Dmitry V. Levin [Tue, 14 Aug 2018 13:43:34 +0000 (13:43 +0000)]
strace.c: introduce struct tcb_wait_data
Introduce a new structure to pass information between next_event(),
restart_delayed_tcb(), and dispatch_event().
This is going to be used by a subsequent change of next_event().
* strace.c (struct tcb_wait_data): New type.
(next_event): Remove parameters, return a pointer
to const struct tcb_wait_data. Return NULL instead of TE_BREAK.
(dispatch_event): Replace all parameters with a pointer
to const struct tcb_wait_data, obtain the trace event, siginfo,
and status from its fields.
(restart_delayed_tcb): Add local struct tcb_wait_data variable
with te field set to TE_RESTART, pass it to dispatch_event().
(main): Remove status and si variables, update next_event()
and dispatch_event() invocations.
Zhibin Li [Wed, 1 Aug 2018 09:53:57 +0000 (17:53 +0800)]
evdev: fix decoding of EVIOCGBIT(0, ...)
There is a comment in drivers/input/evdev.c which says:
/* EV_SYN==0 is EV_CNT, _not_ SYN_CNT, see EVIOCGBIT */
That is, EVIOCGBIT(0, ...) should return a bit mask with supported
event types instead of SYN_* event codes.
* defs.h (evdev_ev): New prototype.
* evdev.c: Include "xlat/evdev_ev.h" and remove "xlat/evdev_sync.h".
(bit_ioctl) <case EV_SYN>: Replace EV_SYN with 0, use evdev_ev
with XT_SORTED in decode_bitset invocation instead.
* ioctl.c: Do not include "xlat/evdev_ev.h".
(evdev_decode_number): Print nr == 0x20 as "0" instead of "EV_SYN".
* tests/ioctl_evdev.c (main): Use 0 instead of EV_SYN in EVIOCGBIT
output.
* xlat/evdev_sync.in: Remove.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org> Fixes: v4.10~89 "Add decoding for evdev ioctls"
* xlat/kvm_cap.in: New file.
* kvm.c: Include "xlat/kvm_cap.h".
(kvm_ioctl_decode_check_extension): New function.
(kvm_ioctl): Use it.
* tests/ioctl_kvm_run_common.c (main): Check decoding
of KVM_CHECK_EXTENSION command.
Signed-off-by: Pierre Marsais <pierre.marsais@lse.epita.fr>
Dmitry V. Levin [Sat, 18 Aug 2018 07:22:47 +0000 (07:22 +0000)]
xlat: workaround V4L2_CID_USER_IMX_BASE Linux kernel ABI breakage
Linux kernel commit v4.18-rc2-106-g421860b9d47053badce4b247576fa48df9ab4c48
has changed the value of V4L2_CID_USER_IMX_BASE constant introduced
by commit v4.13-rc1~141^2~121 because the old value was already used
by V4L2_CID_USER_MAX217X_BASE.
This is of course an ABI breakage that affects Linux kernels starting
with 4.13 and up to 4.18, as well as their LTS derivatives.
Since the imx driver didn't provide any public control ID definitions,
it looks like the best way to handle this situation is to pretend that
the old value of V4L2_CID_USER_IMX_BASE didn't exist.
tests: allow any error code when trying to set time
Lower expectations of particular error code numbers when trying to set
time. This helps to workaround issues with unexpected return codes
in some exotic environments where EPERM is returned instead of EINVAL
(e.g. mock that uses systemd-nspawn).
* tests/clock_xettime.c (main): Do not hardcode EINVAL, print return
codes using sprintrc.
* tests/xettimeofday.c (main): Likewise.
* NEWS: Mention this fix.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
Avoid defining psw_compat_t and s390_compat_regs if their definitions
are provided by kernel. Before Linux commit v3.10-rc1~201^2~11,
these definition were available in <asm/ptrace.h>.
* configure.ac (AC_CHECK_TYPES): Check for s390_compat_regs.
* linux/s390x/arch_regs.c [HAVE_S390_COMPAT_REGS]: Do not define
psw_compat_t and s390_compat_regs.
* NEWS: Mention this fix.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
Dmitry V. Levin [Sun, 5 Aug 2018 11:49:04 +0000 (11:49 +0000)]
strace.spec.in: minor updates here and there
* strace.spec.in: Add descriptive comments to build requirements.
Introduce %maybe_use_defattr and use it instead of %defattr.
Add output of ld --version to the build environment information.
Use printf instead of echo -e.
Dmitry V. Levin [Sun, 5 Aug 2018 11:49:04 +0000 (11:49 +0000)]
strace.spec.in: remove strace64 subpackage
strace64 subpackage used to be installed into ppc32 and sparc32
userspace running by a 64-bit kernel to trace 64-bit processes.
As simultaneous decoding of processes running both in 64-bit and 32-bit
personality is fully supported on every architecture with compat kernel
support for quite some time, one can just install a 64-bit strace and
there is no more need for that strace64 kludge.
Dmitry V. Levin [Sat, 4 Aug 2018 11:45:20 +0000 (11:45 +0000)]
xlat: update IFLA_BRPORT_* constants
* xlat/rtnl_ifla_brport_attrs.in: Add IFLA_BRPORT_GROUP_FWD_MASK
constant introduced by Linux kernel commit v4.15-rc1~84^2~607.
Add IFLA_BRPORT_NEIGH_SUPPRESS constant introduced by Linux kernel
commit v4.15-rc1~84^2~523^2~2.
Add IFLA_BRPORT_ISOLATED constant introduced by Linux kernel commit
v4.18-rc1~114^2~142.
Zhibin Li [Wed, 18 Jul 2018 10:00:58 +0000 (18:00 +0800)]
tests: check decoding of successful quotactl
* tests/quotactl-success.c: New file.
* tests/quotactl-success-v.c: Likewise.
* tests/quotactl-xfs-success.c: Likewise.
* tests/quotactl-xfs-success-v.c: Likewise.
* tests/quotactl-success.test: New test.
* tests/quotactl-success-v.test: Likewise.
* tests/quotactl-xfs-success.test: Likewise.
* tests/quotactl-xfs-success-v.test: Likewise.
* tests/.gitignore: Add quotactl-success, quotactl-success-v,
quotactl-xfs-success and quotactl-xfs-success-v.
* tests/Makefile.am (check_PROGRAMS): Likewise.
(DECODER_TESTS): Add four new tests listed above.
* tests/quotactl.h: [INJECT_RETVAL] (check_quota): Use errstr instead of
sprintrc; check that returned value is equivalent to the injected one;
append "(INJECTED)" string to errstr.
* tests/quotactl.c: Changed condition rc != 0 to rc < 0 in callback
functions and value of opaque pointer addr_cb_arg in order to check
successful injection.
* tests/quotactl-xfs.c: Likewise.
net: enhance decoding of getsockopt(PACKET_STATISTICS)
* net.c (print_tpacket_stats): Change decoder to match the kernel
behaviour: getsockopt syscall accepts any non-negative *optlen and
writes either MIN(sizeof(struct tpacket_stats), *optlen) or
MIN(sizeof(struct tpacket_stats_v3), *optlen) bytes of data.
* net.c: Remove checks for SO_PEERCRED, SO_ATTACH_FILTER, and
SO_ATTACH_REUSEPORT_CBPF macros as they are defined unconditionally
since commit v4.23~199.
* net.c (print_ucred): Rename to print_get_ucred, change decoder
to match the kernel behaviour: getsockopt syscall accepts any
non-negative *optlen and writes MIN(sizeof(struct ucred), *optlen)
bytes of data.
(print_getsockopt): Replace print_ucred with print_get_ucred.
* tests/so_percred.c: Include <string.h>.
(main): Update expected output.
* net.c (print_get_linger): Change decoder to match the kernel
behaviour: getsockopt syscall accepts any non-negative *optlen
and writes MIN(sizeof(struct linger), *optlen) bytes of data.
(print_set_linger): Move after definition of SYS_FUNC(getsockopt).
* tests/so_linger.c: Include <stddef.h> and <string.h>.
(main): Update expected output.
* net.c (print_getsockopt): Apply (ulen < 0 || rlen < 0) check for all
cases. As the kernel neither accepts nor returns a negative length
in case of successful getsockopt syscall invocation, these negative
values must have been forged by userspace.
Workaround additional incompatibility between <sys/stat.h> and
<asm/stat.h> introduced by glibc-2.27.9000-566-gfd70af4.
* tests/xstatx.c (statx, statx_timestamp): Redefine before the inclusion
of <sys/stat.h> to avoid conflicts between definitions provided by the
linux kernel headers and the GNU libc.
Zhibin Li [Thu, 12 Jul 2018 09:32:12 +0000 (17:32 +0800)]
tests: use check_quota for invalid quotactl commands
Use check_quota function for invalid commands instead of manual quotactl
syscall invocations. In this way successful injection can also be checked
for these commands in the next commit.
* tests/quotactl.c (main): Add invalid_cmd_str and invalid_id_str local
variables, remove unused variable rc, use check_quota instead of manual
quotactl syscall invocations.
* tests/quotaclt.c (main): Add invalid_cmd_str variable, remove unused
variable rc, use check_quota instead of manual quotactl syscall invocations.
Zhibin Li [Thu, 12 Jul 2018 09:32:11 +0000 (17:32 +0800)]
tests: enhance test coverage of quotactl syscall parser
* tests/quotactl.c (print_dqfmt): New function.
(main): Use print_dqfmt to print flags of quota format, improve existing
tests and add more tests for Q_GETFMT.
* tests/quotactl-xfs.c: Improve existing tests and add more tests for
Q_XGETQSTAT and Q_XGETQSTATV.
Zhibin Li [Thu, 12 Jul 2018 06:49:42 +0000 (14:49 +0800)]
tests: extend coverage for other fcntl flags
* tests/fcntl-common.c: Include <assert.h>.
(struct fcntl_cmd_check): New structure.
(print_retval_flags, test_other_set_cmd, test_other_get_cmd,
print_flags_getfd, print_flags_getsig, print_flags_getlease,
test_fcntl_others): New functions.
(main): Use test_fcntl_others.
* xlat/bpf_attach_type.in (BPF_CGROUP_UDP4_SENDMSG,
BPF_CGROUP_UDP6_SENDMSG): New constant, introduced by Linux commit
v4.18-rc1~114^2~9^2~20^2~4.
(BPF_LIRC_MODE2): New constant, introduced by Linux commit
v4.18-rc1~114^2~9^2~15^2~1.
* xlat/bpf_commands.in (BPF_BTF_LOAD): New constant, introduced by Linux
commit v4.18-rc1~114^2~417^2~1^2~5.
(BPF_BTF_GET_FD_BY_ID): New constant, introduced by Linux commit
v4.18-rc1~114^2~223^2~21^2~4.
(BPF_TASK_FD_QUERY): New constant, introduced by Linux commit
v4.18-rc1~114^2~148^2~1^2~5.
* xlat/bpf_map_types.in (BPF_MAP_TYPE_XSKMAP): New constant, introduced
by Linux commit v4.18-rc1~114^2~304^2~4^2~8.
(BPF_MAP_TYPE_SOCKHASH): New constant, introduced by Linux commit
v4.18-rc1~114^2~223^2~5^2~2.
* xlat/bpf_prog_types.in (BPF_PROG_TYPE_LWT_SEG6LOCAL): New constant,
introduced by Linux commit v4.18-rc1~114^2~148^2~2^2~1.
(BPF_PROG_TYPE_LIRC_MODE2): New constant, introduced by Linux commit
v4.18-rc1~114^2~9^2~15^2~1.
* tests/bpf.c: Update expected output.
Wire up io_pgetevents and rseq on hppa, microblaze, mips, powerpc, and s390
* linux/hppa/syscallent.h ([350]): Wire up io_pgetevents syscall
introduced by Linux commit v4.18-rc3~7^2~2.
* linux/microblaze/syscallent.h ([399], [400]): Wire up io_pgetevents
and rseq syscalls introduced by Linux commit v4.18-rc3~33^2~2.
* linux/mips/syscallent-n32.h ([6331], [6332]): Wire up rseq and
io_pgetevents syscalls introduced by Linux commits v4.18-rc2~11^2~2
and v4.18-rc2~11^2, respectively.
* linux/mips/syscallent-n64.h ([5327], [5328]): Likewise.
* linux/mips/syscallent-o32.h ([4367], [4368]): Likewise.
* linux/powerpc/syscallent.h ([388]): Wire up io_pgetevents syscall
introduced by Linux commit v4.18-rc3~14^2~2.
* linux/powerpc64/syscallent.h ([388]): Likewise.
* linux/s390/syscallent.h ([382], [383]): Wire up io_pgetevents
and rseq syscalls introduced by Linux commits v4.18-rc4~18^2~1
and v4.18-rc4~18^2, respectively.
* linux/s390x/syscallent.h([382], [383]): Likewise.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
* maint/gen-tag-message.sh: Align the wording of the changes section
header with the one used in NEWS, add a reference to CREDITS in
"Contributors" section.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
* filter_qualify.c (qualify_kvm): Call error_msg_and_die instead of
error_msg in case of invalid -e kvm= argument.
* tests/options-syntax.test: Check it.
* tests/ioctl_kvm_run_common.c: Rename from ioctl_kvm_run.c.
(run_kvm): Parametrize printing of KVM_RUN ioctl with print_KVM_RUN
invocation.
(main): Invoke optional KVM_NO_CPUID_CALLBACK macro when the old kernel
behavior is detected.
* tests/Makefile.am (EXTRA_DIST): Add ioctl_kvm_run_common.c.
* tests/ioctl_kvm_run.c: New file, a wrapper around
ioctl_kvm_run_common.c.
* ioctl_kvm_run_auxstr_vcpu.c: Likewise.
* tests/gen_tests.in (ioctl_kvm_run_auxstr_vcpu): New test.
* tests/pure_executables.list: Add ioctl_kvm_run_auxstr_vcpu.
* tests/.gitignore: Likewise.
Signed-off-by: Masatake YAMATO <yamato@redhat.com> Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
kvm: attach the exit reason of vcpu as auxstr to KVM_RUN output
In KVM, a virtual machine implementation like Qemu can access a vcpu
via ioctl. KVM_RUN is an ioctl command to enter vcpu. The command
returns control for various reasons: needs of device emulation or
consuming time slices are the typical ones. The vmi takes a different
action for the reason.
We, strace users, want to know the reason to understand kvm. This
change prints the reason as auxstr if "-e kvm=vcpu" option is given,
and if strace runs on Linux 4.16.0 or higher, which includes commit e46b469278a59781f9b25ff608af84892963821b, "kvm: embed vcpu id to dentry
of vcpu anon inode."
The way to get the reason is a bit complicated because the ioctl does
not return it to the userspace directly. Instead, the vmi and kvm
communicate via an area of the process virtual memory where the fd of
vcpu is mmap'ed. strace must peek the area to know the reason.
The change does three things: (1) recording the area for the given vcpu
when the target calls VCPU_CREATE to vcpu_info_list per tcb data field,
(2) verifying the data recorded in vcpu_info_list before doing (3), and
(3) decoding the exit reason field of the area.
The change is complicated because there is a case that strace
does not have a chance to do (1) if -p option is used.
In this case, vcpu_info data created in the step (2).
The area has more fields than "exit reason",
dumping them may be implemented in the future.
* defs.h (struct tcb) [HAVE_LINUX_KVM_H]: Add vcpu_info_list field.
[HAVE_LINUX_KVM_H]: (kvm_run_structure_decoder_init,
kvm_vcpu_info_free): New declarations.
* strace.c (usage): Add "kvm" as a new expression for -e option.
(droptcb): Call kvm_vcpu_info_free.
* filter_qualify.c (qualify_kvm): New function calling
kvm_run_structure_decoder_init to enable for attaching the exit
reason to auxstr.
(qual_options): Add "kvm" as an entry.
* xlat/kvm_exit_reason.in: New file.
* kvm.c: Include xmalloc.h and mmap_cache.h.
(dump_kvm_run_structure): New static variable.
(kvm_run_structure_decoder_init): New function.
(vcpu_info): New struct definition representing the 3-tuple: vcpu file
descriptor, id of the vcpu, and mmap'ed entry.
(vcpu_find, vcpu_alloc, vcpu_register, vcpu_getinfo,
kvm_vcpu_info_free): New functions to access tcb's vcpu_info_list
field and vcpu_info data type.
(is_map_for_file, map_len): New helper functions.
(kvm_ioclt_run_attach_auxstr, kvm_ioctl_decode_run): New functions
decoding vcpu exit reason and attaching the decoded data to auxstr
field of tcb.
(kvm_ioctl_create_vcpu): Call vcpu_register to make an entry mapping
a file descriptor and the vcpu id associated with the fd.
(kvm_ioctl): Call kvm_ioctl_decode_run.
Signed-off-by: Masatake YAMATO <yamato@redhat.com> Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
* mmap_cache.c (mmap_cache_search_custom): New function.
* mmap_cache.h (mmap_cache_search_fn): New type.
(mmap_cache_search_custom): New function prototype.
Pierre Marsais [Thu, 28 Jun 2018 02:37:44 +0000 (03:37 +0100)]
tests: check decoding of kvm ioctls related to struct kvm_cpuid2
* tests/ioctl_kvm_run.c [!HAVE_STRUCT_KVM_CPUID2]: Skip the test.
(KVM_MAX_CPUID_ENTRIES): New macro.
(print_cpuid_ioctl): New function.
(main): Add tests for cpuid related ioctl.
Signed-off-by: Pierre Marsais <pierre.marsais@lse.epita.fr>
Reduce false positives by wrapping network-related operations into a loop.
* ci/install-dependencies.sh (retry_if_failed): New function.
Use it to invoke apt-get update, apt-get install, git clone,
add-apt-repository, and pip install.
Dmitry V. Levin [Mon, 18 Jun 2018 22:59:38 +0000 (22:59 +0000)]
Fix build on systems where libc relies on <linux/signal.h>
Commit v4.23~10 has broken the build with Android's libc because it
relies on the UAPI headers (such as <linux/signal.h>) to define various
types, rather than copying them into the libc headers (such as <signal.h>)
like most other Linux C libraries.
* linux/linux/signal.h: Rename to linux/signal.h.in.
* Makefile.am (strace_SOURCES): Remove linux/linux/signal.h.
(EXTRA_DIST): Add linux/signal.h.in.
(DISTCLEANFILES): Add linux/linux/signal.h.
* configure.ac (AC_CACHE_CHECK): Check whether <linux/signal.h>
can be used along with <signal.h>, install linux/linux/signal.h
workaround if they cannot be used together.
* xlat/aio_cmds.in: New file.
* aio.c (tprint_lio_opcode): Change array of structs "cmds" to array of
enums "subs", use printxval_indexn_ex for printing cmd, return sub type
using subs array.
* xlat/sock_tcp_options.in (TCP_ZEROCOPY_RECEIVE): New constant,
introduced by Linux commit v4.18-rc1~114^2~351^2~1.
(TCP_INQ): New constant, introduced by Linux commit v4.18-rc1~114^2~330.
* print_aio_sigset.c: New file.
* Makefile.am (strace_SOURCES): Add it.
* aio.c (print_io_getevents): Rename from SYS_FUNC(io_getevents), add
has_usig flag, print user sigset on exiting if has_usig is set to true.
(SYS_FUNC(io_getevents), SYS_FUNC(io_pgetevents)): New functions,
wrappers for print_io_getevents.
* linux/32/syscallent.h ([292]): Wire up io_pgetevents.
* linux/64/syscallent.h ([292]): Likewise.
* linux/i386/syscallent.h ([385]): Likewise.
* linux/x32/syscallent.h ([333]): Likewise.
* linux/x86_64/syscallent.h ([333]): Likewise.
* NEWS: Mention this improvement.
Co-Authored-by: Dmitry V. Levin <ldv@altlinux.org>
Dmitry V. Levin [Tue, 12 Jun 2018 12:27:22 +0000 (12:27 +0000)]
travis: re-enable testing with the latest mainline kernel headers
As <linux/signal.h> is out of the game now, the Linux commit v4.17-rc3-7-g7a074e9 should not cause harm anymore, so let's enable
testing with the latest mainline kernel headers again.
* .travis.yml (matrix): Do not specify KBRANCH for KHEADERS=torvalds/linux.
projects / strace / log