diff options
| author | Alexei Starovoitov <ast@kernel.org> | 2025-10-27 09:56:28 -0700 |
|---|---|---|
| committer | Alexei Starovoitov <ast@kernel.org> | 2025-10-27 09:56:28 -0700 |
| commit | ff880798de39251aa5e28ab389c16d294b7af658 (patch) | |
| tree | fd8c51839c4790e65e3130c6d54c532b3496d472 /include | |
| parent | e7586577b75f811bd14c12f2dd70afc3ece4756b (diff) | |
| parent | 784cdf931543805120aed1e1c43df8e6fa436a55 (diff) | |
Merge branch 'bpf-introduce-file-dynptr'
Mykyta Yatsenko says:
====================
bpf: Introduce file dynptr
From: Mykyta Yatsenko <yatsenko@meta.com>
This series adds a new dynptr kind, file dynptr, which enables BPF
programs to perform safe reads from files in a structured way.
Initial motivations include:
* Parsing the executable’s ELF to locate thread-local variable symbols
* Capturing stack traces when frame pointers are disabled
By leveraging the existing dynptr abstraction, we reuse the verifier’s
lifetime/size checks and keep the API consistent with existing dynptr
read helpers.
Technical details:
1. Reuses the existing freader library to read files a folio at a time.
2. bpf_dynptr_slice() and bpf_dynptr_read() always copy data from folios
into a program-provided buffer; zero-copy access is intentionally not
supported to keep it simple.
3. Reads may sleep if the requested folios are not in the page cache.
4. Few verifier changes required:
* Support dynptr destruction in kfuncs
* Add kfunc address substitution based on whether the program runs in
a sleepable or non-sleepable context.
Testing:
The final patch adds a selftest that validates BPF program reads the
same data as userspace, page faults are enabled in sleepable context and
disabled in non-sleepable.
Changelog:
---
v4 -> v5
v4: https://lore.kernel.org/all/20251021200334.220542-1-mykyta.yatsenko5@gmail.com/
* Inlined and removed kfunc_call_imm(), run overflow check for call_imm
only if !bpf_jit_supports_far_kfunc_call().
v3 -> v4
v3: https://lore.kernel.org/bpf/20251020222538.932915-1-mykyta.yatsenko5@gmail.com/
* Remove ringbuf usage from selftests
* bpf_dynptr_set_null(ptr) when discarding file dynptr
* call kfunc_call_imm() in specialize_kfunc() only, removed
call from add_kfunc_call()
v2 -> v3
v2: https://lore.kernel.org/bpf/20251015161155.120148-1-mykyta.yatsenko5@gmail.com/
* Add negative tests
* Rewrote tests to use LSM for bpf_get_task_exe_file()
* Move call_imm overflow check into kfunc_call_imm()
v1 -> v2
v1: https://lore.kernel.org/bpf/20251003160416.585080-1-mykyta.yatsenko5@gmail.com/
* Remove ELF parsing selftest
* Expanded u32 -> u64 refactoring, changes in include/uapi/linux/bpf.h
* Removed freader.{c,h}, instead move freader definitions into
buildid.h.
* Small refactoring of the multiple folios reading algorithm
* Directly return error after unmark_stack_slots_dynptr().
* Make kfuncs receive trusted arguments.
* Remove enum bpf_is_sleepable, use bool instead
* Remove unnecessary sorting from specialize_kfunc()
* Remove bool kfunc_in_sleepable_ctx; field from the struct
bpf_insn_aux_data, rely on non_sleepable field introduced by Kumar
* Refactor selftests, do madvise(...MADV_PAGEOUT) for all pages read by
the test
* Introduce the test for non-sleepable case, verify it fails with -EFAULT
====================
Link: https://lore.kernel.org/r/20251026203853.135105-1-mykyta.yatsenko5@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Diffstat (limited to 'include')
| -rw-r--r-- | include/linux/bpf.h | 30 | ||||
| -rw-r--r-- | include/linux/buildid.h | 25 | ||||
| -rw-r--r-- | include/uapi/linux/bpf.h | 8 |
3 files changed, 48 insertions, 15 deletions
diff --git a/include/linux/bpf.h b/include/linux/bpf.h index e53cda0aabb6..a47d67db3be5 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -670,6 +670,9 @@ static inline bool bpf_map_has_internal_structs(struct bpf_map *map) void bpf_map_free_internal_structs(struct bpf_map *map, void *obj); +int bpf_dynptr_from_file_sleepable(struct file *file, u32 flags, + struct bpf_dynptr *ptr__uninit); + extern const struct bpf_map_ops bpf_map_offload_ops; /* bpf_type_flag contains a set of flags that are applicable to the values of @@ -792,12 +795,15 @@ enum bpf_type_flag { /* DYNPTR points to skb_metadata_end()-skb_metadata_len() */ DYNPTR_TYPE_SKB_META = BIT(19 + BPF_BASE_TYPE_BITS), + /* DYNPTR points to file */ + DYNPTR_TYPE_FILE = BIT(20 + BPF_BASE_TYPE_BITS), + __BPF_TYPE_FLAG_MAX, __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, }; #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \ - | DYNPTR_TYPE_XDP | DYNPTR_TYPE_SKB_META) + | DYNPTR_TYPE_XDP | DYNPTR_TYPE_SKB_META | DYNPTR_TYPE_FILE) /* Max number of base types. */ #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) @@ -1385,21 +1391,23 @@ enum bpf_dynptr_type { BPF_DYNPTR_TYPE_XDP, /* Points to skb_metadata_end()-skb_metadata_len() */ BPF_DYNPTR_TYPE_SKB_META, + /* Underlying data is a file */ + BPF_DYNPTR_TYPE_FILE, }; -int bpf_dynptr_check_size(u32 size); -u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr); -const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u32 len); -void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u32 len); +int bpf_dynptr_check_size(u64 size); +u64 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr); +const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u64 len); +void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u64 len); bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr); -int __bpf_dynptr_write(const struct bpf_dynptr_kern *dst, u32 offset, - void *src, u32 len, u64 flags); -void *bpf_dynptr_slice_rdwr(const struct bpf_dynptr *p, u32 offset, - void *buffer__opt, u32 buffer__szk); +int __bpf_dynptr_write(const struct bpf_dynptr_kern *dst, u64 offset, + void *src, u64 len, u64 flags); +void *bpf_dynptr_slice_rdwr(const struct bpf_dynptr *p, u64 offset, + void *buffer__opt, u64 buffer__szk); -static inline int bpf_dynptr_check_off_len(const struct bpf_dynptr_kern *ptr, u32 offset, u32 len) +static inline int bpf_dynptr_check_off_len(const struct bpf_dynptr_kern *ptr, u64 offset, u64 len) { - u32 size = __bpf_dynptr_size(ptr); + u64 size = __bpf_dynptr_size(ptr); if (len > size || offset > size - len) return -E2BIG; diff --git a/include/linux/buildid.h b/include/linux/buildid.h index 014a88c41073..831c1b4b626c 100644 --- a/include/linux/buildid.h +++ b/include/linux/buildid.h @@ -18,4 +18,29 @@ void init_vmlinux_build_id(void); static inline void init_vmlinux_build_id(void) { } #endif +struct freader { + void *buf; + u32 buf_sz; + int err; + union { + struct { + struct file *file; + struct folio *folio; + void *addr; + loff_t folio_off; + bool may_fault; + }; + struct { + const char *data; + u64 data_sz; + }; + }; +}; + +void freader_init_from_file(struct freader *r, void *buf, u32 buf_sz, + struct file *file, bool may_fault); +void freader_init_from_mem(struct freader *r, const char *data, u64 data_sz); +const void *freader_fetch(struct freader *r, loff_t file_off, size_t sz); +void freader_cleanup(struct freader *r); + #endif diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 6829936d33f5..77edd0253989 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -5618,7 +5618,7 @@ union bpf_attr { * Return * *sk* if casting is valid, or **NULL** otherwise. * - * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) + * long bpf_dynptr_from_mem(void *data, u64 size, u64 flags, struct bpf_dynptr *ptr) * Description * Get a dynptr to local memory *data*. * @@ -5661,7 +5661,7 @@ union bpf_attr { * Return * Nothing. Always succeeds. * - * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) + * long bpf_dynptr_read(void *dst, u64 len, const struct bpf_dynptr *src, u64 offset, u64 flags) * Description * Read *len* bytes from *src* into *dst*, starting from *offset* * into *src*. @@ -5671,7 +5671,7 @@ union bpf_attr { * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if * *flags* is not 0. * - * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) + * long bpf_dynptr_write(const struct bpf_dynptr *dst, u64 offset, void *src, u64 len, u64 flags) * Description * Write *len* bytes from *src* into *dst*, starting from *offset* * into *dst*. @@ -5692,7 +5692,7 @@ union bpf_attr { * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). * - * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) + * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u64 offset, u64 len) * Description * Get a pointer to the underlying dynptr data. * |
