The branch stable/13 has been updated by kib: URL: https://cgit.FreeBSD.org/src/commit/?id=3f602aa873122e8c3771135e6d0ed2610ea229b1
commit 3f602aa873122e8c3771135e6d0ed2610ea229b1 Author: Konstantin Belousov <[email protected]> AuthorDate: 2021-10-04 01:29:26 +0000 Commit: Konstantin Belousov <[email protected]> CommitDate: 2021-10-16 13:01:47 +0000 i386: move signal delivery code to exec_machdep.c (cherry picked from commit 4c5bf591522c5449d017b7ea496488c42f847963) --- sys/conf/files.i386 | 1 + sys/i386/i386/exec_machdep.c | 1443 ++++++++++++++++++++++++++++++++++++++++ sys/i386/i386/machdep.c | 1391 -------------------------------------- sys/i386/i386/ptrace_machdep.c | 32 + 4 files changed, 1476 insertions(+), 1391 deletions(-) diff --git a/sys/conf/files.i386 b/sys/conf/files.i386 index 5c65ad53cea4..6a0ca49c0432 100644 --- a/sys/conf/files.i386 +++ b/sys/conf/files.i386 @@ -165,6 +165,7 @@ i386/i386/copyout.c standard i386/i386/db_disasm.c optional ddb i386/i386/db_interface.c optional ddb i386/i386/db_trace.c optional ddb +i386/i386/exec_machdep.c standard i386/i386/elan-mmcr.c optional cpu_elan | cpu_soekris i386/i386/elf_machdep.c standard i386/i386/exception.s standard diff --git a/sys/i386/i386/exec_machdep.c b/sys/i386/i386/exec_machdep.c new file mode 100644 index 000000000000..13628aba7ab1 --- /dev/null +++ b/sys/i386/i386/exec_machdep.c @@ -0,0 +1,1443 @@ +/*- + * SPDX-License-Identifier: BSD-4-Clause + * + * Copyright (c) 2018 The FreeBSD Foundation + * Copyright (c) 1992 Terrence R. Lambert. + * Copyright (c) 1982, 1987, 1990 The Regents of the University of California. + * All rights reserved. + * + * This code is derived from software contributed to Berkeley by + * William Jolitz. + * + * Portions of this software were developed by A. Joseph Koshy under + * sponsorship from the FreeBSD Foundation and Google, Inc. + * + * 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 the University of + * California, Berkeley and its contributors. + * 4. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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. + * + * from: @(#)machdep.c 7.4 (Berkeley) 6/3/91 + */ + +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); + +#include "opt_cpu.h" +#include "opt_ddb.h" +#include "opt_kstack_pages.h" + +#include <sys/param.h> +#include <sys/proc.h> +#include <sys/systm.h> +#include <sys/exec.h> +#include <sys/imgact.h> +#include <sys/kdb.h> +#include <sys/kernel.h> +#include <sys/ktr.h> +#include <sys/linker.h> +#include <sys/lock.h> +#include <sys/malloc.h> +#include <sys/mutex.h> +#include <sys/pcpu.h> +#include <sys/ptrace.h> +#include <sys/rwlock.h> +#include <sys/signalvar.h> +#include <sys/syscallsubr.h> +#include <sys/sysctl.h> +#include <sys/sysent.h> +#include <sys/sysproto.h> +#include <sys/ucontext.h> +#include <sys/vmmeter.h> + +#include <vm/vm.h> +#include <vm/vm_param.h> +#include <vm/vm_extern.h> +#include <vm/vm_kern.h> +#include <vm/vm_page.h> +#include <vm/vm_map.h> +#include <vm/vm_object.h> + +#ifdef DDB +#ifndef KDB +#error KDB must be enabled in order for DDB to work! +#endif +#include <ddb/ddb.h> +#include <ddb/db_sym.h> +#endif + +#include <machine/cpu.h> +#include <machine/cputypes.h> +#include <machine/md_var.h> +#include <machine/pcb.h> +#include <machine/pcb_ext.h> +#include <machine/proc.h> +#include <machine/reg.h> +#include <machine/sigframe.h> +#include <machine/specialreg.h> +#include <machine/sysarch.h> +#include <machine/trap.h> + +static void fpstate_drop(struct thread *td); +static void get_fpcontext(struct thread *td, mcontext_t *mcp, + char *xfpusave, size_t xfpusave_len); +static int set_fpcontext(struct thread *td, mcontext_t *mcp, + char *xfpustate, size_t xfpustate_len); +#ifdef COMPAT_43 +static void osendsig(sig_t catcher, ksiginfo_t *, sigset_t *mask); +#endif +#ifdef COMPAT_FREEBSD4 +static void freebsd4_sendsig(sig_t catcher, ksiginfo_t *, sigset_t *mask); +#endif + +extern struct sysentvec elf32_freebsd_sysvec; + +/* + * Send an interrupt to process. + * + * Stack is set up to allow sigcode stored at top to call routine, + * followed by call to sigreturn routine below. After sigreturn + * resets the signal mask, the stack, and the frame pointer, it + * returns to the user specified pc, psl. + */ +#ifdef COMPAT_43 +static void +osendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) +{ + struct osigframe sf, *fp; + struct proc *p; + struct thread *td; + struct sigacts *psp; + struct trapframe *regs; + int sig; + int oonstack; + + td = curthread; + p = td->td_proc; + PROC_LOCK_ASSERT(p, MA_OWNED); + sig = ksi->ksi_signo; + psp = p->p_sigacts; + mtx_assert(&psp->ps_mtx, MA_OWNED); + regs = td->td_frame; + oonstack = sigonstack(regs->tf_esp); + + /* Allocate space for the signal handler context. */ + if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && + SIGISMEMBER(psp->ps_sigonstack, sig)) { + fp = (struct osigframe *)((uintptr_t)td->td_sigstk.ss_sp + + td->td_sigstk.ss_size - sizeof(struct osigframe)); +#if defined(COMPAT_43) + td->td_sigstk.ss_flags |= SS_ONSTACK; +#endif + } else + fp = (struct osigframe *)regs->tf_esp - 1; + + /* Build the argument list for the signal handler. */ + sf.sf_signum = sig; + sf.sf_scp = (register_t)&fp->sf_siginfo.si_sc; + bzero(&sf.sf_siginfo, sizeof(sf.sf_siginfo)); + if (SIGISMEMBER(psp->ps_siginfo, sig)) { + /* Signal handler installed with SA_SIGINFO. */ + sf.sf_arg2 = (register_t)&fp->sf_siginfo; + sf.sf_siginfo.si_signo = sig; + sf.sf_siginfo.si_code = ksi->ksi_code; + sf.sf_ahu.sf_action = (__osiginfohandler_t *)catcher; + sf.sf_addr = 0; + } else { + /* Old FreeBSD-style arguments. */ + sf.sf_arg2 = ksi->ksi_code; + sf.sf_addr = (register_t)ksi->ksi_addr; + sf.sf_ahu.sf_handler = catcher; + } + mtx_unlock(&psp->ps_mtx); + PROC_UNLOCK(p); + + /* Save most if not all of trap frame. */ + sf.sf_siginfo.si_sc.sc_eax = regs->tf_eax; + sf.sf_siginfo.si_sc.sc_ebx = regs->tf_ebx; + sf.sf_siginfo.si_sc.sc_ecx = regs->tf_ecx; + sf.sf_siginfo.si_sc.sc_edx = regs->tf_edx; + sf.sf_siginfo.si_sc.sc_esi = regs->tf_esi; + sf.sf_siginfo.si_sc.sc_edi = regs->tf_edi; + sf.sf_siginfo.si_sc.sc_cs = regs->tf_cs; + sf.sf_siginfo.si_sc.sc_ds = regs->tf_ds; + sf.sf_siginfo.si_sc.sc_ss = regs->tf_ss; + sf.sf_siginfo.si_sc.sc_es = regs->tf_es; + sf.sf_siginfo.si_sc.sc_fs = regs->tf_fs; + sf.sf_siginfo.si_sc.sc_gs = rgs(); + sf.sf_siginfo.si_sc.sc_isp = regs->tf_isp; + + /* Build the signal context to be used by osigreturn(). */ + sf.sf_siginfo.si_sc.sc_onstack = (oonstack) ? 1 : 0; + SIG2OSIG(*mask, sf.sf_siginfo.si_sc.sc_mask); + sf.sf_siginfo.si_sc.sc_sp = regs->tf_esp; + sf.sf_siginfo.si_sc.sc_fp = regs->tf_ebp; + sf.sf_siginfo.si_sc.sc_pc = regs->tf_eip; + sf.sf_siginfo.si_sc.sc_ps = regs->tf_eflags; + sf.sf_siginfo.si_sc.sc_trapno = regs->tf_trapno; + sf.sf_siginfo.si_sc.sc_err = regs->tf_err; + + /* + * If we're a vm86 process, we want to save the segment registers. + * We also change eflags to be our emulated eflags, not the actual + * eflags. + */ + if (regs->tf_eflags & PSL_VM) { + /* XXX confusing names: `tf' isn't a trapframe; `regs' is. */ + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86 = &td->td_pcb->pcb_ext->ext_vm86; + + sf.sf_siginfo.si_sc.sc_gs = tf->tf_vm86_gs; + sf.sf_siginfo.si_sc.sc_fs = tf->tf_vm86_fs; + sf.sf_siginfo.si_sc.sc_es = tf->tf_vm86_es; + sf.sf_siginfo.si_sc.sc_ds = tf->tf_vm86_ds; + + if (vm86->vm86_has_vme == 0) + sf.sf_siginfo.si_sc.sc_ps = + (tf->tf_eflags & ~(PSL_VIF | PSL_VIP)) | + (vm86->vm86_eflags & (PSL_VIF | PSL_VIP)); + + /* See sendsig() for comments. */ + tf->tf_eflags &= ~(PSL_VM | PSL_NT | PSL_VIF | PSL_VIP); + } + + /* + * Copy the sigframe out to the user's stack. + */ + if (copyout(&sf, fp, sizeof(*fp)) != 0) { + PROC_LOCK(p); + sigexit(td, SIGILL); + } + + regs->tf_esp = (int)fp; + if (p->p_sysent->sv_sigcode_base != 0) { + regs->tf_eip = p->p_sysent->sv_sigcode_base + szsigcode - + szosigcode; + } else { + /* a.out sysentvec does not use shared page */ + regs->tf_eip = p->p_sysent->sv_psstrings - szosigcode; + } + regs->tf_eflags &= ~(PSL_T | PSL_D); + regs->tf_cs = _ucodesel; + regs->tf_ds = _udatasel; + regs->tf_es = _udatasel; + regs->tf_fs = _udatasel; + load_gs(_udatasel); + regs->tf_ss = _udatasel; + PROC_LOCK(p); + mtx_lock(&psp->ps_mtx); +} +#endif /* COMPAT_43 */ + +#ifdef COMPAT_FREEBSD4 +static void +freebsd4_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) +{ + struct sigframe4 sf, *sfp; + struct proc *p; + struct thread *td; + struct sigacts *psp; + struct trapframe *regs; + int sig; + int oonstack; + + td = curthread; + p = td->td_proc; + PROC_LOCK_ASSERT(p, MA_OWNED); + sig = ksi->ksi_signo; + psp = p->p_sigacts; + mtx_assert(&psp->ps_mtx, MA_OWNED); + regs = td->td_frame; + oonstack = sigonstack(regs->tf_esp); + + /* Save user context. */ + bzero(&sf, sizeof(sf)); + sf.sf_uc.uc_sigmask = *mask; + sf.sf_uc.uc_stack = td->td_sigstk; + sf.sf_uc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) + ? ((oonstack) ? SS_ONSTACK : 0) : SS_DISABLE; + sf.sf_uc.uc_mcontext.mc_onstack = (oonstack) ? 1 : 0; + sf.sf_uc.uc_mcontext.mc_gs = rgs(); + bcopy(regs, &sf.sf_uc.uc_mcontext.mc_fs, sizeof(*regs)); + bzero(sf.sf_uc.uc_mcontext.mc_fpregs, + sizeof(sf.sf_uc.uc_mcontext.mc_fpregs)); + bzero(sf.sf_uc.uc_mcontext.__spare__, + sizeof(sf.sf_uc.uc_mcontext.__spare__)); + bzero(sf.sf_uc.__spare__, sizeof(sf.sf_uc.__spare__)); + + /* Allocate space for the signal handler context. */ + if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack && + SIGISMEMBER(psp->ps_sigonstack, sig)) { + sfp = (struct sigframe4 *)((uintptr_t)td->td_sigstk.ss_sp + + td->td_sigstk.ss_size - sizeof(struct sigframe4)); +#if defined(COMPAT_43) + td->td_sigstk.ss_flags |= SS_ONSTACK; +#endif + } else + sfp = (struct sigframe4 *)regs->tf_esp - 1; + + /* Build the argument list for the signal handler. */ + sf.sf_signum = sig; + sf.sf_ucontext = (register_t)&sfp->sf_uc; + bzero(&sf.sf_si, sizeof(sf.sf_si)); + if (SIGISMEMBER(psp->ps_siginfo, sig)) { + /* Signal handler installed with SA_SIGINFO. */ + sf.sf_siginfo = (register_t)&sfp->sf_si; + sf.sf_ahu.sf_action = (__siginfohandler_t *)catcher; + + /* Fill in POSIX parts */ + sf.sf_si.si_signo = sig; + sf.sf_si.si_code = ksi->ksi_code; + sf.sf_si.si_addr = ksi->ksi_addr; + } else { + /* Old FreeBSD-style arguments. */ + sf.sf_siginfo = ksi->ksi_code; + sf.sf_addr = (register_t)ksi->ksi_addr; + sf.sf_ahu.sf_handler = catcher; + } + mtx_unlock(&psp->ps_mtx); + PROC_UNLOCK(p); + + /* + * If we're a vm86 process, we want to save the segment registers. + * We also change eflags to be our emulated eflags, not the actual + * eflags. + */ + if (regs->tf_eflags & PSL_VM) { + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86 = &td->td_pcb->pcb_ext->ext_vm86; + + sf.sf_uc.uc_mcontext.mc_gs = tf->tf_vm86_gs; + sf.sf_uc.uc_mcontext.mc_fs = tf->tf_vm86_fs; + sf.sf_uc.uc_mcontext.mc_es = tf->tf_vm86_es; + sf.sf_uc.uc_mcontext.mc_ds = tf->tf_vm86_ds; + + if (vm86->vm86_has_vme == 0) + sf.sf_uc.uc_mcontext.mc_eflags = + (tf->tf_eflags & ~(PSL_VIF | PSL_VIP)) | + (vm86->vm86_eflags & (PSL_VIF | PSL_VIP)); + + /* + * Clear PSL_NT to inhibit T_TSSFLT faults on return from + * syscalls made by the signal handler. This just avoids + * wasting time for our lazy fixup of such faults. PSL_NT + * does nothing in vm86 mode, but vm86 programs can set it + * almost legitimately in probes for old cpu types. + */ + tf->tf_eflags &= ~(PSL_VM | PSL_NT | PSL_VIF | PSL_VIP); + } + + /* + * Copy the sigframe out to the user's stack. + */ + if (copyout(&sf, sfp, sizeof(*sfp)) != 0) { + PROC_LOCK(p); + sigexit(td, SIGILL); + } + + regs->tf_esp = (int)sfp; + regs->tf_eip = p->p_sysent->sv_sigcode_base + szsigcode - + szfreebsd4_sigcode; + regs->tf_eflags &= ~(PSL_T | PSL_D); + regs->tf_cs = _ucodesel; + regs->tf_ds = _udatasel; + regs->tf_es = _udatasel; + regs->tf_fs = _udatasel; + regs->tf_ss = _udatasel; + PROC_LOCK(p); + mtx_lock(&psp->ps_mtx); +} +#endif /* COMPAT_FREEBSD4 */ + +void +sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) +{ + struct sigframe sf, *sfp; + struct proc *p; + struct thread *td; + struct sigacts *psp; + char *sp; + struct trapframe *regs; + struct segment_descriptor *sdp; + char *xfpusave; + size_t xfpusave_len; + int sig; + int oonstack; + + td = curthread; + p = td->td_proc; + PROC_LOCK_ASSERT(p, MA_OWNED); + sig = ksi->ksi_signo; + psp = p->p_sigacts; + mtx_assert(&psp->ps_mtx, MA_OWNED); +#ifdef COMPAT_FREEBSD4 + if (SIGISMEMBER(psp->ps_freebsd4, sig)) { + freebsd4_sendsig(catcher, ksi, mask); + return; + } +#endif +#ifdef COMPAT_43 + if (SIGISMEMBER(psp->ps_osigset, sig)) { + osendsig(catcher, ksi, mask); + return; + } +#endif + regs = td->td_frame; + oonstack = sigonstack(regs->tf_esp); + + if (cpu_max_ext_state_size > sizeof(union savefpu) && use_xsave) { + xfpusave_len = cpu_max_ext_state_size - sizeof(union savefpu); + xfpusave = __builtin_alloca(xfpusave_len); + } else { + xfpusave_len = 0; + xfpusave = NULL; + } + + /* Save user context. */ + bzero(&sf, sizeof(sf)); + sf.sf_uc.uc_sigmask = *mask; + sf.sf_uc.uc_stack = td->td_sigstk; + sf.sf_uc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) + ? ((oonstack) ? SS_ONSTACK : 0) : SS_DISABLE; + sf.sf_uc.uc_mcontext.mc_onstack = (oonstack) ? 1 : 0; + sf.sf_uc.uc_mcontext.mc_gs = rgs(); + bcopy(regs, &sf.sf_uc.uc_mcontext.mc_fs, sizeof(*regs)); + sf.sf_uc.uc_mcontext.mc_len = sizeof(sf.sf_uc.uc_mcontext); /* magic */ + get_fpcontext(td, &sf.sf_uc.uc_mcontext, xfpusave, xfpusave_len); + fpstate_drop(td); + /* + * Unconditionally fill the fsbase and gsbase into the mcontext. + */ + sdp = &td->td_pcb->pcb_fsd; + sf.sf_uc.uc_mcontext.mc_fsbase = sdp->sd_hibase << 24 | + sdp->sd_lobase; + sdp = &td->td_pcb->pcb_gsd; + sf.sf_uc.uc_mcontext.mc_gsbase = sdp->sd_hibase << 24 | + sdp->sd_lobase; + bzero(sf.sf_uc.uc_mcontext.mc_spare2, + sizeof(sf.sf_uc.uc_mcontext.mc_spare2)); + + /* Allocate space for the signal handler context. */ + if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack && + SIGISMEMBER(psp->ps_sigonstack, sig)) { + sp = (char *)td->td_sigstk.ss_sp + td->td_sigstk.ss_size; +#if defined(COMPAT_43) + td->td_sigstk.ss_flags |= SS_ONSTACK; +#endif + } else + sp = (char *)regs->tf_esp - 128; + if (xfpusave != NULL) { + sp -= xfpusave_len; + sp = (char *)((unsigned int)sp & ~0x3F); + sf.sf_uc.uc_mcontext.mc_xfpustate = (register_t)sp; + } + sp -= sizeof(struct sigframe); + + /* Align to 16 bytes. */ + sfp = (struct sigframe *)((unsigned int)sp & ~0xF); + + /* Build the argument list for the signal handler. */ + sf.sf_signum = sig; + sf.sf_ucontext = (register_t)&sfp->sf_uc; + bzero(&sf.sf_si, sizeof(sf.sf_si)); + if (SIGISMEMBER(psp->ps_siginfo, sig)) { + /* Signal handler installed with SA_SIGINFO. */ + sf.sf_siginfo = (register_t)&sfp->sf_si; + sf.sf_ahu.sf_action = (__siginfohandler_t *)catcher; + + /* Fill in POSIX parts */ + sf.sf_si = ksi->ksi_info; + sf.sf_si.si_signo = sig; /* maybe a translated signal */ + } else { + /* Old FreeBSD-style arguments. */ + sf.sf_siginfo = ksi->ksi_code; + sf.sf_addr = (register_t)ksi->ksi_addr; + sf.sf_ahu.sf_handler = catcher; + } + mtx_unlock(&psp->ps_mtx); + PROC_UNLOCK(p); + + /* + * If we're a vm86 process, we want to save the segment registers. + * We also change eflags to be our emulated eflags, not the actual + * eflags. + */ + if (regs->tf_eflags & PSL_VM) { + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86 = &td->td_pcb->pcb_ext->ext_vm86; + + sf.sf_uc.uc_mcontext.mc_gs = tf->tf_vm86_gs; + sf.sf_uc.uc_mcontext.mc_fs = tf->tf_vm86_fs; + sf.sf_uc.uc_mcontext.mc_es = tf->tf_vm86_es; + sf.sf_uc.uc_mcontext.mc_ds = tf->tf_vm86_ds; + + if (vm86->vm86_has_vme == 0) + sf.sf_uc.uc_mcontext.mc_eflags = + (tf->tf_eflags & ~(PSL_VIF | PSL_VIP)) | + (vm86->vm86_eflags & (PSL_VIF | PSL_VIP)); + + /* + * Clear PSL_NT to inhibit T_TSSFLT faults on return from + * syscalls made by the signal handler. This just avoids + * wasting time for our lazy fixup of such faults. PSL_NT + * does nothing in vm86 mode, but vm86 programs can set it + * almost legitimately in probes for old cpu types. + */ + tf->tf_eflags &= ~(PSL_VM | PSL_NT | PSL_VIF | PSL_VIP); + } + + /* + * Copy the sigframe out to the user's stack. + */ + if (copyout(&sf, sfp, sizeof(*sfp)) != 0 || + (xfpusave != NULL && copyout(xfpusave, + (void *)sf.sf_uc.uc_mcontext.mc_xfpustate, xfpusave_len) + != 0)) { + PROC_LOCK(p); + sigexit(td, SIGILL); + } + + regs->tf_esp = (int)sfp; + regs->tf_eip = p->p_sysent->sv_sigcode_base; + if (regs->tf_eip == 0) + regs->tf_eip = p->p_sysent->sv_psstrings - szsigcode; + regs->tf_eflags &= ~(PSL_T | PSL_D); + regs->tf_cs = _ucodesel; + regs->tf_ds = _udatasel; + regs->tf_es = _udatasel; + regs->tf_fs = _udatasel; + regs->tf_ss = _udatasel; + PROC_LOCK(p); + mtx_lock(&psp->ps_mtx); +} + +/* + * System call to cleanup state after a signal has been taken. Reset + * signal mask and stack state from context left by sendsig (above). + * Return to previous pc and psl as specified by context left by + * sendsig. Check carefully to make sure that the user has not + * modified the state to gain improper privileges. + */ +#ifdef COMPAT_43 +int +osigreturn(struct thread *td, struct osigreturn_args *uap) +{ + struct osigcontext sc; + struct trapframe *regs; + struct osigcontext *scp; + int eflags, error; + ksiginfo_t ksi; + + regs = td->td_frame; + error = copyin(uap->sigcntxp, &sc, sizeof(sc)); + if (error != 0) + return (error); + scp = ≻ + eflags = scp->sc_ps; + if (eflags & PSL_VM) { + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86; + + /* + * if pcb_ext == 0 or vm86_inited == 0, the user hasn't + * set up the vm86 area, and we can't enter vm86 mode. + */ + if (td->td_pcb->pcb_ext == 0) + return (EINVAL); + vm86 = &td->td_pcb->pcb_ext->ext_vm86; + if (vm86->vm86_inited == 0) + return (EINVAL); + + /* Go back to user mode if both flags are set. */ + if ((eflags & PSL_VIP) && (eflags & PSL_VIF)) { + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + } + + if (vm86->vm86_has_vme) { + eflags = (tf->tf_eflags & ~VME_USERCHANGE) | + (eflags & VME_USERCHANGE) | PSL_VM; + } else { + vm86->vm86_eflags = eflags; /* save VIF, VIP */ + eflags = (tf->tf_eflags & ~VM_USERCHANGE) | + (eflags & VM_USERCHANGE) | PSL_VM; + } + tf->tf_vm86_ds = scp->sc_ds; + tf->tf_vm86_es = scp->sc_es; + tf->tf_vm86_fs = scp->sc_fs; + tf->tf_vm86_gs = scp->sc_gs; + tf->tf_ds = _udatasel; + tf->tf_es = _udatasel; + tf->tf_fs = _udatasel; + } else { + /* + * Don't allow users to change privileged or reserved flags. + */ + if (!EFL_SECURE(eflags, regs->tf_eflags)) { + return (EINVAL); + } + + /* + * Don't allow users to load a valid privileged %cs. Let the + * hardware check for invalid selectors, excess privilege in + * other selectors, invalid %eip's and invalid %esp's. + */ + if (!CS_SECURE(scp->sc_cs)) { + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_trapno = T_PROTFLT; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + return (EINVAL); + } + regs->tf_ds = scp->sc_ds; + regs->tf_es = scp->sc_es; + regs->tf_fs = scp->sc_fs; + } + + /* Restore remaining registers. */ + regs->tf_eax = scp->sc_eax; + regs->tf_ebx = scp->sc_ebx; + regs->tf_ecx = scp->sc_ecx; + regs->tf_edx = scp->sc_edx; + regs->tf_esi = scp->sc_esi; + regs->tf_edi = scp->sc_edi; + regs->tf_cs = scp->sc_cs; + regs->tf_ss = scp->sc_ss; + regs->tf_isp = scp->sc_isp; + regs->tf_ebp = scp->sc_fp; + regs->tf_esp = scp->sc_sp; + regs->tf_eip = scp->sc_pc; + regs->tf_eflags = eflags; + +#if defined(COMPAT_43) + if (scp->sc_onstack & 1) + td->td_sigstk.ss_flags |= SS_ONSTACK; + else + td->td_sigstk.ss_flags &= ~SS_ONSTACK; +#endif + kern_sigprocmask(td, SIG_SETMASK, (sigset_t *)&scp->sc_mask, NULL, + SIGPROCMASK_OLD); + return (EJUSTRETURN); +} +#endif /* COMPAT_43 */ + +#ifdef COMPAT_FREEBSD4 +int +freebsd4_sigreturn(struct thread *td, struct freebsd4_sigreturn_args *uap) +{ + struct ucontext4 uc; + struct trapframe *regs; + struct ucontext4 *ucp; + int cs, eflags, error; + ksiginfo_t ksi; + + error = copyin(uap->sigcntxp, &uc, sizeof(uc)); + if (error != 0) + return (error); + ucp = &uc; + regs = td->td_frame; + eflags = ucp->uc_mcontext.mc_eflags; + if (eflags & PSL_VM) { + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86; + + /* + * if pcb_ext == 0 or vm86_inited == 0, the user hasn't + * set up the vm86 area, and we can't enter vm86 mode. + */ + if (td->td_pcb->pcb_ext == 0) + return (EINVAL); + vm86 = &td->td_pcb->pcb_ext->ext_vm86; + if (vm86->vm86_inited == 0) + return (EINVAL); + + /* Go back to user mode if both flags are set. */ + if ((eflags & PSL_VIP) && (eflags & PSL_VIF)) { + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + } + if (vm86->vm86_has_vme) { + eflags = (tf->tf_eflags & ~VME_USERCHANGE) | + (eflags & VME_USERCHANGE) | PSL_VM; + } else { + vm86->vm86_eflags = eflags; /* save VIF, VIP */ + eflags = (tf->tf_eflags & ~VM_USERCHANGE) | + (eflags & VM_USERCHANGE) | PSL_VM; + } + bcopy(&ucp->uc_mcontext.mc_fs, tf, sizeof(struct trapframe)); + tf->tf_eflags = eflags; + tf->tf_vm86_ds = tf->tf_ds; + tf->tf_vm86_es = tf->tf_es; + tf->tf_vm86_fs = tf->tf_fs; + tf->tf_vm86_gs = ucp->uc_mcontext.mc_gs; + tf->tf_ds = _udatasel; + tf->tf_es = _udatasel; + tf->tf_fs = _udatasel; + } else { + /* + * Don't allow users to change privileged or reserved flags. + */ + if (!EFL_SECURE(eflags, regs->tf_eflags)) { + uprintf( + "pid %d (%s): freebsd4_sigreturn eflags = 0x%x\n", + td->td_proc->p_pid, td->td_name, eflags); + return (EINVAL); + } + + /* + * Don't allow users to load a valid privileged %cs. Let the + * hardware check for invalid selectors, excess privilege in + * other selectors, invalid %eip's and invalid %esp's. + */ + cs = ucp->uc_mcontext.mc_cs; + if (!CS_SECURE(cs)) { + uprintf("pid %d (%s): freebsd4_sigreturn cs = 0x%x\n", + td->td_proc->p_pid, td->td_name, cs); + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_trapno = T_PROTFLT; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + return (EINVAL); + } + + bcopy(&ucp->uc_mcontext.mc_fs, regs, sizeof(*regs)); + } + +#if defined(COMPAT_43) + if (ucp->uc_mcontext.mc_onstack & 1) + td->td_sigstk.ss_flags |= SS_ONSTACK; + else + td->td_sigstk.ss_flags &= ~SS_ONSTACK; +#endif + kern_sigprocmask(td, SIG_SETMASK, &ucp->uc_sigmask, NULL, 0); + return (EJUSTRETURN); +} +#endif /* COMPAT_FREEBSD4 */ + +int +sys_sigreturn(struct thread *td, struct sigreturn_args *uap) +{ + ucontext_t uc; + struct proc *p; + struct trapframe *regs; + ucontext_t *ucp; + char *xfpustate; + size_t xfpustate_len; + int cs, eflags, error, ret; + ksiginfo_t ksi; + + p = td->td_proc; + + error = copyin(uap->sigcntxp, &uc, sizeof(uc)); + if (error != 0) + return (error); + ucp = &uc; + if ((ucp->uc_mcontext.mc_flags & ~_MC_FLAG_MASK) != 0) { + uprintf("pid %d (%s): sigreturn mc_flags %x\n", p->p_pid, + td->td_name, ucp->uc_mcontext.mc_flags); + return (EINVAL); + } + regs = td->td_frame; + eflags = ucp->uc_mcontext.mc_eflags; + if (eflags & PSL_VM) { + struct trapframe_vm86 *tf = (struct trapframe_vm86 *)regs; + struct vm86_kernel *vm86; + + /* + * if pcb_ext == 0 or vm86_inited == 0, the user hasn't + * set up the vm86 area, and we can't enter vm86 mode. + */ + if (td->td_pcb->pcb_ext == 0) + return (EINVAL); + vm86 = &td->td_pcb->pcb_ext->ext_vm86; + if (vm86->vm86_inited == 0) + return (EINVAL); + + /* Go back to user mode if both flags are set. */ + if ((eflags & PSL_VIP) && (eflags & PSL_VIF)) { + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + } + + if (vm86->vm86_has_vme) { + eflags = (tf->tf_eflags & ~VME_USERCHANGE) | + (eflags & VME_USERCHANGE) | PSL_VM; + } else { + vm86->vm86_eflags = eflags; /* save VIF, VIP */ + eflags = (tf->tf_eflags & ~VM_USERCHANGE) | + (eflags & VM_USERCHANGE) | PSL_VM; + } + bcopy(&ucp->uc_mcontext.mc_fs, tf, sizeof(struct trapframe)); + tf->tf_eflags = eflags; + tf->tf_vm86_ds = tf->tf_ds; + tf->tf_vm86_es = tf->tf_es; + tf->tf_vm86_fs = tf->tf_fs; + tf->tf_vm86_gs = ucp->uc_mcontext.mc_gs; + tf->tf_ds = _udatasel; + tf->tf_es = _udatasel; + tf->tf_fs = _udatasel; + } else { + /* + * Don't allow users to change privileged or reserved flags. + */ + if (!EFL_SECURE(eflags, regs->tf_eflags)) { + uprintf("pid %d (%s): sigreturn eflags = 0x%x\n", + td->td_proc->p_pid, td->td_name, eflags); + return (EINVAL); + } + + /* + * Don't allow users to load a valid privileged %cs. Let the + * hardware check for invalid selectors, excess privilege in + * other selectors, invalid %eip's and invalid %esp's. + */ + cs = ucp->uc_mcontext.mc_cs; + if (!CS_SECURE(cs)) { + uprintf("pid %d (%s): sigreturn cs = 0x%x\n", + td->td_proc->p_pid, td->td_name, cs); + ksiginfo_init_trap(&ksi); + ksi.ksi_signo = SIGBUS; + ksi.ksi_code = BUS_OBJERR; + ksi.ksi_trapno = T_PROTFLT; + ksi.ksi_addr = (void *)regs->tf_eip; + trapsignal(td, &ksi); + return (EINVAL); + } + + if ((uc.uc_mcontext.mc_flags & _MC_HASFPXSTATE) != 0) { + xfpustate_len = uc.uc_mcontext.mc_xfpustate_len; + if (xfpustate_len > cpu_max_ext_state_size - + sizeof(union savefpu)) { + uprintf( + "pid %d (%s): sigreturn xfpusave_len = 0x%zx\n", + p->p_pid, td->td_name, xfpustate_len); + return (EINVAL); + } + xfpustate = __builtin_alloca(xfpustate_len); + error = copyin( + (const void *)uc.uc_mcontext.mc_xfpustate, + xfpustate, xfpustate_len); + if (error != 0) { + uprintf( + "pid %d (%s): sigreturn copying xfpustate failed\n", + p->p_pid, td->td_name); + return (error); + } + } else { + xfpustate = NULL; + xfpustate_len = 0; + } + ret = set_fpcontext(td, &ucp->uc_mcontext, xfpustate, + xfpustate_len); + if (ret != 0) + return (ret); + bcopy(&ucp->uc_mcontext.mc_fs, regs, sizeof(*regs)); + } + +#if defined(COMPAT_43) + if (ucp->uc_mcontext.mc_onstack & 1) + td->td_sigstk.ss_flags |= SS_ONSTACK; + else + td->td_sigstk.ss_flags &= ~SS_ONSTACK; +#endif + + kern_sigprocmask(td, SIG_SETMASK, &ucp->uc_sigmask, NULL, 0); + return (EJUSTRETURN); +} + +/* + * Reset the hardware debug registers if they were in use. + * They won't have any meaning for the newly exec'd process. + */ +void +x86_clear_dbregs(struct pcb *pcb) +{ + if ((pcb->pcb_flags & PCB_DBREGS) == 0) + return; + + pcb->pcb_dr0 = 0; + pcb->pcb_dr1 = 0; + pcb->pcb_dr2 = 0; + pcb->pcb_dr3 = 0; + pcb->pcb_dr6 = 0; + pcb->pcb_dr7 = 0; + + if (pcb == curpcb) { + /* + * Clear the debug registers on the running CPU, + * otherwise they will end up affecting the next + * process we switch to. + */ + reset_dbregs(); + } + pcb->pcb_flags &= ~PCB_DBREGS; +} + +#ifdef COMPAT_43 +static void +setup_priv_lcall_gate(struct proc *p) +{ + struct i386_ldt_args uap; + union descriptor desc; + u_int lcall_addr; + + bzero(&uap, sizeof(uap)); + uap.start = 0; + uap.num = 1; + lcall_addr = p->p_sysent->sv_psstrings - sz_lcall_tramp; + bzero(&desc, sizeof(desc)); + desc.sd.sd_type = SDT_MEMERA; + desc.sd.sd_dpl = SEL_UPL; + desc.sd.sd_p = 1; + desc.sd.sd_def32 = 1; + desc.sd.sd_gran = 1; + desc.sd.sd_lolimit = 0xffff; + desc.sd.sd_hilimit = 0xf; + desc.sd.sd_lobase = lcall_addr; + desc.sd.sd_hibase = lcall_addr >> 24; + i386_set_ldt(curthread, &uap, &desc); +} +#endif + +/* + * Reset registers to default values on exec. + */ +void +exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack) +{ + struct trapframe *regs; + struct pcb *pcb; + register_t saved_eflags; + + regs = td->td_frame; + pcb = td->td_pcb; *** 1965 LINES SKIPPED ***
