tests: check decoding of IFLA_EVENT netlink attribute

* tests/nlattr_ifinfomsg.c (main): Add a check for IFLA_EVENT attribute
decoding.

xlat: handle xlat constants that are not identifiers

Extend xlat generator to support such input lines as
IPV4_DEVCONF_FORWARDING-1 0

This is going to be used later in IFLA_AF_SPEC decoding.

* xlat/gen.sh (cond_def): If the xlat value is not an identifier name,
extract an identifier name from the beginning of xlat value.

xlat.c: handle NULL xlat in lookup routines as incremental search

This is going to be used later in IFLA_AF_SPEC decoding.

* xlat.c (xlookup): Add static variable pos, store xlat there when it is
non-NULL, use pos for lookup.
(xlat_search, xlat_idx): Add static variables pos and memb_left, store
xlat and nmemb there when xlat is non-NULL, use them for lookup.
(printxvals_ex): Store the last non-NULL xlat in static variable "last".
If the first xlat argument is NULL, use "last" instead.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

xlat.c: cleanup printxvals_ex a bit

* xlat.c (printxvals_ex): Change "str" variable visibility to the whole
function, jump to va_end instead of duplicating call and return, return
!!str.

Leverage the fact that inet_protocols xlat is sorted

* defs.h (inet_protocols): Move up, add description.
(inet_protocols_size): New declaration.
* net.c (inet_protocols_size): New constant variable.
* nlattr.c (decode_nla_ip_proto): Specify xlat_size, set xt to XT_SORTED.

Print arp_hardware_types as a sorted xlat

* rtnl_link.c (DECL_NETLINK_ROUTE_DECODER(decode_ifinfomsg)): Use
PRINT_FIELD_XVAL_SORTED_SIZED instead of PRINT_FIELD_XVAL for
arp_hardware_types.
* sock.c (print_ifreq): Likewise.

Provide a value for arp_hardware_types_size array

* defs.h (arp_hardware_types_size): New declaration.
* sockaddr.c (arp_hardware_types_size): New variable.

print_fields.h: introduce PRINT_FIELD_XVAL_SORTED_SIZED

For those cases when array definition is not readily available
and should be provided separately.

* print_fields.h (PRINT_FIELD_XVAL_SORTED_SIZED): New macro.

tests/opipe.test: remove unneeded run_prog call

There is no need to invoke grep using run_prog.

* tests/opipe.test: Remove run_prog decoration for the grep invocation.

tests: check BPF_OBJ_GET_INFO_BY_FD decoding

* tests/bpf-obj_get_info_by_fd-prog-v.c: New file.
* tests/bpf-obj_get_info_by_fd-prog.c: Likewise.
* tests/bpf-obj_get_info_by_fd-v.c: Likewise.
* tests/bpf-obj_get_info_by_fd.c: Likewise.
* configure.ac (AC_CHECK_HEADERS): Check for struct bpf_insn,
struct bpf_map_info, and struct bpf_prog_info.
* tests/pure_executables.list: Add bpf-obj_get_info_by_fd,
bpf-obj_get_info_by_fd-v, bpf-obj_get_info_by_fd-prog, and
bpf-obj_get_info_by_fd-prog-v.
* tests/.gitignore: Likewise.
* tests/gen_tests.in (bpf-obj_get_info_by_fd,
bpf-obj_get_info_by_fd-v, bpf-obj_get_info_by_fd-prog,
bpf-obj_get_info_by_fd-prog-v): New tests.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: enhance BPF_OBJ_GET_INFO_BY_FD decoder further

* bpf.c (print_ebpf_prog): Remove "decode" argument, all users updated.
(print_bpf_prog_info): Use the minimum of xlated_prog_len values on
entering and exiting syscall as the size of xlated_prog_insns array
to be decoded.  Likewise, use the minimum of nr_map_ids values on entering
and exiting syscall as the size of map_ids array to be decoded.

bpf: enhance BPF_OBJ_GET_INFO_BY_FD decoder

* bpf.c (print_bpf_map_info): Print only those fields that are
present in the structure, as it has been growing over time.
(print_bpf_prog_info): Likewise.  Print map_ids field as an array
of uint32_t instead of uint64_t; print "name" field.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: move handling of the saved state to BPF_OBJ_GET_INFO_BY_FD decoder

As it is needed for proper info_len printing on exiting.

* bpf.c (print_bpf_obj_info_fn): Change return type to void, add "saved"
argument.
(struct obj_get_info_saved): New type definition.
(print_bpf_map_info, print_bpf_prog_info): Change return type to void,
add "saved" argument, do not handle private tcb data allocation and
obtainment.
(print_bpf_obj_info_addr): Change return type to void, add tcb argument,
print the address only on exiting.
(print_bpf_obj_info): Change return type to void, add "saved" argument.
(BEGIN_BPF_CMD_DECODER(BPF_OBJ_GET_INFO_BY_FD)): Handle private data
allocation/saving/restoring, use it for storing/printing info_len field.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: print log_buf field as a string

* bpf.c (BEGIN_BPF_CMD_DECODER(BPF_PROG_LOAD)): Print log_buf field
as a string.
* tests/bpf.c: Update expected output.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: print xlated_prog_insns only when xlated_prog_len had not been 0

If xlated_prog_len is zero on entering syscall and non-zero on exiting,
the program itself is not written into xlated_prog_insns.

* bpf.c (print_ebpf_prog): Add decode argument, print the address only
if decode is false.
(BEGIN_BPF_CMD_DECODER(BPF_PROG_LOAD)): Specify decode == true argument
to print_ebpf_prog.
(print_bpf_prog_info): Specify saved xlated_prog_len as decode argument
to print_ebpf_prog.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: move print_big_u64_addr inside print_ebpf_prog

* bpf.c (print_ebpf_prog): Add static qualifier.  Change the type
of "addr" argument from kernel_ulong_t to uint64_t, print it using
print_big_u64_addr.
(BEGIN_BPF_CMD_DECODER(BPF_PROG_LOAD), print_bpf_prog_info): Do not call
print_big_u64_addr before print_ebpf_prog.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf_attr.h: add ATTRIBUTE_ALIGNED(8) to all fields of type uint64_t

Our policy is that all fields of type uint64_t in this header file
must have this attribute.

This shouldn't cause any contradictions with <linux/bpf.h>
unless the latter is buggy.

By word "buggy" I mean containing such changes as Linux kernel commit
v4.16-rc1~123^2~109^2~5^2~4.

* bpf_attr.h (struct BPF_MAP_UPDATE_ELEM_struct): Add
ATTRIBUTE_ALIGNED(8) to "flags" field.
(struct bpf_prog_info_struct): Add ATTRIBUTE_ALIGNED(8) to "load_time"
field.

tests: cast printf arguments to short when %h format qualifier is used

Reported by clang.

* tests/mknod.c (main): Cast printf argument for %#03ho format to short.
* tests/mknodat.c (main): Likewise.
* tests/net-sockaddr.c (main): Cast printf argument for %hu format
to short.
* tests/semop.c (main): Likewise.

xlat/evdev_keycode.in: remove constants with duplicating values

Those are used for describing input device category and not specific
button.

Reported by clang.

* xlat/evdev_keycode.in (BTN_MOUSE, BTN_JOYSTICK, BTN_GAMEPAD, BTN_DIGI,
BTN_WHEEL, BTN_TRIGGER_HAPPY): Remove.

configure.ac: include <sys/socket.h> before <linux/rtnetlink.h>

This is needed for old Linux UAPI headers that do no contain commit
v3.1-rc2~19^2~29.

* configure.ac: Include <sys/socket.h> before <linux/rtnetlink.h>
in AC_CHECK_* invocations.

configure.ac: check for linux/rtnetlink.h instead of linux/if_link.h for declarations

Historically, type definitions and declarations related to netlink route
interface attributes were in linux/rtnetlink.h, later they were moved
elsewhere (first linux/if.h, then linux/if_link.h) with the inclusion
of the corresponding header file in linux/rtnetlink.h.

* configure.ac: Use <linux/rtnetlink.h> instead of <linux/if_link.h>
in checks for interface netlink attributes-related declarations
and definitions.

m4: move the gawk script from m4/gen_bpf_attr_m4.sh into a separate file

As it turns out, old versions of gawk do not support -e option.

* m4/gen_bpf_attr_m4.sh: Move the gawk script ...
* m4/gen_bpf_attr_m4.awk: ... here.

s390: wire up kexec_file_load syscall

* linux/s390/syscallent.h ([381]): Add kexec_file_load syscall.
* linux/s390/syscallent.h ([381]): Likewise.

tests: check short read of struct flock/flock64

* tests/fcntl-common.c (test_flock_einval): Check short read
of struct flock.
(test_flock64_einval): Check short read of struct flock64.

tests: extend struct flock/flock64 decoding checks

* tests/fcntl-common.c (test_flock_einval, test_flock): Create
struct_kernel_flock object using tail_alloc.
(test_flock64_einval): Create struct_kernel_flock64 object using
tail_alloc.
* tests/fcntl64.c (test_flock64_lk64): Likewise.

tests: extend F_OFD_* decoding checks

Check that non-faulty F_OFD_* fcntl operations are decoded properly.

* tests/fcntl-common.c: Include <string.h>.
(test_flock64_ofd): New function.
(test_flock64): Invoke it, remove [F_OFD_SETLK].

tests/fcntl-common.c: define and initialize errstr

Initialize errstr inside invoke_test_syscall so that its users
won't have to go into trouble of caring about clobbering errno.

* tests/fcntl-common.c (errstr): New variable.
(invoke_test_syscall): Initialize it.
(test_flock_einval, test_flock64_einval, test_flock,
test_f_owner_ex_type_pid): Use it.
* tests/fcntl.c (test_flock64_undecoded): Likewise.
* tests/fcntl64.c (test_flock64_lk64): Likewise.

tests/fcntl-common.c: add fd argument to invoke_test_syscall

Extend invoke_test_syscall for later use with different file
descriptors.

* tests/fcntl-common.c (invoke_test_syscall): Add fd argument, pass it
to TEST_SYSCALL_NR syscall.  All users updated.

tests: move F_OFD_SETLK* checks from fcntl64.c to fcntl-common.c

This change enables F_OFD_SETLK* tests for fcntl as well.

* tests/fcntl64.c (test_flock64): Rename to test_flock64_lk64.
[F_OFD_SETLK]: Remove.
(test_flock64_einval): Move ...
* tests/fcntl-common.c: ... here.
(test_flock64_lk64): New protype.
(test_flock64_einval, test_flock64): New functions.
* tests/fcntl.c (test_flock64_einval): Rename to test_flock64_undecoded.
(TEST_FLOCK64_UNDECODED): New macro.
(test_flock64): Rename to test_flock64_lk64, replace TEST_FLOCK64_EINVAL
with TEST_FLOCK64_UNDECODED.

tests: move F_OWNER_* tests to fcntl-common.c

This change enables F_OWNER_* tests for fcntl64 as well.

* tests/fcntl.c: Move main, [TEST_FLOCK64_EINVAL],
and [TEST_F_OWNER_EX] ...
* tests/fcntl-common.c: ... here.
(test_flock64): New prototype.
* tests/fcntl64.c (main): Remove.

tests: rename struct_flock.c to fcntl-common.c

It is not limited to struct flock, more fcntl tests are going to land
in this file.

* tests/struct_flock.c: Rename to fcntl-common.c.
* tests/Makefile.am (EXTRA_DIST): Rename struct_flock.c
to fcntl-common.c.
* tests/fcntl.c: Likewise.
* tests/fcntl64.c: Likewise.

tests: check decoding of new FRA_* netlink attributes

* configure.ac (AC_CHECK_TYPES): Check for struct fib_rule_port_range
in <linux/fib_rules.h>.
* tests/nlattr_fib_rule_hdr.c: Include <linux/in.h>.
(FRA_PROTOCOL, FRA_IP_PROTO, FRA_SPORT_RANGE, FRA_DPORT_RANGE): New
macro constants.
[!HAVE_STRUCT_FIB_RULE_PORT_RANGE] (struct fib_rule_port_range): New
type.
(main): Check decoding of new FRA_* netlink attributes.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

rtnl_rule: decode new FRA_* attributes

* nlattr.h (DECL_NLA(ip_proto), DECL_NLA(rt_proto)): New declarations.
* nlattr.c (decode_nla_ip_proto): New function.
* rtnl_route.c (decode_nla_rt_proto): Likewise.
* rtnl_rule.c (decode_rule_port_rang): Likewise.
(fib_rule_hdr_nla_decoders) <[FRA_PROTOCOL]>: New attribute, introduced
by Linux commit v4.17-rc1~148^2~371.
(fib_rule_hdr_nla_decoders) <[FRA_IP_PROTO], [FRA_SPORT_RANGE],
[FRA_DPORT_RANGE]>: New attributes, introduced by Linux commit
v4.17-rc1~148^2~328^2~4.
* xlat/rtnl_rule_attrs.in (FRA_PROTOCOL, FRA_IP_PROTO, FRA_SPORT_RANGE,
FRA_DPORT_RANGE): New constants.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

nlattr: introduce decode_nla_xval helper

* nlattr.c (decode_nla_xval): New function.
* nlattr.h: Include "xlat.h".
(struct decode_nla_xlat_opts): New type.
(DECL_NLA(xval)): New declaration.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

Move definitions of xlat_type and xlat_style from defs.h to xlat.h

* defs.h (enum xlat_type, enum xlat_style, XLAT_STYLE_FORMAT_SHIFT,
XLAT_STYLE_VERBOSITY_MASK, XLAT_STYLE_FORMAT_MASK, XLAT_STYLE_SPEC_BITS,
XLAT_STYLE_MASK): Move ...
* xlat.h: ... here.

travis: switch from gcc-7 to gcc-8

* .travis.yml (matrix): Change gcc-7 to gcc-8.

tests: use syscall() for ioctl() calls with oversized arguments

* tests/ioctl_inotify.c (sys_ioctl): New function.
(main): Use sys_ioctl for ioctl calls with oversized arguments.
* tests/ioctl_loop.c: Likewise.
* tests/ioctl_perf.c: Likewise.

print_ifindex: respect xlat style settings

Print interface indices according to the preferred xlat style settings.

* print_ifindex.c (INI_PFX, INI_SFX): New helper macros.
[HAVE_IF_INDEXTONAME] (get_ifname): New function.
[HAVE_IF_INDEXTONAME] (print_ifindex): Implement as a wrapper around
get_ifname and print_xlat_ex.

tests/test_nlattr.h: fix pointer calculation in TEST_NLATTR_ARRAY, take 2

The old code works only when an array contains exactly two items.

* tests/test_nlattr.h (TEST_NLATTR_ARRAY): Fix pointer address
for the case of incomplete read.

nlattr: print index names in netlink meminfo array

* xlat/netlink_sk_meminfo_indices.in: New file.
* nlattr.c: Include xlat/netlink_sk_meminfo_indices.h.
(decode_nla_meminfo): Use print_array_ex, pass
netlink_sk_meminfo_indices as index xlat table.
* tests/nlattr_inet_diag_msg.c: Update expected output.

Introduce ARRSZ_PAIR macro

A shorthand similar to ARG_STR in its purpose: it provides and ability
to pass an array an its size without variable name duplication.

* macros.h (ARRSZ_PAIR): New macro.

print_array: add support for printing array indices

* defs.h (XLAT_STYLE_SPEC_BITS, XLAT_STYLE_MASK): New macro constants.
(tfetch_mem_fn, print_fn): New typedefs.
(enum print_array_flag_bits, enum print_array_flags): New enumerations.
(print_array_ex): Rename from print_array, add flags, index_xlat,
index_xlat_size, and index_dflt arguments.
(print_array): New static inline function, a thin wrapper around
print_array_ex.
util.c: Include "xlat.h".
(print_array): Rename to print_array_ex, add flags, index_xlat,
index_xlat_size, and index_dflt arguments.  Print array indices
according to the style settings specified by flags if PAF_PRINT_INDICES
is set.

Enhance printing of unfetchable object addresses in sequences and arrays

* btrfs.c (btrfs_print_tree_search): Print the address of unfetchable
object inside the sequence using printaddr_comment.
* msghdr.c (decode_msg_control): Likewise.
* execve.c (printargv): Print the address of unfetchable object inside
the array using printaddr_comment.
* netlink.c (fetch_nlmsghdr): Add in_array argument.  When in_array
is true, print the address of unfetchable object using
printaddr_comment.
* nlattr.c (fetch_nlattr): Likewise.
* tests/execve.c (main): Update expected output.
* tests/execveat.c (main): Likewise.
* tests/msg_control.c (test_scm_rights1, test_scm_rights2,
test_scm_security): Likewise.
* tests/netlink_protocol.c (send_query): Likewise.
* tests/nlattr.c (test_nlattr): Likewise.

print_array: enhance printing of unfetchable object addresses

When umoven_func invocation fails to fetch data, it prints the faulty
address.  If this happens to a subsequent umoven_func invocation,
the printed address may be undistinguishable from a valid data printed
by print_func, e.g. when the data is printed in a numeric form like
[0x1, 0x2, 0x3, 0xdefaced].

Fix this source of confusion by moving the printing of the faulty
address from umoven_func to print_array itself.  This change renames
umoven_func to tfetch_mem_func and changes its semantics, so that
 - tfetch_mem_func never prints anything;
 - tfetch_mem_func returns true if the fetch succeeded,
   and false otherwise.

* defs.h (print_array): Replace umoven_func argument with
tfetch_mem_func.
* util.c (print_array): Replace umoven_func argument with
tfetch_mem_func, document expected tfetch_mem_func return value
semantics.  When tfetch_mem_func returns false, print either addr
or "... /* addr */" depending on the context (inside the array or not).
* bpf.c (print_ebpf_prog, print_bpf_prog_info,
BEGIN_BPF_CMD_DECODER(BPF_PROG_QUERY)): Replace umoven_or_printaddr
argument of print_array with tfetch_mem.
* bpf_filter.c (print_bpf_fprog): Likewise.
* btrfs.c (btrfs_print_logical_ino_container,
btrfs_print_ino_path_container, btrfs_print_qgroup_inherit,
btrfs_ioctl): Likewise.
* dm.c (dm_decode_dm_target_deps): Likewise.
* epoll.c (epoll_wait_common): Likewise.
* file_ioctl.c (file_ioctl): Likewise.
* ipc_sem.c (tprint_sembuf_array): Likewise.
* kexec.c (print_kexec_segments): Likewise.
* mem.c (SYS_FUNC(subpage_prot)): Likewise.
* net.c (print_getsockopt): Likewise.
* netlink.c (decode_nlmsgerr_attr_cookie): Likewise.
* netlink_netlink_diag.c (decode_netlink_diag_groups): Likewise.
* netlink_packet_diag.c (decode_packet_diag_mclist): Likewise.
* netlink_unix_diag.c (decode_unix_diag_inode): Likewise.
* nlattr.c (decode_nla_meminfo): Likewise.
* numa.c (print_nodemask, SYS_FUNC(move_pages),
* perf_ioctl.c (perf_ioctl_query_bpf): Likewise.
* poll.c (decode_poll_entering): Likewise.
* printsiginfo.c (print_siginfo_array): Likewise.
* rtnl_tc.c (decode_tca_stab_data): Likewise.
* sock.c (decode_ifconf): Likewise.
* uid.c (print_groups): Likewise.
* io.c (SYS_FUNC(io_submit), SYS_FUNC(io_getevents)): Replace
umoven_or_printaddr argument of print_array with tfetch_mem.
(tprint_iov_upto): Replace umoven_or_printaddr_ignore_syserror
with tfetch_mem_ignore_syserror.
* v4l2.c (print_v4l2_format_fmt): Replace umoven_or_printaddr argument
of print_array with tfetch_mem.
(print_v4l2_ext_controls): Replace umoven_or_printaddr_ignore_syserror
with tfetch_mem_ignore_syserror.
* mmsghdr.c (fetch_struct_mmsghdr_or_printaddr): Rename
to fetch_struct_mmsghdr_for_print, do not print address, return bool.
(decode_mmsgvec): Replace fetch_struct_mmsghdr_or_printaddr
with fetch_struct_mmsghdr_for_print.
* tests/aio.c (main): Update expected output.
* tests/bpf.c (print_BPF_PROG_QUERY_attr5): Likewise.
* tests/ioctl_perf-success.c (main): Likewise.
* tests/ioctl_v4l2.c (main): Update expected output.
* tests/kexec_load.c (main): Likewise.
* tests/mmsg_name.c (test_mmsg_name): Update expected output.
* tests/move_pages.c (print_page_array, print_node_array): Likewise.
* tests/poll.c (print_pollfd_array_entering): Likewise.
* tests/preadv-pwritev.c (main): Likewise.
* tests/preadv2-pwritev2.c (dumpio): Likewise.
* tests/process_vm_readv_writev.c (print_iov): Likewise.
* tests/pwritev.c (print_iovec): Likewise.
* tests/readv.c (main): Likewise.
* tests/seccomp-filter-v.c
* tests/semop.c (main): Likewise.
* tests/set_mempolicy.c (print_nodes): Likewise.
* tests/setgroups.c (main): Likewise.
* tests/test_nlattr.h (print_nlattr) Likewise.

Co-Authored-by: Eugene Syromyatnikov <evgsyr@gmail.com>

Introduce tfetch_* family of functions and macros

These new functions are going to be used instead of umove_or_printaddr*
as umoven_func argument of print_array.

* defs.h (tfetch_mem64, tfetch_mem64_ignore_syserror): New
function prototypes.
(tfetch_mem, tfetch_mem_ignore_syserror): New static inline
wrappers.
(tfetch_obj): New macro wrapper around tfetch_mem.
* util.c (tfetch_mem64, tfetch_mem64_ignore_syserror): New
functions.
(umoven_or_printaddr64, umoven_or_printaddr64_ignore_syserror): Use
them.

Introduce printaddr_comment

* defs.h (printaddr_comment): New static inline wrapper around
tprintf_comment.

Turn printaddr into a thin wrapper around printaddr64

* util.c (printaddr): Move ...
* defs.h: ... here, add "static inline" qualifiers.

defs.h: add comments about umove* return values

* defs.h (umoven, umoven_or_printaddr64,
umoven_or_printaddr64_ignore_syserror, umovestr): Add comments about
return values.

tests: use TAIL_ALLOC_OBJECT_CONST_PTR where appropriate

* tests/kcmp.c (main): Use TAIL_ALLOC_OBJECT_CONST_PTR.
* tests/modify_ldt.c (main): Likewise.
* tests/netlink_protocol.c (send_query): Likewise.
* tests/ptrace.c (test_peeksiginfo, main): Likewise.
* tests/s390_guarded_storage.c (main): Likewise.
* tests/s390_sthyi.c (main): Likewise.
* tests/xet_thread_area_x86.c (main): Likewise.

bpf: decode bpf_attr.info field used by BPF_OBJ_GET_INFO_BY_FD command

* bpf_attr.h [!BPF_TAG_SIZE] (BPF_TAG_SIZE): Define.
[BPF_TAG_SIZE]: Check that BPF_TAG_SIZE is 8.
(struct bpf_map_info_struct, struct bpf_prog_info_struct): New
structures.
(bpf_map_info_struct_size, expected_bpf_map_info_struct_size,
bpf_prog_info_struct_size, expected_bpf_prog_info_struct_size): New
macro constants.
* bpf.c (print_bpf_obj_info_fn): New type.
(print_bpf_map_info, print_bpf_prog_info, fetch_bpf_obj_info,
print_bpf_obj_info_addr, print_bpf_obj_info): New functions.
(BEGIN_BPF_CMD_DECODER(BPF_OBJ_GET_INFO_BY_FD)): Print bpf_fd and info_len
fields only on entering, call print_bpf_obj_info for printing info field.
* tests/bpf.c (BPF_OBJ_GET_INFO_BY_FD_checks): Print info field
as a pointer.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: add support for checking structures outside union bpf_attr

struct bpf_prog_info and bpf_map_info need essentially the same handling
as union bpf_attr.

* gen_bpf_attr_check.sh: Derive type_name from $struct if it doesn't
start with "BPF_", derive TYPE_NAME from type_name, use them in code
generation.
* m4/gen_bpf_attr_m4.sh: Rewrite parsing/generation code into awk,
add support for structures outside union bpf_attr.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

bpf: remove bpf_priv_data

As various bpf commands need different data to save between entering
and exiting of the syscall, do not emulate union bpf_attr approach.

* bpf.c (struct bpf_priv_data): Remove.
(DECL_BPF_CMD_DECODER): Remove "struct bpf_priv_data *" argument.
(BEGIN_BPF_CMD_DECODER(BPF_PROG_QUERY)): Use set_tcb_priv_ulong
and get_tcb_priv_ulong to keep bpf_attr.query.prog_cnt field.
(SYS_FUNC(bpf)): Remove priv variable, do not initialize it,
and do not pass it to bpf_cmd_decoders.

tests: add union bpf_attr.query.prog_ids field decoding checks

* tests/bpf-success-v.c: New file.
* tests/bpf-success.c: Likewise.
* tests/bpf-success-v.test: New test.
* tests/bpf-success.test: Likewise.
* tests/.gitignore: Add bpf-success and bpf-success-v.
* tests/Makefile.am (check_PROGRAMS): Likewise.
(DECODER_TESTS): Add bpf-success.test and bpf-success-v.test.
* tests/bpf.c [INJECT_RETVAL] (sys_bpf): Check that retuned value
is equivalent to the injected one; append "(INJECTED)" string to errstr.
(prog_load_ids, prog_load_ids_ptr): New variables.
(init_BPF_PROG_QUERY_attr4, print_BPF_PROG_QUERY_attr4,
init_BPF_PROG_QUERY_attr5, print_BPF_PROG_QUERY_attr5): New functions.
(BPF_PROG_QUERY_checks): Drop "const" qualifier, add new checks.

bpf: BPF_PROG_QUERY bpf_attr.query.prog_ids printing fixes

* bpf.c (BEGIN_BPF_CMD_DECODER(BPF_PROG_QUERY)): Use
print_uint32_array_member instead of print_uint64_array_member for
printing prog_ids elements (and change the type of prog_id_buf from
uint64_t to uint32_t), as they are 32-bit; use common trick with
print_big_u64_addr for handling possible ambiguity with prog_ids
address.
* tests/bpf.c (BIG_ADDR_MAYBE): New macro.
(BPF_PROG_QUERY_checks): Update expected output.

bpf: add support for instruction decoding

* xlat/ebpf_regs.in: New file.
* bpf.c (struct ebpf_insn, struct ebpf_insns_data): New type definitions.
(print_ebpf_insn, print_ebpf_prog): New functions.
(DEF_BPF_CMD_DECODER(BPF_PROG_LOAD)): Use them.

tests: check decoding of successful PERF_EVENT_IOC_{ID,QUERY_BPF} ioctls

As these have some argument output on the exit path.

* tests/ioctl_perf-success.c: New file.
* tests/ioctl_perf-success.test: New test.
* tests/Makefile.am (DECODER_TESTS): Add it.
(check_PROGRAMS): Add ioctl_perf-success.
* tests/.gitignore: Likewise.
* tests/ioctl_perf.c: Remove TODO.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

Implement PERF_EVENT_IOC_* decoding

* perf.c (fetch_perf_event_attr, print_perf_event_attr): Remove "static"
qualifier.
* defs.h (fetch_perf_event_attr, print_perf_event_attr): New
declarations.
* perf_event_struct.h (struct perf_event_query_bpf): New type
definition.
* perf_ioctl.c: New file.
* Makefile.am (strace_SOURCES): Add it.
* xlat/perf_ioctl_cmds.in: New file.
* xlat/perf_ioctl_flags.in: Likewise.
* ioctl.c (ioctl_decode) <case '$'>: Call perf_ioctl.
* tests/gen_tests.in (ioctl_perf): New test.
* tests/ioctl_perf.c: New file.
* tests/pure_executables.list: ioctl_perf.
* tests/.gitignore: Likewise.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

tests: fix ipc_msgbuf.test

* tests/ipc_msgbuf.test: Fix -a argument.

xlat: update IPPROTO_* constants

* xlat/inet_protocols.in (IPPROTO_GGP, IPPROTO_HELLO, IPPROTO_ND):
Remove, not defined in Linux.
(IPPROTO_BEETPH): New constant, introduced by Linux commit
v2.6.19-rc1~33^2.
(IPPROTO_L2TP): New constant, introduced by Linux commit
v2.6.35-rc1~473^2~594.
(IPPROTO_MH): New constant, introduced by Linux commit
v2.6.19-rc1~1272^2~157.
Add default values to constants, sort the file.

xlat: add fallback definitions to RTPROT_* constants

* xlat/routing_protocols.in: Add default values to constants.

xlat: add SECCOMP_FILTER_FLAG_SPEC_ALLOW

* xlat/seccomp_filter_flags.in (SECCOMP_FILTER_FLAG_SPEC_ALLOW): New
constant, introduced by Linux commit v4.17-rc3-24-g00a02d0.
* tests/seccomp-filter.c: Update expected output.
* tests/seccomp-filter-v.c: Likewise.

prctl: add PR_GET_SPECULATION_CTRL/PR_SET_SPECULATION_CTRL decoding

* xlat/pr_spec_cmds.in: New fille.
* xlat/pr_spec_get_store_bypass_flags.in: Likewise.
* xlat/pr_spec_set_store_bypass_flags.in: Likewise.
* xlat/prctl_options.in (PR_GET_SPECULATION_CTRL,
PR_SET_SPECULATION_CTRL): New constants, introduced by Linux commit
v4.17-rc3-15-gb617cfc.
* prctl.c (SYS_FUNC(prctl)) <case PR_GET_SPECULATION_CTRL,
case PR_SET_SPECULATION_CTRL>: Implement decoding of new prctl options.
* tests/Makefile.am (check_PROGRAMS): Add prctl-spec-inject.
(DECODER_TESTS): Add prctl-spec-inject.test.
* tests/prctl-spec-inject.c: New file.
* tests/prctl-spec-inject.test: New test.
* tests/.gitignore: Add prctl-spec-inject.

tests: add keyctl variants with different xlat verbosity levels

* tests/keyctl-Xabbrev.c: New file.
* tests/keyctl-Xraw.c: Likewise.
* tests/keyctl-Xverbose.c: Likewise.
* tests/keyctl.c (XARG_STR): New macro.
(do_keyctl): Print command in accordance with XLAT_RAW/XLAT_VERBOSE settings.
(main): Update expected output.
* tests/pure_executables.list: Add keyctl-Xabbrev, keyctl-Xraw,
and keyctl-Xverbose.
* tests/.gitignore: Likewise.
* tests/gen_tests.in (keyctl-Xabbrev, keyctl-Xraw, keyctl-Xverbose): New
tests.

keyctl: print KEYCTL_SET_REQKEY_KEYRING argument as an integer

* xlat/key_reqkeys.in: Add #val_type int.
* keyctl.c (SYS_FUNC(keyctl)) <case KEYCTL_SET_REQKEY_KEYRING>: Print
arg2 as an integer.
* tests/keyctl.c: Update expected output.

keyctl: respect xlat style when printing errno

* keyctl.c (keyctl_reject_key): Use print_xlat_ex for printing error
argument.

syscall.c: unify error value printing

There's little reason not to decode error code for syscalls printed in
raw; moreover, it creates inconsistencies for unknown syscall printing.

Another issue was with error numbers unknown to strace: previously, they
were printed as "... = -1 1234 (Unknown error 1234)" which looked kinda
weird.

* syscall.c (print_err_ret): New function.
(syscall_exiting_trace): Remove u_error variable, use tcp->u_error
instead. Remove u_error_str variable.
(syscall_exiting_trace) <if (raw(tcp))>: Call print_err_ret if u_error
is non-zero.
(syscall_exiting_trace) <if (!(sys_res & RVAL_NONE) && tcp->u_error)>:
Use print_err_ret for default error printing.
* tests/nsyscalls.c [!LINUX_MIPSO32] (test_syscall): Update expected
output.
* tests/qual_fault.c (invoke) <if (is_raw)>: Likewise.
* strace.1.in (.SH DESCIPTION): Update unknown syscall example.

ioctl: implement INOTIFY_IOC_SETNEXTWD decoding

* inotify_ioctl.c: New file.
* Makefile.am (strace_SOURCES): Add it.
* defs.h (DECL_IOCTL(inotify)): New declaration.
* ioctl.c (ioctl_decode) <case 'I'>: Call inotify_ioctl.
* tests/ioctl_inotify.c: New file.
* tests/.gitignore: Add ioctl_inotify.
* tests/pure_executables.list: Likewise.
* tests/gen_tests.in (ioctl_inotify): New test.

tests/test_nlattr.h: avoid testing cropped objects of size 1

* tests/test_nlattr.h (TEST_NLATTR_OBJECT_EX_): Check
"len < sizeof(obj_)" case only if "sizeof(obj_) > 1".

tests: pass index to print_elem_ callback in TEST_NLATTR_ARRAY

* tests/test_nlattr.h (TEST_NLATTR_ARRAY): Pass array index to the
print_elem_ callback.
* tests/nlattr_inet_diag_msg.c (print_uint): Add index argument.
* tests/nlattr_netlink_diag_msg.c (print_xlong): Likewise.
* tests/nlattr_packet_diag_msg.c (print_packet_diag_mclist,
print_sock_filter): Likewise.
* tests/nlattr_unix_diag_msg.c (print_uint): Likewise.

tests: further cleanup bpf big address checks

Replace BIG_ADDR_IS_64BIT and BIG_ADDR_IS_32BIT with BIG_ADDR().

* tests/bpf.c (BIG_ADDR): New macro.
(BIG_ADDR_IS_64BIT, BIG_ADDR_IS_32BIT): Remove.
(BPF_PROG_LOAD_checks, BPF_OBJ_PIN_checks,
BPF_RAW_TRACEPOINT_OPEN_checks): Use BIG_ADDR() instead of
BIG_ADDR_IS_64BIT and BIG_ADDR_IS_32BIT.

tests: cleanup bpf big address checks

Introduce BIG_ADDR_IS_64BIT and BIG_ADDR_IS_32BIT to reduce redundancy.

* tests/bpf.c (BIG_ADDR_IS_64BIT, BIG_ADDR_IS_32BIT): New macros.
(BPF_PROG_LOAD_checks, BPF_OBJ_PIN_checks,
BPF_RAW_TRACEPOINT_OPEN_checks): Use them.

bpf: add support for BPF_RAW_TRACEPOINT_OPEN command decoding

* bpf_attr.h (struct BPF_RAW_TRACEPOINT_OPEN_struct): New type
definition.
(BPF_RAW_TRACEPOINT_OPEN_struct_size,
expected_BPF_RAW_TRACEPOINT_OPEN_struct_size): New macros.
* xlat/bpf_commands.in (BPF_RAW_TRACEPOINT_OPEN): New constant,
introduced by Linux commit v4.17-rc1~148^2~19^2~4^2~3.
* xlat/bpf_prog_types.in (BPF_PROG_TYPE_RAW_TRACEPOINT): Likewise.
* bpf.c (BEGIN_BPF_CMD_DECODER(BPF_RAW_TRACEPOINT_OPEN)): New bpf
command decoder.
(SYS_FUNC(bpf)) <bpf_cmd_decoders[]>: Add
BPF_CMD_ENTRY(BPF_RAW_TRACEPOINT_OPEN).
* tests/bpf.c (union bpf_attr_data): Add
BPF_ATTR_DATA_FIELD(BPF_RAW_TRACEPOINT_OPEN).
(BPF_PROG_LOAD_checks): Update.
(init_BPF_RAW_TRACEPOINT_attr2): New function.
(BPF_RAW_TRACEPOINT_OPEN_checks): New checks array.
(main) <checks>: Add CHK(BPF_RAW_TRACEPOINT_OPEN).

bpf: add support for decoding struct bpf_attr.expected_attach_type field

* bpf_attr.h (struct BPF_PROG_LOAD_struct): Add expected_attach_type
field.
(BPF_PROG_LOAD_struct_size): Update to offsetofend of
expected_attach_type field.
(expected_BPF_PROG_LOAD_struct_size): Update to 72.
* bpf.c (BEGIN_BPF_CMD_DECODER(BPF_PROG_LOAD)): Decode
expected_attach_type field.
* tests/bpf.c: Update expected output.

nlattr: do not assume that SK_MEMINFO_VARS is constant

SK_MEMINFO_VARS changes over time (as it was in Linux commit
v4.7-rc1~154^2~354^2~4, for example), so we cannot use it
for sanity checks.

* nlattr.c (print_meminfo): Remove.
(decode_nla_meminfo): Use generic print_uint32_array_member element
printer callback.
* tests/nlattr_inet_diag_msg.c (main): Update expected output.

Introduce print_uint32_array_member helper function

Analogous to print_int32_array_member, this helper is going to be used
in various decoders, including BPF_PROG_QUERY, PERF_EVENT_IOC_*,
and netlink meminfo.

* defs.h (print_uint32_array_member): New declaration.
* util.c (print_uint32_array_member): New function.

Introduce print_int32_array_member helper function

Analogous to print_uint64_array_member, this helper is going to be used
for IFLA_AF_SPEC decoding.

* defs.h (print_int32_array_member): New declaration.
* util.c (print_int32_array_member): New function.

defs.h: move print_uint64_array_member and print_array declarations down

In preparation for the forthcoming changes.

tests/test_nlattr.h: fix pointer calculation in TEST_NLATTR_ARRAY

The old code works only when an array contains exactly two items.

* tests/test_nlattr.h (TEST_NLATTR_ARRAY): Fix pointer address
for the case of incomplete read.

sockaddr: add support for sockaddr_l2 without l2_bdaddr_type

Apparently, struct sockaddr_l2 changed over time: the l2_bdaddr_type
field was added only by commit v3.5-rc1~35^2~3^2~18^2^2~39.

* configure.ac (AC_CHECK_HEADERS): Move bluetooth/bluetooth.h check into
a separate check.
(AC_CHECK_HEADERS([bluetooth/bluetooth.h])): New check, also checks for
struct sockaddr_l2.l2_bdaddr_type member in case of success.
* sockaddr.c (print_sockaddr_data_bt) <case offsetof(struct sockaddr_l2,
l2_bdaddr_type)>: Add to sizeof(struct sockaddr_l2) case. Print
l2_bdaddr_type field only if the provided sockaddr is big enough.
* tests/net-sockaddr.c (check_l2): Put l2_bdaddr_type usage/printing
under HAVE_STRUCT_SOCKADDR_L2_L2_BDADDR_TYPE ifdef, add a check for the
sockaddr_l2 structure without the l2_bdaddr_type field.

sock: guess ifr_hwaddr size in SIOCSIFHWADDR/SIOCGIFHWADDR

And make the output more structured.

* sock.c: Include xlat/arp_hardware_types.h in XLAT_MACROS_ONLY mode.
(print_ifreq) <case SIOCSIFHWADDR, case SIOCGIFHWADDR>: Introduce
hwaddr_sizes array, print ifr_hwaddr.sa_family, print ifr_hwaddr.sa_data
using hwaddr_sizes and PRINT_FIELD_MAC_SZ.

Simplify errnoent and signalent

Remove personality support for errnoent and signalent as
there is nothing personality-specific in these files.

* linux/aarch64/errnoent1.h: Remove.
* linux/aarch64/signalent1.h: Likewise.
* linux/powerpc64/errnoent1.h: Likewise.
* linux/powerpc64/signalent1.h: Likewise.
* linux/riscv/errnoent1.h: Likewise.
* linux/riscv/signalent1.h: Likewise.
* linux/s390x/errnoent1.h: Likewise.
* linux/s390x/signalent1.h: Likewise.
* linux/sparc64/errnoent1.h: Likewise.
* linux/sparc64/signalent1.h: Likewise.
* linux/tile/errnoent1.h: Likewise.
* linux/tile/signalent1.h: Likewise.
* linux/x32/errnoent1.h: Likewise.
* linux/x32/signalent1.h: Likewise.
* linux/x86_64/errnoent1.h: Likewise.
* linux/x86_64/errnoent2.h: Likewise.
* linux/x86_64/signalent1.h: Likewise.
* linux/x86_64/signalent2.h: Likewise.
* Makefile.am (EXTRA_DIST): Remove them.
* defs.h (errnoent0): Rename to errnoent.
(signalent0): Rename signalent.
(nerrnos, nsignals): Change to const variables unconditionally.
* syscall.c (errnoent0): Rename to errnoent.
(signalent0): Rename signalent.
(nerrnos, nsignals): Change to const variables unconditionally.
(nerrnos0, nsignals0): Remove.
[SUPPORTED_PERSONALITIES > 1] (errnoent1, signalent1, nerrnos1,
nsignals1): Remove.
[SUPPORTED_PERSONALITIES > 2] (errnoent2, signalent2, nerrnos2,
nsignals2): Likewise.
(set_personality): Do not assign errnoent, signalent, nerrnos,
and nsignals.

strace.1.in: mention -Xraw in -eraw description

* strace.1.in (\fB\-e\ raw\fR=\,\fIset\fR): Mention -X raw.

strace.1.in: display -k option description only if it is enabled

Leverage the fact that manpage is generated and add condition statements
around mentions of -k option.

* strace.1.in (.SH SYNOPSIS): Put option list inside an
".if '@ENABLE_STACKTRACE_TRUE@'#' .ig end_unwind_opt ... .end_unwind_opt"
condition; add a second list of options without -k and put it inside
".if '@ENABLE_STACKTRACE_FALSE@'#' .ig end_no_unwind_opt ... .end_no_unwind_opt"
condition.
(.SS Output format) <-k>: Put it inside
".if '@ENABLE_STACKTRACE_TRUE@'#' .ig end_unwind ... .end_unwind"
condition, remove the mention of --enable-stacktrace configuration
option requirement.

strace.1.in: format a reference to "open" syscall using bold font

* strace.1.in (.SH DESCRIPTION): Format a reference to "open" syscall
using bold font.

tests/qual_fault.test: add -efu arguments to the interpreter

* tests/qual_fault.test (#!/bin/sh): Add -efu arguments.

tests/qual_fault.test: swap expected/output files in match_diff calls

match_diff expects the first argument as strace output and the second
argument as expected output.

* tests/qual_fault.test (check_fault_injection): Swap arguments
in match_diff calls.

defs.h: convert some macro wrappers into static inline functions

* defs.h (pathtrace_select, pathtrace_match, printxval_searchn,
printxval_indexn, sprintxval, sprintflags, printnum_slong,
printnum_ulong, printnum_ptr, printnum_kptr): Convert macro wrappers
into static inline functions.

xlat: add support for xlat_styles in printxval_dispatch

* defs.h (printxval_dispatch_ex): Rename from printxval_dispatch, add
style argument.
(printxval_dispatch): New static inline function, a thin wrapper around
printxval_dispatch_ex.
* xlat.c (printxval_dispatch): Rename to printxval_dispatch_ex, add
style argument, handle it.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

Make printxval_dispatch publicly available

* evdev.c (enum xlat_type, printxval_dispatch): Move elsewhere.
* defs.h (enum_xlat_type): Move from evdev.c.
(printxval_dispatch): New declaration.
* xlat.c (printxval_dispatch): Move from evdev.c, drop static qualifier.

Revert "evdev: remove XT_NORMAL"

This reverts commit cc52da10c6f13feba577a83088f34916c6350165
in preparation for printxval_dispatch generalisation.

* evdev.c (enum xlat_type): Add XT_NORMAL back.
(printxval_dispatch): Handle XT_NORMAL again.

defs.h: introduce is_bigendian macro

This macro is going to be used in inline checks.

* defs.h (is_bigendian): New macro constant.

util.c: make memory allocation error messages more informative

As those are pretty unusual, let's add some additional information
that can be reported by users.

* util.c (print_quoted_string_ex, dumpstr): Add reasons for memory
allocation errors.
(sizeof_iov): Change type from size_t to unsigned int.
(dumpiov_upto): Rewrite size initialisation similarly to the way it's
done in print_quoted_string_ex, add reasons for memory allocation errors.

Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>

strace.1.in: fix -A description

* strace.1.in (.SS Output format) <-A>: Add missing .B.

strace.1.in: document x32 64-bit syscall peculiarity

* strace.1.in (.SH NOTES): Document "#64" suffix for 64-bit syscalls on
x32.

linux/x32/syscallent.h: change 64-bit syscall designation

The current way of designation of 64-bit specific syscalls in x32
personality not only conflicts with fault injection specification
syntax, but also makes it impossible to specify such syscalls at all
(since everything that starts from a digit is considered a syscall
number specification).  So, let's change prefix to suffix and use
octothorpe instead of colon as a separator.

* linux/x32/syscallent.h (13, 15, 16, 19, 20, 45, 46, 47, 54, 55, 59,
101, 127, 128. 129, 131, 134, 156, 174, 177, 178, 180, 205, 206, 209,
211, 214, 215, 222, 236, 244, 246, 247, 273, 274, 278, 279, 295, 296,
297, 299, 307, 310, 311, 32, 327, 328): Change "64:" prefix to "#64"
suffix in syscall name.

Closes: https://github.com/strace/strace/issues/36

xlat: add BPF_PROG_TYPE_SK_MSG

* xlat/bpf_prog_types.in (BPF_PROG_TYPE_SK_MSG): New constant,
introduced by Linux commit v4.17-rc1~148^2~156^2~1^2~13.
* tests/bpf.c (BPF_PROG_LOAD_checks): Update expected output.

xlat: update bpf_attach_type constants

* xlat/bpf_attach_type.in (BPF_SK_MSG_VERDICT): New constant,
introduced by Linux commit v4.17-rc1~148^2~156^2~1^2~13.
(BPF_CGROUP_INET4_BIND, BPF_CGROUP_INET6_BIND): New constants,
introduced by Linux commit v4.17-rc1~148^2~19^2^2~6.
(BPF_CGROUP_INET4_CONNECT, BPF_CGROUP_INET6_CONNECT): New constants,
introduced by Linux commit v4.17-rc1~148^2~19^2^2~3.
(BPF_CGROUP_INET4_POST_BIND, BPF_CGROUP_INET6_POST_BIND): New constants,
introduced by Linux commit v4.17-rc1~148^2~19^2^2~1.
* tests/bpf.c (BPF_PROG_QUERY_checks): Update expected output.

xlat/sock_rds_options.in: make it sorted

Define SO_TIMESTAMP twice, for PA-RISC and all the other architectures.

* xlat/sock_rds_options.in [!__hppa__] (SO_TIMESTAMP): Define to 29.
[__hppa__] (SO_TIMESTAMP): Define to 0x4012.
* net.c (print_sockopt_fd_level_name) <case SOL_RDS>: Use
printxval_search instead of printxval.

Update ioctl entries from linux v4.17

* linux/32/ioctls_inc_align16.h: Update from linux v4.17-rc5
using ioctls_gen.sh.
* linux/32/ioctls_inc_align32.h: Likewise.
* linux/32/ioctls_inc_align64.h: Likewise.
* linux/64/ioctls_inc.h: Likewise.
* linux/x32/ioctls_inc0.h: Likewise.
* linux/i386/ioctls_arch0.h: Likewise.
* linux/x86_64/ioctls_arch0.h: Likewise.
* NEWS: Mention this.

maint: update for linux 4.17-rc5

* maint/ioctls_sym.sh: (x86_list): Add KVM_HYPERV_EVENTFD.