper vma by provided function pointer and reads its build id with the new kfunc.
On user side the test uses readelf to get
test_progsbuild id and compares it with the one read from iterator.Signed-off-by Jiri Olsa <[email protected]>--- .../selftests/bpf/prog_tests/bpf_iter.c | 44 +++++++++++++++++++ .../selftests/bpf/progs/bpf_iter_build_id.c | 41 +++++++++++++++++ 2 files changed, 85 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/bpf_iter_build_id.ctools/testing/selftests/bpf/prog_tests/bpf_iter.c +0
-0@@ -33,6 +33,9 @@
<...>
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1569
Quoted 5 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +4
-0+static void test_task_vma_build_id (void ) +{ +struct bpf_iter_build_id *skel; +char buf[BUILDID_STR_SIZE ] = {};
Jiri, do you mean buf[BUILDID_STR_SIZE + 1]? I see you have buf[BUILDID_STR_SIZE] = 0; below.
Quoted 2 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +2
-0+int iter_fd, len; +char *build_id;
<...>
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1589
Quoted 3 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +3
-0+ +while ((len =read (iter_fd, buf,sizeof (buf))) >0 ) + ;
I think you need to pass 'buf + len' to read(), otherwise the last iteration will overwrite the content read from the previous iterations.
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1594
Quoted 5 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +5
-0+ buf[BUILDID_STR_SIZE ] =0 ; + +/* Read build_id via readelf to compare with iterator buf. */ +if (!ASSERT_OK (read_self_buildid (&build_id),"read_buildid" )) + goto exit;
We need to close iter_fd before going to exit.
Quoted 10 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +10
-0+ +ASSERT_STREQ (buf, build_id,"build_id_match" ); +ASSERT_GT (skel->data->size,0 ,"size" ); + +free (build_id); +close (iter_fd); +exit: +bpf_iter_build_id__destroy (skel); +} +
<...>
Quoted 3 lines
--
2.38.1
Re: [PATCHv4 bpf-next 4/4] selftests/bpf: Add bpf_vma_build_id_parse task vma iterator test
Quoted 26 lines · Hao Luo
Quoted 23 lines · Jiri Olsa
Adding tests for using new
bpf_vma_build_id_parsekfunc intask_vmaiterator program.On bpf program side the iterator filters test proccess and proper vma by provided function pointer and reads its build id with the new kfunc.
On user side the test uses readelf to get
test_progsbuild id and compares it with the one read from iterator.Signed-off-by Jiri Olsa <[email protected]>--- .../selftests/bpf/prog_tests/bpf_iter.c | 44 +++++++++++++++++++ .../selftests/bpf/progs/bpf_iter_build_id.c | 41 +++++++++++++++++ 2 files changed, 85 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/bpf_iter_build_id.ctools/testing/selftests/bpf/prog_tests/bpf_iter.c +0
-0@@ -33,6 +33,9 @@<...>
Quoted 1 line
Quoted 7 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1569
Quoted 4 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +4
-0+static void test_task_vma_build_id (void ) +{ +struct bpf_iter_build_id *skel; +char buf[BUILDID_STR_SIZE ] = {};Jiri, do you mean buf[BUILDID_STR_SIZE + 1]? I see you have buf[BUILDID_STR_SIZE] = 0; below.
ugh, that plus one ended up in the BUILDID_STR_SIZE define, will fix
Quoted 11 lines
Quoted 3 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +2
-0+int iter_fd, len; +char *build_id;<...>
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1589
Quoted 3 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +3
-0+ +while ((len =read (iter_fd, buf,sizeof (buf))) >0 ) + ;I think you need to pass 'buf + len' to
read(), otherwise the last iteration will overwrite the content read from the previous iterations.
ok
Quoted 8 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c static void test_task_vma_offset(void) · line 1594
Quoted 6 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +5
-0+ buf[BUILDID_STR_SIZE ] =0 ; + +/* Read build_id via readelf to compare with iterator buf. */ +if (!ASSERT_OK (read_self_buildid (&build_id),"read_buildid" )) + goto exit;We need to close
iter_fdbefore going to exit.
right, will fix
thanks,
jirka
Quoted 15 lines
Quoted 11 lines
tools/testing/selftests/bpf/prog_tests/bpf_iter.c +10
-0+ +ASSERT_STREQ (buf, build_id,"build_id_match" ); +ASSERT_GT (skel->data->size,0 ,"size" ); + +free (build_id); +close (iter_fd); +exit: +bpf_iter_build_id__destroy (skel); +} +<...>
Quoted 3 lines
--
2.38.1