freebsd-skq/usr.bin/truss/setup.c

592 lines
14 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1997-12-06 05:23:12 +00:00
/*
* Various setup functions for truss. Not the cleanest-written code,
* I'm afraid.
*/
#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/sysctl.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 <assert.h>
#include <err.h>
#include <errno.h>
#include <signal.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 <stdint.h>
1997-12-06 05:23:12 +00:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sysdecode.h>
#include <time.h>
1997-12-06 05:23:12 +00:00
#include <unistd.h>
#include "truss.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 "syscall.h"
#include "extern.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
SET_DECLARE(procabi, struct procabi);
static sig_atomic_t detaching;
1997-12-06 05:23:12 +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
static void new_proc(struct trussinfo *, pid_t);
1997-12-06 05:23:12 +00:00
/*
* setup_and_wait() is called to start a process. All it really does
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
* is fork(), enable tracing in the child, and then exec the given
* command. At that point, the child process stops, and the parent
* can wake up and deal with it.
1997-12-06 05:23:12 +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
void
setup_and_wait(struct trussinfo *info, char *command[])
{
pid_t pid;
pid = vfork();
if (pid == -1)
err(1, "fork failed");
if (pid == 0) { /* Child */
ptrace(PT_TRACE_ME, 0, 0, 0);
execvp(command[0], command);
err(1, "execvp %s", command[0]);
}
/* Only in the parent here */
if (waitpid(pid, NULL, 0) < 0)
err(1, "unexpect stop in waitpid");
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
new_proc(info, pid);
1997-12-06 05:23:12 +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
* start_tracing is called to attach to an existing process.
1997-12-06 05:23:12 +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
void
start_tracing(struct trussinfo *info, pid_t pid)
{
int ret, retry;
retry = 10;
do {
ret = ptrace(PT_ATTACH, pid, NULL, 0);
usleep(200);
} while (ret && retry-- > 0);
if (ret)
err(1, "can not attach to target process");
if (waitpid(pid, NULL, 0) < 0)
err(1, "Unexpect stop in waitpid");
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
new_proc(info, pid);
1997-12-06 05:23:12 +00:00
}
/*
* Restore a process back to it's pre-truss state.
* Called for SIGINT, SIGTERM, SIGQUIT. This only
* applies if truss was told to monitor an already-existing
* process.
*/
void
restore_proc(int signo __unused)
{
detaching = 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
static void
detach_proc(pid_t pid)
{
1997-12-06 05:23:12 +00:00
/* stop the child so that we can detach */
kill(pid, SIGSTOP);
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 (waitpid(pid, NULL, 0) < 0)
err(1, "Unexpected stop in waitpid");
if (ptrace(PT_DETACH, pid, (caddr_t)1, 0) < 0)
err(1, "Can not detach the process");
kill(pid, SIGCONT);
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
}
/*
* Determine the ABI. This is called after every exec, and when
* a process is first monitored.
*/
static struct procabi *
find_abi(pid_t pid)
{
struct procabi **pabi;
size_t len;
int error;
int mib[4];
char progt[32];
len = sizeof(progt);
mib[0] = CTL_KERN;
mib[1] = KERN_PROC;
mib[2] = KERN_PROC_SV_NAME;
mib[3] = pid;
error = sysctl(mib, 4, progt, &len, NULL, 0);
if (error != 0)
err(2, "can not get sysvec name");
SET_FOREACH(pabi, procabi) {
if (strcmp((*pabi)->type, progt) == 0)
return (*pabi);
}
warnx("ABI %s for pid %ld is not supported", progt, (long)pid);
return (NULL);
}
static void
new_proc(struct trussinfo *info, pid_t pid)
{
struct procinfo *np;
/*
* If this happens it means there is a bug in truss. Unfortunately
* this will kill any processes are attached to.
*/
LIST_FOREACH(np, &info->proclist, entries) {
if (np->pid == pid)
errx(1, "Duplicate process for pid %ld", (long)pid);
}
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 (info->flags & FOLLOWFORKS)
if (ptrace(PT_FOLLOW_FORK, pid, NULL, 1) == -1)
err(1, "Unable to follow forks for pid %ld", (long)pid);
np = calloc(1, sizeof(struct procinfo));
np->pid = pid;
np->abi = find_abi(pid);
SLIST_INIT(&np->threadlist);
LIST_INSERT_HEAD(&info->proclist, np, entries);
}
static void
free_proc(struct procinfo *p)
{
struct threadinfo *t, *t2;
SLIST_FOREACH_SAFE(t, &p->threadlist, entries, t2) {
free(t);
}
LIST_REMOVE(p, entries);
free(p);
}
static void
detach_all_procs(struct trussinfo *info)
{
struct procinfo *p, *p2;
LIST_FOREACH_SAFE(p, &info->proclist, entries, p2) {
detach_proc(p->pid);
free_proc(p);
}
}
static struct procinfo *
find_proc(struct trussinfo *info, pid_t pid)
{
struct procinfo *np;
LIST_FOREACH(np, &info->proclist, entries) {
if (np->pid == pid)
return (np);
}
return (NULL);
1997-12-06 05:23:12 +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
* Change curthread member based on (pid, lwpid).
* If it is a new thread, create a threadinfo structure.
*/
static void
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
find_thread(struct trussinfo *info, pid_t pid, lwpid_t lwpid)
{
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
struct procinfo *np;
struct threadinfo *nt;
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
np = find_proc(info, pid);
assert(np != NULL);
SLIST_FOREACH(nt, &np->threadlist, entries) {
if (nt->tid == lwpid) {
info->curthread = nt;
return;
}
}
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
nt = calloc(1, sizeof(struct threadinfo));
if (nt == NULL)
err(1, "calloc() failed");
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
nt->proc = np;
nt->tid = lwpid;
SLIST_INSERT_HEAD(&np->threadlist, nt, entries);
info->curthread = nt;
}
/*
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
* When a process exits, it no longer has any threads left. However,
* the main loop expects a valid curthread. In cases when a thread
* triggers the termination (e.g. calling exit or triggering a fault)
* we would ideally use that thread. However, if a process is killed
* by a signal sent from another process then there is no "correct"
* thread. We just punt and use the first thread.
*/
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
static void
find_exit_thread(struct trussinfo *info, pid_t pid)
{
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
struct procinfo *np;
struct threadinfo *nt;
np = find_proc(info, pid);
assert(np != NULL);
if (SLIST_EMPTY(&np->threadlist)) {
/*
* If an existing process exits right after we attach
* to it but before it posts any events, there won't
* be any threads. Create a dummy thread and set its
* "before" time to the global start time.
*/
nt = calloc(1, sizeof(struct threadinfo));
if (nt == NULL)
err(1, "calloc() failed");
nt->proc = np;
nt->tid = 0;
SLIST_INSERT_HEAD(&np->threadlist, nt, entries);
nt->before = info->start_time;
}
info->curthread = SLIST_FIRST(&np->threadlist);
}
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
static void
alloc_syscall(struct threadinfo *t, struct ptrace_lwpinfo *pl)
{
u_int i;
assert(t->in_syscall == 0);
assert(t->cs.number == 0);
assert(t->cs.name == NULL);
assert(t->cs.nargs == 0);
for (i = 0; i < nitems(t->cs.s_args); i++)
assert(t->cs.s_args[i] == NULL);
memset(t->cs.args, 0, sizeof(t->cs.args));
t->cs.number = pl->pl_syscall_code;
t->in_syscall = 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
static void
free_syscall(struct threadinfo *t)
{
u_int i;
for (i = 0; i < t->cs.nargs; i++)
free(t->cs.s_args[i]);
memset(&t->cs, 0, sizeof(t->cs));
t->in_syscall = 0;
}
static void
enter_syscall(struct trussinfo *info, struct ptrace_lwpinfo *pl)
{
struct threadinfo *t;
struct syscall *sc;
u_int i, narg;
t = info->curthread;
alloc_syscall(t, pl);
narg = MIN(pl->pl_syscall_narg, nitems(t->cs.args));
if (narg != 0 && t->proc->abi->fetch_args(info, narg) != 0) {
free_syscall(t);
return;
}
t->cs.name = sysdecode_syscallname(t->proc->abi->abi, t->cs.number);
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 (t->cs.name == NULL)
fprintf(info->outfile, "-- UNKNOWN %s SYSCALL %d --\n",
t->proc->abi->type, t->cs.number);
sc = get_syscall(t->cs.name, narg);
t->cs.nargs = sc->nargs;
assert(sc->nargs <= nitems(t->cs.s_args));
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
t->cs.sc = sc;
/*
* At this point, we set up the system call arguments.
* We ignore any OUT ones, however -- those are arguments that
* are set by the system call, and so are probably meaningless
* now. This doesn't currently support arguments that are
* passed in *and* out, however.
*/
if (t->cs.name != NULL) {
#if DEBUG
fprintf(stderr, "syscall %s(", t->cs.name);
#endif
for (i = 0; i < t->cs.nargs; i++) {
#if DEBUG
fprintf(stderr, "0x%lx%s", sc ?
t->cs.args[sc->args[i].offset] : t->cs.args[i],
i < (t->cs.nargs - 1) ? "," : "");
#endif
if (!(sc->args[i].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
t->cs.s_args[i] = print_arg(&sc->args[i],
t->cs.args, 0, info);
}
}
#if DEBUG
fprintf(stderr, ")\n");
#endif
}
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
clock_gettime(CLOCK_REALTIME, &t->before);
}
static void
exit_syscall(struct trussinfo *info, struct ptrace_lwpinfo *pl)
{
struct threadinfo *t;
struct procinfo *p;
struct syscall *sc;
long retval[2];
u_int i;
int errorp;
t = info->curthread;
if (!t->in_syscall)
return;
clock_gettime(CLOCK_REALTIME, &t->after);
p = t->proc;
if (p->abi->fetch_retval(info, retval, &errorp) < 0) {
free_syscall(t);
return;
}
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
sc = t->cs.sc;
/*
* Here, we only look for arguments that have OUT masked in --
* otherwise, they were handled in enter_syscall().
*/
for (i = 0; i < sc->nargs; i++) {
char *temp;
if (sc->args[i].type & OUT) {
/*
* If an error occurred, then don't bother
* getting the data; it may not be valid.
*/
if (errorp) {
asprintf(&temp, "0x%lx",
t->cs.args[sc->args[i].offset]);
} else {
temp = print_arg(&sc->args[i],
t->cs.args, retval, info);
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
}
t->cs.s_args[i] = temp;
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_syscall_ret(info, errorp, retval);
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
free_syscall(t);
/*
* If the process executed a new image, check the ABI. If the
* new ABI isn't supported, stop tracing this process.
*/
if (pl->pl_flags & PL_FLAG_EXEC) {
p->abi = find_abi(p->pid);
if (p->abi == NULL) {
if (ptrace(PT_DETACH, p->pid, (caddr_t)1, 0) < 0)
err(1, "Can not detach the process");
free_proc(p);
}
}
}
static void
report_exit(struct trussinfo *info, siginfo_t *si)
{
struct timespec timediff;
if (info->flags & FOLLOWFORKS)
fprintf(info->outfile, "%5d: ", si->si_pid);
clock_gettime(CLOCK_REALTIME, &info->curthread->after);
if (info->flags & ABSOLUTETIMESTAMPS) {
timespecsubt(&info->curthread->after, &info->start_time,
&timediff);
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
if (info->flags & RELATIVETIMESTAMPS) {
timespecsubt(&info->curthread->after, &info->curthread->before,
&timediff);
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
if (si->si_code == CLD_EXITED)
fprintf(info->outfile, "process exit, rval = %u\n",
si->si_status);
else
fprintf(info->outfile, "process killed, signal = %u%s\n",
si->si_status, si->si_code == CLD_DUMPED ?
" (core dumped)" : "");
}
static void
report_new_child(struct trussinfo *info, pid_t pid)
{
struct timespec timediff;
clock_gettime(CLOCK_REALTIME, &info->curthread->after);
assert(info->flags & FOLLOWFORKS);
fprintf(info->outfile, "%5d: ", pid);
if (info->flags & ABSOLUTETIMESTAMPS) {
timespecsubt(&info->curthread->after, &info->start_time,
&timediff);
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
if (info->flags & RELATIVETIMESTAMPS) {
timediff.tv_sec = 0;
timediff.tv_nsec = 0;
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
fprintf(info->outfile, "<new process>\n");
}
static void
report_signal(struct trussinfo *info, siginfo_t *si)
{
struct timespec timediff;
char *signame;
if (info->flags & FOLLOWFORKS)
fprintf(info->outfile, "%5d: ", si->si_pid);
if (info->flags & ABSOLUTETIMESTAMPS) {
timespecsubt(&info->curthread->after, &info->start_time,
&timediff);
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
if (info->flags & RELATIVETIMESTAMPS) {
timespecsubt(&info->curthread->after, &info->curthread->before,
&timediff);
fprintf(info->outfile, "%jd.%09ld ", (intmax_t)timediff.tv_sec,
timediff.tv_nsec);
}
signame = strsig(si->si_status);
fprintf(info->outfile, "SIGNAL %u (%s)\n", si->si_status,
signame == NULL ? "?" : signame);
}
/*
* Wait for events until all the processes have exited or truss has been
* asked to stop.
*/
void
eventloop(struct trussinfo *info)
{
struct ptrace_lwpinfo pl;
siginfo_t si;
int pending_signal;
while (!LIST_EMPTY(&info->proclist)) {
if (detaching) {
detach_all_procs(info);
return;
}
if (waitid(P_ALL, 0, &si, WTRAPPED | WEXITED) == -1) {
if (errno == EINTR)
continue;
err(1, "Unexpected error from waitid");
}
assert(si.si_signo == SIGCHLD);
switch (si.si_code) {
case CLD_EXITED:
case CLD_KILLED:
case CLD_DUMPED:
find_exit_thread(info, si.si_pid);
if ((info->flags & COUNTONLY) == 0)
report_exit(info, &si);
free_proc(info->curthread->proc);
info->curthread = NULL;
break;
case CLD_TRAPPED:
if (ptrace(PT_LWPINFO, si.si_pid, (caddr_t)&pl,
sizeof(pl)) == -1)
err(1, "ptrace(PT_LWPINFO)");
if (pl.pl_flags & PL_FLAG_CHILD) {
new_proc(info, si.si_pid);
assert(LIST_FIRST(&info->proclist)->abi !=
NULL);
}
find_thread(info, si.si_pid, pl.pl_lwpid);
if (si.si_status == SIGTRAP &&
(pl.pl_flags & (PL_FLAG_SCE|PL_FLAG_SCX)) != 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
if (pl.pl_flags & PL_FLAG_SCE)
enter_syscall(info, &pl);
else if (pl.pl_flags & PL_FLAG_SCX)
exit_syscall(info, &pl);
pending_signal = 0;
} else if (pl.pl_flags & PL_FLAG_CHILD) {
if ((info->flags & COUNTONLY) == 0)
report_new_child(info, si.si_pid);
pending_signal = 0;
} else {
if ((info->flags & NOSIGS) == 0)
report_signal(info, &si);
pending_signal = si.si_status;
}
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
ptrace(PT_SYSCALL, si.si_pid, (caddr_t)1,
pending_signal);
break;
case CLD_STOPPED:
errx(1, "waitid reported CLD_STOPPED");
case CLD_CONTINUED:
break;
}
}
}