diff options
| -rw-r--r-- | Documentation/filesystems/proc.rst | 4 | ||||
| -rw-r--r-- | fs/coredump.c | 114 | ||||
| -rw-r--r-- | include/linux/coredump.h | 4 | ||||
| -rw-r--r-- | include/uapi/linux/coredump.h | 70 | ||||
| -rw-r--r-- | tools/include/uapi/linux/coredump.h | 70 | ||||
| -rw-r--r-- | tools/testing/selftests/coredump/coredump_socket_protocol_test.c | 1232 | ||||
| -rw-r--r-- | tools/testing/selftests/coredump/coredump_test_helpers.c | 279 | ||||
| -rw-r--r-- | tools/testing/selftests/coredump/coredump_test_helpers.h | 26 |
8 files changed, 1278 insertions, 521 deletions
diff --git a/Documentation/filesystems/proc.rst b/Documentation/filesystems/proc.rst index c102b62023cd..fc59c98acca1 100644 --- a/Documentation/filesystems/proc.rst +++ b/Documentation/filesystems/proc.rst @@ -1963,6 +1963,10 @@ For example:: $ echo 0x7 > /proc/self/coredump_filter $ ./some_program +If the coredump socket protocol is used a coredump server can select memory +types to include dynamically. See COREDUMP_MEMORY_TYPES in +include/uapi/linux/coredump.h. + 3.5 /proc/<pid>/mountinfo - Information about mounts -------------------------------------------------------- diff --git a/fs/coredump.c b/fs/coredump.c index dcaf998e54f6..33befe6b51c2 100644 --- a/fs/coredump.c +++ b/fs/coredump.c @@ -759,10 +759,39 @@ static inline bool coredump_sock_send(struct file *file, struct coredump_req *re return ret == sizeof(*req); } -static_assert(sizeof(struct coredump_req) == COREDUMP_REQ_SIZE_VER0); -static_assert(sizeof(struct coredump_ack) == COREDUMP_ACK_SIZE_VER0); +static_assert(sizeof(struct coredump_req) == COREDUMP_REQ_SIZE_VER1); +static_assert(sizeof(struct coredump_ack) == COREDUMP_ACK_SIZE_VER1); static_assert(sizeof(enum coredump_mark) == sizeof(__u32)); +/* Every memory type this kernel knows. */ +#define COREDUMP_MEMORY_ALL \ + (COREDUMP_MEMORY_ANON_PRIVATE | COREDUMP_MEMORY_ANON_SHARED | \ + COREDUMP_MEMORY_FILE_PRIVATE | COREDUMP_MEMORY_FILE_SHARED | \ + COREDUMP_MEMORY_ELF_HEADERS | \ + COREDUMP_MEMORY_HUGETLB_PRIVATE | COREDUMP_MEMORY_HUGETLB_SHARED | \ + COREDUMP_MEMORY_DAX_PRIVATE | COREDUMP_MEMORY_DAX_SHARED) + +#define COREDUMP_MEMORY_TYPE_BIT(mmf) BIT((mmf) - MMF_DUMP_FILTER_SHIFT) +static_assert(COREDUMP_MEMORY_ALL == (MMF_DUMP_FILTER_MASK >> MMF_DUMP_FILTER_SHIFT)); +static_assert(COREDUMP_MEMORY_ANON_PRIVATE == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_ANON_PRIVATE)); +static_assert(COREDUMP_MEMORY_ANON_SHARED == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_ANON_SHARED)); +static_assert(COREDUMP_MEMORY_FILE_PRIVATE == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_MAPPED_PRIVATE)); +static_assert(COREDUMP_MEMORY_FILE_SHARED == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_MAPPED_SHARED)); +static_assert(COREDUMP_MEMORY_ELF_HEADERS == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_ELF_HEADERS)); +static_assert(COREDUMP_MEMORY_HUGETLB_PRIVATE == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_HUGETLB_PRIVATE)); +static_assert(COREDUMP_MEMORY_HUGETLB_SHARED == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_HUGETLB_SHARED)); +static_assert(COREDUMP_MEMORY_DAX_PRIVATE == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_DAX_PRIVATE)); +static_assert(COREDUMP_MEMORY_DAX_SHARED == + COREDUMP_MEMORY_TYPE_BIT(MMF_DUMP_DAX_SHARED)); + static inline bool coredump_sock_mark(struct file *file, enum coredump_mark mark) { struct msghdr msg = { .msg_flags = MSG_NOSIGNAL }; @@ -804,11 +833,14 @@ static inline void coredump_sock_shutdown(struct file *file) static bool coredump_sock_request(struct core_name *cn, struct coredump_params *cprm) { struct coredump_req req = { - .size = sizeof(struct coredump_req), - .mask = COREDUMP_KERNEL | COREDUMP_USERSPACE | - COREDUMP_REJECT | COREDUMP_WAIT | - COREDUMP_RECORDS | COREDUMP_SPARSE, - .size_ack = sizeof(struct coredump_ack), + .size = sizeof(struct coredump_req), + .mask = COREDUMP_KERNEL | COREDUMP_USERSPACE | + COREDUMP_REJECT | COREDUMP_WAIT | + COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES, + .size_ack = sizeof(struct coredump_ack), + .memory_types = cprm->memory_types, + .memory_types_mask = COREDUMP_MEMORY_ALL, }; struct coredump_ack ack = {}; ssize_t usize; @@ -873,6 +905,30 @@ static bool coredump_sock_request(struct core_name *cn, struct coredump_params * return false; } + if (ack.mask & COREDUMP_MEMORY_TYPES) { + /* The memory types need the whole field. */ + if (usize < COREDUMP_ACK_SIZE_VER1) { + coredump_sock_mark(cprm->file, COREDUMP_MARK_MINSIZE); + return false; + } + + /* The memory types only select what the kernel writes. */ + if (!(ack.mask & COREDUMP_KERNEL)) { + coredump_sock_mark(cprm->file, COREDUMP_MARK_CONFLICTING); + return false; + } + + /* Refuse unknown memory types. */ + if (ack.memory_types & ~req.memory_types_mask) { + coredump_sock_mark(cprm->file, COREDUMP_MARK_UNSUPPORTED); + return false; + } + } else if (ack.memory_types) { + /* Like @spare the field must be zero when it isn't used. */ + coredump_sock_mark(cprm->file, COREDUMP_MARK_UNSUPPORTED); + return false; + } + /* Record header scratch; a bvec can't point at the stack. */ if (ack.mask & COREDUMP_RECORDS) { cprm->record_hdr = kmalloc_obj(*cprm->record_hdr); @@ -880,6 +936,10 @@ static bool coredump_sock_request(struct core_name *cn, struct coredump_params * return false; } + /* The server's selection replaces the task's entirely. */ + if (ack.mask & COREDUMP_MEMORY_TYPES) + cprm->memory_types = ack.memory_types; + cprm->mask = ack.mask; return coredump_sock_mark(cprm->file, COREDUMP_MARK_REQACK); } @@ -1190,6 +1250,10 @@ static void do_coredump(struct core_name *cn, struct coredump_params *cprm, } } +#define COREDUMP_TASK_MEMORY_TYPES(mm) \ + ((__mm_flags_get_word((mm)) & MMF_DUMP_FILTER_MASK) >> \ + MMF_DUMP_FILTER_SHIFT) + void vfs_coredump(const kernel_siginfo_t *siginfo) { size_t *argv __free(kfree) = NULL; @@ -1201,8 +1265,8 @@ void vfs_coredump(const kernel_siginfo_t *siginfo) struct coredump_params cprm = { .siginfo = siginfo, .limit = rlimit(RLIMIT_CORE), - /* Snapshot MMF_DUMP_FILTER_* (unlocked) and dumpable for the dump. */ - .mm_flags = __mm_flags_get_word(mm), + /* Snapshot the memory types (unlocked) and dumpable for the dump. */ + .memory_types = COREDUMP_TASK_MEMORY_TYPES(mm), .dumpable = task_exec_state_get_dumpable(current), .vma_meta = NULL, .cpu = raw_smp_processor_id(), @@ -1736,15 +1800,15 @@ static bool always_dump_vma(struct vm_area_struct *vma) } #define DUMP_SIZE_MAYBE_ELFHDR_PLACEHOLDER 1 +#define COREDUMP_MEMORY_TYPE_INCLUDE(types, type) \ + ((types) & COREDUMP_MEMORY_##type) /* * Decide how much of @vma's contents should be included in a core dump. */ static unsigned long vma_dump_size(struct vm_area_struct *vma, - unsigned long mm_flags) + u64 memory_types) { -#define FILTER(type) (mm_flags & (1UL << MMF_DUMP_##type)) - /* always dump the vdso and vsyscall sections */ if (always_dump_vma(vma)) goto whole; @@ -1754,18 +1818,22 @@ static unsigned long vma_dump_size(struct vm_area_struct *vma, /* support for DAX */ if (vma_is_dax(vma)) { - if ((vma->vm_flags & VM_SHARED) && FILTER(DAX_SHARED)) + if ((vma->vm_flags & VM_SHARED) && + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, DAX_SHARED)) goto whole; - if (!(vma->vm_flags & VM_SHARED) && FILTER(DAX_PRIVATE)) + if (!(vma->vm_flags & VM_SHARED) && + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, DAX_PRIVATE)) goto whole; return 0; } /* Hugetlb memory check */ if (is_vm_hugetlb_page(vma)) { - if ((vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_SHARED)) + if ((vma->vm_flags & VM_SHARED) && + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, HUGETLB_SHARED)) goto whole; - if (!(vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_PRIVATE)) + if (!(vma->vm_flags & VM_SHARED) && + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, HUGETLB_PRIVATE)) goto whole; return 0; } @@ -1777,25 +1845,27 @@ static unsigned long vma_dump_size(struct vm_area_struct *vma, /* By default, dump shared memory if mapped from an anonymous file. */ if (vma->vm_flags & VM_SHARED) { if (file_inode(vma->vm_file)->i_nlink == 0 ? - FILTER(ANON_SHARED) : FILTER(MAPPED_SHARED)) + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, ANON_SHARED) : + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, FILE_SHARED)) goto whole; return 0; } /* Dump segments that have been written to. */ - if ((!IS_ENABLED(CONFIG_MMU) || vma->anon_vma) && FILTER(ANON_PRIVATE)) + if ((!IS_ENABLED(CONFIG_MMU) || vma->anon_vma) && + COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, ANON_PRIVATE)) goto whole; if (vma->vm_file == NULL) return 0; - if (FILTER(MAPPED_PRIVATE)) + if (COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, FILE_PRIVATE)) goto whole; /* * If this is the beginning of an executable file mapping, * dump the first page to aid in determining what was mapped here. */ - if (FILTER(ELF_HEADERS) && + if (COREDUMP_MEMORY_TYPE_INCLUDE(memory_types, ELF_HEADERS) && vma->vm_pgoff == 0 && (vma->vm_flags & VM_READ)) { if ((READ_ONCE(file_inode(vma->vm_file)->i_mode) & 0111) != 0) return PAGE_SIZE; @@ -1811,8 +1881,6 @@ static unsigned long vma_dump_size(struct vm_area_struct *vma, return DUMP_SIZE_MAYBE_ELFHDR_PLACEHOLDER; } -#undef FILTER - return 0; whole: @@ -1897,7 +1965,7 @@ static bool dump_vma_snapshot(struct coredump_params *cprm) m->start = vma->vm_start; m->end = vma->vm_end; m->flags = vma->vm_flags; - m->dump_size = vma_dump_size(vma, cprm->mm_flags); + m->dump_size = vma_dump_size(vma, cprm->memory_types); m->pgoff = vma->vm_pgoff; m->file = vma->vm_file; if (m->file) diff --git a/include/linux/coredump.h b/include/linux/coredump.h index b252bb2843b3..74af57b9406b 100644 --- a/include/linux/coredump.h +++ b/include/linux/coredump.h @@ -32,8 +32,8 @@ struct coredump_params { const kernel_siginfo_t *siginfo; struct file *file; unsigned long limit; - /* MMF_DUMP_FILTER_* bits, snapshot of mm->flags at dump start. */ - unsigned long mm_flags; + /* COREDUMP_MEMORY_* types to dump, the task's or the server's. */ + u64 memory_types; /* Snapshot of dumpable at dump start. */ enum task_dumpable dumpable; int cpu; diff --git a/include/uapi/linux/coredump.h b/include/uapi/linux/coredump.h index f3771861ca48..6d0c53b534ea 100644 --- a/include/uapi/linux/coredump.h +++ b/include/uapi/linux/coredump.h @@ -16,6 +16,9 @@ * requires COREDUMP_KERNEL * @COREDUMP_SPARSE: describe the holes in the coredump as zero records * instead of transferring them; requires COREDUMP_RECORDS + * @COREDUMP_MEMORY_TYPES: dump the memory types in + * coredump_ack->memory_types instead of the ones + * the task selected; requires COREDUMP_KERNEL */ enum { COREDUMP_KERNEL = (1ULL << 0), @@ -24,6 +27,37 @@ enum { COREDUMP_WAIT = (1ULL << 3), COREDUMP_RECORDS = (1ULL << 4), COREDUMP_SPARSE = (1ULL << 5), + COREDUMP_MEMORY_TYPES = (1ULL << 6), +}; + +/** + * coredump memory types + * @COREDUMP_MEMORY_ANON_PRIVATE: anonymous private memory + * @COREDUMP_MEMORY_ANON_SHARED: anonymous shared memory + * @COREDUMP_MEMORY_FILE_PRIVATE: file-backed private memory + * @COREDUMP_MEMORY_FILE_SHARED: file-backed shared memory + * @COREDUMP_MEMORY_ELF_HEADERS: the first page of a file-backed private + * mapping that starts an ELF file + * @COREDUMP_MEMORY_HUGETLB_PRIVATE: hugetlb private memory + * @COREDUMP_MEMORY_HUGETLB_SHARED: hugetlb shared memory + * @COREDUMP_MEMORY_DAX_PRIVATE: DAX private memory + * @COREDUMP_MEMORY_DAX_SHARED: DAX shared memory + * + * A bitmask of memory types a coredump may request to be included. New + * memory type bits must ensure that they do not steal memory from an + * existing one so a coredump server will continue to get the same + * coredumps even if a new bit is introduced. + */ +enum { + COREDUMP_MEMORY_ANON_PRIVATE = (1ULL << 0), + COREDUMP_MEMORY_ANON_SHARED = (1ULL << 1), + COREDUMP_MEMORY_FILE_PRIVATE = (1ULL << 2), + COREDUMP_MEMORY_FILE_SHARED = (1ULL << 3), + COREDUMP_MEMORY_ELF_HEADERS = (1ULL << 4), + COREDUMP_MEMORY_HUGETLB_PRIVATE = (1ULL << 5), + COREDUMP_MEMORY_HUGETLB_SHARED = (1ULL << 6), + COREDUMP_MEMORY_DAX_PRIVATE = (1ULL << 7), + COREDUMP_MEMORY_DAX_SHARED = (1ULL << 8), }; /** @@ -31,6 +65,8 @@ enum { * @size: size of struct coredump_req * @size_ack: known size of struct coredump_ack on this kernel * @mask: supported features + * @memory_types: the memory types the task selected + * @memory_types_mask: the memory types this kernel knows * * When a coredump happens the kernel will connect to the coredump * socket and send a coredump request to the coredump server. The @size @@ -49,15 +85,27 @@ enum { * struct coredump_ack the kernel knows. Userspace may only send up to * coredump_req->size_ack bytes to the kernel and must set * coredump_ack->size accordingly. + * + * @memory_types is set to the default memory types that are included in + * the coredump. This can be overridden by raising bits in + * coredump_ack->memory_types. + * + * @memory_types_mask contains a bitmask of all memory types the kernel + * knows about. A coredump server may only raise bits in + * coredump_ack->memory_types that are raised in + * coredump_req->memory_types_mask. */ struct coredump_req { __u32 size; __u32 size_ack; __u64 mask; + __u64 memory_types; + __u64 memory_types_mask; }; enum { COREDUMP_REQ_SIZE_VER0 = 16U, /* size of first published struct */ + COREDUMP_REQ_SIZE_VER1 = 32U, /* memory_types and memory_types_mask added */ }; /** @@ -65,6 +113,8 @@ enum { * @size: size of the struct * @spare: unused * @mask: features kernel is supposed to use + * @memory_types: memory types to dump, only with COREDUMP_MEMORY_TYPES + * in @mask * * The @size member must be set to the size of struct coredump_ack. It * may never exceed what the kernel returned in coredump_req->size_ack @@ -74,15 +124,30 @@ enum { * The @mask member must be set to the features the coredump server * wants the kernel to use. Only bits the kernel returned in * coredump_req->mask may be set. + * + * If COREDUMP_MEMORY_TYPES is raised in @mask the kernel dumps the + * memory types set in the @memory_types mask. Zero is valid and dumps + * no memory apart from the mappings that are always dumped. + * + * Note that memory a task excluded via MADV_DONTDUMP is always left + * out. A coredump server wanting to add or drop memory types instead of + * outright replacing it should simply copy coredump_req->memory_types + * and then mask off or raise types as needed. + * + * Note that @memory_types must be zero if COREDUMP_MEMORY_TYPES isn't + * raised. COREDUMP_MEMORY_TYPES requires COREDUMP_KERNEL and an ack of + * at least COREDUMP_ACK_SIZE_VER1 bytes. */ struct coredump_ack { __u32 size; __u32 spare; __u64 mask; + __u64 memory_types; }; enum { COREDUMP_ACK_SIZE_VER0 = 16U, /* size of first published struct */ + COREDUMP_ACK_SIZE_VER1 = 24U, /* memory_types added */ }; /** @@ -90,11 +155,12 @@ enum { * * The kernel will place a single byte on the coredump socket. The * markers notify userspace whether the coredump ack succeeded or - * failed. + * failed. After any marker other than COREDUMP_MARK_REQACK the kernel + * closes the connection and no coredump is generated. * * @COREDUMP_MARK_MINSIZE: the provided coredump_ack size was too small * @COREDUMP_MARK_MAXSIZE: the provided coredump_ack size was too big - * @COREDUMP_MARK_UNSUPPORTED: the provided coredump_ack mask was invalid + * @COREDUMP_MARK_UNSUPPORTED: the provided coredump_ack mask or memory types were invalid * @COREDUMP_MARK_CONFLICTING: the provided coredump_ack mask has conflicting options * @COREDUMP_MARK_REQACK: the coredump request and ack was successful * @__COREDUMP_MARK_MAX: the maximum coredump mark value diff --git a/tools/include/uapi/linux/coredump.h b/tools/include/uapi/linux/coredump.h index f3771861ca48..6d0c53b534ea 100644 --- a/tools/include/uapi/linux/coredump.h +++ b/tools/include/uapi/linux/coredump.h @@ -16,6 +16,9 @@ * requires COREDUMP_KERNEL * @COREDUMP_SPARSE: describe the holes in the coredump as zero records * instead of transferring them; requires COREDUMP_RECORDS + * @COREDUMP_MEMORY_TYPES: dump the memory types in + * coredump_ack->memory_types instead of the ones + * the task selected; requires COREDUMP_KERNEL */ enum { COREDUMP_KERNEL = (1ULL << 0), @@ -24,6 +27,37 @@ enum { COREDUMP_WAIT = (1ULL << 3), COREDUMP_RECORDS = (1ULL << 4), COREDUMP_SPARSE = (1ULL << 5), + COREDUMP_MEMORY_TYPES = (1ULL << 6), +}; + +/** + * coredump memory types + * @COREDUMP_MEMORY_ANON_PRIVATE: anonymous private memory + * @COREDUMP_MEMORY_ANON_SHARED: anonymous shared memory + * @COREDUMP_MEMORY_FILE_PRIVATE: file-backed private memory + * @COREDUMP_MEMORY_FILE_SHARED: file-backed shared memory + * @COREDUMP_MEMORY_ELF_HEADERS: the first page of a file-backed private + * mapping that starts an ELF file + * @COREDUMP_MEMORY_HUGETLB_PRIVATE: hugetlb private memory + * @COREDUMP_MEMORY_HUGETLB_SHARED: hugetlb shared memory + * @COREDUMP_MEMORY_DAX_PRIVATE: DAX private memory + * @COREDUMP_MEMORY_DAX_SHARED: DAX shared memory + * + * A bitmask of memory types a coredump may request to be included. New + * memory type bits must ensure that they do not steal memory from an + * existing one so a coredump server will continue to get the same + * coredumps even if a new bit is introduced. + */ +enum { + COREDUMP_MEMORY_ANON_PRIVATE = (1ULL << 0), + COREDUMP_MEMORY_ANON_SHARED = (1ULL << 1), + COREDUMP_MEMORY_FILE_PRIVATE = (1ULL << 2), + COREDUMP_MEMORY_FILE_SHARED = (1ULL << 3), + COREDUMP_MEMORY_ELF_HEADERS = (1ULL << 4), + COREDUMP_MEMORY_HUGETLB_PRIVATE = (1ULL << 5), + COREDUMP_MEMORY_HUGETLB_SHARED = (1ULL << 6), + COREDUMP_MEMORY_DAX_PRIVATE = (1ULL << 7), + COREDUMP_MEMORY_DAX_SHARED = (1ULL << 8), }; /** @@ -31,6 +65,8 @@ enum { * @size: size of struct coredump_req * @size_ack: known size of struct coredump_ack on this kernel * @mask: supported features + * @memory_types: the memory types the task selected + * @memory_types_mask: the memory types this kernel knows * * When a coredump happens the kernel will connect to the coredump * socket and send a coredump request to the coredump server. The @size @@ -49,15 +85,27 @@ enum { * struct coredump_ack the kernel knows. Userspace may only send up to * coredump_req->size_ack bytes to the kernel and must set * coredump_ack->size accordingly. + * + * @memory_types is set to the default memory types that are included in + * the coredump. This can be overridden by raising bits in + * coredump_ack->memory_types. + * + * @memory_types_mask contains a bitmask of all memory types the kernel + * knows about. A coredump server may only raise bits in + * coredump_ack->memory_types that are raised in + * coredump_req->memory_types_mask. */ struct coredump_req { __u32 size; __u32 size_ack; __u64 mask; + __u64 memory_types; + __u64 memory_types_mask; }; enum { COREDUMP_REQ_SIZE_VER0 = 16U, /* size of first published struct */ + COREDUMP_REQ_SIZE_VER1 = 32U, /* memory_types and memory_types_mask added */ }; /** @@ -65,6 +113,8 @@ enum { * @size: size of the struct * @spare: unused * @mask: features kernel is supposed to use + * @memory_types: memory types to dump, only with COREDUMP_MEMORY_TYPES + * in @mask * * The @size member must be set to the size of struct coredump_ack. It * may never exceed what the kernel returned in coredump_req->size_ack @@ -74,15 +124,30 @@ enum { * The @mask member must be set to the features the coredump server * wants the kernel to use. Only bits the kernel returned in * coredump_req->mask may be set. + * + * If COREDUMP_MEMORY_TYPES is raised in @mask the kernel dumps the + * memory types set in the @memory_types mask. Zero is valid and dumps + * no memory apart from the mappings that are always dumped. + * + * Note that memory a task excluded via MADV_DONTDUMP is always left + * out. A coredump server wanting to add or drop memory types instead of + * outright replacing it should simply copy coredump_req->memory_types + * and then mask off or raise types as needed. + * + * Note that @memory_types must be zero if COREDUMP_MEMORY_TYPES isn't + * raised. COREDUMP_MEMORY_TYPES requires COREDUMP_KERNEL and an ack of + * at least COREDUMP_ACK_SIZE_VER1 bytes. */ struct coredump_ack { __u32 size; __u32 spare; __u64 mask; + __u64 memory_types; }; enum { COREDUMP_ACK_SIZE_VER0 = 16U, /* size of first published struct */ + COREDUMP_ACK_SIZE_VER1 = 24U, /* memory_types added */ }; /** @@ -90,11 +155,12 @@ enum { * * The kernel will place a single byte on the coredump socket. The * markers notify userspace whether the coredump ack succeeded or - * failed. + * failed. After any marker other than COREDUMP_MARK_REQACK the kernel + * closes the connection and no coredump is generated. * * @COREDUMP_MARK_MINSIZE: the provided coredump_ack size was too small * @COREDUMP_MARK_MAXSIZE: the provided coredump_ack size was too big - * @COREDUMP_MARK_UNSUPPORTED: the provided coredump_ack mask was invalid + * @COREDUMP_MARK_UNSUPPORTED: the provided coredump_ack mask or memory types were invalid * @COREDUMP_MARK_CONFLICTING: the provided coredump_ack mask has conflicting options * @COREDUMP_MARK_REQACK: the coredump request and ack was successful * @__COREDUMP_MARK_MAX: the maximum coredump mark value diff --git a/tools/testing/selftests/coredump/coredump_socket_protocol_test.c b/tools/testing/selftests/coredump/coredump_socket_protocol_test.c index daff908232a2..f5c9bad87546 100644 --- a/tools/testing/selftests/coredump/coredump_socket_protocol_test.c +++ b/tools/testing/selftests/coredump/coredump_socket_protocol_test.c @@ -508,91 +508,89 @@ out: wait_and_check_coredump_server(pid_coredump_server, _metadata, self); } -TEST_F(coredump, socket_request_invalid_flag_combination) +/* An ack the kernel must refuse and how. */ +struct refused_ack { + /* The ack, and how many bytes of it the server sends before it hangs up. */ + struct coredump_ack ack; + size_t bytes; + /* The marker the kernel answers with, or none if @no_marker. */ + enum coredump_mark mark; + bool no_marker; +}; + +/* Send @refused, expect the kernel to refuse it and hang up. */ +static void check_refused_ack(struct __test_metadata *const _metadata, + FIXTURE_DATA(coredump) *self, + const struct refused_ack *refused) { - int pidfd, ret, status; + int pidfd, status; pid_t pid, pid_coredump_server; struct pidfd_info info = {}; int ipc_sockets[2]; char c; + ASSERT_EQ(socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets), 0); ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); - ret = socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); - ASSERT_EQ(ret, 0); - pid_coredump_server = fork(); ASSERT_GE(pid_coredump_server, 0); if (pid_coredump_server == 0) { - struct coredump_req req = {}; int fd_server = -1, fd_coredump = -1, fd_peer_pidfd = -1; int exit_code = EXIT_FAILURE; + struct coredump_req req = {}; close(ipc_sockets[0]); fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); - if (fd_server < 0) { - fprintf(stderr, "socket_request_invalid_flag_combination: create_and_listen_unix_socket failed: %m\n"); + if (fd_server < 0) goto out; - } - if (write_nointr(ipc_sockets[1], "1", 1) < 0) { - fprintf(stderr, "socket_request_invalid_flag_combination: write_nointr to ipc socket failed: %m\n"); + if (write_nointr(ipc_sockets[1], "1", 1) < 0) goto out; - } close(ipc_sockets[1]); fd_coredump = accept4(fd_server, NULL, NULL, SOCK_CLOEXEC); - if (fd_coredump < 0) { - fprintf(stderr, "socket_request_invalid_flag_combination: accept4 failed: %m\n"); + if (fd_coredump < 0) goto out; - } fd_peer_pidfd = get_peer_pidfd(fd_coredump); - if (fd_peer_pidfd < 0) { - fprintf(stderr, "socket_request_invalid_flag_combination: get_peer_pidfd failed\n"); + if (fd_peer_pidfd < 0) goto out; - } - if (!get_pidfd_info(fd_peer_pidfd, &info)) { - fprintf(stderr, "socket_request_invalid_flag_combination: get_pidfd_info failed\n"); + /* The task shows as dumping while it waits for the ack. */ + if (!get_pidfd_info(fd_peer_pidfd, &info)) goto out; - } - if (!(info.mask & PIDFD_INFO_COREDUMP)) { - fprintf(stderr, "socket_request_invalid_flag_combination: PIDFD_INFO_COREDUMP not set in mask\n"); + if (!(info.mask & PIDFD_INFO_COREDUMP) || + !(info.coredump_mask & PIDFD_COREDUMPED)) { + fprintf(stderr, "Peer isn't marked as dumping\n"); goto out; } - if (!(info.coredump_mask & PIDFD_COREDUMPED)) { - fprintf(stderr, "socket_request_invalid_flag_combination: PIDFD_COREDUMPED not set in coredump_mask\n"); + if (!read_coredump_req(fd_coredump, &req)) goto out; - } - if (!read_coredump_req(fd_coredump, &req)) { - fprintf(stderr, "socket_request_invalid_flag_combination: read_coredump_req failed\n"); + if (!check_coredump_req(&req)) goto out; - } - if (!check_coredump_req(&req)) { - fprintf(stderr, "socket_request_invalid_flag_combination: check_coredump_req failed\n"); + if (!send_coredump_ack_bytes(fd_coredump, &refused->ack, + refused->bytes)) goto out; - } - if (!send_coredump_ack(fd_coredump, &req, - COREDUMP_KERNEL | COREDUMP_REJECT | COREDUMP_WAIT, 0)) { - fprintf(stderr, "socket_request_invalid_flag_combination: send_coredump_ack failed\n"); + /* Nothing more to say. A server that died looks the same. */ + if (shutdown(fd_coredump, SHUT_WR)) goto out; - } - if (!read_marker(fd_coredump, COREDUMP_MARK_CONFLICTING)) { - fprintf(stderr, "socket_request_invalid_flag_combination: read_marker COREDUMP_MARK_CONFLICTING failed\n"); + if (!refused->no_marker && + !read_marker(fd_coredump, refused->mark)) + goto out; + + /* The kernel hangs up after a refusal, marker or not. */ + if (!read_hangup(fd_coredump)) goto out; - } exit_code = EXIT_SUCCESS; - fprintf(stderr, "socket_request_invalid_flag_combination: completed successfully\n"); out: if (fd_peer_pidfd >= 0) close(fd_peer_pidfd); @@ -627,362 +625,72 @@ out: wait_and_check_coredump_server(pid_coredump_server, _metadata, self); } -TEST_F(coredump, socket_request_unknown_flag) +/* Ack @ack_mask, expect the kernel to refuse it as conflicting. */ +static void check_conflicting_ack(struct __test_metadata *const _metadata, + FIXTURE_DATA(coredump) *self, __u64 ack_mask) { - int pidfd, ret, status; - pid_t pid, pid_coredump_server; - struct pidfd_info info = {}; - int ipc_sockets[2]; - char c; - - ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); - - ret = socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); - ASSERT_EQ(ret, 0); - - pid_coredump_server = fork(); - ASSERT_GE(pid_coredump_server, 0); - if (pid_coredump_server == 0) { - struct coredump_req req = {}; - int fd_server = -1, fd_coredump = -1, fd_peer_pidfd = -1; - int exit_code = EXIT_FAILURE; - - close(ipc_sockets[0]); - - fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); - if (fd_server < 0) { - fprintf(stderr, "socket_request_unknown_flag: create_and_listen_unix_socket failed: %m\n"); - goto out; - } - - if (write_nointr(ipc_sockets[1], "1", 1) < 0) { - fprintf(stderr, "socket_request_unknown_flag: write_nointr to ipc socket failed: %m\n"); - goto out; - } - - close(ipc_sockets[1]); - - fd_coredump = accept4(fd_server, NULL, NULL, SOCK_CLOEXEC); - if (fd_coredump < 0) { - fprintf(stderr, "socket_request_unknown_flag: accept4 failed: %m\n"); - goto out; - } - - fd_peer_pidfd = get_peer_pidfd(fd_coredump); - if (fd_peer_pidfd < 0) { - fprintf(stderr, "socket_request_unknown_flag: get_peer_pidfd failed\n"); - goto out; - } - - if (!get_pidfd_info(fd_peer_pidfd, &info)) { - fprintf(stderr, "socket_request_unknown_flag: get_pidfd_info failed\n"); - goto out; - } - - if (!(info.mask & PIDFD_INFO_COREDUMP)) { - fprintf(stderr, "socket_request_unknown_flag: PIDFD_INFO_COREDUMP not set in mask\n"); - goto out; - } - - if (!(info.coredump_mask & PIDFD_COREDUMPED)) { - fprintf(stderr, "socket_request_unknown_flag: PIDFD_COREDUMPED not set in coredump_mask\n"); - goto out; - } - - if (!read_coredump_req(fd_coredump, &req)) { - fprintf(stderr, "socket_request_unknown_flag: read_coredump_req failed\n"); - goto out; - } - - if (!check_coredump_req(&req)) { - fprintf(stderr, "socket_request_unknown_flag: check_coredump_req failed\n"); - goto out; - } - - if (!send_coredump_ack(fd_coredump, &req, (1ULL << 63), 0)) { - fprintf(stderr, "socket_request_unknown_flag: send_coredump_ack failed\n"); - goto out; - } - - if (!read_marker(fd_coredump, COREDUMP_MARK_UNSUPPORTED)) { - fprintf(stderr, "socket_request_unknown_flag: read_marker COREDUMP_MARK_UNSUPPORTED failed\n"); - goto out; - } - - exit_code = EXIT_SUCCESS; - fprintf(stderr, "socket_request_unknown_flag: completed successfully\n"); -out: - if (fd_peer_pidfd >= 0) - close(fd_peer_pidfd); - if (fd_coredump >= 0) - close(fd_coredump); - if (fd_server >= 0) - close(fd_server); - _exit(exit_code); - } - self->pid_coredump_server = pid_coredump_server; - - EXPECT_EQ(close(ipc_sockets[1]), 0); - ASSERT_EQ(read_nointr(ipc_sockets[0], &c, 1), 1); - EXPECT_EQ(close(ipc_sockets[0]), 0); - - pid = fork(); - ASSERT_GE(pid, 0); - if (pid == 0) - crashing_child(); - - pidfd = sys_pidfd_open(pid, 0); - ASSERT_GE(pidfd, 0); - - waitpid(pid, &status, 0); - ASSERT_TRUE(WIFSIGNALED(status)); - ASSERT_FALSE(WCOREDUMP(status)); + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .mask = ack_mask, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_CONFLICTING, + }; + + check_refused_ack(_metadata, self, &refused); +} - ASSERT_TRUE(get_pidfd_info(pidfd, &info)); - ASSERT_GT((info.mask & PIDFD_INFO_COREDUMP), 0); - ASSERT_GT((info.coredump_mask & PIDFD_COREDUMPED), 0); +/* More than one of KERNEL, USERSPACE and REJECT. */ +TEST_F(coredump, socket_request_invalid_flag_combination) +{ + check_conflicting_ack(_metadata, self, + COREDUMP_KERNEL | COREDUMP_REJECT | COREDUMP_WAIT); +} - wait_and_check_coredump_server(pid_coredump_server, _metadata, self); +/* A flag the kernel didn't advertise in coredump_req->mask. */ +TEST_F(coredump, socket_request_unknown_flag) +{ + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .mask = 1ULL << 63, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_UNSUPPORTED, + }; + + check_refused_ack(_metadata, self, &refused); } +/* An ack smaller than the first published struct. */ TEST_F(coredump, socket_request_invalid_size_small) { - int pidfd, ret, status; - pid_t pid, pid_coredump_server; - struct pidfd_info info = {}; - int ipc_sockets[2]; - char c; - - ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); - - ret = socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); - ASSERT_EQ(ret, 0); - - pid_coredump_server = fork(); - ASSERT_GE(pid_coredump_server, 0); - if (pid_coredump_server == 0) { - struct coredump_req req = {}; - int fd_server = -1, fd_coredump = -1, fd_peer_pidfd = -1; - int exit_code = EXIT_FAILURE; - - close(ipc_sockets[0]); - - fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); - if (fd_server < 0) { - fprintf(stderr, "socket_request_invalid_size_small: create_and_listen_unix_socket failed: %m\n"); - goto out; - } - - if (write_nointr(ipc_sockets[1], "1", 1) < 0) { - fprintf(stderr, "socket_request_invalid_size_small: write_nointr to ipc socket failed: %m\n"); - goto out; - } - - close(ipc_sockets[1]); - - fd_coredump = accept4(fd_server, NULL, NULL, SOCK_CLOEXEC); - if (fd_coredump < 0) { - fprintf(stderr, "socket_request_invalid_size_small: accept4 failed: %m\n"); - goto out; - } - - fd_peer_pidfd = get_peer_pidfd(fd_coredump); - if (fd_peer_pidfd < 0) { - fprintf(stderr, "socket_request_invalid_size_small: get_peer_pidfd failed\n"); - goto out; - } - - if (!get_pidfd_info(fd_peer_pidfd, &info)) { - fprintf(stderr, "socket_request_invalid_size_small: get_pidfd_info failed\n"); - goto out; - } - - if (!(info.mask & PIDFD_INFO_COREDUMP)) { - fprintf(stderr, "socket_request_invalid_size_small: PIDFD_INFO_COREDUMP not set in mask\n"); - goto out; - } - - if (!(info.coredump_mask & PIDFD_COREDUMPED)) { - fprintf(stderr, "socket_request_invalid_size_small: PIDFD_COREDUMPED not set in coredump_mask\n"); - goto out; - } - - if (!read_coredump_req(fd_coredump, &req)) { - fprintf(stderr, "socket_request_invalid_size_small: read_coredump_req failed\n"); - goto out; - } - - if (!check_coredump_req(&req)) { - fprintf(stderr, "socket_request_invalid_size_small: check_coredump_req failed\n"); - goto out; - } - - if (!send_coredump_ack(fd_coredump, &req, - COREDUMP_REJECT | COREDUMP_WAIT, - COREDUMP_ACK_SIZE_VER0 / 2)) { - fprintf(stderr, "socket_request_invalid_size_small: send_coredump_ack failed\n"); - goto out; - } - - if (!read_marker(fd_coredump, COREDUMP_MARK_MINSIZE)) { - fprintf(stderr, "socket_request_invalid_size_small: read_marker COREDUMP_MARK_MINSIZE failed\n"); - goto out; - } - - exit_code = EXIT_SUCCESS; - fprintf(stderr, "socket_request_invalid_size_small: completed successfully\n"); -out: - if (fd_peer_pidfd >= 0) - close(fd_peer_pidfd); - if (fd_coredump >= 0) - close(fd_coredump); - if (fd_server >= 0) - close(fd_server); - _exit(exit_code); - } - self->pid_coredump_server = pid_coredump_server; - - EXPECT_EQ(close(ipc_sockets[1]), 0); - ASSERT_EQ(read_nointr(ipc_sockets[0], &c, 1), 1); - EXPECT_EQ(close(ipc_sockets[0]), 0); - - pid = fork(); - ASSERT_GE(pid, 0); - if (pid == 0) - crashing_child(); - - pidfd = sys_pidfd_open(pid, 0); - ASSERT_GE(pidfd, 0); - - waitpid(pid, &status, 0); - ASSERT_TRUE(WIFSIGNALED(status)); - ASSERT_FALSE(WCOREDUMP(status)); - - ASSERT_TRUE(get_pidfd_info(pidfd, &info)); - ASSERT_GT((info.mask & PIDFD_INFO_COREDUMP), 0); - ASSERT_GT((info.coredump_mask & PIDFD_COREDUMPED), 0); - - wait_and_check_coredump_server(pid_coredump_server, _metadata, self); + struct refused_ack refused = { + .ack = { + .size = COREDUMP_ACK_SIZE_VER0 / 2, + .mask = COREDUMP_REJECT | COREDUMP_WAIT, + }, + .bytes = COREDUMP_ACK_SIZE_VER0 / 2, + .mark = COREDUMP_MARK_MINSIZE, + }; + + check_refused_ack(_metadata, self, &refused); } +/* An ack bigger than the kernel said it accepts. */ TEST_F(coredump, socket_request_invalid_size_large) { - int pidfd, ret, status; - pid_t pid, pid_coredump_server; - struct pidfd_info info = {}; - int ipc_sockets[2]; - char c; - - ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); - - ret = socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets); - ASSERT_EQ(ret, 0); - - pid_coredump_server = fork(); - ASSERT_GE(pid_coredump_server, 0); - if (pid_coredump_server == 0) { - struct coredump_req req = {}; - int fd_server = -1, fd_coredump = -1, fd_peer_pidfd = -1; - int exit_code = EXIT_FAILURE; - - close(ipc_sockets[0]); - - fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); - if (fd_server < 0) { - fprintf(stderr, "socket_request_invalid_size_large: create_and_listen_unix_socket failed: %m\n"); - goto out; - } - - if (write_nointr(ipc_sockets[1], "1", 1) < 0) { - fprintf(stderr, "socket_request_invalid_size_large: write_nointr to ipc socket failed: %m\n"); - goto out; - } - - close(ipc_sockets[1]); - - fd_coredump = accept4(fd_server, NULL, NULL, SOCK_CLOEXEC); - if (fd_coredump < 0) { - fprintf(stderr, "socket_request_invalid_size_large: accept4 failed: %m\n"); - goto out; - } - - fd_peer_pidfd = get_peer_pidfd(fd_coredump); - if (fd_peer_pidfd < 0) { - fprintf(stderr, "socket_request_invalid_size_large: get_peer_pidfd failed\n"); - goto out; - } - - if (!get_pidfd_info(fd_peer_pidfd, &info)) { - fprintf(stderr, "socket_request_invalid_size_large: get_pidfd_info failed\n"); - goto out; - } - - if (!(info.mask & PIDFD_INFO_COREDUMP)) { - fprintf(stderr, "socket_request_invalid_size_large: PIDFD_INFO_COREDUMP not set in mask\n"); - goto out; - } - - if (!(info.coredump_mask & PIDFD_COREDUMPED)) { - fprintf(stderr, "socket_request_invalid_size_large: PIDFD_COREDUMPED not set in coredump_mask\n"); - goto out; - } - - if (!read_coredump_req(fd_coredump, &req)) { - fprintf(stderr, "socket_request_invalid_size_large: read_coredump_req failed\n"); - goto out; - } - - if (!check_coredump_req(&req)) { - fprintf(stderr, "socket_request_invalid_size_large: check_coredump_req failed\n"); - goto out; - } - - if (!send_coredump_ack(fd_coredump, &req, - COREDUMP_REJECT | COREDUMP_WAIT, - COREDUMP_ACK_SIZE_VER0 + PAGE_SIZE)) { - fprintf(stderr, "socket_request_invalid_size_large: send_coredump_ack failed\n"); - goto out; - } - - if (!read_marker(fd_coredump, COREDUMP_MARK_MAXSIZE)) { - fprintf(stderr, "socket_request_invalid_size_large: read_marker COREDUMP_MARK_MAXSIZE failed\n"); - goto out; - } - - exit_code = EXIT_SUCCESS; - fprintf(stderr, "socket_request_invalid_size_large: completed successfully\n"); -out: - if (fd_peer_pidfd >= 0) - close(fd_peer_pidfd); - if (fd_coredump >= 0) - close(fd_coredump); - if (fd_server >= 0) - close(fd_server); - _exit(exit_code); - } - self->pid_coredump_server = pid_coredump_server; - - EXPECT_EQ(close(ipc_sockets[1]), 0); - ASSERT_EQ(read_nointr(ipc_sockets[0], &c, 1), 1); - EXPECT_EQ(close(ipc_sockets[0]), 0); - - pid = fork(); - ASSERT_GE(pid, 0); - if (pid == 0) - crashing_child(); - - pidfd = sys_pidfd_open(pid, 0); - ASSERT_GE(pidfd, 0); - - waitpid(pid, &status, 0); - ASSERT_TRUE(WIFSIGNALED(status)); - ASSERT_FALSE(WCOREDUMP(status)); - - ASSERT_TRUE(get_pidfd_info(pidfd, &info)); - ASSERT_GT((info.mask & PIDFD_INFO_COREDUMP), 0); - ASSERT_GT((info.coredump_mask & PIDFD_COREDUMPED), 0); - - wait_and_check_coredump_server(pid_coredump_server, _metadata, self); + struct refused_ack refused = { + .ack = { + .size = COREDUMP_ACK_SIZE_VER0 + PAGE_SIZE, + .mask = COREDUMP_REJECT | COREDUMP_WAIT, + }, + .bytes = COREDUMP_ACK_SIZE_VER0 + PAGE_SIZE, + .mark = COREDUMP_MARK_MAXSIZE, + }; + + check_refused_ack(_metadata, self, &refused); } /* @@ -2038,27 +1746,59 @@ out: EXPECT_EQ(close(fd_core_file), 0); } -/* Ack @ack_mask, expect the kernel to refuse it as conflicting. */ -static void check_conflicting_ack(struct __test_metadata *const _metadata, - FIXTURE_DATA(coredump) *self, __u64 ack_mask) +/* COREDUMP_RECORDS applies to a coredump the kernel writes, nothing else. */ +TEST_F(coredump, socket_request_records_without_kernel) +{ + check_conflicting_ack(_metadata, self, COREDUMP_USERSPACE | COREDUMP_RECORDS); +} + +/* A zero record can't exist outside a record stream. */ +TEST_F(coredump, socket_request_sparse_without_records) +{ + check_conflicting_ack(_metadata, self, COREDUMP_KERNEL | COREDUMP_SPARSE); +} + +/* What the server reports back about the coredump it decided to take. */ +struct stream_choice { + bool sparse; + ssize_t received; + off_t size; + ssize_t vm_size; +}; + +/* + * The kernel blocks in the coredump request until the ack arrives, so a + * coredump server gets to look at the task before it commits to a + * stream. Take the record stream only for a task whose mappings are + * worth it and the plain byte stream for everything else. + */ +static void check_stream_choice(struct __test_metadata *const _metadata, + FIXTURE_DATA(coredump) *self, bool big, + struct stream_choice *choice) { int pidfd, status; pid_t pid, pid_coredump_server; struct pidfd_info info = {}; int ipc_sockets[2]; + int pipefds[2]; char c; ASSERT_EQ(socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets), 0); + ASSERT_EQ(pipe(pipefds), 0); ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); pid_coredump_server = fork(); ASSERT_GE(pid_coredump_server, 0); if (pid_coredump_server == 0) { int fd_server = -1, fd_coredump = -1, fd_peer_pidfd = -1; + int fd_file = -1; int exit_code = EXIT_FAILURE; struct coredump_req req = {}; + struct stream_choice got = {}; + __u64 mask; close(ipc_sockets[0]); + close(pipefds[0]); fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); if (fd_server < 0) @@ -2077,20 +1817,66 @@ static void check_conflicting_ack(struct __test_metadata *const _metadata, if (fd_peer_pidfd < 0) goto out; + /* + * The reassembled coredump is bigger than the mapping the + * child made, so keep it on the detached tmpfs and sparse. + */ + fd_file = open_coredump_tmpfile(self->fd_tmpfs_detached); + if (fd_file < 0) + goto out; + if (!read_coredump_req(fd_coredump, &req)) goto out; if (!check_coredump_req(&req)) goto out; - if (!send_coredump_ack(fd_coredump, &req, ack_mask, 0)) + /* + * Nothing is on the wire yet and the kernel is waiting for + * the ack, so there is all the time in the world to look at + * the task and decide what to ask it for. + */ + got.vm_size = peer_vm_size(fd_peer_pidfd); + if (got.vm_size < 0) + goto out; + got.sparse = got.vm_size >= SPARSE_STREAM_THRESHOLD; + + fprintf(stderr, "Peer maps %zd bytes, asking for %s\n", + got.vm_size, + got.sparse ? "a sparse record stream" : "a byte stream"); + + mask = COREDUMP_KERNEL | COREDUMP_WAIT; + if (got.sparse) + mask |= COREDUMP_RECORDS | COREDUMP_SPARSE; + + if (!send_coredump_ack(fd_coredump, &req, mask, 0)) + goto out; + + if (!read_marker(fd_coredump, COREDUMP_MARK_REQACK)) + goto out; + + if (got.sparse) { + got.received = recv_coredump_records(fd_coredump, fd_file, + &got.size, NULL, -1); + } else { + got.received = recv_coredump_bytes(fd_coredump, fd_file); + got.size = got.received; + } + if (got.received < 0) goto out; - if (!read_marker(fd_coredump, COREDUMP_MARK_CONFLICTING)) + /* Either way a debugger has to see an ordinary core file. */ + if (!is_elf_core(fd_file)) + goto out; + + if (write_nointr(pipefds[1], &got, sizeof(got)) != sizeof(got)) goto out; exit_code = EXIT_SUCCESS; out: + close(pipefds[1]); + if (fd_file >= 0) + close(fd_file); if (fd_peer_pidfd >= 0) close(fd_peer_pidfd); if (fd_coredump >= 0) @@ -2102,20 +1888,25 @@ out: self->pid_coredump_server = pid_coredump_server; EXPECT_EQ(close(ipc_sockets[1]), 0); + EXPECT_EQ(close(pipefds[1]), 0); ASSERT_EQ(read_nointr(ipc_sockets[0], &c, 1), 1); EXPECT_EQ(close(ipc_sockets[0]), 0); pid = fork(); ASSERT_GE(pid, 0); if (pid == 0) - crashing_child(); + crashing_child_sparse(big ? SPARSE_MAPPING_SIZE : PAGE_SIZE); pidfd = sys_pidfd_open(pid, 0); ASSERT_GE(pidfd, 0); waitpid(pid, &status, 0); ASSERT_TRUE(WIFSIGNALED(status)); - ASSERT_FALSE(WCOREDUMP(status)); + ASSERT_TRUE(WCOREDUMP(status)); + + ASSERT_EQ(read_nointr(pipefds[0], choice, sizeof(*choice)), + sizeof(*choice)); + EXPECT_EQ(close(pipefds[0]), 0); ASSERT_TRUE(get_pidfd_info(pidfd, &info)); ASSERT_GT((info.mask & PIDFD_INFO_COREDUMP), 0); @@ -2124,45 +1915,131 @@ out: wait_and_check_coredump_server(pid_coredump_server, _metadata, self); } -/* COREDUMP_RECORDS applies to a coredump the kernel writes, nothing else. */ -TEST_F(coredump, socket_request_records_without_kernel) +/* A task with little mapped isn't worth a record stream. */ +TEST_F(coredump, socket_request_stream_choice_small) { - check_conflicting_ack(_metadata, self, COREDUMP_USERSPACE | COREDUMP_RECORDS); + struct stream_choice choice = {}; + + check_stream_choice(_metadata, self, false, &choice); + + ASSERT_LT(choice.vm_size, (ssize_t)SPARSE_STREAM_THRESHOLD); + ASSERT_FALSE(choice.sparse); + ASSERT_GT(choice.received, 0); } -/* A zero record can't exist outside a record stream. */ -TEST_F(coredump, socket_request_sparse_without_records) +/* A task sitting on a big mapping is. */ +TEST_F(coredump, socket_request_stream_choice_large) { - check_conflicting_ack(_metadata, self, COREDUMP_KERNEL | COREDUMP_SPARSE); + struct stream_choice choice = {}; + + check_stream_choice(_metadata, self, true, &choice); + + ASSERT_GE(choice.vm_size, (ssize_t)SPARSE_STREAM_THRESHOLD); + ASSERT_TRUE(choice.sparse); + ASSERT_GT(choice.size, (off_t)SPARSE_MAPPING_SIZE); + + /* The holes didn't have to go over the socket. */ + ASSERT_LT(choice.received, choice.size / 8); } -/* What the server reports back about the coredump it decided to take. */ -struct stream_choice { - bool sparse; - ssize_t received; - off_t size; - ssize_t vm_size; +/* What a coredump server was built with. */ +struct server_build { + /* sizeof(struct coredump_req) and sizeof(struct coredump_ack) back then. */ + size_t req_size; + size_t ack_size; + /* The features it raises if the kernel offers them. */ + __u64 wants; + /* Its policy: what it drops from and adds to the task's selection. */ + __u64 drop; + __u64 add; +}; + +/* A server from when the structs were first published: kernel-written dumps. */ +static const struct server_build server_build_ver0 = { + .req_size = COREDUMP_REQ_SIZE_VER0, + .ack_size = COREDUMP_ACK_SIZE_VER0, + .wants = COREDUMP_KERNEL, +}; + +/* A server built against this header: no shared memory, always the ELF headers. */ +static const struct server_build server_build_ver1 = { + .req_size = sizeof(struct coredump_req), + .ack_size = sizeof(struct coredump_ack), + .wants = COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES, + .drop = COREDUMP_MEMORY_ANON_SHARED | COREDUMP_MEMORY_FILE_SHARED, + .add = COREDUMP_MEMORY_ELF_HEADERS, }; /* - * The kernel blocks in the coredump request until the ack arrives, so a - * coredump server gets to look at the task before it commits to a - * stream. Take the record stream only for a task whose mappings are - * worth it and the plain byte stream for everything else. + * Build the ack the way a server does: from what the kernel offers, what + * this build implements, and what fits in the ack the kernel accepts. + * Fields the build never read are zero and never consulted. */ -static void check_stream_choice(struct __test_metadata *const _metadata, - FIXTURE_DATA(coredump) *self, bool big, - struct stream_choice *choice) +static void negotiate(const struct coredump_req *req, + const struct server_build *build, + struct coredump_ack *ack) +{ + __u64 offered = req->mask & build->wants; + + memset(ack, 0, sizeof(*ack)); + ack->size = build->ack_size < req->size_ack ? build->ack_size : req->size_ack; + /* These builds only ever have the kernel write the coredump. */ + ack->mask = COREDUMP_KERNEL; + + /* Sparse needs records, records need the kernel to write. */ + if (offered & COREDUMP_RECORDS) { + ack->mask |= COREDUMP_RECORDS; + if (offered & COREDUMP_SPARSE) + ack->mask |= COREDUMP_SPARSE; + } + + /* The memory types need an ack that carries them. */ + if ((offered & COREDUMP_MEMORY_TYPES) && ack->size >= COREDUMP_ACK_SIZE_VER1) { + ack->mask |= COREDUMP_MEMORY_TYPES; + /* Start from the task's selection; only advertised types pass. */ + ack->memory_types = (req->memory_types & ~build->drop) | build->add; + ack->memory_types &= req->memory_types_mask; + } +} + +/* What a memory types test asks of the kernel and what it expects back. */ +struct memory_choice { + /* Memory types the crashing child selects, or FILTER_TASK_INHERIT. */ + __u64 task_filter; + /* Negotiate the ack as this server build, NULL to send it as given. */ + const struct server_build *build; + /* The ack, or what the negotiation must arrive at. */ + __u64 mask; + __u64 memory_types; + size_t size_ack; + /* The shared mapping is in the coredump with all of its memory. */ + bool shared_dumped; + /* No memory at all. Pull a page from /proc/<pid>/mem instead. */ + bool skeleton; +}; + +/* A skeleton still carries the vdso and friends, nothing bigger. */ +#define SKELETON_DATA_PAGES 16 + +/* + * The crashing child maps shared anonymous memory and tells the server + * where. The server acks with @choice and checks whether that mapping's + * segment in the coredump carries its memory. + */ +static void check_memory_dump(struct __test_metadata *const _metadata, + FIXTURE_DATA(coredump) *self, + const struct memory_choice *choice) { int pidfd, status; pid_t pid, pid_coredump_server; struct pidfd_info info = {}; int ipc_sockets[2]; - int pipefds[2]; + int addr_pipe[2]; char c; ASSERT_EQ(socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC, 0, ipc_sockets), 0); - ASSERT_EQ(pipe(pipefds), 0); + ASSERT_EQ(pipe(addr_pipe), 0); ASSERT_TRUE(set_core_pattern("@@/tmp/coredump.socket")); pid_coredump_server = fork(); @@ -2172,11 +2049,21 @@ static void check_stream_choice(struct __test_metadata *const _metadata, int fd_file = -1; int exit_code = EXIT_FAILURE; struct coredump_req req = {}; - struct stream_choice got = {}; - __u64 mask; + struct coredump_ack ack = { + .size = choice->size_ack, + .mask = choice->mask, + .memory_types = choice->memory_types, + }; + /* How much of the request this server reads. */ + size_t req_size = choice->build ? choice->build->req_size : sizeof(req); + __u64 task_filter; + ElfW(Phdr) segment; + ssize_t received; + off_t size; + char *addr; close(ipc_sockets[0]); - close(pipefds[0]); + close(addr_pipe[1]); fd_server = create_and_listen_unix_socket("/tmp/coredump.socket"); if (fd_server < 0) @@ -2195,64 +2082,127 @@ static void check_stream_choice(struct __test_metadata *const _metadata, if (fd_peer_pidfd < 0) goto out; - /* - * The reassembled coredump is bigger than the mapping the - * child made, so keep it on the detached tmpfs and sparse. - */ fd_file = open_coredump_tmpfile(self->fd_tmpfs_detached); if (fd_file < 0) goto out; - if (!read_coredump_req(fd_coredump, &req)) + if (!read_coredump_req_sized(fd_coredump, &req, req_size)) goto out; - if (!check_coredump_req(&req)) + if (!peer_coredump_filter(fd_peer_pidfd, &task_filter)) goto out; - /* - * Nothing is on the wire yet and the kernel is waiting for - * the ack, so there is all the time in the world to look at - * the task and decide what to ask it for. - */ - got.vm_size = peer_vm_size(fd_peer_pidfd); - if (got.vm_size < 0) + /* A build from before the memory types never read that far. */ + if (req_size >= COREDUMP_REQ_SIZE_VER1) { + if (!check_coredump_req(&req)) + goto out; + + /* The request reports the memory types the task selected. */ + if (req.memory_types != task_filter) { + fprintf(stderr, "Request reports 0x%llx, task selected 0x%llx\n", + (unsigned long long)req.memory_types, + (unsigned long long)task_filter); + goto out; + } + } + + if (choice->task_filter != FILTER_TASK_INHERIT && + task_filter != choice->task_filter) { + fprintf(stderr, "Task selected 0x%llx, child asked for 0x%llx\n", + (unsigned long long)task_filter, + (unsigned long long)choice->task_filter); goto out; - got.sparse = got.vm_size >= SPARSE_STREAM_THRESHOLD; + } - fprintf(stderr, "Peer maps %zd bytes, asking for %s\n", - got.vm_size, - got.sparse ? "a sparse record stream" : "a byte stream"); + /* The child sent the address of its mapping before it crashed. */ + if (read_nointr(addr_pipe[0], &addr, sizeof(addr)) != sizeof(addr)) + goto out; - mask = COREDUMP_KERNEL | COREDUMP_WAIT; - if (got.sparse) - mask |= COREDUMP_RECORDS | COREDUMP_SPARSE; + /* A server build negotiates its ack and must arrive at the choice. */ + if (choice->build) { + negotiate(&req, choice->build, &ack); - if (!send_coredump_ack(fd_coredump, &req, mask, 0)) + if (ack.size != choice->size_ack || ack.mask != choice->mask || + ack.memory_types != choice->memory_types) { + fprintf(stderr, + "Negotiated %u bytes, mask 0x%llx, types 0x%llx\n", + ack.size, (unsigned long long)ack.mask, + (unsigned long long)ack.memory_types); + goto out; + } + } + + if (!send_coredump_ack_types(fd_coredump, &req, ack.mask, + ack.memory_types, ack.size)) goto out; if (!read_marker(fd_coredump, COREDUMP_MARK_REQACK)) goto out; - if (got.sparse) { - got.received = recv_coredump_records(fd_coredump, fd_file, - &got.size, NULL, -1); - } else { - got.received = recv_coredump_bytes(fd_coredump, fd_file); - got.size = got.received; - } - if (got.received < 0) + if (ack.mask & COREDUMP_RECORDS) + received = recv_coredump_records(fd_coredump, fd_file, + &size, NULL, -1); + else + received = recv_coredump_bytes(fd_coredump, fd_file); + if (received < 0) goto out; - /* Either way a debugger has to see an ordinary core file. */ if (!is_elf_core(fd_file)) goto out; - if (write_nointr(pipefds[1], &got, sizeof(got)) != sizeof(got)) + /* A dump ending in holes or empty segments must still be whole. */ + if (!check_coredump_extent(fd_file)) + goto out; + + if (!find_coredump_segment(fd_file, (__u64)(uintptr_t)addr, &segment)) goto out; + if (segment.p_memsz != MEMORY_MAPPING_SIZE) { + fprintf(stderr, "Segment spans %llu bytes, the mapping %u\n", + (unsigned long long)segment.p_memsz, + MEMORY_MAPPING_SIZE); + goto out; + } + + if (segment.p_filesz != (choice->shared_dumped ? segment.p_memsz : 0)) { + fprintf(stderr, "Segment carries %llu bytes, expected %s of them\n", + (unsigned long long)segment.p_filesz, + choice->shared_dumped ? "all" : "none"); + goto out; + } + + if (choice->skeleton) { + __u64 data, notes, data_max; + char buf[PAGE_SIZE]; + + if (!sum_coredump_segments(fd_file, &data, ¬es)) + goto out; + + data_max = SKELETON_DATA_PAGES * sysconf(_SC_PAGESIZE); + if (!notes || data > data_max) { + fprintf(stderr, "Skeleton has %llu note and %llu memory bytes\n", + (unsigned long long)notes, + (unsigned long long)data); + goto out; + } + + /* The task is parked in COREDUMP_WAIT with its memory. */ + if (peer_read_mem(fd_peer_pidfd, (__u64)(uintptr_t)addr, + buf, sizeof(buf)) != sizeof(buf)) + goto out; + + if (buf[0] != 'x') { + fprintf(stderr, "Pulled memory lacks the child's mark\n"); + goto out; + } + + fprintf(stderr, "Skeleton of %zd bytes, pulled %zu bytes of memory\n", + received, sizeof(buf)); + } + exit_code = EXIT_SUCCESS; out: - close(pipefds[1]); + close(addr_pipe[0]); if (fd_file >= 0) close(fd_file); if (fd_peer_pidfd >= 0) @@ -2266,14 +2216,15 @@ out: self->pid_coredump_server = pid_coredump_server; EXPECT_EQ(close(ipc_sockets[1]), 0); - EXPECT_EQ(close(pipefds[1]), 0); + EXPECT_EQ(close(addr_pipe[0]), 0); ASSERT_EQ(read_nointr(ipc_sockets[0], &c, 1), 1); EXPECT_EQ(close(ipc_sockets[0]), 0); pid = fork(); ASSERT_GE(pid, 0); if (pid == 0) - crashing_child_sparse(big ? SPARSE_MAPPING_SIZE : PAGE_SIZE); + crashing_child_memory(choice->task_filter, addr_pipe[1]); + EXPECT_EQ(close(addr_pipe[1]), 0); pidfd = sys_pidfd_open(pid, 0); ASSERT_GE(pidfd, 0); @@ -2282,10 +2233,6 @@ out: ASSERT_TRUE(WIFSIGNALED(status)); ASSERT_TRUE(WCOREDUMP(status)); - ASSERT_EQ(read_nointr(pipefds[0], choice, sizeof(*choice)), - sizeof(*choice)); - EXPECT_EQ(close(pipefds[0]), 0); - ASSERT_TRUE(get_pidfd_info(pidfd, &info)); ASSERT_GT((info.mask & PIDFD_INFO_COREDUMP), 0); ASSERT_GT((info.coredump_mask & PIDFD_COREDUMPED), 0); @@ -2293,31 +2240,376 @@ out: wait_and_check_coredump_server(pid_coredump_server, _metadata, self); } -/* A task with little mapped isn't worth a record stream. */ -TEST_F(coredump, socket_request_stream_choice_small) +/* Without COREDUMP_MEMORY_TYPES the task's own selection decides. */ +TEST_F(coredump, socket_request_memory_types_task_includes) { - struct stream_choice choice = {}; + struct memory_choice choice = { + .task_filter = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .mask = COREDUMP_KERNEL, + .shared_dumped = true, + }; + + check_memory_dump(_metadata, self, &choice); +} - check_stream_choice(_metadata, self, false, &choice); +TEST_F(coredump, socket_request_memory_types_task_excludes) +{ + struct memory_choice choice = { + .task_filter = 0, + .mask = COREDUMP_KERNEL, + .shared_dumped = false, + }; - ASSERT_LT(choice.vm_size, (ssize_t)SPARSE_STREAM_THRESHOLD); - ASSERT_FALSE(choice.sparse); - ASSERT_GT(choice.received, 0); + check_memory_dump(_metadata, self, &choice); } -/* A task sitting on a big mapping is. */ -TEST_F(coredump, socket_request_stream_choice_large) +/* The server drops a memory type the task would have dumped. */ +TEST_F(coredump, socket_request_memory_types_restricts) { - struct stream_choice choice = {}; + struct memory_choice choice = { + .task_filter = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .mask = COREDUMP_KERNEL | COREDUMP_MEMORY_TYPES, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE, + .shared_dumped = false, + }; + + check_memory_dump(_metadata, self, &choice); +} - check_stream_choice(_metadata, self, true, &choice); +/* The server adds a memory type the task had excluded. */ +TEST_F(coredump, socket_request_memory_types_widens) +{ + struct memory_choice choice = { + .task_filter = 0, + .mask = COREDUMP_KERNEL | COREDUMP_MEMORY_TYPES, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .shared_dumped = true, + }; + + check_memory_dump(_metadata, self, &choice); +} - ASSERT_GE(choice.vm_size, (ssize_t)SPARSE_STREAM_THRESHOLD); - ASSERT_TRUE(choice.sparse); - ASSERT_GT(choice.size, (off_t)SPARSE_MAPPING_SIZE); +/* The memory types decide what goes into a record stream just the same. */ +TEST_F(coredump, socket_request_memory_types_records) +{ + struct memory_choice choice = { + .task_filter = FILTER_TASK_INHERIT, + .mask = COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE, + .shared_dumped = false, + }; + + check_memory_dump(_metadata, self, &choice); +} - /* The holes didn't have to go over the socket. */ - ASSERT_LT(choice.received, choice.size / 8); +/* + * An empty selection leaves a skeleton: every program header and every note + * but no memory. A server that wants to pick the memory itself reads it + * from /proc/<pid>/mem while the task waits for it to finish. + */ +TEST_F(coredump, socket_request_memory_types_skeleton) +{ + struct memory_choice choice = { + .task_filter = FILTER_TASK_INHERIT, + .mask = COREDUMP_KERNEL | COREDUMP_WAIT | COREDUMP_MEMORY_TYPES, + .memory_types = 0, + .shared_dumped = false, + .skeleton = true, + }; + + check_memory_dump(_metadata, self, &choice); +} + +/* A memory type the kernel didn't advertise in memory_types_mask. */ +TEST_F(coredump, socket_request_memory_types_unknown_bit) +{ + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .mask = COREDUMP_KERNEL | COREDUMP_MEMORY_TYPES, + .memory_types = 1ULL << 63, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_UNSUPPORTED, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* The memory types must be zero unless COREDUMP_MEMORY_TYPES is raised. */ +TEST_F(coredump, socket_request_memory_types_stale_field) +{ + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .mask = COREDUMP_KERNEL, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_UNSUPPORTED, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* COREDUMP_MEMORY_TYPES needs an ack that has the memory types. */ +TEST_F(coredump, socket_request_memory_types_short_ack) +{ + struct refused_ack refused = { + .ack = { + .size = COREDUMP_ACK_SIZE_VER0, + .mask = COREDUMP_KERNEL | COREDUMP_MEMORY_TYPES, + }, + .bytes = COREDUMP_ACK_SIZE_VER0, + .mark = COREDUMP_MARK_MINSIZE, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* The memory types select what the kernel writes, nothing else. */ +TEST_F(coredump, socket_request_memory_types_without_kernel) +{ + check_conflicting_ack(_metadata, self, COREDUMP_USERSPACE | COREDUMP_MEMORY_TYPES); +} + +/* + * A server built with the first structs reads the request it knows, + * discards the rest and acks with the ack it knows. It raises nothing + * it wasn't built for and the kernel dumps what the task selected. + */ +TEST_F(coredump, socket_request_negotiate_ver0) +{ + struct memory_choice choice = { + .task_filter = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .build = &server_build_ver0, + .mask = COREDUMP_KERNEL, + .memory_types = 0, + .size_ack = COREDUMP_ACK_SIZE_VER0, + .shared_dumped = true, + }; + + check_memory_dump(_metadata, self, &choice); +} + +/* + * A server built against this header takes every feature the kernel + * offers, drops shared memory from what the task selected and adds the + * ELF headers. + */ +TEST_F(coredump, socket_request_negotiate_ver1) +{ + struct memory_choice choice = { + .task_filter = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .build = &server_build_ver1, + .mask = COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ELF_HEADERS, + .size_ack = COREDUMP_ACK_SIZE_VER1, + .shared_dumped = false, + }; + + check_memory_dump(_metadata, self, &choice); +} + +/* An ack that picks none of KERNEL, USERSPACE and REJECT. */ +TEST_F(coredump, socket_request_no_mode) +{ + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .mask = COREDUMP_WAIT, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_CONFLICTING, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* @spare must be zero, like every field that isn't in use. */ +TEST_F(coredump, socket_request_spare) +{ + struct refused_ack refused = { + .ack = { + .size = sizeof(struct coredump_ack), + .spare = 1, + .mask = COREDUMP_KERNEL, + }, + .bytes = sizeof(struct coredump_ack), + .mark = COREDUMP_MARK_UNSUPPORTED, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* An ack size is a byte count. One that ends inside a field is valid. */ +#define ACK_SIZE_BETWEEN (COREDUMP_ACK_SIZE_VER0 + sizeof(__u32)) + +/* Any size from VER0 up to what the kernel accepts works without memory types. */ +TEST_F(coredump, socket_request_ack_size_between) +{ + struct memory_choice choice = { + .task_filter = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .mask = COREDUMP_KERNEL, + .size_ack = ACK_SIZE_BETWEEN, + .shared_dumped = true, + }; + + check_memory_dump(_metadata, self, &choice); +} + +/* The memory types need the whole field, not the part that happens to fit. */ +TEST_F(coredump, socket_request_memory_types_ack_size_between) +{ + struct refused_ack refused = { + .ack = { + .size = ACK_SIZE_BETWEEN, + .mask = COREDUMP_KERNEL | COREDUMP_MEMORY_TYPES, + }, + .bytes = ACK_SIZE_BETWEEN, + .mark = COREDUMP_MARK_MINSIZE, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* A server that hangs up without acking gets no marker and no coredump. */ +TEST_F(coredump, socket_request_server_hangs_up) +{ + struct refused_ack refused = { + .bytes = 0, + .no_marker = true, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* A server that hangs up in the middle of its ack looks the same. */ +TEST_F(coredump, socket_request_ack_truncated) +{ + struct refused_ack refused = { + .ack = { + .size = COREDUMP_ACK_SIZE_VER0, + .mask = COREDUMP_KERNEL, + }, + .bytes = COREDUMP_ACK_SIZE_VER0 / 2, + .no_marker = true, + }; + + check_refused_ack(_metadata, self, &refused); +} + +/* + * The kernels a server built against this header can't meet here: + * negotiate() against their requests, no coredump involved. + */ + +/* The request of a kernel with the first structs and features. */ +static const struct coredump_req req_ver0 = { + .size = COREDUMP_REQ_SIZE_VER0, + .size_ack = COREDUMP_ACK_SIZE_VER0, + .mask = COREDUMP_KERNEL | COREDUMP_USERSPACE | + COREDUMP_REJECT | COREDUMP_WAIT, +}; + +/* The request of this kernel. */ +static const struct coredump_req req_ver1 = { + .size = COREDUMP_REQ_SIZE_VER1, + .size_ack = COREDUMP_ACK_SIZE_VER1, + .mask = COREDUMP_KERNEL | COREDUMP_USERSPACE | + COREDUMP_REJECT | COREDUMP_WAIT | + COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES, + .memory_types = COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ANON_SHARED, + .memory_types_mask = TEST_MEMORY_ALL, +}; + +/* A kernel with the first structs gets the first ack and nothing newer. */ +TEST(negotiate_ver0_kernel) +{ + struct coredump_ack ack; + + negotiate(&req_ver0, &server_build_ver1, &ack); + ASSERT_EQ(ack.size, COREDUMP_ACK_SIZE_VER0); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL); + ASSERT_EQ(ack.memory_types, 0); +} + +/* A kernel with records and sparse but the first structs: both, no types. */ +TEST(negotiate_sparse_kernel) +{ + struct coredump_req req = req_ver0; + struct coredump_ack ack; + + req.mask |= COREDUMP_RECORDS | COREDUMP_SPARSE; + negotiate(&req, &server_build_ver1, &ack); + ASSERT_EQ(ack.size, COREDUMP_ACK_SIZE_VER0); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE); + ASSERT_EQ(ack.memory_types, 0); +} + +/* Records without sparse: sparse isn't raised on its own. */ +TEST(negotiate_records_without_sparse) +{ + struct coredump_req req = req_ver0; + struct coredump_ack ack; + + req.mask |= COREDUMP_RECORDS; + negotiate(&req, &server_build_ver1, &ack); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL | COREDUMP_RECORDS); +} + +/* + * A feature whose ack field lies past what the kernel accepts can't be + * raised. No kernel offers the memory types without the room for them, so a + * request that does stands in for a feature newer than this header. + */ +TEST(negotiate_types_need_room) +{ + struct coredump_req req = req_ver0; + struct coredump_ack ack; + + req.mask |= COREDUMP_MEMORY_TYPES; + negotiate(&req, &server_build_ver1, &ack); + ASSERT_EQ(ack.size, COREDUMP_ACK_SIZE_VER0); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL); + ASSERT_EQ(ack.memory_types, 0); +} + +/* This kernel: the policy applied to the task's selection. */ +TEST(negotiate_ver1_kernel) +{ + struct coredump_ack ack; + + negotiate(&req_ver1, &server_build_ver1, &ack); + ASSERT_EQ(ack.size, COREDUMP_ACK_SIZE_VER1); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES); + ASSERT_EQ(ack.memory_types, COREDUMP_MEMORY_ANON_PRIVATE | + COREDUMP_MEMORY_ELF_HEADERS); +} + +/* A kernel that doesn't know a type the policy adds isn't asked for it. */ +TEST(negotiate_unknown_type) +{ + struct coredump_req req = req_ver1; + struct coredump_ack ack; + + req.memory_types_mask &= ~(__u64)COREDUMP_MEMORY_ELF_HEADERS; + negotiate(&req, &server_build_ver1, &ack); + ASSERT_EQ(ack.mask, COREDUMP_KERNEL | COREDUMP_RECORDS | COREDUMP_SPARSE | + COREDUMP_MEMORY_TYPES); + ASSERT_EQ(ack.memory_types, COREDUMP_MEMORY_ANON_PRIVATE); } TEST_HARNESS_MAIN diff --git a/tools/testing/selftests/coredump/coredump_test_helpers.c b/tools/testing/selftests/coredump/coredump_test_helpers.c index d7cc448eeaf4..4e36e3e4fb78 100644 --- a/tools/testing/selftests/coredump/coredump_test_helpers.c +++ b/tools/testing/selftests/coredump/coredump_test_helpers.c @@ -17,6 +17,7 @@ #include <sys/ioctl.h> #include <sys/mman.h> #include <sys/socket.h> +#include <sys/stat.h> #include <sys/types.h> #include <sys/un.h> #include <sys/wait.h> @@ -73,6 +74,48 @@ void crashing_child_sparse(size_t size) *(volatile int *)NULL = 0; } +/* Select @types through the caller's own /proc/self/coredump_filter. */ +static bool set_coredump_filter(__u64 types) +{ + char buf[32]; + int fd, len; + bool ok; + + fd = open("/proc/self/coredump_filter", O_WRONLY | O_CLOEXEC); + if (fd < 0) + return false; + + len = snprintf(buf, sizeof(buf), "0x%llx", (unsigned long long)types); + ok = write_nointr(fd, buf, len) == len; + close(fd); + return ok; +} + +/* + * Map shared anonymous memory, touch it, tell the server where it is and + * crash. A @task_filter other than FILTER_TASK_INHERIT is selected first. + */ +void crashing_child_memory(__u64 task_filter, int fd_addr) +{ + char *p; + + if (task_filter != FILTER_TASK_INHERIT && !set_coredump_filter(task_filter)) + _exit(EXIT_FAILURE); + + p = mmap(NULL, MEMORY_MAPPING_SIZE, PROT_READ | PROT_WRITE, + MAP_SHARED | MAP_ANONYMOUS, -1, 0); + if (p == MAP_FAILED) + _exit(EXIT_FAILURE); + p[0] = 'x'; + + if (write_nointr(fd_addr, &p, sizeof(p)) != sizeof(p)) + _exit(EXIT_FAILURE); + close(fd_addr); + + /* crash on purpose */ + *(volatile int *)NULL = 0; +} + /* Sink a reassembled record stream is handed to, record by record. */ struct coredump_record_sink { /* @len bytes of coredump data that belong at @offset. */ @@ -916,6 +959,82 @@ static const ElfW(Phdr) *find_segment(const ElfW(Phdr) *phdr, size_t nr, return NULL; } +/* The PT_LOAD segment @vaddr falls into. */ +bool find_coredump_segment(int fd, __u64 vaddr, ElfW(Phdr) *segment) +{ + const ElfW(Phdr) *found; + ElfW(Phdr) *phdr; + size_t nr; + + phdr = read_phdrs(fd, &nr); + if (!phdr) + return false; + + found = find_segment(phdr, nr, vaddr); + if (found) + *segment = *found; + else + fprintf(stderr, "%s: no segment for 0x%llx\n", __func__, + (unsigned long long)vaddr); + + free(phdr); + return found; +} + +/* How many bytes the PT_LOAD and the PT_NOTE segments of @fd carry. */ +bool sum_coredump_segments(int fd, __u64 *data, __u64 *notes) +{ + ElfW(Phdr) *phdr; + size_t nr, i; + + phdr = read_phdrs(fd, &nr); + if (!phdr) + return false; + + *data = 0; + *notes = 0; + for (i = 0; i < nr; i++) { + if (phdr[i].p_type == PT_LOAD) + *data += phdr[i].p_filesz; + else if (phdr[i].p_type == PT_NOTE) + *notes += phdr[i].p_filesz; + } + + free(phdr); + return true; +} + +/* The coredump in @fd is at least as long as every segment it declares. */ +bool check_coredump_extent(int fd) +{ + ElfW(Phdr) *phdr; + struct stat st; + size_t nr, i; + bool ok = true; + + if (fstat(fd, &st)) { + fprintf(stderr, "%s: fstat: %m\n", __func__); + return false; + } + + phdr = read_phdrs(fd, &nr); + if (!phdr) + return false; + + for (i = 0; i < nr; i++) { + if (phdr[i].p_offset + phdr[i].p_filesz <= (__u64)st.st_size) + continue; + fprintf(stderr, "%s: segment %zu ends at %llu, the coredump at %llu\n", + __func__, i, + (unsigned long long)(phdr[i].p_offset + phdr[i].p_filesz), + (unsigned long long)st.st_size); + ok = false; + } + + free(phdr); + return ok; +} + /* The next stretch of memory the segments cover, split ones merged back. */ static bool next_range(const ElfW(Phdr) *phdr, size_t nr, size_t *i, __u64 *start, __u64 *end) @@ -1250,6 +1369,62 @@ ssize_t peer_vm_size(int fd_peer_pidfd) /* Protocol helper functions */ +/* The peer's /proc/<pid>/coredump_filter, which is in memory types. */ +bool peer_coredump_filter(int fd_peer_pidfd, __u64 *memory_types) +{ + struct pidfd_info info = {}; + unsigned long value; + char path[64]; + FILE *f; + int ret; + + if (!get_pidfd_info(fd_peer_pidfd, &info)) + return false; + + snprintf(path, sizeof(path), "/proc/%d/coredump_filter", info.pid); + f = fopen(path, "r"); + if (!f) { + fprintf(stderr, "%s: %s: %m\n", __func__, path); + return false; + } + + ret = fscanf(f, "%lx", &value); + fclose(f); + if (ret != 1) { + fprintf(stderr, "%s: %s: no value\n", __func__, path); + return false; + } + + *memory_types = value; + return true; +} + +/* Read @len bytes at @addr from the peer's /proc/<pid>/mem. */ +ssize_t peer_read_mem(int fd_peer_pidfd, __u64 addr, void *buf, size_t len) +{ + struct pidfd_info info = {}; + char path[64]; + ssize_t ret; + int fd; + + if (!get_pidfd_info(fd_peer_pidfd, &info)) + return -1; + + snprintf(path, sizeof(path), "/proc/%d/mem", info.pid); + fd = open(path, O_RDONLY | O_CLOEXEC); + if (fd < 0) { + fprintf(stderr, "%s: %s: %m\n", __func__, path); + return -1; + } + + ret = pread(fd, buf, len, addr); + if (ret < 0) + fprintf(stderr, "%s: %s at 0x%llx: %m\n", __func__, path, + (unsigned long long)addr); + close(fd); + return ret; +} + ssize_t recv_marker(int fd) { enum coredump_mark mark = COREDUMP_MARK_REQACK; @@ -1292,10 +1467,34 @@ bool read_marker(int fd, enum coredump_mark mark) return ret == mark; } -bool read_coredump_req(int fd, struct coredump_req *req) +/* + * The kernel hung up without sending anything more: end of stream, or a + * reset if it refused the ack on its peeked size and never read it. + */ +bool read_hangup(int fd) { ssize_t ret; - size_t field_size, user_size, known_size, kernel_size, remaining_size; + char c; + + ret = recv(fd, &c, sizeof(c), MSG_WAITALL); + if (ret == 0) { + fprintf(stderr, "Kernel closed the connection\n"); + return true; + } + if (ret < 0 && errno == ECONNRESET) { + fprintf(stderr, "Kernel closed the connection with the ack unread\n"); + return true; + } + + fprintf(stderr, "%s: expected a hangup, got %zd: %m\n", __func__, ret); + return false; +} + +/* Read the request as a server built with a @user_size byte struct does. */ +bool read_coredump_req_sized(int fd, struct coredump_req *req, size_t user_size) +{ + ssize_t ret; + size_t field_size, known_size, kernel_size, remaining_size; memset(req, 0, sizeof(*req)); field_size = sizeof(req->size); @@ -1303,25 +1502,24 @@ bool read_coredump_req(int fd, struct coredump_req *req) /* Peek the size of the coredump request. */ ret = recv(fd, req, field_size, MSG_PEEK | MSG_WAITALL); if (ret != field_size) { - fprintf(stderr, "read_coredump_req: peek failed (got %zd, expected %zu): %m\n", + fprintf(stderr, "%s: peek failed (got %zd, expected %zu): %m\n", __func__, ret, field_size); return false; } kernel_size = req->size; if (kernel_size < COREDUMP_REQ_SIZE_VER0) { - fprintf(stderr, "read_coredump_req: kernel_size %zu < min %d\n", + fprintf(stderr, "%s: kernel_size %zu < min %d\n", __func__, kernel_size, COREDUMP_REQ_SIZE_VER0); return false; } if (kernel_size >= PAGE_SIZE) { - fprintf(stderr, "read_coredump_req: kernel_size %zu >= PAGE_SIZE %d\n", + fprintf(stderr, "%s: kernel_size %zu >= PAGE_SIZE %d\n", __func__, kernel_size, PAGE_SIZE); return false; } /* Consume as much of the request as we know about. */ - user_size = sizeof(struct coredump_req); known_size = user_size < kernel_size ? user_size : kernel_size; ret = recv(fd, req, known_size, MSG_WAITALL); if (ret != known_size) @@ -1354,8 +1552,13 @@ bool read_coredump_req(int fd, struct coredump_req *req) return true; } -bool send_coredump_ack(int fd, const struct coredump_req *req, - __u64 mask, size_t size_ack) +bool read_coredump_req(int fd, struct coredump_req *req) +{ + return read_coredump_req_sized(fd, req, sizeof(*req)); +} + +/* Send @len bytes of @ack as they are, more than the struct if asked to. */ +bool send_coredump_ack_bytes(int fd, const struct coredump_ack *ack, size_t len) { ssize_t ret; /* @@ -1367,34 +1570,60 @@ bool send_coredump_ack(int fd, const struct coredump_req *req, char buffer[PAGE_SIZE]; } large_ack = {}; - if (!size_ack) - size_ack = sizeof(struct coredump_ack) < req->size_ack ? - sizeof(struct coredump_ack) : - req->size_ack; - large_ack.ack.mask = mask; - large_ack.ack.size = size_ack; - ret = send(fd, &large_ack, size_ack, MSG_NOSIGNAL); - if (ret != size_ack) { + if (len > sizeof(large_ack)) + return false; + + large_ack.ack = *ack; + ret = send(fd, &large_ack, len, MSG_NOSIGNAL); + if (ret != len) { fprintf(stderr, "%s: short send %zd: %m\n", __func__, ret); return false; } - fprintf(stderr, "Sent coredump ack with size %zu and mask 0x%llx\n", - size_ack, (unsigned long long)mask); + fprintf(stderr, "Sent %zu bytes of coredump ack: size %u, mask 0x%llx, types 0x%llx\n", + len, ack->size, (unsigned long long)ack->mask, + (unsigned long long)ack->memory_types); return true; } +bool send_coredump_ack_types(int fd, const struct coredump_req *req, + __u64 mask, __u64 memory_types, size_t size_ack) +{ + struct coredump_ack ack = { + .mask = mask, + .memory_types = memory_types, + }; + + if (!size_ack) + size_ack = sizeof(struct coredump_ack) < req->size_ack ? + sizeof(struct coredump_ack) : + req->size_ack; + ack.size = size_ack; + return send_coredump_ack_bytes(fd, &ack, size_ack); +} + +bool send_coredump_ack(int fd, const struct coredump_req *req, + __u64 mask, size_t size_ack) +{ + return send_coredump_ack_types(fd, req, mask, 0, size_ack); +} + /* Every option the kernel is expected to advertise in coredump_req->mask. */ #define TEST_REQ_MASK_ALL \ (COREDUMP_KERNEL | COREDUMP_USERSPACE | \ COREDUMP_REJECT | COREDUMP_WAIT | \ - COREDUMP_RECORDS | COREDUMP_SPARSE) + COREDUMP_RECORDS | COREDUMP_SPARSE | COREDUMP_MEMORY_TYPES) bool check_coredump_req(const struct coredump_req *req) { - if (req->size < COREDUMP_REQ_SIZE_VER0) { - fprintf(stderr, "%s: size %u below minimum %d\n", - __func__, req->size, COREDUMP_REQ_SIZE_VER0); + if (req->size != COREDUMP_REQ_SIZE_VER1) { + fprintf(stderr, "%s: size %u, expected %d\n", + __func__, req->size, COREDUMP_REQ_SIZE_VER1); + return false; + } + if (req->size_ack != COREDUMP_ACK_SIZE_VER1) { + fprintf(stderr, "%s: size_ack %u, expected %d\n", + __func__, req->size_ack, COREDUMP_ACK_SIZE_VER1); return false; } if (req->mask != TEST_REQ_MASK_ALL) { @@ -1403,6 +1632,12 @@ bool check_coredump_req(const struct coredump_req *req) (unsigned long long)TEST_REQ_MASK_ALL); return false; } + if (req->memory_types_mask != TEST_MEMORY_ALL) { + fprintf(stderr, "%s: memory_types_mask 0x%llx, expected 0x%llx\n", + __func__, (unsigned long long)req->memory_types_mask, + (unsigned long long)TEST_MEMORY_ALL); + return false; + } return true; } diff --git a/tools/testing/selftests/coredump/coredump_test_helpers.h b/tools/testing/selftests/coredump/coredump_test_helpers.h index 97ad5cfeae92..3f2f87837558 100644 --- a/tools/testing/selftests/coredump/coredump_test_helpers.h +++ b/tools/testing/selftests/coredump/coredump_test_helpers.h @@ -3,6 +3,7 @@ #ifndef __COREDUMP_TEST_HELPERS_H #define __COREDUMP_TEST_HELPERS_H +#include <link.h> #include <stdbool.h> #include <sys/types.h> #include <linux/coredump.h> @@ -21,10 +22,30 @@ /* A task mapping at least this much is worth a record stream. */ #define SPARSE_STREAM_THRESHOLD (SPARSE_MAPPING_SIZE / 2) +/* Size of the shared anonymous mapping the memory types tests map. */ +#define MEMORY_MAPPING_SIZE (4 * 1024 * 1024) + +/* Leave the coredump_filter the crashing child inherited alone. */ +#define FILTER_TASK_INHERIT ((__u64)-1) + +/* Every memory type the kernel is expected to advertise. */ +#define TEST_MEMORY_ALL \ + (COREDUMP_MEMORY_ANON_PRIVATE | COREDUMP_MEMORY_ANON_SHARED | \ + COREDUMP_MEMORY_FILE_PRIVATE | COREDUMP_MEMORY_FILE_SHARED | \ + COREDUMP_MEMORY_ELF_HEADERS | \ + COREDUMP_MEMORY_HUGETLB_PRIVATE | COREDUMP_MEMORY_HUGETLB_SHARED | \ + COREDUMP_MEMORY_DAX_PRIVATE | COREDUMP_MEMORY_DAX_SHARED) + /* Shared helper function declarations */ void *do_nothing(void *arg); void crashing_child(void); void crashing_child_sparse(size_t size); +void crashing_child_memory(__u64 task_filter, int fd_addr); +bool find_coredump_segment(int fd, __u64 vaddr, ElfW(Phdr) *segment); +bool sum_coredump_segments(int fd, __u64 *data, __u64 *notes); +bool check_coredump_extent(int fd); +bool peer_coredump_filter(int fd_peer_pidfd, __u64 *memory_types); +ssize_t peer_read_mem(int fd_peer_pidfd, __u64 addr, void *buf, size_t len); ssize_t recv_coredump_records(int fd_coredump, int fd_core_file, off_t *coredump_size, bool *truncated, int fd_peer_pidfd); @@ -43,9 +64,14 @@ bool get_pidfd_info(int fd_peer_pidfd, struct pidfd_info *info); /* Protocol helper function declarations */ ssize_t recv_marker(int fd); bool read_marker(int fd, enum coredump_mark mark); +bool read_hangup(int fd); bool read_coredump_req(int fd, struct coredump_req *req); +bool read_coredump_req_sized(int fd, struct coredump_req *req, size_t user_size); bool send_coredump_ack(int fd, const struct coredump_req *req, __u64 mask, size_t size_ack); +bool send_coredump_ack_types(int fd, const struct coredump_req *req, + __u64 mask, __u64 memory_types, size_t size_ack); +bool send_coredump_ack_bytes(int fd, const struct coredump_ack *ack, size_t len); bool check_coredump_req(const struct coredump_req *req); int open_coredump_tmpfile(int fd_tmpfs_detached); void process_coredump_worker(int fd_coredump, int fd_peer_pidfd, int fd_core_file); |