freebsd-dev/usr.bin/truss/syscalls.c

2401 lines
73 KiB
C
Raw Normal View History

/*
* Copyright 1997 Sean Eric Fagan
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Sean Eric Fagan
* 4. Neither the name of the author may be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1997-12-06 05:23:12 +00:00
/*
* This file has routines used to print out system calls and their
* arguments.
*/
#include <sys/capsicum.h>
#include <sys/types.h>
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#include <sys/event.h>
#include <sys/ioccom.h>
#include <sys/mount.h>
#include <sys/ptrace.h>
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#include <sys/resource.h>
#include <sys/socket.h>
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#include <sys/stat.h>
#include <sys/un.h>
#include <sys/wait.h>
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#include <machine/sysarch.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <assert.h>
#include <ctype.h>
#include <err.h>
2004-03-23 09:04:06 +00:00
#include <fcntl.h>
#include <poll.h>
#include <sched.h>
#include <signal.h>
#include <stdbool.h>
1997-12-06 05:23:12 +00:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sysdecode.h>
1997-12-06 05:23:12 +00:00
#include <unistd.h>
#include <vis.h>
#include <contrib/cloudabi/cloudabi_types_common.h>
#include "truss.h"
#include "extern.h"
1997-12-06 05:23:12 +00:00
#include "syscall.h"
/*
* This should probably be in its own file, sorted alphabetically.
1997-12-06 05:23:12 +00:00
*/
static struct syscall decoded_syscalls[] = {
/* Native ABI */
{ .name = "__acl_aclcheck_fd", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_aclcheck_file", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_aclcheck_link", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_delete_fd", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Acltype, 1 } } },
{ .name = "__acl_delete_file", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Acltype, 1 } } },
{ .name = "__acl_delete_link", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Acltype, 1 } } },
{ .name = "__acl_get_fd", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_get_file", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_get_link", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_set_fd", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_set_file", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__acl_set_link", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Acltype, 1 }, { Ptr, 2 } } },
{ .name = "__cap_rights_get", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { CapRights | OUT, 2 } } },
{ .name = "__getcwd", .ret_type = 1, .nargs = 2,
.args = { { Name | OUT, 0 }, { Int, 1 } } },
{ .name = "_umtx_op", .ret_type = 1, .nargs = 5,
.args = { { Ptr, 0 }, { Umtxop, 1 }, { LongHex, 2 }, { Ptr, 3 },
{ Ptr, 4 } } },
{ .name = "accept", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Sockaddr | OUT, 1 }, { Ptr | OUT, 2 } } },
{ .name = "access", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Accessmode, 1 } } },
{ .name = "bind", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Sockaddr | IN, 1 }, { Socklent, 2 } } },
{ .name = "bindat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Int, 1 }, { Sockaddr | IN, 2 },
{ Int, 3 } } },
{ .name = "break", .ret_type = 1, .nargs = 1,
.args = { { Ptr, 0 } } },
{ .name = "cap_fcntls_get", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CapFcntlRights | OUT, 1 } } },
{ .name = "cap_fcntls_limit", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CapFcntlRights, 1 } } },
{ .name = "cap_getmode", .ret_type = 1, .nargs = 1,
.args = { { PUInt | OUT, 0 } } },
{ .name = "cap_rights_limit", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CapRights, 1 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "chdir", .ret_type = 1, .nargs = 1,
.args = { { Name, 0 } } },
{ .name = "chflags", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { FileFlags, 1 } } },
{ .name = "chflagsat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { FileFlags, 2 },
{ Atflags, 3 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "chmod", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Octal, 1 } } },
{ .name = "chown", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Int, 1 }, { Int, 2 } } },
{ .name = "chroot", .ret_type = 1, .nargs = 1,
.args = { { Name, 0 } } },
{ .name = "clock_gettime", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Timespec | OUT, 1 } } },
{ .name = "close", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "connect", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Sockaddr | IN, 1 }, { Socklent, 2 } } },
{ .name = "connectat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Int, 1 }, { Sockaddr | IN, 2 },
{ Int, 3 } } },
{ .name = "dup", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "dup2", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Int, 1 } } },
{ .name = "eaccess", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Accessmode, 1 } } },
{ .name = "execve", .ret_type = 1, .nargs = 3,
.args = { { Name | IN, 0 }, { ExecArgs | IN, 1 },
{ ExecEnv | IN, 2 } } },
{ .name = "exit", .ret_type = 0, .nargs = 1,
.args = { { Hex, 0 } } },
{ .name = "extattr_delete_fd", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Extattrnamespace, 1 }, { Name, 2 } } },
{ .name = "extattr_delete_file", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 } } },
{ .name = "extattr_delete_link", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 } } },
{ .name = "extattr_get_fd", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | OUT, 3 }, { Sizet, 4 } } },
{ .name = "extattr_get_file", .ret_type = 1, .nargs = 5,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | OUT, 3 }, { Sizet, 4 } } },
{ .name = "extattr_get_link", .ret_type = 1, .nargs = 5,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | OUT, 3 }, { Sizet, 4 } } },
{ .name = "extattr_list_fd", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { Extattrnamespace, 1 }, { BinString | OUT, 2 },
{ Sizet, 3 } } },
{ .name = "extattr_list_file", .ret_type = 1, .nargs = 4,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { BinString | OUT, 2 },
{ Sizet, 3 } } },
{ .name = "extattr_list_link", .ret_type = 1, .nargs = 4,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { BinString | OUT, 2 },
{ Sizet, 3 } } },
{ .name = "extattr_set_fd", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | IN, 3 }, { Sizet, 4 } } },
{ .name = "extattr_set_file", .ret_type = 1, .nargs = 5,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | IN, 3 }, { Sizet, 4 } } },
{ .name = "extattr_set_link", .ret_type = 1, .nargs = 5,
.args = { { Name, 0 }, { Extattrnamespace, 1 }, { Name, 2 },
{ BinString | IN, 3 }, { Sizet, 4 } } },
{ .name = "extattrctl", .ret_type = 1, .nargs = 5,
.args = { { Name, 0 }, { Hex, 1 }, { Name, 2 },
{ Extattrnamespace, 3 }, { Name, 4 } } },
{ .name = "faccessat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { Accessmode, 2 },
{ Atflags, 3 } } },
{ .name = "fchflags", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { FileFlags, 1 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "fchmod", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Octal, 1 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "fchmodat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name, 1 }, { Octal, 2 }, { Atflags, 3 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "fchown", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { Int, 2 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "fchownat", .ret_type = 1, .nargs = 5,
.args = { { Atfd, 0 }, { Name, 1 }, { Int, 2 }, { Int, 3 },
{ Atflags, 4 } } },
{ .name = "fcntl", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Fcntl, 1 }, { Fcntlflag, 2 } } },
2017-03-18 18:26:56 +00:00
{ .name = "flock", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Flockop, 1 } } },
{ .name = "fstat", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Stat | OUT, 1 } } },
{ .name = "fstatat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { Stat | OUT, 2 },
{ Atflags, 3 } } },
{ .name = "fstatfs", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { StatFs | OUT, 1 } } },
{ .name = "ftruncate", .ret_type = 1, .nargs = 2,
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
.args = { { Int | IN, 0 }, { QuadHex | IN, 1 } } },
{ .name = "futimens", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Timespec2 | IN, 1 } } },
{ .name = "futimes", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Timeval2 | IN, 1 } } },
{ .name = "futimesat", .ret_type = 1, .nargs = 3,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { Timeval2 | IN, 2 } } },
{ .name = "getdirentries", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | OUT, 1 }, { Int, 2 },
{ PQuadHex | OUT, 3 } } },
{ .name = "getfsstat", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Long, 1 }, { Getfsstatmode, 2 } } },
{ .name = "getitimer", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Itimerval | OUT, 2 } } },
{ .name = "getpeername", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Sockaddr | OUT, 1 }, { Ptr | OUT, 2 } } },
{ .name = "getpgid", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "getpriority", .ret_type = 1, .nargs = 2,
.args = { { Priowhich, 0 }, { Int, 1 } } },
{ .name = "getrlimit", .ret_type = 1, .nargs = 2,
.args = { { Resource, 0 }, { Rlimit | OUT, 1 } } },
{ .name = "getrusage", .ret_type = 1, .nargs = 2,
.args = { { RusageWho, 0 }, { Rusage | OUT, 1 } } },
{ .name = "getsid", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "getsockname", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Sockaddr | OUT, 1 }, { Ptr | OUT, 2 } } },
2017-05-03 12:18:09 +00:00
{ .name = "getsockopt", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { Sockoptlevel, 1 }, { Sockoptname, 2 },
{ Ptr | OUT, 3 }, { Ptr | OUT, 4 } } },
{ .name = "gettimeofday", .ret_type = 1, .nargs = 2,
.args = { { Timeval | OUT, 0 }, { Ptr, 1 } } },
{ .name = "ioctl", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Ioctl, 1 }, { Ptr, 2 } } },
{ .name = "kevent", .ret_type = 1, .nargs = 6,
.args = { { Int, 0 }, { Kevent, 1 }, { Int, 2 }, { Kevent | OUT, 3 },
{ Int, 4 }, { Timespec, 5 } } },
{ .name = "kill", .ret_type = 1, .nargs = 2,
.args = { { Int | IN, 0 }, { Signal | IN, 1 } } },
{ .name = "kldfind", .ret_type = 1, .nargs = 1,
.args = { { Name | IN, 0 } } },
{ .name = "kldfirstmod", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "kldload", .ret_type = 1, .nargs = 1,
.args = { { Name | IN, 0 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "kldnext", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "kldstat", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Ptr, 1 } } },
{ .name = "kldsym", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Kldsymcmd, 1 }, { Ptr, 2 } } },
{ .name = "kldunload", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "kldunloadf", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Kldunloadflags, 1 } } },
{ .name = "kse_release", .ret_type = 0, .nargs = 1,
.args = { { Timespec, 0 } } },
{ .name = "lchflags", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { FileFlags, 1 } } },
{ .name = "lchmod", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Octal, 1 } } },
{ .name = "lchown", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Int, 1 }, { Int, 2 } } },
{ .name = "link", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Name, 1 } } },
{ .name = "linkat", .ret_type = 1, .nargs = 5,
.args = { { Atfd, 0 }, { Name, 1 }, { Atfd, 2 }, { Name, 3 },
{ Atflags, 4 } } },
2017-05-03 09:09:34 +00:00
{ .name = "listen", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Int, 1 } } },
{ .name = "lseek", .ret_type = 2, .nargs = 3,
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
.args = { { Int, 0 }, { QuadHex, 1 }, { Whence, 2 } } },
{ .name = "lstat", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Stat | OUT, 1 } } },
{ .name = "lutimes", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Timeval2 | IN, 1 } } },
2017-03-19 00:31:21 +00:00
{ .name = "madvise", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Sizet, 1 }, { Madvice, 2 } } },
2017-06-08 04:45:13 +00:00
{ .name = "minherit", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Sizet, 1 }, { Minherit, 2 } } },
{ .name = "mkdir", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Octal, 1 } } },
{ .name = "mkdirat", .ret_type = 1, .nargs = 3,
.args = { { Atfd, 0 }, { Name, 1 }, { Octal, 2 } } },
{ .name = "mkfifo", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Octal, 1 } } },
{ .name = "mkfifoat", .ret_type = 1, .nargs = 3,
.args = { { Atfd, 0 }, { Name, 1 }, { Octal, 2 } } },
{ .name = "mknod", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Octal, 1 }, { Int, 2 } } },
{ .name = "mknodat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name, 1 }, { Octal, 2 }, { Int, 3 } } },
{ .name = "mlock", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Sizet, 1 } } },
{ .name = "mlockall", .ret_type = 1, .nargs = 1,
.args = { { Mlockall, 0 } } },
{ .name = "mmap", .ret_type = 1, .nargs = 6,
.args = { { Ptr, 0 }, { Sizet, 1 }, { Mprot, 2 }, { Mmapflags, 3 },
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
{ Int, 4 }, { QuadHex, 5 } } },
2015-10-05 18:11:30 +00:00
{ .name = "modfind", .ret_type = 1, .nargs = 1,
.args = { { Name | IN, 0 } } },
{ .name = "mount", .ret_type = 1, .nargs = 4,
.args = { { Name, 0 }, { Name, 1 }, { Mountflags, 2 }, { Ptr, 3 } } },
{ .name = "mprotect", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Sizet, 1 }, { Mprot, 2 } } },
2017-06-08 08:10:57 +00:00
{ .name = "msync", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Sizet, 1 }, { Msync, 2 } } },
{ .name = "munlock", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Sizet, 1 } } },
{ .name = "munmap", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Sizet, 1 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "nanosleep", .ret_type = 1, .nargs = 1,
.args = { { Timespec, 0 } } },
{ .name = "nmount", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { UInt, 1 }, { Mountflags, 2 } } },
{ .name = "open", .ret_type = 1, .nargs = 3,
.args = { { Name | IN, 0 }, { Open, 1 }, { Octal, 2 } } },
{ .name = "openat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { Open, 2 },
{ Octal, 3 } } },
{ .name = "pathconf", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Pathconf, 1 } } },
{ .name = "pipe", .ret_type = 1, .nargs = 1,
.args = { { PipeFds | OUT, 0 } } },
{ .name = "pipe2", .ret_type = 1, .nargs = 2,
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
.args = { { Ptr, 0 }, { Pipe2, 1 } } },
{ .name = "poll", .ret_type = 1, .nargs = 3,
.args = { { Pollfd, 0 }, { Int, 1 }, { Int, 2 } } },
{ .name = "posix_fadvise", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { QuadHex, 1 }, { QuadHex, 2 },
{ Fadvice, 3 } } },
{ .name = "posix_openpt", .ret_type = 1, .nargs = 1,
.args = { { Open, 0 } } },
{ .name = "pread", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | OUT, 1 }, { Sizet, 2 },
{ QuadHex, 3 } } },
{ .name = "procctl", .ret_type = 1, .nargs = 4,
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
.args = { { Idtype, 0 }, { Quad, 1 }, { Procctl, 2 }, { Ptr, 3 } } },
{ .name = "ptrace", .ret_type = 1, .nargs = 4,
.args = { { Ptraceop, 0 }, { Int, 1 }, { Ptr, 2 }, { Int, 3 } } },
{ .name = "pwrite", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | IN, 1 }, { Sizet, 2 },
{ QuadHex, 3 } } },
2017-06-10 00:53:00 +00:00
{ .name = "quotactl", .ret_type = 1, .nargs = 4,
.args = { { Name, 0 }, { Quotactlcmd, 1 }, { Int, 2 }, { Ptr, 3 } } },
{ .name = "read", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { BinString | OUT, 1 }, { Sizet, 2 } } },
{ .name = "readlink", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Readlinkres | OUT, 1 }, { Sizet, 2 } } },
{ .name = "readlinkat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name, 1 }, { Readlinkres | OUT, 2 },
{ Sizet, 3 } } },
{ .name = "reboot", .ret_type = 1, .nargs = 1,
.args = { { Reboothowto, 0 } } },
{ .name = "recvfrom", .ret_type = 1, .nargs = 6,
.args = { { Int, 0 }, { BinString | OUT, 1 }, { Sizet, 2 },
{ Msgflags, 3 }, { Sockaddr | OUT, 4 },
{ Ptr | OUT, 5 } } },
{ .name = "recvmsg", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Ptr, 1 }, { Msgflags, 2 } } },
{ .name = "rename", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Name, 1 } } },
{ .name = "renameat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name, 1 }, { Atfd, 2 }, { Name, 3 } } },
{ .name = "rfork", .ret_type = 1, .nargs = 1,
.args = { { Rforkflags, 0 } } },
{ .name = "rmdir", .ret_type = 1, .nargs = 1,
.args = { { Name, 0 } } },
{ .name = "rtprio", .ret_type = 1, .nargs = 3,
.args = { { Rtpriofunc, 0 }, { Int, 1 }, { Ptr, 2 } } },
{ .name = "rtprio_thread", .ret_type = 1, .nargs = 3,
.args = { { Rtpriofunc, 0 }, { Int, 1 }, { Ptr, 2 } } },
{ .name = "sched_get_priority_max", .ret_type = 1, .nargs = 1,
.args = { { Schedpolicy, 0 } } },
{ .name = "sched_get_priority_min", .ret_type = 1, .nargs = 1,
.args = { { Schedpolicy, 0 } } },
{ .name = "sched_getparam", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Schedparam | OUT, 1 } } },
{ .name = "sched_getscheduler", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "sched_rr_get_interval", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Timespec | OUT, 1 } } },
{ .name = "sched_setparam", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Schedparam, 1 } } },
{ .name = "sched_setscheduler", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Schedpolicy, 1 }, { Schedparam, 2 } } },
{ .name = "sctp_generic_recvmsg", .ret_type = 1, .nargs = 7,
.args = { { Int, 0 }, { Ptr | IN, 1 }, { Int, 2 },
{ Sockaddr | OUT, 3 }, { Ptr | OUT, 4 }, { Ptr | OUT, 5 },
{ Ptr | OUT, 6 } } },
{ .name = "sctp_generic_sendmsg", .ret_type = 1, .nargs = 7,
.args = { { Int, 0 }, { BinString | IN, 1 }, { Int, 2 },
{ Sockaddr | IN, 3 }, { Socklent, 4 }, { Ptr | IN, 5 },
{ Msgflags, 6 } } },
{ .name = "select", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { Fd_set, 1 }, { Fd_set, 2 }, { Fd_set, 3 },
{ Timeval, 4 } } },
{ .name = "sendmsg", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Ptr, 1 }, { Msgflags, 2 } } },
{ .name = "sendto", .ret_type = 1, .nargs = 6,
.args = { { Int, 0 }, { BinString | IN, 1 }, { Sizet, 2 },
{ Msgflags, 3 }, { Sockaddr | IN, 4 },
{ Socklent | IN, 5 } } },
{ .name = "setitimer", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Itimerval, 1 }, { Itimerval | OUT, 2 } } },
{ .name = "setpriority", .ret_type = 1, .nargs = 3,
.args = { { Priowhich, 0 }, { Int, 1 }, { Int, 2 } } },
{ .name = "setrlimit", .ret_type = 1, .nargs = 2,
.args = { { Resource, 0 }, { Rlimit | IN, 1 } } },
2017-05-03 12:18:09 +00:00
{ .name = "setsockopt", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { Sockoptlevel, 1 }, { Sockoptname, 2 },
{ Ptr | IN, 3 }, { Socklent, 4 } } },
{ .name = "shutdown", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Shutdown, 1 } } },
{ .name = "sigaction", .ret_type = 1, .nargs = 3,
.args = { { Signal, 0 }, { Sigaction | IN, 1 },
{ Sigaction | OUT, 2 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "sigpending", .ret_type = 1, .nargs = 1,
.args = { { Sigset | OUT, 0 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "sigprocmask", .ret_type = 1, .nargs = 3,
.args = { { Sigprocmask, 0 }, { Sigset, 1 }, { Sigset | OUT, 2 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "sigqueue", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Signal, 1 }, { LongHex, 2 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "sigreturn", .ret_type = 1, .nargs = 1,
.args = { { Ptr, 0 } } },
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
{ .name = "sigsuspend", .ret_type = 1, .nargs = 1,
.args = { { Sigset | IN, 0 } } },
{ .name = "sigtimedwait", .ret_type = 1, .nargs = 3,
.args = { { Sigset | IN, 0 }, { Ptr, 1 }, { Timespec | IN, 2 } } },
{ .name = "sigwait", .ret_type = 1, .nargs = 2,
.args = { { Sigset | IN, 0 }, { Ptr, 1 } } },
{ .name = "sigwaitinfo", .ret_type = 1, .nargs = 2,
.args = { { Sigset | IN, 0 }, { Ptr, 1 } } },
{ .name = "socket", .ret_type = 1, .nargs = 3,
2017-05-03 09:23:13 +00:00
.args = { { Sockdomain, 0 }, { Socktype, 1 }, { Sockprotocol, 2 } } },
{ .name = "stat", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Stat | OUT, 1 } } },
{ .name = "statfs", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { StatFs | OUT, 1 } } },
{ .name = "symlink", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Name, 1 } } },
{ .name = "symlinkat", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Atfd, 1 }, { Name, 2 } } },
{ .name = "sysarch", .ret_type = 1, .nargs = 2,
.args = { { Sysarch, 0 }, { Ptr, 1 } } },
{ .name = "thr_kill", .ret_type = 1, .nargs = 2,
.args = { { Long, 0 }, { Signal, 1 } } },
{ .name = "thr_self", .ret_type = 1, .nargs = 1,
.args = { { Ptr, 0 } } },
{ .name = "truncate", .ret_type = 1, .nargs = 2,
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
.args = { { Name | IN, 0 }, { QuadHex | IN, 1 } } },
#if 0
/* Does not exist */
{ .name = "umount", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Int, 2 } } },
#endif
{ .name = "unlink", .ret_type = 1, .nargs = 1,
.args = { { Name, 0 } } },
{ .name = "unlinkat", .ret_type = 1, .nargs = 3,
.args = { { Atfd, 0 }, { Name, 1 }, { Atflags, 2 } } },
{ .name = "unmount", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Mountflags, 1 } } },
{ .name = "utimensat", .ret_type = 1, .nargs = 4,
.args = { { Atfd, 0 }, { Name | IN, 1 }, { Timespec2 | IN, 2 },
{ Atflags, 3 } } },
{ .name = "utimes", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Timeval2 | IN, 1 } } },
{ .name = "utrace", .ret_type = 1, .nargs = 1,
.args = { { Utrace, 0 } } },
{ .name = "wait4", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { ExitStatus | OUT, 1 }, { Waitoptions, 2 },
{ Rusage | OUT, 3 } } },
{ .name = "wait6", .ret_type = 1, .nargs = 6,
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
.args = { { Idtype, 0 }, { Quad, 1 }, { ExitStatus | OUT, 2 },
{ Waitoptions, 3 }, { Rusage | OUT, 4 }, { Ptr, 5 } } },
{ .name = "write", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { BinString | IN, 1 }, { Sizet, 2 } } },
/* Linux ABI */
{ .name = "linux_access", .ret_type = 1, .nargs = 2,
.args = { { Name, 0 }, { Accessmode, 1 } } },
{ .name = "linux_execve", .ret_type = 1, .nargs = 3,
.args = { { Name | IN, 0 }, { ExecArgs | IN, 1 },
{ ExecEnv | IN, 2 } } },
{ .name = "linux_lseek", .ret_type = 2, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { Whence, 2 } } },
{ .name = "linux_mkdir", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Int, 1 } } },
{ .name = "linux_newfstat", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Ptr | OUT, 1 } } },
{ .name = "linux_newstat", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Ptr | OUT, 1 } } },
{ .name = "linux_open", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Hex, 1 }, { Octal, 2 } } },
{ .name = "linux_readlink", .ret_type = 1, .nargs = 3,
.args = { { Name, 0 }, { Name | OUT, 1 }, { Sizet, 2 } } },
{ .name = "linux_socketcall", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { LinuxSockArgs, 1 } } },
{ .name = "linux_stat64", .ret_type = 1, .nargs = 2,
.args = { { Name | IN, 0 }, { Ptr | OUT, 1 } } },
/* CloudABI system calls. */
{ .name = "cloudabi_sys_clock_res_get", .ret_type = 1, .nargs = 1,
.args = { { CloudABIClockID, 0 } } },
{ .name = "cloudabi_sys_clock_time_get", .ret_type = 1, .nargs = 2,
.args = { { CloudABIClockID, 0 }, { CloudABITimestamp, 1 } } },
{ .name = "cloudabi_sys_condvar_signal", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { CloudABIMFlags, 1 }, { UInt, 2 } } },
{ .name = "cloudabi_sys_fd_close", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "cloudabi_sys_fd_create1", .ret_type = 1, .nargs = 1,
.args = { { CloudABIFileType, 0 } } },
{ .name = "cloudabi_sys_fd_create2", .ret_type = 1, .nargs = 2,
.args = { { CloudABIFileType, 0 }, { PipeFds | OUT, 0 } } },
{ .name = "cloudabi_sys_fd_datasync", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "cloudabi_sys_fd_dup", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "cloudabi_sys_fd_replace", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_fd_seek", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { CloudABIWhence, 2 } } },
{ .name = "cloudabi_sys_fd_stat_get", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CloudABIFDStat | OUT, 1 } } },
{ .name = "cloudabi_sys_fd_stat_put", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { CloudABIFDStat | IN, 1 },
{ ClouduABIFDSFlags, 2 } } },
{ .name = "cloudabi_sys_fd_sync", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "cloudabi_sys_file_advise", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { Int, 1 }, { Int, 2 },
{ CloudABIAdvice, 3 } } },
{ .name = "cloudabi_sys_file_allocate", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { Int, 2 } } },
{ .name = "cloudabi_sys_file_create", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { BinString | IN, 1 },
{ CloudABIFileType, 3 } } },
{ .name = "cloudabi_sys_file_link", .ret_type = 1, .nargs = 4,
.args = { { CloudABILookup, 0 }, { BinString | IN, 1 },
{ Int, 3 }, { BinString | IN, 4 } } },
{ .name = "cloudabi_sys_file_open", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | IN, 1 },
{ CloudABIOFlags, 3 }, { CloudABIFDStat | IN, 4 } } },
{ .name = "cloudabi_sys_file_readdir", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | OUT, 1 }, { Int, 2 },
{ Int, 3 } } },
{ .name = "cloudabi_sys_file_readlink", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | IN, 1 },
{ BinString | OUT, 3 }, { Int, 4 } } },
{ .name = "cloudabi_sys_file_rename", .ret_type = 1, .nargs = 4,
.args = { { Int, 0 }, { BinString | IN, 1 },
{ Int, 3 }, { BinString | IN, 4 } } },
{ .name = "cloudabi_sys_file_stat_fget", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CloudABIFileStat | OUT, 1 } } },
{ .name = "cloudabi_sys_file_stat_fput", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { CloudABIFileStat | IN, 1 },
{ CloudABIFSFlags, 2 } } },
{ .name = "cloudabi_sys_file_stat_get", .ret_type = 1, .nargs = 3,
.args = { { CloudABILookup, 0 }, { BinString | IN, 1 },
{ CloudABIFileStat | OUT, 3 } } },
{ .name = "cloudabi_sys_file_stat_put", .ret_type = 1, .nargs = 4,
.args = { { CloudABILookup, 0 }, { BinString | IN, 1 },
{ CloudABIFileStat | IN, 3 }, { CloudABIFSFlags, 4 } } },
{ .name = "cloudabi_sys_file_symlink", .ret_type = 1, .nargs = 3,
.args = { { BinString | IN, 0 },
{ Int, 2 }, { BinString | IN, 3 } } },
{ .name = "cloudabi_sys_file_unlink", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { BinString | IN, 1 },
{ CloudABIULFlags, 3 } } },
{ .name = "cloudabi_sys_lock_unlock", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { CloudABIMFlags, 1 } } },
{ .name = "cloudabi_sys_mem_advise", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Int, 1 }, { CloudABIAdvice, 2 } } },
{ .name = "cloudabi_sys_mem_lock", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_mem_map", .ret_type = 1, .nargs = 6,
.args = { { Ptr, 0 }, { Int, 1 }, { CloudABIMProt, 2 },
{ CloudABIMFlags, 3 }, { Int, 4 }, { Int, 5 } } },
{ .name = "cloudabi_sys_mem_protect", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Int, 1 }, { CloudABIMProt, 2 } } },
{ .name = "cloudabi_sys_mem_sync", .ret_type = 1, .nargs = 3,
.args = { { Ptr, 0 }, { Int, 1 }, { CloudABIMSFlags, 2 } } },
{ .name = "cloudabi_sys_mem_unlock", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_mem_unmap", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_proc_exec", .ret_type = 1, .nargs = 5,
.args = { { Int, 0 }, { BinString | IN, 1 }, { Int, 2 },
{ IntArray, 3 }, { Int, 4 } } },
{ .name = "cloudabi_sys_proc_exit", .ret_type = 1, .nargs = 1,
.args = { { Int, 0 } } },
{ .name = "cloudabi_sys_proc_fork", .ret_type = 1, .nargs = 0 },
{ .name = "cloudabi_sys_proc_raise", .ret_type = 1, .nargs = 1,
.args = { { CloudABISignal, 0 } } },
{ .name = "cloudabi_sys_random_get", .ret_type = 1, .nargs = 2,
.args = { { BinString | OUT, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_sock_accept", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CloudABISockStat | OUT, 1 } } },
{ .name = "cloudabi_sys_sock_bind", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { BinString | IN, 2 } } },
{ .name = "cloudabi_sys_sock_connect", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { Int, 1 }, { BinString | IN, 2 } } },
{ .name = "cloudabi_sys_sock_listen", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { Int, 1 } } },
{ .name = "cloudabi_sys_sock_shutdown", .ret_type = 1, .nargs = 2,
.args = { { Int, 0 }, { CloudABISDFlags, 1 } } },
{ .name = "cloudabi_sys_sock_stat_get", .ret_type = 1, .nargs = 3,
.args = { { Int, 0 }, { CloudABISockStat | OUT, 1 },
{ CloudABISSFlags, 2 } } },
{ .name = "cloudabi_sys_thread_exit", .ret_type = 1, .nargs = 2,
.args = { { Ptr, 0 }, { CloudABIMFlags, 1 } } },
{ .name = "cloudabi_sys_thread_yield", .ret_type = 1, .nargs = 0 },
{ .name = 0 },
1997-12-06 05:23:12 +00:00
};
static STAILQ_HEAD(, syscall) syscalls;
1997-12-06 05:23:12 +00:00
/* Xlat idea taken from strace */
struct xlat {
int val;
const char *str;
};
#define X(a) { a, #a },
#define XEND { 0, NULL }
static struct xlat kevent_filters[] = {
X(EVFILT_READ) X(EVFILT_WRITE) X(EVFILT_AIO) X(EVFILT_VNODE)
X(EVFILT_PROC) X(EVFILT_SIGNAL) X(EVFILT_TIMER)
X(EVFILT_PROCDESC) X(EVFILT_FS) X(EVFILT_LIO) X(EVFILT_USER)
X(EVFILT_SENDFILE) XEND
};
static struct xlat kevent_flags[] = {
X(EV_ADD) X(EV_DELETE) X(EV_ENABLE) X(EV_DISABLE) X(EV_ONESHOT)
X(EV_CLEAR) X(EV_RECEIPT) X(EV_DISPATCH) X(EV_FORCEONESHOT)
X(EV_DROP) X(EV_FLAG1) X(EV_ERROR) X(EV_EOF) XEND
};
static struct xlat kevent_user_ffctrl[] = {
X(NOTE_FFNOP) X(NOTE_FFAND) X(NOTE_FFOR) X(NOTE_FFCOPY)
XEND
};
static struct xlat kevent_rdwr_fflags[] = {
X(NOTE_LOWAT) X(NOTE_FILE_POLL) XEND
};
static struct xlat kevent_vnode_fflags[] = {
X(NOTE_DELETE) X(NOTE_WRITE) X(NOTE_EXTEND) X(NOTE_ATTRIB)
X(NOTE_LINK) X(NOTE_RENAME) X(NOTE_REVOKE) XEND
};
static struct xlat kevent_proc_fflags[] = {
X(NOTE_EXIT) X(NOTE_FORK) X(NOTE_EXEC) X(NOTE_TRACK) X(NOTE_TRACKERR)
X(NOTE_CHILD) XEND
};
static struct xlat kevent_timer_fflags[] = {
X(NOTE_SECONDS) X(NOTE_MSECONDS) X(NOTE_USECONDS) X(NOTE_NSECONDS)
XEND
};
static struct xlat poll_flags[] = {
X(POLLSTANDARD) X(POLLIN) X(POLLPRI) X(POLLOUT) X(POLLERR)
X(POLLHUP) X(POLLNVAL) X(POLLRDNORM) X(POLLRDBAND)
X(POLLWRBAND) X(POLLINIGNEOF) XEND
};
static struct xlat sigaction_flags[] = {
X(SA_ONSTACK) X(SA_RESTART) X(SA_RESETHAND) X(SA_NOCLDSTOP)
X(SA_NODEFER) X(SA_NOCLDWAIT) X(SA_SIGINFO) XEND
};
static struct xlat pathconf_arg[] = {
X(_PC_LINK_MAX) X(_PC_MAX_CANON) X(_PC_MAX_INPUT)
X(_PC_NAME_MAX) X(_PC_PATH_MAX) X(_PC_PIPE_BUF)
X(_PC_CHOWN_RESTRICTED) X(_PC_NO_TRUNC) X(_PC_VDISABLE)
X(_PC_ASYNC_IO) X(_PC_PRIO_IO) X(_PC_SYNC_IO)
X(_PC_ALLOC_SIZE_MIN) X(_PC_FILESIZEBITS)
X(_PC_REC_INCR_XFER_SIZE) X(_PC_REC_MAX_XFER_SIZE)
X(_PC_REC_MIN_XFER_SIZE) X(_PC_REC_XFER_ALIGN)
X(_PC_SYMLINK_MAX) X(_PC_ACL_EXTENDED) X(_PC_ACL_PATH_MAX)
X(_PC_CAP_PRESENT) X(_PC_INF_PRESENT) X(_PC_MAC_PRESENT)
X(_PC_ACL_NFS4) X(_PC_MIN_HOLE_SIZE) XEND
};
static struct xlat at_flags[] = {
X(AT_EACCESS) X(AT_SYMLINK_NOFOLLOW) X(AT_SYMLINK_FOLLOW)
X(AT_REMOVEDIR) XEND
};
static struct xlat sysarch_ops[] = {
#if defined(__i386__) || defined(__amd64__)
X(I386_GET_LDT) X(I386_SET_LDT) X(I386_GET_IOPERM) X(I386_SET_IOPERM)
X(I386_VM86) X(I386_GET_FSBASE) X(I386_SET_FSBASE) X(I386_GET_GSBASE)
X(I386_SET_GSBASE) X(I386_GET_XFPUSTATE) X(AMD64_GET_FSBASE)
X(AMD64_SET_FSBASE) X(AMD64_GET_GSBASE) X(AMD64_SET_GSBASE)
X(AMD64_GET_XFPUSTATE)
#endif
XEND
};
static struct xlat linux_socketcall_ops[] = {
X(LINUX_SOCKET) X(LINUX_BIND) X(LINUX_CONNECT) X(LINUX_LISTEN)
X(LINUX_ACCEPT) X(LINUX_GETSOCKNAME) X(LINUX_GETPEERNAME)
X(LINUX_SOCKETPAIR) X(LINUX_SEND) X(LINUX_RECV) X(LINUX_SENDTO)
X(LINUX_RECVFROM) X(LINUX_SHUTDOWN) X(LINUX_SETSOCKOPT)
X(LINUX_GETSOCKOPT) X(LINUX_SENDMSG) X(LINUX_RECVMSG)
XEND
};
#undef X
#define X(a) { CLOUDABI_##a, #a },
static struct xlat cloudabi_advice[] = {
X(ADVICE_DONTNEED) X(ADVICE_NOREUSE) X(ADVICE_NORMAL)
X(ADVICE_RANDOM) X(ADVICE_SEQUENTIAL) X(ADVICE_WILLNEED)
XEND
};
static struct xlat cloudabi_clockid[] = {
X(CLOCK_MONOTONIC) X(CLOCK_PROCESS_CPUTIME_ID)
X(CLOCK_REALTIME) X(CLOCK_THREAD_CPUTIME_ID)
XEND
};
static struct xlat cloudabi_errno[] = {
X(E2BIG) X(EACCES) X(EADDRINUSE) X(EADDRNOTAVAIL)
X(EAFNOSUPPORT) X(EAGAIN) X(EALREADY) X(EBADF) X(EBADMSG)
X(EBUSY) X(ECANCELED) X(ECHILD) X(ECONNABORTED) X(ECONNREFUSED)
X(ECONNRESET) X(EDEADLK) X(EDESTADDRREQ) X(EDOM) X(EDQUOT)
X(EEXIST) X(EFAULT) X(EFBIG) X(EHOSTUNREACH) X(EIDRM) X(EILSEQ)
X(EINPROGRESS) X(EINTR) X(EINVAL) X(EIO) X(EISCONN) X(EISDIR)
X(ELOOP) X(EMFILE) X(EMLINK) X(EMSGSIZE) X(EMULTIHOP)
X(ENAMETOOLONG) X(ENETDOWN) X(ENETRESET) X(ENETUNREACH)
X(ENFILE) X(ENOBUFS) X(ENODEV) X(ENOENT) X(ENOEXEC) X(ENOLCK)
X(ENOLINK) X(ENOMEM) X(ENOMSG) X(ENOPROTOOPT) X(ENOSPC)
X(ENOSYS) X(ENOTCONN) X(ENOTDIR) X(ENOTEMPTY) X(ENOTRECOVERABLE)
X(ENOTSOCK) X(ENOTSUP) X(ENOTTY) X(ENXIO) X(EOVERFLOW)
X(EOWNERDEAD) X(EPERM) X(EPIPE) X(EPROTO) X(EPROTONOSUPPORT)
X(EPROTOTYPE) X(ERANGE) X(EROFS) X(ESPIPE) X(ESRCH) X(ESTALE)
X(ETIMEDOUT) X(ETXTBSY) X(EXDEV) X(ENOTCAPABLE)
XEND
};
static struct xlat cloudabi_fdflags[] = {
X(FDFLAG_APPEND) X(FDFLAG_DSYNC) X(FDFLAG_NONBLOCK)
X(FDFLAG_RSYNC) X(FDFLAG_SYNC)
XEND
};
static struct xlat cloudabi_fdsflags[] = {
X(FDSTAT_FLAGS) X(FDSTAT_RIGHTS)
XEND
};
static struct xlat cloudabi_filetype[] = {
X(FILETYPE_UNKNOWN) X(FILETYPE_BLOCK_DEVICE)
X(FILETYPE_CHARACTER_DEVICE) X(FILETYPE_DIRECTORY)
X(FILETYPE_FIFO) X(FILETYPE_POLL) X(FILETYPE_PROCESS)
X(FILETYPE_REGULAR_FILE) X(FILETYPE_SHARED_MEMORY)
X(FILETYPE_SOCKET_DGRAM) X(FILETYPE_SOCKET_SEQPACKET)
X(FILETYPE_SOCKET_STREAM) X(FILETYPE_SYMBOLIC_LINK)
XEND
};
static struct xlat cloudabi_fsflags[] = {
X(FILESTAT_ATIM) X(FILESTAT_ATIM_NOW) X(FILESTAT_MTIM)
X(FILESTAT_MTIM_NOW) X(FILESTAT_SIZE)
XEND
};
static struct xlat cloudabi_mflags[] = {
X(MAP_ANON) X(MAP_FIXED) X(MAP_PRIVATE) X(MAP_SHARED)
XEND
};
static struct xlat cloudabi_mprot[] = {
X(PROT_EXEC) X(PROT_WRITE) X(PROT_READ)
XEND
};
static struct xlat cloudabi_msflags[] = {
X(MS_ASYNC) X(MS_INVALIDATE) X(MS_SYNC)
XEND
};
static struct xlat cloudabi_oflags[] = {
X(O_CREAT) X(O_DIRECTORY) X(O_EXCL) X(O_TRUNC)
XEND
};
static struct xlat cloudabi_sa_family[] = {
X(AF_UNSPEC) X(AF_INET) X(AF_INET6) X(AF_UNIX)
XEND
};
static struct xlat cloudabi_sdflags[] = {
X(SHUT_RD) X(SHUT_WR)
XEND
};
static struct xlat cloudabi_signal[] = {
X(SIGABRT) X(SIGALRM) X(SIGBUS) X(SIGCHLD) X(SIGCONT) X(SIGFPE)
X(SIGHUP) X(SIGILL) X(SIGINT) X(SIGKILL) X(SIGPIPE) X(SIGQUIT)
X(SIGSEGV) X(SIGSTOP) X(SIGSYS) X(SIGTERM) X(SIGTRAP) X(SIGTSTP)
X(SIGTTIN) X(SIGTTOU) X(SIGURG) X(SIGUSR1) X(SIGUSR2)
X(SIGVTALRM) X(SIGXCPU) X(SIGXFSZ)
XEND
};
static struct xlat cloudabi_ssflags[] = {
X(SOCKSTAT_CLEAR_ERROR)
XEND
};
static struct xlat cloudabi_ssstate[] = {
X(SOCKSTATE_ACCEPTCONN)
XEND
};
static struct xlat cloudabi_ulflags[] = {
X(UNLINK_REMOVEDIR)
XEND
};
static struct xlat cloudabi_whence[] = {
X(WHENCE_CUR) X(WHENCE_END) X(WHENCE_SET)
XEND
};
#undef X
#undef XEND
/*
* Searches an xlat array for a value, and returns it if found. Otherwise
* return a string representation.
*/
static const char *
lookup(struct xlat *xlat, int val, int base)
{
static char tmp[16];
for (; xlat->str != NULL; xlat++)
if (xlat->val == val)
return (xlat->str);
switch (base) {
case 8:
sprintf(tmp, "0%o", val);
break;
case 16:
sprintf(tmp, "0x%x", val);
break;
case 10:
sprintf(tmp, "%u", val);
break;
default:
errx(1,"Unknown lookup base");
break;
}
return (tmp);
}
static const char *
xlookup(struct xlat *xlat, int val)
{
return (lookup(xlat, val, 16));
}
2015-08-19 20:02:03 +00:00
/*
* Searches an xlat array containing bitfield values. Remaining bits
* set after removing the known ones are printed at the end:
* IN|0x400.
*/
static char *
xlookup_bits(struct xlat *xlat, int val)
{
int len, rem;
static char str[512];
len = 0;
rem = val;
for (; xlat->str != NULL; xlat++) {
if ((xlat->val & rem) == xlat->val) {
2015-08-19 20:02:03 +00:00
/*
* Don't print the "all-bits-zero" string unless all
* bits are really zero.
*/
if (xlat->val == 0 && val != 0)
continue;
len += sprintf(str + len, "%s|", xlat->str);
rem &= ~(xlat->val);
}
}
2015-08-19 20:02:03 +00:00
/*
* If we have leftover bits or didn't match anything, print
* the remainder.
*/
if (rem || len == 0)
len += sprintf(str + len, "0x%x", rem);
if (len && str[len - 1] == '|')
len--;
str[len] = 0;
return (str);
}
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
static void
print_integer_arg(const char *(*decoder)(int), FILE *fp, int value)
{
const char *str;
str = decoder(value);
if (str != NULL)
fputs(str, fp);
else
fprintf(fp, "%d", value);
}
static void
print_mask_arg(bool (*decoder)(FILE *, int, int *), FILE *fp, int value)
{
int rem;
if (!decoder(fp, value, &rem))
fprintf(fp, "0x%x", rem);
else if (rem != 0)
fprintf(fp, "|0x%x", rem);
}
static void
print_mask_arg32(bool (*decoder)(FILE *, uint32_t, uint32_t *), FILE *fp,
uint32_t value)
{
uint32_t rem;
if (!decoder(fp, value, &rem))
fprintf(fp, "0x%x", rem);
else if (rem != 0)
fprintf(fp, "|0x%x", rem);
}
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
#ifndef __LP64__
/*
* Add argument padding to subsequent system calls afater a Quad
* syscall arguments as needed. This used to be done by hand in the
* decoded_syscalls table which was ugly and error prone. It is
* simpler to do the fixup of offsets at initalization time than when
* decoding arguments.
*/
static void
quad_fixup(struct syscall *sc)
{
int offset, prev;
u_int i;
offset = 0;
prev = -1;
for (i = 0; i < sc->nargs; i++) {
/* This arg type is a dummy that doesn't use offset. */
if ((sc->args[i].type & ARG_MASK) == PipeFds)
continue;
assert(prev < sc->args[i].offset);
prev = sc->args[i].offset;
sc->args[i].offset += offset;
switch (sc->args[i].type & ARG_MASK) {
case Quad:
case QuadHex:
#ifdef __powerpc__
/*
* 64-bit arguments on 32-bit powerpc must be
* 64-bit aligned. If the current offset is
* not aligned, the calling convention inserts
* a 32-bit pad argument that should be skipped.
*/
if (sc->args[i].offset % 2 == 1) {
sc->args[i].offset++;
offset++;
}
#endif
offset++;
default:
break;
}
}
}
#endif
void
init_syscalls(void)
{
struct syscall *sc;
STAILQ_INIT(&syscalls);
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
for (sc = decoded_syscalls; sc->name != NULL; sc++) {
#ifndef __LP64__
quad_fixup(sc);
#endif
STAILQ_INSERT_HEAD(&syscalls, sc, entries);
Automate the handling of QUAD_ALIGN and QUAD_SLOTS. Previously, the offset in a system call description specified the array index of the start of a system call argument. For most system call arguments this was the same as the index of the argument in the function signature. 64-bit arguments (off_t and id_t values) passed on 32-bit platforms use two slots in the array however. This was handled by adding (QUAD_SLOTS - 1) to the slot indicies of any subsequent arguments after a 64-bit argument (though written as ("{ Quad, 1 }, { Int, 1 + QUAD_SLOTS }" rather than "{ Quad, 1 }, { Int, 2 + QUAD_SLOTS - 1 }"). If a system call contained multiple 64-bit arguments (such as posix_fadvise()), then additional arguments would need to use 'QUAD_SLOTS * 2' but remember to subtract 2 from the initial number, etc. In addition, 32-bit powerpc requires 64-bit arguments to be 64-bit aligned, so if the effective index in the array of a 64-bit argument is odd, it needs QUAD_ALIGN added to the current and any subsequent slots. However, if the effective index in the array of a 64-bit argument was even, QUAD_ALIGN was omitted. This approach was messy and error prone. This commit replaces it with automated pre-processing of the system call table to do fixups for 64-bit argument offsets. The offset in a system call description now indicates the index of an argument in the associated function call's signature. A fixup function is run against each decoded system call description during startup on 32-bit platforms. The fixup function maintains an 'offset' value which holds an offset to be added to each remaining system call argument's index. Initially offset is 0. When a 64-bit system call argument is encountered, the offset is first aligned to a 64-bit boundary (only on powerpc) and then incremented to account for the second argument slot used by the argument. This modified 'offset' is then applied to any remaining arguments. This approach does require a few things that were not previously required: 1) Each system call description must now list arguments in ascending order (existing ones all do) without using duplicate slots in the register array. A new assert() should catch any future descriptions which violate this rule. 2) A system call description is still permitted to omit arguments (though none currently do), but if the call accepts 64-bit arguments those cannot be omitted or incorrect results will be displated on 32-bit systems. Tested on: amd64 and i386
2017-03-15 23:08:11 +00:00
}
}
static struct syscall *
find_syscall(struct procabi *abi, u_int number)
{
struct extra_syscall *es;
if (number < nitems(abi->syscalls))
return (abi->syscalls[number]);
STAILQ_FOREACH(es, &abi->extra_syscalls, entries) {
if (es->number == number)
return (es->sc);
}
return (NULL);
}
static void
add_syscall(struct procabi *abi, u_int number, struct syscall *sc)
{
struct extra_syscall *es;
if (number < nitems(abi->syscalls)) {
assert(abi->syscalls[number] == NULL);
abi->syscalls[number] = sc;
} else {
es = malloc(sizeof(*es));
es->sc = sc;
es->number = number;
STAILQ_INSERT_TAIL(&abi->extra_syscalls, es, entries);
}
}
1997-12-06 05:23:12 +00:00
/*
* If/when the list gets big, it might be desirable to do it
* as a hash table or binary search.
*/
struct syscall *
get_syscall(struct threadinfo *t, u_int number, u_int nargs)
{
struct syscall *sc;
const char *name;
char *new_name;
u_int i;
sc = find_syscall(t->proc->abi, number);
if (sc != NULL)
return (sc);
name = sysdecode_syscallname(t->proc->abi->abi, number);
if (name == NULL) {
asprintf(&new_name, "#%d", number);
name = new_name;
} else
new_name = NULL;
STAILQ_FOREACH(sc, &syscalls, entries) {
if (strcmp(name, sc->name) == 0) {
add_syscall(t->proc->abi, number, sc);
free(new_name);
return (sc);
}
}
/* It is unknown. Add it into the list. */
#if DEBUG
fprintf(stderr, "unknown syscall %s -- setting args to %d\n", name,
nargs);
#endif
sc = calloc(1, sizeof(struct syscall));
sc->name = name;
if (new_name != NULL)
sc->unknown = true;
sc->ret_type = 1;
sc->nargs = nargs;
for (i = 0; i < nargs; i++) {
sc->args[i].offset = i;
/* Treat all unknown arguments as LongHex. */
sc->args[i].type = LongHex;
1997-12-06 05:23:12 +00:00
}
STAILQ_INSERT_HEAD(&syscalls, sc, entries);
add_syscall(t->proc->abi, number, sc);
return (sc);
1997-12-06 05:23:12 +00:00
}
/*
* Copy a fixed amount of bytes from the process.
*/
static int
get_struct(pid_t pid, void *offset, void *buf, int len)
{
struct ptrace_io_desc iorequest;
iorequest.piod_op = PIOD_READ_D;
iorequest.piod_offs = offset;
iorequest.piod_addr = buf;
iorequest.piod_len = len;
if (ptrace(PT_IO, pid, (caddr_t)&iorequest, 0) < 0)
return (-1);
return (0);
}
#define MAXSIZE 4096
1997-12-06 05:23:12 +00:00
/*
* Copy a string from the process. Note that it is
* expected to be a C string, but if max is set, it will
* only get that much.
*/
static char *
get_string(pid_t pid, void *addr, int max)
{
struct ptrace_io_desc iorequest;
char *buf, *nbuf;
size_t offset, size, totalsize;
offset = 0;
if (max)
size = max + 1;
else {
/* Read up to the end of the current page. */
size = PAGE_SIZE - ((uintptr_t)addr % PAGE_SIZE);
if (size > MAXSIZE)
size = MAXSIZE;
}
totalsize = size;
buf = malloc(totalsize);
if (buf == NULL)
return (NULL);
for (;;) {
iorequest.piod_op = PIOD_READ_D;
iorequest.piod_offs = (char *)addr + offset;
iorequest.piod_addr = buf + offset;
iorequest.piod_len = size;
if (ptrace(PT_IO, pid, (caddr_t)&iorequest, 0) < 0) {
free(buf);
return (NULL);
}
if (memchr(buf + offset, '\0', size) != NULL)
return (buf);
offset += size;
if (totalsize < MAXSIZE && max == 0) {
size = MAXSIZE - totalsize;
if (size > PAGE_SIZE)
size = PAGE_SIZE;
nbuf = realloc(buf, totalsize + size);
if (nbuf == NULL) {
buf[totalsize - 1] = '\0';
return (buf);
}
buf = nbuf;
totalsize += size;
} else {
buf[totalsize - 1] = '\0';
return (buf);
1997-12-06 05:23:12 +00:00
}
}
}
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
static const char *
strsig2(int sig)
{
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
static char tmp[32];
const char *signame;
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
signame = sysdecode_signal(sig);
if (signame == NULL) {
snprintf(tmp, sizeof(tmp), "%d", sig);
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
signame = tmp;
}
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
return (signame);
}
1997-12-06 05:23:12 +00:00
static void
print_kevent(FILE *fp, struct kevent *ke, int input)
{
switch (ke->filter) {
case EVFILT_READ:
case EVFILT_WRITE:
case EVFILT_VNODE:
case EVFILT_PROC:
case EVFILT_TIMER:
case EVFILT_PROCDESC:
fprintf(fp, "%ju", (uintmax_t)ke->ident);
break;
case EVFILT_SIGNAL:
fputs(strsig2(ke->ident), fp);
break;
default:
fprintf(fp, "%p", (void *)ke->ident);
}
fprintf(fp, ",%s,%s,", xlookup(kevent_filters, ke->filter),
xlookup_bits(kevent_flags, ke->flags));
switch (ke->filter) {
case EVFILT_READ:
case EVFILT_WRITE:
fputs(xlookup_bits(kevent_rdwr_fflags, ke->fflags), fp);
break;
case EVFILT_VNODE:
fputs(xlookup_bits(kevent_vnode_fflags, ke->fflags), fp);
break;
case EVFILT_PROC:
case EVFILT_PROCDESC:
fputs(xlookup_bits(kevent_proc_fflags, ke->fflags), fp);
break;
case EVFILT_TIMER:
fputs(xlookup_bits(kevent_timer_fflags, ke->fflags), fp);
break;
case EVFILT_USER: {
int ctrl, data;
ctrl = ke->fflags & NOTE_FFCTRLMASK;
2015-09-30 00:08:24 +00:00
data = ke->fflags & NOTE_FFLAGSMASK;
if (input) {
fputs(xlookup(kevent_user_ffctrl, ctrl), fp);
if (ke->fflags & NOTE_TRIGGER)
fputs("|NOTE_TRIGGER", fp);
if (data != 0)
fprintf(fp, "|%#x", data);
} else {
fprintf(fp, "%#x", data);
}
break;
}
default:
fprintf(fp, "%#x", ke->fflags);
}
fprintf(fp, ",%#jx,%p", (uintmax_t)ke->data, ke->udata);
}
static void
print_utrace(FILE *fp, void *utrace_addr, size_t len)
{
unsigned char *utrace_buffer;
fprintf(fp, "{ ");
if (sysdecode_utrace(fp, utrace_addr, len)) {
fprintf(fp, " }");
return;
}
utrace_buffer = utrace_addr;
fprintf(fp, "%zu:", len);
while (len--)
fprintf(fp, " %02x", *utrace_buffer++);
fprintf(fp, " }");
}
1997-12-06 05:23:12 +00:00
/*
* Converts a syscall argument into a string. Said string is
2015-08-19 20:02:03 +00:00
* allocated via malloc(), so needs to be free()'d. sc is
1997-12-06 05:23:12 +00:00
* a pointer to the syscall description (see above); args is
* an array of all of the system call arguments.
*/
char *
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
print_arg(struct syscall_args *sc, unsigned long *args, long *retval,
struct trussinfo *trussinfo)
{
FILE *fp;
char *tmp;
size_t tmplen;
pid_t pid;
fp = open_memstream(&tmp, &tmplen);
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
pid = trussinfo->curthread->proc->pid;
switch (sc->type & ARG_MASK) {
case Hex:
fprintf(fp, "0x%x", (int)args[sc->offset]);
break;
case Octal:
fprintf(fp, "0%o", (int)args[sc->offset]);
break;
case Int:
fprintf(fp, "%d", (int)args[sc->offset]);
break;
case UInt:
fprintf(fp, "%u", (unsigned int)args[sc->offset]);
break;
case PUInt: {
unsigned int val;
if (get_struct(pid, (void *)args[sc->offset], &val,
sizeof(val)) == 0)
fprintf(fp, "{ %u }", val);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case LongHex:
fprintf(fp, "0x%lx", args[sc->offset]);
break;
case Long:
fprintf(fp, "%ld", args[sc->offset]);
break;
case Sizet:
fprintf(fp, "%zu", (size_t)args[sc->offset]);
break;
case Name: {
/* NULL-terminated string. */
char *tmp2;
2015-08-19 20:02:03 +00:00
tmp2 = get_string(pid, (void*)args[sc->offset], 0);
fprintf(fp, "\"%s\"", tmp2);
free(tmp2);
break;
}
case BinString: {
2015-08-19 20:02:03 +00:00
/*
* Binary block of data that might have printable characters.
* XXX If type|OUT, assume that the length is the syscall's
* return value. Otherwise, assume that the length of the block
* is in the next syscall argument.
*/
int max_string = trussinfo->strsize;
2015-08-19 20:02:03 +00:00
char tmp2[max_string + 1], *tmp3;
int len;
int truncated = 0;
if (sc->type & OUT)
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
len = retval[0];
else
len = args[sc->offset + 1];
2015-08-19 20:02:03 +00:00
/*
* Don't print more than max_string characters, to avoid word
* wrap. If we have to truncate put some ... after the string.
*/
if (len > max_string) {
len = max_string;
truncated = 1;
}
if (len && get_struct(pid, (void*)args[sc->offset], &tmp2, len)
!= -1) {
tmp3 = malloc(len * 4 + 1);
while (len) {
if (strvisx(tmp3, tmp2, len,
VIS_CSTYLE|VIS_TAB|VIS_NL) <= max_string)
break;
len--;
truncated = 1;
}
fprintf(fp, "\"%s\"%s", tmp3, truncated ?
"..." : "");
free(tmp3);
} else {
fprintf(fp, "0x%lx", args[sc->offset]);
}
break;
}
case ExecArgs:
case ExecEnv:
case StringArray: {
uintptr_t addr;
union {
char *strarray[0];
char buf[PAGE_SIZE];
} u;
char *string;
size_t len;
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
u_int first, i;
/*
* Only parse argv[] and environment arrays from exec calls
* if requested.
*/
if (((sc->type & ARG_MASK) == ExecArgs &&
(trussinfo->flags & EXECVEARGS) == 0) ||
((sc->type & ARG_MASK) == ExecEnv &&
(trussinfo->flags & EXECVEENVS) == 0)) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
2015-09-30 00:08:24 +00:00
/*
* Read a page of pointers at a time. Punt if the top-level
* pointer is not aligned. Note that the first read is of
* a partial page.
*/
addr = args[sc->offset];
if (addr % sizeof(char *) != 0) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
len = PAGE_SIZE - (addr & PAGE_MASK);
if (get_struct(pid, (void *)addr, u.buf, len) == -1) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
fputc('[', fp);
first = 1;
i = 0;
while (u.strarray[i] != NULL) {
string = get_string(pid, u.strarray[i], 0);
fprintf(fp, "%s \"%s\"", first ? "" : ",", string);
free(string);
first = 0;
i++;
if (i == len / sizeof(char *)) {
addr += len;
len = PAGE_SIZE;
if (get_struct(pid, (void *)addr, u.buf, len) ==
-1) {
fprintf(fp, ", <inval>");
break;
}
i = 0;
}
}
fputs(" ]", fp);
break;
}
#ifdef __LP64__
case Quad:
fprintf(fp, "%ld", args[sc->offset]);
break;
case QuadHex:
fprintf(fp, "0x%lx", args[sc->offset]);
break;
#else
case Quad:
case QuadHex: {
unsigned long long ll;
2015-08-19 20:02:03 +00:00
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#if _BYTE_ORDER == _LITTLE_ENDIAN
ll = (unsigned long long)args[sc->offset + 1] << 32 |
args[sc->offset];
#else
ll = (unsigned long long)args[sc->offset] << 32 |
args[sc->offset + 1];
#endif
if ((sc->type & ARG_MASK) == Quad)
fprintf(fp, "%lld", ll);
else
fprintf(fp, "0x%llx", ll);
break;
}
#endif
case PQuadHex: {
uint64_t val;
if (get_struct(pid, (void *)args[sc->offset], &val,
sizeof(val)) == 0)
fprintf(fp, "{ 0x%jx }", (uintmax_t)val);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Ptr:
fprintf(fp, "0x%lx", args[sc->offset]);
break;
case Readlinkres: {
char *tmp2;
2015-08-19 20:02:03 +00:00
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
if (retval[0] == -1)
break;
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
tmp2 = get_string(pid, (void*)args[sc->offset], retval[0]);
fprintf(fp, "\"%s\"", tmp2);
free(tmp2);
break;
}
case Ioctl: {
2015-08-19 20:02:03 +00:00
const char *temp;
unsigned long cmd;
cmd = args[sc->offset];
temp = sysdecode_ioctlname(cmd);
if (temp)
fputs(temp, fp);
else {
fprintf(fp, "0x%lx { IO%s%s 0x%lx('%c'), %lu, %lu }",
2015-08-19 20:02:03 +00:00
cmd, cmd & IOC_OUT ? "R" : "",
cmd & IOC_IN ? "W" : "", IOCGROUP(cmd),
isprint(IOCGROUP(cmd)) ? (char)IOCGROUP(cmd) : '?',
cmd & 0xFF, IOCPARM_LEN(cmd));
}
break;
}
case Timespec: {
struct timespec ts;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &ts,
sizeof(ts)) != -1)
fprintf(fp, "{ %jd.%09ld }", (intmax_t)ts.tv_sec,
ts.tv_nsec);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Timespec2: {
struct timespec ts[2];
const char *sep;
unsigned int i;
if (get_struct(pid, (void *)args[sc->offset], &ts, sizeof(ts))
!= -1) {
fputs("{ ", fp);
sep = "";
for (i = 0; i < nitems(ts); i++) {
fputs(sep, fp);
sep = ", ";
switch (ts[i].tv_nsec) {
case UTIME_NOW:
fprintf(fp, "UTIME_NOW");
break;
case UTIME_OMIT:
fprintf(fp, "UTIME_OMIT");
break;
default:
fprintf(fp, "%jd.%09ld",
(intmax_t)ts[i].tv_sec,
ts[i].tv_nsec);
break;
}
}
fputs(" }", fp);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Timeval: {
struct timeval tv;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &tv, sizeof(tv))
!= -1)
fprintf(fp, "{ %jd.%06ld }", (intmax_t)tv.tv_sec,
tv.tv_usec);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Timeval2: {
struct timeval tv[2];
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &tv, sizeof(tv))
!= -1)
fprintf(fp, "{ %jd.%06ld, %jd.%06ld }",
(intmax_t)tv[0].tv_sec, tv[0].tv_usec,
(intmax_t)tv[1].tv_sec, tv[1].tv_usec);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Itimerval: {
struct itimerval itv;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &itv,
sizeof(itv)) != -1)
fprintf(fp, "{ %jd.%06ld, %jd.%06ld }",
(intmax_t)itv.it_interval.tv_sec,
itv.it_interval.tv_usec,
(intmax_t)itv.it_value.tv_sec,
itv.it_value.tv_usec);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case LinuxSockArgs:
{
struct linux_socketcall_args largs;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], (void *)&largs,
sizeof(largs)) != -1)
fprintf(fp, "{ %s, 0x%lx }",
lookup(linux_socketcall_ops, largs.what, 10),
(long unsigned int)largs.args);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Pollfd: {
/*
* XXX: A Pollfd argument expects the /next/ syscall argument
* to be the number of fds in the array. This matches the poll
* syscall.
*/
struct pollfd *pfd;
int numfds = args[sc->offset + 1];
size_t bytes = sizeof(struct pollfd) * numfds;
int i;
if ((pfd = malloc(bytes)) == NULL)
err(1, "Cannot malloc %zu bytes for pollfd array",
bytes);
if (get_struct(pid, (void *)args[sc->offset], pfd, bytes)
!= -1) {
fputs("{", fp);
for (i = 0; i < numfds; i++) {
fprintf(fp, " %d/%s", pfd[i].fd,
xlookup_bits(poll_flags, pfd[i].events));
}
fputs(" }", fp);
} else {
fprintf(fp, "0x%lx", args[sc->offset]);
}
free(pfd);
break;
}
case Fd_set: {
/*
* XXX: A Fd_set argument expects the /first/ syscall argument
* to be the number of fds in the array. This matches the
* select syscall.
*/
fd_set *fds;
int numfds = args[0];
size_t bytes = _howmany(numfds, _NFDBITS) * _NFDBITS;
int i;
if ((fds = malloc(bytes)) == NULL)
err(1, "Cannot malloc %zu bytes for fd_set array",
bytes);
if (get_struct(pid, (void *)args[sc->offset], fds, bytes)
!= -1) {
fputs("{", fp);
for (i = 0; i < numfds; i++) {
if (FD_ISSET(i, fds))
fprintf(fp, " %d", i);
}
fputs(" }", fp);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
free(fds);
break;
}
case Signal:
fputs(strsig2(args[sc->offset]), fp);
break;
case Sigset: {
long sig;
sigset_t ss;
int i, first;
sig = args[sc->offset];
if (get_struct(pid, (void *)args[sc->offset], (void *)&ss,
sizeof(ss)) == -1) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
fputs("{ ", fp);
first = 1;
for (i = 1; i < sys_nsig; i++) {
if (sigismember(&ss, i)) {
fprintf(fp, "%s%s", !first ? "|" : "",
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
strsig2(i));
first = 0;
}
}
if (!first)
fputc(' ', fp);
fputc('}', fp);
break;
}
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
case Sigprocmask:
print_integer_arg(sysdecode_sigprocmask_how, fp,
args[sc->offset]);
break;
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
case Fcntlflag:
2015-08-19 20:02:03 +00:00
/* XXX: Output depends on the value of the previous argument. */
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
if (sysdecode_fcntl_arg_p(args[sc->offset - 1]))
sysdecode_fcntl_arg(fp, args[sc->offset - 1],
args[sc->offset], 16);
break;
case Open:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_mask_arg(sysdecode_open_flags, fp, args[sc->offset]);
break;
case Fcntl:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_fcntl_cmd, fp, args[sc->offset]);
break;
case Mprot:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_mask_arg(sysdecode_mmap_prot, fp, args[sc->offset]);
break;
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
case Mmapflags:
print_mask_arg(sysdecode_mmap_flags, fp, args[sc->offset]);
break;
case Whence:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_whence, fp, args[sc->offset]);
break;
case Sockdomain:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_socketdomain, fp, args[sc->offset]);
break;
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
case Socktype:
print_mask_arg(sysdecode_socket_type, fp, args[sc->offset]);
break;
case Shutdown:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_shutdown_how, fp, args[sc->offset]);
break;
case Resource:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_rlimit, fp, args[sc->offset]);
break;
case RusageWho:
print_integer_arg(sysdecode_getrusage_who, fp, args[sc->offset]);
break;
case Pathconf:
fputs(xlookup(pathconf_arg, args[sc->offset]), fp);
break;
case Rforkflags:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_mask_arg(sysdecode_rfork_flags, fp, args[sc->offset]);
break;
case Sockaddr: {
char addr[64];
struct sockaddr_in *lsin;
struct sockaddr_in6 *lsin6;
struct sockaddr_un *sun;
struct sockaddr *sa;
socklen_t len;
u_char *q;
if (args[sc->offset] == 0) {
fputs("NULL", fp);
break;
}
/*
* Extract the address length from the next argument. If
* this is an output sockaddr (OUT is set), then the
* next argument is a pointer to a socklen_t. Otherwise
* the next argument contains a socklen_t by value.
*/
if (sc->type & OUT) {
if (get_struct(pid, (void *)args[sc->offset + 1],
&len, sizeof(len)) == -1) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
} else
len = args[sc->offset + 1];
/* If the length is too small, just bail. */
if (len < sizeof(*sa)) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
sa = calloc(1, len);
if (get_struct(pid, (void *)args[sc->offset], sa, len) == -1) {
free(sa);
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
switch (sa->sa_family) {
case AF_INET:
if (len < sizeof(*lsin))
goto sockaddr_short;
lsin = (struct sockaddr_in *)(void *)sa;
2015-08-19 20:02:03 +00:00
inet_ntop(AF_INET, &lsin->sin_addr, addr, sizeof(addr));
fprintf(fp, "{ AF_INET %s:%d }", addr,
htons(lsin->sin_port));
break;
case AF_INET6:
if (len < sizeof(*lsin6))
goto sockaddr_short;
lsin6 = (struct sockaddr_in6 *)(void *)sa;
inet_ntop(AF_INET6, &lsin6->sin6_addr, addr,
2015-08-19 20:02:03 +00:00
sizeof(addr));
fprintf(fp, "{ AF_INET6 [%s]:%d }", addr,
htons(lsin6->sin6_port));
break;
case AF_UNIX:
sun = (struct sockaddr_un *)sa;
fprintf(fp, "{ AF_UNIX \"%.*s\" }",
(int)(len - offsetof(struct sockaddr_un, sun_path)),
sun->sun_path);
break;
default:
sockaddr_short:
fprintf(fp,
"{ sa_len = %d, sa_family = %d, sa_data = {",
(int)sa->sa_len, (int)sa->sa_family);
for (q = (u_char *)sa->sa_data;
q < (u_char *)sa + len; q++)
fprintf(fp, "%s 0x%02x",
q == (u_char *)sa->sa_data ? "" : ",",
*q);
fputs(" } }", fp);
}
free(sa);
break;
2004-03-23 09:04:06 +00:00
}
case Sigaction: {
struct sigaction sa;
if (get_struct(pid, (void *)args[sc->offset], &sa, sizeof(sa))
!= -1) {
fputs("{ ", fp);
if (sa.sa_handler == SIG_DFL)
fputs("SIG_DFL", fp);
else if (sa.sa_handler == SIG_IGN)
fputs("SIG_IGN", fp);
else
fprintf(fp, "%p", sa.sa_handler);
fprintf(fp, " %s ss_t }",
xlookup_bits(sigaction_flags, sa.sa_flags));
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Kevent: {
/*
2015-08-19 20:02:03 +00:00
* XXX XXX: The size of the array is determined by either the
* next syscall argument, or by the syscall return value,
* depending on which argument number we are. This matches the
* kevent syscall, but luckily that's the only syscall that uses
* them.
*/
struct kevent *ke;
int numevents = -1;
size_t bytes;
int i;
if (sc->offset == 1)
numevents = args[sc->offset+1];
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
else if (sc->offset == 3 && retval[0] != -1)
numevents = retval[0];
if (numevents >= 0) {
bytes = sizeof(struct kevent) * numevents;
if ((ke = malloc(bytes)) == NULL)
err(1,
"Cannot malloc %zu bytes for kevent array",
bytes);
} else
ke = NULL;
if (numevents >= 0 && get_struct(pid, (void *)args[sc->offset],
ke, bytes) != -1) {
fputc('{', fp);
for (i = 0; i < numevents; i++) {
fputc(' ', fp);
print_kevent(fp, &ke[i], sc->offset == 1);
}
fputs(" }", fp);
} else {
fprintf(fp, "0x%lx", args[sc->offset]);
}
free(ke);
break;
}
case Stat: {
struct stat st;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &st, sizeof(st))
!= -1) {
char mode[12];
2015-08-19 20:02:03 +00:00
strmode(st.st_mode, mode);
fprintf(fp,
"{ mode=%s,inode=%ju,size=%jd,blksize=%ld }", mode,
(uintmax_t)st.st_ino, (intmax_t)st.st_size,
(long)st.st_blksize);
} else {
fprintf(fp, "0x%lx", args[sc->offset]);
}
break;
}
case StatFs: {
unsigned int i;
struct statfs buf;
2015-10-03 19:08:36 +00:00
if (get_struct(pid, (void *)args[sc->offset], &buf,
sizeof(buf)) != -1) {
char fsid[17];
bzero(fsid, sizeof(fsid));
if (buf.f_fsid.val[0] != 0 || buf.f_fsid.val[1] != 0) {
for (i = 0; i < sizeof(buf.f_fsid); i++)
snprintf(&fsid[i*2],
sizeof(fsid) - (i*2), "%02x",
((u_char *)&buf.f_fsid)[i]);
}
fprintf(fp,
"{ fstypename=%s,mntonname=%s,mntfromname=%s,"
"fsid=%s }", buf.f_fstypename, buf.f_mntonname,
buf.f_mntfromname, fsid);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Rusage: {
struct rusage ru;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &ru, sizeof(ru))
!= -1) {
fprintf(fp,
"{ u=%jd.%06ld,s=%jd.%06ld,in=%ld,out=%ld }",
(intmax_t)ru.ru_utime.tv_sec, ru.ru_utime.tv_usec,
(intmax_t)ru.ru_stime.tv_sec, ru.ru_stime.tv_usec,
ru.ru_inblock, ru.ru_oublock);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Rlimit: {
struct rlimit rl;
2015-08-19 20:02:03 +00:00
if (get_struct(pid, (void *)args[sc->offset], &rl, sizeof(rl))
!= -1) {
fprintf(fp, "{ cur=%ju,max=%ju }",
rl.rlim_cur, rl.rlim_max);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case ExitStatus: {
int status;
if (get_struct(pid, (void *)args[sc->offset], &status,
sizeof(status)) != -1) {
fputs("{ ", fp);
if (WIFCONTINUED(status))
fputs("CONTINUED", fp);
else if (WIFEXITED(status))
fprintf(fp, "EXITED,val=%d",
WEXITSTATUS(status));
else if (WIFSIGNALED(status))
fprintf(fp, "SIGNALED,sig=%s%s",
strsig2(WTERMSIG(status)),
WCOREDUMP(status) ? ",cored" : "");
else
fprintf(fp, "STOPPED,sig=%s",
strsig2(WTERMSIG(status)));
fputs(" }", fp);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Waitoptions:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_mask_arg(sysdecode_wait6_options, fp, args[sc->offset]);
break;
case Idtype:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_idtype, fp, args[sc->offset]);
break;
case Procctl:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_procctl_cmd, fp, args[sc->offset]);
break;
case Umtxop:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_umtx_op, fp, args[sc->offset]);
break;
case Atfd:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_integer_arg(sysdecode_atfd, fp, args[sc->offset]);
break;
case Atflags:
fputs(xlookup_bits(at_flags, args[sc->offset]), fp);
break;
case Accessmode:
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
print_mask_arg(sysdecode_access_mode, fp, args[sc->offset]);
break;
case Sysarch:
fputs(xlookup(sysarch_ops, args[sc->offset]), fp);
break;
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
case PipeFds:
/*
* The pipe() system call in the kernel returns its
* two file descriptors via return values. However,
* the interface exposed by libc is that pipe()
* accepts a pointer to an array of descriptors.
* Format the output to match the libc API by printing
* the returned file descriptors as a fake argument.
*
* Overwrite the first retval to signal a successful
* return as well.
*/
fprintf(fp, "{ %ld, %ld }", retval[0], retval[1]);
retval[0] = 0;
break;
case Utrace: {
size_t len;
void *utrace_addr;
len = args[sc->offset + 1];
utrace_addr = calloc(1, len);
if (get_struct(pid, (void *)args[sc->offset],
(void *)utrace_addr, len) != -1)
print_utrace(fp, utrace_addr, len);
else
fprintf(fp, "0x%lx", args[sc->offset]);
free(utrace_addr);
break;
}
case IntArray: {
int descriptors[16];
unsigned long i, ndescriptors;
bool truncated;
ndescriptors = args[sc->offset + 1];
truncated = false;
if (ndescriptors > nitems(descriptors)) {
ndescriptors = nitems(descriptors);
truncated = true;
}
if (get_struct(pid, (void *)args[sc->offset],
descriptors, ndescriptors * sizeof(descriptors[0])) != -1) {
fprintf(fp, "{");
for (i = 0; i < ndescriptors; i++)
fprintf(fp, i == 0 ? " %d" : ", %d",
descriptors[i]);
fprintf(fp, truncated ? ", ... }" : " }");
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
Move mksubr from kdump into libsysdecode. Restructure this script so that it generates a header of tables instead of a source file. The tables are included in a flags.c source file which provides functions to decode various system call arguments. For functions that decode an enumeration, the function returns a pointer to a string for known values and NULL for unknown values. For functions that do more complex decoding (typically of a bitmask), the function accepts a pointer to a FILE object (open_memstream() can be used as a string builder) to which decoded values are written. If the function operates on a bitmask, the function returns true if any bits were decoded or false if the entire value was valid. Additionally, the third argument accepts a pointer to a value to which any undecoded bits are stored. This pointer can be NULL if the caller doesn't care about remaining bits. Convert kdump over to using decoder functions from libsysdecode instead of mksubr. truss also uses decoders from libsysdecode instead of private lookup tables, though lookup tables for objects not decoded by kdump remain in truss for now. Eventually most of these tables should move into libsysdecode as the automated table generation approach from mksubr is less stale than the static tables in truss. Some changes have been made to truss and kdump output: - The flags passed to open() are now properly decoded in that one of O_RDONLY, O_RDWR, O_WRONLY, or O_EXEC is always included in a decoded mask. - Optional arguments to open(), openat(), and fcntl() are only printed in kdump if they exist (e.g. the mode is only printed for open() if O_CREAT is set in the flags). - Print argument to F_GETLK/SETLK/SETLKW in kdump as a pointer, not int. - Include all procctl() commands. - Correctly decode pipe2() flags in truss by not assuming full open()-like flags with O_RDONLY, etc. - Decode file flags passed to *chflags() as file flags (UF_* and SF_*) rather than as a file mode. - Fix decoding of quotactl() commands by splitting out the two command components instead of assuming the raw command value matches the primary command component. In addition, truss and kdump now build without triggering any warnings. All of the sysdecode manpages now include the required headers in the synopsis. Reviewed by: kib (several older versions), wblock (manpages) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7847
2016-10-17 22:37:07 +00:00
case Pipe2:
print_mask_arg(sysdecode_pipe2_flags, fp, args[sc->offset]);
break;
case CapFcntlRights: {
uint32_t rights;
if (sc->type & OUT) {
if (get_struct(pid, (void *)args[sc->offset], &rights,
sizeof(rights)) == -1) {
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
} else
rights = args[sc->offset];
print_mask_arg32(sysdecode_cap_fcntlrights, fp, rights);
break;
}
case Fadvice:
print_integer_arg(sysdecode_fadvice, fp, args[sc->offset]);
break;
case FileFlags: {
fflags_t rem;
if (!sysdecode_fileflags(fp, args[sc->offset], &rem))
fprintf(fp, "0x%x", rem);
else if (rem != 0)
fprintf(fp, "|0x%x", rem);
break;
}
2017-03-18 18:26:56 +00:00
case Flockop:
print_mask_arg(sysdecode_flock_operation, fp, args[sc->offset]);
break;
case Getfsstatmode:
print_integer_arg(sysdecode_getfsstat_mode, fp,
args[sc->offset]);
break;
case Kldsymcmd:
print_integer_arg(sysdecode_kldsym_cmd, fp, args[sc->offset]);
break;
case Kldunloadflags:
print_integer_arg(sysdecode_kldunload_flags, fp,
args[sc->offset]);
break;
2017-03-19 00:31:21 +00:00
case Madvice:
print_integer_arg(sysdecode_madvice, fp, args[sc->offset]);
break;
case Socklent:
fprintf(fp, "%u", (socklen_t)args[sc->offset]);
break;
2017-05-03 09:23:13 +00:00
case Sockprotocol: {
const char *temp;
int domain, protocol;
2017-05-03 09:23:13 +00:00
domain = args[sc->offset - 2];
2017-05-03 09:23:13 +00:00
protocol = args[sc->offset];
if (protocol == 0) {
fputs("0", fp);
} else {
temp = sysdecode_socket_protocol(domain, protocol);
if (temp) {
fputs(temp, fp);
} else {
fprintf(fp, "%d", protocol);
}
2017-05-03 09:23:13 +00:00
}
break;
}
2017-05-03 12:18:09 +00:00
case Sockoptlevel:
print_integer_arg(sysdecode_sockopt_level, fp,
args[sc->offset]);
break;
case Sockoptname: {
const char *temp;
int level, name;
level = args[sc->offset - 1];
name = args[sc->offset];
temp = sysdecode_sockopt_name(level, name);
if (temp) {
fputs(temp, fp);
} else {
fprintf(fp, "%d", name);
}
break;
}
case Msgflags:
print_mask_arg(sysdecode_msg_flags, fp, args[sc->offset]);
break;
case CapRights: {
cap_rights_t rights;
if (get_struct(pid, (void *)args[sc->offset], &rights,
sizeof(rights)) != -1) {
fputs("{ ", fp);
sysdecode_cap_rights(fp, &rights);
fputs(" }", fp);
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case Acltype:
print_integer_arg(sysdecode_acltype, fp, args[sc->offset]);
break;
case Extattrnamespace:
print_integer_arg(sysdecode_extattrnamespace, fp,
args[sc->offset]);
break;
2017-06-08 04:45:13 +00:00
case Minherit:
print_integer_arg(sysdecode_minherit_inherit, fp,
args[sc->offset]);
break;
case Mlockall:
print_mask_arg(sysdecode_mlockall_flags, fp, args[sc->offset]);
break;
case Mountflags:
print_mask_arg(sysdecode_mount_flags, fp, args[sc->offset]);
break;
2017-06-08 08:10:57 +00:00
case Msync:
print_mask_arg(sysdecode_msync_flags, fp, args[sc->offset]);
break;
case Priowhich:
print_integer_arg(sysdecode_prio_which, fp, args[sc->offset]);
break;
case Ptraceop:
print_integer_arg(sysdecode_ptrace_request, fp,
args[sc->offset]);
break;
2017-06-10 00:53:00 +00:00
case Quotactlcmd:
if (!sysdecode_quotactl_cmd(fp, args[sc->offset]))
fprintf(fp, "%#x", (int)args[sc->offset]);
break;
case Reboothowto:
print_mask_arg(sysdecode_reboot_howto, fp, args[sc->offset]);
break;
case Rtpriofunc:
print_integer_arg(sysdecode_rtprio_function, fp,
args[sc->offset]);
break;
case Schedpolicy:
print_integer_arg(sysdecode_scheduler_policy, fp,
args[sc->offset]);
break;
case Schedparam: {
struct sched_param sp;
if (get_struct(pid, (void *)args[sc->offset], &sp,
sizeof(sp)) != -1)
fprintf(fp, "{ %d }", sp.sched_priority);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case CloudABIAdvice:
fputs(xlookup(cloudabi_advice, args[sc->offset]), fp);
break;
case CloudABIClockID:
fputs(xlookup(cloudabi_clockid, args[sc->offset]), fp);
break;
case ClouduABIFDSFlags:
fputs(xlookup_bits(cloudabi_fdsflags, args[sc->offset]), fp);
break;
case CloudABIFDStat: {
cloudabi_fdstat_t fds;
if (get_struct(pid, (void *)args[sc->offset], &fds, sizeof(fds))
!= -1) {
fprintf(fp, "{ %s, ",
xlookup(cloudabi_filetype, fds.fs_filetype));
fprintf(fp, "%s, ... }",
xlookup_bits(cloudabi_fdflags, fds.fs_flags));
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case CloudABIFileStat: {
cloudabi_filestat_t fsb;
if (get_struct(pid, (void *)args[sc->offset], &fsb, sizeof(fsb))
!= -1)
2016-10-11 17:11:17 +00:00
fprintf(fp, "{ %s, %ju }",
xlookup(cloudabi_filetype, fsb.st_filetype),
2016-10-11 17:11:17 +00:00
(uintmax_t)fsb.st_size);
else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case CloudABIFileType:
fputs(xlookup(cloudabi_filetype, args[sc->offset]), fp);
break;
case CloudABIFSFlags:
fputs(xlookup_bits(cloudabi_fsflags, args[sc->offset]), fp);
break;
case CloudABILookup:
if ((args[sc->offset] & CLOUDABI_LOOKUP_SYMLINK_FOLLOW) != 0)
fprintf(fp, "%d|LOOKUP_SYMLINK_FOLLOW",
(int)args[sc->offset]);
else
fprintf(fp, "%d", (int)args[sc->offset]);
break;
case CloudABIMFlags:
fputs(xlookup_bits(cloudabi_mflags, args[sc->offset]), fp);
break;
case CloudABIMProt:
fputs(xlookup_bits(cloudabi_mprot, args[sc->offset]), fp);
break;
case CloudABIMSFlags:
fputs(xlookup_bits(cloudabi_msflags, args[sc->offset]), fp);
break;
case CloudABIOFlags:
fputs(xlookup_bits(cloudabi_oflags, args[sc->offset]), fp);
break;
case CloudABISDFlags:
fputs(xlookup_bits(cloudabi_sdflags, args[sc->offset]), fp);
break;
case CloudABISignal:
fputs(xlookup(cloudabi_signal, args[sc->offset]), fp);
break;
case CloudABISockStat: {
cloudabi_sockstat_t ss;
if (get_struct(pid, (void *)args[sc->offset], &ss, sizeof(ss))
!= -1) {
fprintf(fp, "{ %s, ", xlookup(
cloudabi_sa_family, ss.ss_sockname.sa_family));
fprintf(fp, "%s, ", xlookup(
cloudabi_sa_family, ss.ss_peername.sa_family));
fprintf(fp, "%s, ", xlookup(
cloudabi_errno, ss.ss_error));
fprintf(fp, "%s }", xlookup_bits(
cloudabi_ssstate, ss.ss_state));
} else
fprintf(fp, "0x%lx", args[sc->offset]);
break;
}
case CloudABISSFlags:
fputs(xlookup_bits(cloudabi_ssflags, args[sc->offset]), fp);
break;
case CloudABITimestamp:
fprintf(fp, "%lu.%09lus", args[sc->offset] / 1000000000,
args[sc->offset] % 1000000000);
break;
case CloudABIULFlags:
fputs(xlookup_bits(cloudabi_ulflags, args[sc->offset]), fp);
break;
case CloudABIWhence:
fputs(xlookup(cloudabi_whence, args[sc->offset]), fp);
break;
default:
errx(1, "Invalid argument type %d\n", sc->type & ARG_MASK);
}
fclose(fp);
return (tmp);
1997-12-06 05:23:12 +00:00
}
/*
* Print (to outfile) the system call and its arguments.
1997-12-06 05:23:12 +00:00
*/
void
print_syscall(struct trussinfo *trussinfo)
{
struct threadinfo *t;
const char *name;
char **s_args;
int i, len, nargs;
t = trussinfo->curthread;
name = t->cs.sc->name;
nargs = t->cs.nargs;
s_args = t->cs.s_args;
len = print_line_prefix(trussinfo);
len += fprintf(trussinfo->outfile, "%s(", name);
for (i = 0; i < nargs; i++) {
if (s_args[i] != NULL)
len += fprintf(trussinfo->outfile, "%s", s_args[i]);
else
len += fprintf(trussinfo->outfile,
"<missing argument>");
len += fprintf(trussinfo->outfile, "%s", i < (nargs - 1) ?
"," : "");
}
len += fprintf(trussinfo->outfile, ")");
for (i = 0; i < 6 - (len / 8); i++)
fprintf(trussinfo->outfile, "\t");
}
void
print_syscall_ret(struct trussinfo *trussinfo, int errorp, long *retval)
{
struct timespec timediff;
struct threadinfo *t;
struct syscall *sc;
2016-02-23 20:00:55 +00:00
int error;
t = trussinfo->curthread;
sc = t->cs.sc;
if (trussinfo->flags & COUNTONLY) {
timespecsubt(&t->after, &t->before, &timediff);
timespecadd(&sc->time, &timediff, &sc->time);
sc->ncalls++;
if (errorp)
sc->nerror++;
return;
}
print_syscall(trussinfo);
fflush(trussinfo->outfile);
if (retval == NULL) {
/*
* This system call resulted in the current thread's exit,
* so there is no return value or error to display.
*/
fprintf(trussinfo->outfile, "\n");
return;
}
2016-02-23 20:00:55 +00:00
if (errorp) {
error = sysdecode_abi_to_freebsd_errno(t->proc->abi->abi,
retval[0]);
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
fprintf(trussinfo->outfile, " ERR#%ld '%s'\n", retval[0],
2016-02-23 20:00:55 +00:00
error == INT_MAX ? "Unknown error" : strerror(error));
}
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#ifndef __LP64__
else if (sc->ret_type == 2) {
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
off_t off;
#if _BYTE_ORDER == _LITTLE_ENDIAN
off = (off_t)retval[1] << 32 | retval[0];
#else
off = (off_t)retval[0] << 32 | retval[1];
#endif
fprintf(trussinfo->outfile, " = %jd (0x%jx)\n", (intmax_t)off,
(intmax_t)off);
}
Several changes to truss. - Refactor the interface between the ABI-independent code and the ABI-specific backends. The backends now provide smaller hooks to fetch system call arguments and return values. The rest of the system call entry and exit handling that was previously duplicated among all the backends has been moved to one place. - Merge the loop when waiting for an event with the loop for handling stops. This also means not emulating a procfs-like interface on top of ptrace(). Instead, use a single event loop that fetches process events via waitid(). Among other things this allows us to report the full 32-bit exit value. - Use PT_FOLLOW_FORK to follow new child processes instead of forking a new truss process for each new child. This allows one truss process to monitor a tree of processes and truss -c should now display one total for the entire tree instead of separate summaries per process. - Use the recently added fields to ptrace_lwpinfo to determine the current system call number and argument count. The latter is especially useful and fixes a regression since the conversion from procfs. truss now generally prints the correct number of arguments for most system calls rather than printing extra arguments for any call not listed in the table in syscalls.c. - Actually check the new ABI when processes call exec. The comments claimed that this happened but it was not being done (perhaps this was another regression in the conversion to ptrace()). If the new ABI after exec is not supported, truss detaches from the process. If truss does not support the ABI for a newly executed process the process is killed before it returns from exec. - Along with the refactor, teach the various ABI-specific backends to fetch both return values, not just the first. Use this to properly report the full 64-bit return value from lseek(). In addition, the handler for "pipe" now pulls the pair of descriptors out of the return values (which is the true kernel system call interface) but displays them as an argument (which matches the interface exported by libc). - Each ABI handler adds entries to a linker set rather than requiring a statically defined table of handlers in main.c. - The arm and mips system call fetching code was changed to follow the same pattern as amd64 (and the in-kernel handler) of fetching register arguments first and then reading any remaining arguments from the stack. This should fix indirect system call arguments on at least arm. - The mipsn32 and n64 ABIs will now look for arguments in A4 through A7. - Use register %ebp for the 6th system call argument for Linux/i386 ABIs to match the in-kernel argument fetch code. - For powerpc binaries on a powerpc64 system, fetch the extra arguments on the stack as 32-bit values that are then copied into the 64-bit argument array instead of reading the 32-bit values directly into the 64-bit array. Reviewed by: kib (earlier version) Tested on: amd64 (FreeBSD/amd64 & i386), i386, arm (earlier version) Tested on: powerpc64 (FreeBSD/powerpc64 & powerpc) MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D3575
2015-09-30 19:13:32 +00:00
#endif
else
fprintf(trussinfo->outfile, " = %ld (0x%lx)\n", retval[0],
retval[0]);
1997-12-06 05:23:12 +00:00
}
void
print_summary(struct trussinfo *trussinfo)
{
struct timespec total = {0, 0};
struct syscall *sc;
int ncall, nerror;
fprintf(trussinfo->outfile, "%-20s%15s%8s%8s\n",
"syscall", "seconds", "calls", "errors");
ncall = nerror = 0;
STAILQ_FOREACH(sc, &syscalls, entries)
if (sc->ncalls) {
fprintf(trussinfo->outfile, "%-20s%5jd.%09ld%8d%8d\n",
sc->name, (intmax_t)sc->time.tv_sec,
sc->time.tv_nsec, sc->ncalls, sc->nerror);
timespecadd(&total, &sc->time, &total);
ncall += sc->ncalls;
nerror += sc->nerror;
}
fprintf(trussinfo->outfile, "%20s%15s%8s%8s\n",
"", "-------------", "-------", "-------");
fprintf(trussinfo->outfile, "%-20s%5jd.%09ld%8d%8d\n",
"", (intmax_t)total.tv_sec, total.tv_nsec, ncall, nerror);
}