2 * Copyright (c) 1994, Sean Eric Fagan
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by Sean Eric Fagan.
16 * 4. The name of the author may not be used to endorse or promote products
17 * derived from this software without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * $FreeBSD: src/sys/kern/sys_process.c,v 1.51.2.6 2003/01/08 03:06:45 kan Exp $
32 * $DragonFly: src/sys/kern/sys_process.c,v 1.30 2007/02/19 01:14:23 corecode Exp $
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/sysproto.h>
40 #include <sys/vnode.h>
41 #include <sys/ptrace.h>
47 #include <vm/vm_map.h>
48 #include <vm/vm_page.h>
51 #include <vfs/procfs/procfs.h>
53 #include <sys/thread2.h>
54 #include <sys/mplock2.h>
55 #include <sys/spinlock2.h>
57 /* use the equivalent procfs code */
60 pread (struct proc *procp, unsigned int addr, unsigned int *retval) {
65 int page_offset; /* offset into page */
66 vm_offset_t pageno; /* page number */
67 vm_map_entry_t out_entry;
72 /* Map page into kernel space */
74 map = &procp->p_vmspace->vm_map;
76 page_offset = addr - trunc_page(addr);
77 pageno = trunc_page(addr);
80 rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry,
81 &object, &pindex, &out_prot, &wired);
83 if (rv != KERN_SUCCESS)
86 vm_map_lookup_done (tmap, out_entry, 0);
88 /* Find space in kernel_map for the page we're interested in */
89 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
93 VM_PROT_ALL, VM_PROT_ALL,
97 vm_object_reference (object);
99 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
102 bcopy ((caddr_t)kva + page_offset,
103 retval, sizeof *retval);
105 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
112 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) {
117 int page_offset; /* offset into page */
118 vm_offset_t pageno; /* page number */
119 vm_map_entry_t out_entry;
123 boolean_t fix_prot = 0;
125 /* Map page into kernel space */
127 map = &procp->p_vmspace->vm_map;
129 page_offset = addr - trunc_page(addr);
130 pageno = trunc_page(addr);
133 * Check the permissions for the area we're interested in.
136 if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE,
137 VM_PROT_WRITE) == FALSE) {
139 * If the page was not writable, we make it so.
140 * XXX It is possible a page may *not* be read/executable,
141 * if a process changes that!
144 /* The page isn't writable, so let's try making it so... */
145 if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE,
146 VM_PROT_ALL, 0)) != KERN_SUCCESS)
147 return EFAULT; /* I guess... */
151 * Now we need to get the page. out_entry, out_prot, wired, and
152 * single_use aren't used. One would think the vm code would be
153 * a *bit* nicer... We use tmap because vm_map_lookup() can
154 * change the map argument.
158 rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry,
159 &object, &pindex, &out_prot, &wired);
160 if (rv != KERN_SUCCESS) {
165 * Okay, we've got the page. Let's release tmap.
168 vm_map_lookup_done (tmap, out_entry, 0);
171 * Fault the page in...
174 rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE);
175 if (rv != KERN_SUCCESS)
178 /* Find space in kernel_map for the page we're interested in */
179 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
183 VM_PROT_ALL, VM_PROT_ALL,
186 vm_object_reference (object);
188 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
190 bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum);
192 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
196 vm_map_protect (map, pageno, pageno + PAGE_SIZE,
197 VM_PROT_READ|VM_PROT_EXECUTE, 0);
203 * Process debugging system call.
208 sys_ptrace(struct ptrace_args *uap)
210 struct proc *p = curproc;
213 * XXX this obfuscation is to reduce stack usage, but the register
214 * structs may be too large to put on the stack anyway.
217 struct ptrace_io_desc piod;
234 error = copyin(uap->addr, &r.reg, sizeof r.reg);
237 error = copyin(uap->addr, &r.fpreg, sizeof r.fpreg);
241 error = copyin(uap->addr, &r.dbreg, sizeof r.dbreg);
245 error = copyin(uap->addr, &r.piod, sizeof r.piod);
254 error = kern_ptrace(p, uap->req, uap->pid, addr, uap->data,
255 &uap->sysmsg_result);
262 (void)copyout(&r.piod, uap->addr, sizeof r.piod);
265 error = copyout(&r.reg, uap->addr, sizeof r.reg);
268 error = copyout(&r.fpreg, uap->addr, sizeof r.fpreg);
272 error = copyout(&r.dbreg, uap->addr, sizeof r.dbreg);
281 kern_ptrace(struct proc *curp, int req, pid_t pid, void *addr, int data, int *res)
287 struct ptrace_io_desc *piod;
292 if (req == PT_TRACE_ME) {
295 if ((p = pfind(pid)) == NULL)
298 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred))
301 /* Can't trace a process that's currently exec'ing. */
302 if ((p->p_flag & P_INEXEC) != 0)
315 if (p->p_pid == curp->p_pid)
319 if (p->p_flag & P_TRACED)
322 if (curp->p_flag & P_TRACED)
323 for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr)
327 /* not owned by you, has done setuid (unless you're root) */
328 if ((p->p_ucred->cr_ruid != curp->p_ucred->cr_ruid) ||
329 (p->p_flag & P_SUGID)) {
330 if ((error = priv_check_cred(curp->p_ucred, PRIV_ROOT, 0)) != 0)
334 /* can't trace init when securelevel > 0 */
335 if (securelevel > 0 && p->p_pid == 1)
368 /* not being traced... */
369 if ((p->p_flag & P_TRACED) == 0)
372 /* not being traced by YOU */
373 if (p->p_pptr != curp)
376 /* not currently stopped */
377 if (p->p_stat != SSTOP ||
378 (p->p_flag & P_WAITED) == 0) {
390 lp = FIRST_LWP_IN_PROC(p);
393 * Single step fixup ala procfs
399 * Actually do the requests
406 /* set my trace flag and "owner" so it can read/write me */
407 p->p_flag |= P_TRACED;
408 p->p_oppid = p->p_pptr->p_pid;
412 /* security check done above */
413 p->p_flag |= P_TRACED;
414 p->p_oppid = p->p_pptr->p_pid;
415 if (p->p_pptr != curp)
416 proc_reparent(p, curp);
418 goto sendsig; /* in PT_CONTINUE below */
423 /* Zero means do not send any signal */
424 if (data < 0 || data > _SIG_MAXSIG)
429 if (req == PT_STEP) {
430 if ((error = ptrace_single_step (lp))) {
436 if (addr != (void *)1) {
437 if ((error = ptrace_set_pc (lp,
438 (u_long)(uintfptr_t)addr))) {
445 if (req == PT_DETACH) {
446 /* reset process parent */
447 if (p->p_oppid != p->p_pptr->p_pid) {
450 pp = pfind(p->p_oppid);
451 proc_reparent(p, pp ? pp : initproc);
454 p->p_flag &= ~(P_TRACED | P_WAITED);
457 /* should we send SIGCHLD? */
462 * Deliver or queue signal. If the process is stopped
463 * force it to be SACTIVE again.
466 if (p->p_stat == SSTOP) {
468 lp->lwp_flag |= LWP_BREAKTSLEEP;
483 * NOTE! uio_offset represents the offset in the target
484 * process. The iov is in the current process (the guy
485 * making the ptrace call) so uio_td must be the current
486 * process (though for a SYSSPACE transfer it doesn't
490 /* write = 0 set above */
491 iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp;
492 iov.iov_len = sizeof(int);
495 uio.uio_offset = (off_t)(uintptr_t)addr;
496 uio.uio_resid = sizeof(int);
497 uio.uio_segflg = UIO_SYSSPACE;
498 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
499 uio.uio_td = curthread;
500 error = procfs_domem(curp, lp, NULL, &uio);
501 if (uio.uio_resid != 0) {
503 * XXX procfs_domem() doesn't currently return ENOSPC,
504 * so I think write() can bogusly return 0.
505 * XXX what happens for short writes? We don't want
506 * to write partial data.
507 * XXX procfs_domem() returns EPERM for other invalid
508 * addresses. Convert this to EINVAL. Does this
509 * clobber returns of EPERM for other reasons?
511 if (error == 0 || error == ENOSPC || error == EPERM)
512 error = EINVAL; /* EOF */
520 * NOTE! uio_offset represents the offset in the target
521 * process. The iov is in the current process (the guy
522 * making the ptrace call) so uio_td must be the current
526 iov.iov_base = piod->piod_addr;
527 iov.iov_len = piod->piod_len;
530 uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs;
531 uio.uio_resid = piod->piod_len;
532 uio.uio_segflg = UIO_USERSPACE;
533 uio.uio_td = curthread;
534 switch (piod->piod_op) {
537 uio.uio_rw = UIO_READ;
541 uio.uio_rw = UIO_WRITE;
546 error = procfs_domem(curp, lp, NULL, &uio);
547 piod->piod_len -= uio.uio_resid;
552 goto sendsig; /* in PT_CONTINUE above */
558 #endif /* PT_SETREGS */
561 /* write = 0 above */
562 #endif /* PT_SETREGS */
563 #if defined(PT_SETREGS) || defined(PT_GETREGS)
564 if (!procfs_validregs(lp)) /* no P_SYSTEM procs please */
568 iov.iov_len = sizeof(struct reg);
572 uio.uio_resid = sizeof(struct reg);
573 uio.uio_segflg = UIO_SYSSPACE;
574 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
575 uio.uio_td = curthread;
576 return (procfs_doregs(curp, lp, NULL, &uio));
578 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */
584 #endif /* PT_SETFPREGS */
587 /* write = 0 above */
588 #endif /* PT_SETFPREGS */
589 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
590 if (!procfs_validfpregs(lp)) /* no P_SYSTEM procs please */
594 iov.iov_len = sizeof(struct fpreg);
598 uio.uio_resid = sizeof(struct fpreg);
599 uio.uio_segflg = UIO_SYSSPACE;
600 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
601 uio.uio_td = curthread;
602 return (procfs_dofpregs(curp, lp, NULL, &uio));
604 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */
610 #endif /* PT_SETDBREGS */
613 /* write = 0 above */
614 #endif /* PT_SETDBREGS */
615 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS)
616 if (!procfs_validdbregs(lp)) /* no P_SYSTEM procs please */
620 iov.iov_len = sizeof(struct dbreg);
624 uio.uio_resid = sizeof(struct dbreg);
625 uio.uio_segflg = UIO_SYSSPACE;
626 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
627 uio.uio_td = curthread;
628 return (procfs_dodbregs(curp, lp, NULL, &uio));
630 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */
640 trace_req(struct proc *p)
648 * Stop a process because of a procfs event. Stay stopped until p->p_step
649 * is cleared (cleared by PIOCCONT in procfs).
654 stopevent(struct proc *p, unsigned int event, unsigned int val)
657 * Set event info. Recheck p_stops in case we are
658 * racing a close() on procfs.
660 spin_lock_wr(&p->p_spin);
661 if ((p->p_stops & event) == 0) {
662 spin_unlock_wr(&p->p_spin);
668 tsleep_interlock(&p->p_step, 0);
669 spin_unlock_wr(&p->p_spin);
672 * Wakeup any PIOCWAITing procs and wait for p_step to
677 tsleep(&p->p_step, PINTERLOCKED, "stopevent", 0);
678 spin_lock_wr(&p->p_spin);
679 if (p->p_step == 0) {
680 spin_unlock_wr(&p->p_spin);
683 tsleep_interlock(&p->p_step, 0);
684 spin_unlock_wr(&p->p_spin);