Index: head/sys/i386/i386/vm_machdep.c =================================================================== --- head/sys/i386/i386/vm_machdep.c (revision 273990) +++ head/sys/i386/i386/vm_machdep.c (revision 273991) @@ -1,864 +1,863 @@ /*- * Copyright (c) 1982, 1986 The Regents of the University of California. * Copyright (c) 1989, 1990 William Jolitz * Copyright (c) 1994 John Dyson * All rights reserved. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department, and William Jolitz. * * 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: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$ */ #include __FBSDID("$FreeBSD$"); #include "opt_isa.h" #include "opt_npx.h" #include "opt_reset.h" #include "opt_cpu.h" #include "opt_xbox.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef CPU_ELAN #include #endif #include #include #include #include #include #include #ifdef XEN #include #endif #ifdef PC98 #include #else #include #endif #ifdef XBOX #include #endif #ifndef NSFBUFS #define NSFBUFS (512 + maxusers * 16) #endif _Static_assert(OFFSETOF_CURTHREAD == offsetof(struct pcpu, pc_curthread), "OFFSETOF_CURTHREAD does not correspond with offset of pc_curthread."); _Static_assert(OFFSETOF_CURPCB == offsetof(struct pcpu, pc_curpcb), "OFFSETOF_CURPCB does not correspond with offset of pc_curpcb."); static void cpu_reset_real(void); #ifdef SMP static void cpu_reset_proxy(void); static u_int cpu_reset_proxyid; static volatile u_int cpu_reset_proxy_active; #endif -extern int _ucodesel, _udatasel; /* * Finish a fork operation, with process p2 nearly set up. * Copy and update the pcb, set up the stack so that the child * ready to run and return to user mode. */ void cpu_fork(td1, p2, td2, flags) register struct thread *td1; register struct proc *p2; struct thread *td2; int flags; { register struct proc *p1; struct pcb *pcb2; struct mdproc *mdp2; p1 = td1->td_proc; if ((flags & RFPROC) == 0) { if ((flags & RFMEM) == 0) { /* unshare user LDT */ struct mdproc *mdp1 = &p1->p_md; struct proc_ldt *pldt, *pldt1; mtx_lock_spin(&dt_lock); if ((pldt1 = mdp1->md_ldt) != NULL && pldt1->ldt_refcnt > 1) { pldt = user_ldt_alloc(mdp1, pldt1->ldt_len); if (pldt == NULL) panic("could not copy LDT"); mdp1->md_ldt = pldt; set_user_ldt(mdp1); user_ldt_deref(pldt1); } else mtx_unlock_spin(&dt_lock); } return; } /* Ensure that td1's pcb is up to date. */ if (td1 == curthread) td1->td_pcb->pcb_gs = rgs(); #ifdef DEV_NPX critical_enter(); if (PCPU_GET(fpcurthread) == td1) npxsave(td1->td_pcb->pcb_save); critical_exit(); #endif /* Point the pcb to the top of the stack */ pcb2 = (struct pcb *)(td2->td_kstack + td2->td_kstack_pages * PAGE_SIZE) - 1; td2->td_pcb = pcb2; /* Copy td1's pcb */ bcopy(td1->td_pcb, pcb2, sizeof(*pcb2)); /* Properly initialize pcb_save */ pcb2->pcb_save = &pcb2->pcb_user_save; /* Point mdproc and then copy over td1's contents */ mdp2 = &p2->p_md; bcopy(&p1->p_md, mdp2, sizeof(*mdp2)); /* * Create a new fresh stack for the new process. * Copy the trap frame for the return to user mode as if from a * syscall. This copies most of the user mode register values. * The -16 is so we can expand the trapframe if we go to vm86. */ td2->td_frame = (struct trapframe *)((caddr_t)td2->td_pcb - 16) - 1; bcopy(td1->td_frame, td2->td_frame, sizeof(struct trapframe)); td2->td_frame->tf_eax = 0; /* Child returns zero */ td2->td_frame->tf_eflags &= ~PSL_C; /* success */ td2->td_frame->tf_edx = 1; /* * If the parent process has the trap bit set (i.e. a debugger had * single stepped the process to the system call), we need to clear * the trap flag from the new frame unless the debugger had set PF_FORK * on the parent. Otherwise, the child will receive a (likely * unexpected) SIGTRAP when it executes the first instruction after * returning to userland. */ if ((p1->p_pfsflags & PF_FORK) == 0) td2->td_frame->tf_eflags &= ~PSL_T; /* * Set registers for trampoline to user mode. Leave space for the * return address on stack. These are the kernel mode register values. */ #ifdef PAE pcb2->pcb_cr3 = vtophys(vmspace_pmap(p2->p_vmspace)->pm_pdpt); #else pcb2->pcb_cr3 = vtophys(vmspace_pmap(p2->p_vmspace)->pm_pdir); #endif pcb2->pcb_edi = 0; pcb2->pcb_esi = (int)fork_return; /* fork_trampoline argument */ pcb2->pcb_ebp = 0; pcb2->pcb_esp = (int)td2->td_frame - sizeof(void *); pcb2->pcb_ebx = (int)td2; /* fork_trampoline argument */ pcb2->pcb_eip = (int)fork_trampoline; pcb2->pcb_psl = PSL_KERNEL; /* ints disabled */ /*- * pcb2->pcb_dr*: cloned above. * pcb2->pcb_savefpu: cloned above. * pcb2->pcb_flags: cloned above. * pcb2->pcb_onfault: cloned above (always NULL here?). * pcb2->pcb_gs: cloned above. * pcb2->pcb_ext: cleared below. */ /* * XXX don't copy the i/o pages. this should probably be fixed. */ pcb2->pcb_ext = 0; /* Copy the LDT, if necessary. */ mtx_lock_spin(&dt_lock); if (mdp2->md_ldt != NULL) { if (flags & RFMEM) { mdp2->md_ldt->ldt_refcnt++; } else { mdp2->md_ldt = user_ldt_alloc(mdp2, mdp2->md_ldt->ldt_len); if (mdp2->md_ldt == NULL) panic("could not copy LDT"); } } mtx_unlock_spin(&dt_lock); /* Setup to release spin count in fork_exit(). */ td2->td_md.md_spinlock_count = 1; /* * XXX XEN need to check on PSL_USER is handled */ td2->td_md.md_saved_flags = PSL_KERNEL | PSL_I; /* * Now, cpu_switch() can schedule the new process. * pcb_esp is loaded pointing to the cpu_switch() stack frame * containing the return address when exiting cpu_switch. * This will normally be to fork_trampoline(), which will have * %ebx loaded with the new proc's pointer. fork_trampoline() * will set up a stack to call fork_return(p, frame); to complete * the return to user-mode. */ } /* * Intercept the return address from a freshly forked process that has NOT * been scheduled yet. * * This is needed to make kernel threads stay in kernel mode. */ void cpu_set_fork_handler(td, func, arg) struct thread *td; void (*func)(void *); void *arg; { /* * Note that the trap frame follows the args, so the function * is really called like this: func(arg, frame); */ td->td_pcb->pcb_esi = (int) func; /* function */ td->td_pcb->pcb_ebx = (int) arg; /* first arg */ } void cpu_exit(struct thread *td) { /* * If this process has a custom LDT, release it. Reset pc->pcb_gs * and %gs before we free it in case they refer to an LDT entry. */ mtx_lock_spin(&dt_lock); if (td->td_proc->p_md.md_ldt) { td->td_pcb->pcb_gs = _udatasel; load_gs(_udatasel); user_ldt_free(td); } else mtx_unlock_spin(&dt_lock); } void cpu_thread_exit(struct thread *td) { #ifdef DEV_NPX critical_enter(); if (td == PCPU_GET(fpcurthread)) npxdrop(); critical_exit(); #endif /* Disable any hardware breakpoints. */ if (td->td_pcb->pcb_flags & PCB_DBREGS) { reset_dbregs(); td->td_pcb->pcb_flags &= ~PCB_DBREGS; } } void cpu_thread_clean(struct thread *td) { struct pcb *pcb; pcb = td->td_pcb; if (pcb->pcb_ext != NULL) { /* if (pcb->pcb_ext->ext_refcount-- == 1) ?? */ /* * XXX do we need to move the TSS off the allocated pages * before freeing them? (not done here) */ kmem_free(kernel_arena, (vm_offset_t)pcb->pcb_ext, ctob(IOPAGES + 1)); pcb->pcb_ext = NULL; } } void cpu_thread_swapin(struct thread *td) { } void cpu_thread_swapout(struct thread *td) { } void cpu_thread_alloc(struct thread *td) { td->td_pcb = (struct pcb *)(td->td_kstack + td->td_kstack_pages * PAGE_SIZE) - 1; td->td_frame = (struct trapframe *)((caddr_t)td->td_pcb - 16) - 1; td->td_pcb->pcb_ext = NULL; td->td_pcb->pcb_save = &td->td_pcb->pcb_user_save; } void cpu_thread_free(struct thread *td) { cpu_thread_clean(td); } void cpu_set_syscall_retval(struct thread *td, int error) { switch (error) { case 0: td->td_frame->tf_eax = td->td_retval[0]; td->td_frame->tf_edx = td->td_retval[1]; td->td_frame->tf_eflags &= ~PSL_C; break; case ERESTART: /* * Reconstruct pc, assuming lcall $X,y is 7 bytes, int * 0x80 is 2 bytes. We saved this in tf_err. */ td->td_frame->tf_eip -= td->td_frame->tf_err; break; case EJUSTRETURN: break; default: if (td->td_proc->p_sysent->sv_errsize) { if (error >= td->td_proc->p_sysent->sv_errsize) error = -1; /* XXX */ else error = td->td_proc->p_sysent->sv_errtbl[error]; } td->td_frame->tf_eax = error; td->td_frame->tf_eflags |= PSL_C; break; } } /* * Initialize machine state (pcb and trap frame) for a new thread about to * upcall. Put enough state in the new thread's PCB to get it to go back * userret(), where we can intercept it again to set the return (upcall) * Address and stack, along with those from upcals that are from other sources * such as those generated in thread_userret() itself. */ void cpu_set_upcall(struct thread *td, struct thread *td0) { struct pcb *pcb2; /* Point the pcb to the top of the stack. */ pcb2 = td->td_pcb; /* * Copy the upcall pcb. This loads kernel regs. * Those not loaded individually below get their default * values here. */ bcopy(td0->td_pcb, pcb2, sizeof(*pcb2)); pcb2->pcb_flags &= ~(PCB_NPXINITDONE | PCB_NPXUSERINITDONE | PCB_KERNNPX); pcb2->pcb_save = &pcb2->pcb_user_save; /* * Create a new fresh stack for the new thread. */ bcopy(td0->td_frame, td->td_frame, sizeof(struct trapframe)); /* If the current thread has the trap bit set (i.e. a debugger had * single stepped the process to the system call), we need to clear * the trap flag from the new frame. Otherwise, the new thread will * receive a (likely unexpected) SIGTRAP when it executes the first * instruction after returning to userland. */ td->td_frame->tf_eflags &= ~PSL_T; /* * Set registers for trampoline to user mode. Leave space for the * return address on stack. These are the kernel mode register values. */ pcb2->pcb_edi = 0; pcb2->pcb_esi = (int)fork_return; /* trampoline arg */ pcb2->pcb_ebp = 0; pcb2->pcb_esp = (int)td->td_frame - sizeof(void *); /* trampoline arg */ pcb2->pcb_ebx = (int)td; /* trampoline arg */ pcb2->pcb_eip = (int)fork_trampoline; pcb2->pcb_psl &= ~(PSL_I); /* interrupts must be disabled */ pcb2->pcb_gs = rgs(); /* * If we didn't copy the pcb, we'd need to do the following registers: * pcb2->pcb_cr3: cloned above. * pcb2->pcb_dr*: cloned above. * pcb2->pcb_savefpu: cloned above. * pcb2->pcb_flags: cloned above. * pcb2->pcb_onfault: cloned above (always NULL here?). * pcb2->pcb_gs: cloned above. * pcb2->pcb_ext: cleared below. */ pcb2->pcb_ext = NULL; /* Setup to release spin count in fork_exit(). */ td->td_md.md_spinlock_count = 1; td->td_md.md_saved_flags = PSL_KERNEL | PSL_I; } /* * Set that machine state for performing an upcall that has to * be done in thread_userret() so that those upcalls generated * in thread_userret() itself can be done as well. */ void cpu_set_upcall_kse(struct thread *td, void (*entry)(void *), void *arg, stack_t *stack) { /* * Do any extra cleaning that needs to be done. * The thread may have optional components * that are not present in a fresh thread. * This may be a recycled thread so make it look * as though it's newly allocated. */ cpu_thread_clean(td); /* * Set the trap frame to point at the beginning of the uts * function. */ td->td_frame->tf_ebp = 0; td->td_frame->tf_esp = (((int)stack->ss_sp + stack->ss_size - 4) & ~0x0f) - 4; td->td_frame->tf_eip = (int)entry; /* * Pass the address of the mailbox for this kse to the uts * function as a parameter on the stack. */ suword((void *)(td->td_frame->tf_esp + sizeof(void *)), (int)arg); } int cpu_set_user_tls(struct thread *td, void *tls_base) { struct segment_descriptor sd; uint32_t base; /* * Construct a descriptor and store it in the pcb for * the next context switch. Also store it in the gdt * so that the load of tf_fs into %fs will activate it * at return to userland. */ base = (uint32_t)tls_base; sd.sd_lobase = base & 0xffffff; sd.sd_hibase = (base >> 24) & 0xff; sd.sd_lolimit = 0xffff; /* 4GB limit, wraps around */ sd.sd_hilimit = 0xf; sd.sd_type = SDT_MEMRWA; sd.sd_dpl = SEL_UPL; sd.sd_p = 1; sd.sd_xx = 0; sd.sd_def32 = 1; sd.sd_gran = 1; critical_enter(); /* set %gs */ td->td_pcb->pcb_gsd = sd; if (td == curthread) { PCPU_GET(fsgs_gdt)[1] = sd; load_gs(GSEL(GUGS_SEL, SEL_UPL)); } critical_exit(); return (0); } /* * Convert kernel VA to physical address */ vm_paddr_t kvtop(void *addr) { vm_paddr_t pa; pa = pmap_kextract((vm_offset_t)addr); if (pa == 0) panic("kvtop: zero page frame"); return (pa); } #ifdef SMP static void cpu_reset_proxy() { cpuset_t tcrp; cpu_reset_proxy_active = 1; while (cpu_reset_proxy_active == 1) ; /* Wait for other cpu to see that we've started */ CPU_SETOF(cpu_reset_proxyid, &tcrp); stop_cpus(tcrp); printf("cpu_reset_proxy: Stopped CPU %d\n", cpu_reset_proxyid); DELAY(1000000); cpu_reset_real(); } #endif void cpu_reset() { #ifdef XBOX if (arch_i386_is_xbox) { /* Kick the PIC16L, it can reboot the box */ pic16l_reboot(); for (;;); } #endif #ifdef SMP cpuset_t map; u_int cnt; if (smp_started) { map = all_cpus; CPU_CLR(PCPU_GET(cpuid), &map); CPU_NAND(&map, &stopped_cpus); if (!CPU_EMPTY(&map)) { printf("cpu_reset: Stopping other CPUs\n"); stop_cpus(map); } if (PCPU_GET(cpuid) != 0) { cpu_reset_proxyid = PCPU_GET(cpuid); cpustop_restartfunc = cpu_reset_proxy; cpu_reset_proxy_active = 0; printf("cpu_reset: Restarting BSP\n"); /* Restart CPU #0. */ /* XXX: restart_cpus(1 << 0); */ CPU_SETOF(0, &started_cpus); wmb(); cnt = 0; while (cpu_reset_proxy_active == 0 && cnt < 10000000) cnt++; /* Wait for BSP to announce restart */ if (cpu_reset_proxy_active == 0) printf("cpu_reset: Failed to restart BSP\n"); enable_intr(); cpu_reset_proxy_active = 2; while (1); /* NOTREACHED */ } DELAY(1000000); } #endif cpu_reset_real(); /* NOTREACHED */ } static void cpu_reset_real() { struct region_descriptor null_idt; #ifndef PC98 int b; #endif disable_intr(); #ifdef XEN if (smp_processor_id() == 0) HYPERVISOR_shutdown(SHUTDOWN_reboot); else HYPERVISOR_shutdown(SHUTDOWN_poweroff); #endif #ifdef CPU_ELAN if (elan_mmcr != NULL) elan_mmcr->RESCFG = 1; #endif if (cpu == CPU_GEODE1100) { /* Attempt Geode's own reset */ outl(0xcf8, 0x80009044ul); outl(0xcfc, 0xf); } #ifdef PC98 /* * Attempt to do a CPU reset via CPU reset port. */ if ((inb(0x35) & 0xa0) != 0xa0) { outb(0x37, 0x0f); /* SHUT0 = 0. */ outb(0x37, 0x0b); /* SHUT1 = 0. */ } outb(0xf0, 0x00); /* Reset. */ #else #if !defined(BROKEN_KEYBOARD_RESET) /* * Attempt to do a CPU reset via the keyboard controller, * do not turn off GateA20, as any machine that fails * to do the reset here would then end up in no man's land. */ outb(IO_KBD + 4, 0xFE); DELAY(500000); /* wait 0.5 sec to see if that did it */ #endif /* * Attempt to force a reset via the Reset Control register at * I/O port 0xcf9. Bit 2 forces a system reset when it * transitions from 0 to 1. Bit 1 selects the type of reset * to attempt: 0 selects a "soft" reset, and 1 selects a * "hard" reset. We try a "hard" reset. The first write sets * bit 1 to select a "hard" reset and clears bit 2. The * second write forces a 0 -> 1 transition in bit 2 to trigger * a reset. */ outb(0xcf9, 0x2); outb(0xcf9, 0x6); DELAY(500000); /* wait 0.5 sec to see if that did it */ /* * Attempt to force a reset via the Fast A20 and Init register * at I/O port 0x92. Bit 1 serves as an alternate A20 gate. * Bit 0 asserts INIT# when set to 1. We are careful to only * preserve bit 1 while setting bit 0. We also must clear bit * 0 before setting it if it isn't already clear. */ b = inb(0x92); if (b != 0xff) { if ((b & 0x1) != 0) outb(0x92, b & 0xfe); outb(0x92, b | 0x1); DELAY(500000); /* wait 0.5 sec to see if that did it */ } #endif /* PC98 */ printf("No known reset method worked, attempting CPU shutdown\n"); DELAY(1000000); /* wait 1 sec for printf to complete */ /* Wipe the IDT. */ null_idt.rd_limit = 0; null_idt.rd_base = 0; lidt(&null_idt); /* "good night, sweet prince .... " */ breakpoint(); /* NOTREACHED */ while(1); } /* * Get an sf_buf from the freelist. May block if none are available. */ void sf_buf_map(struct sf_buf *sf, int flags) { pt_entry_t opte, *ptep; /* * Update the sf_buf's virtual-to-physical mapping, flushing the * virtual address from the TLB. Since the reference count for * the sf_buf's old mapping was zero, that mapping is not * currently in use. Consequently, there is no need to exchange * the old and new PTEs atomically, even under PAE. */ ptep = vtopte(sf->kva); opte = *ptep; #ifdef XEN PT_SET_MA(sf->kva, xpmap_ptom(VM_PAGE_TO_PHYS(sf->m)) | pgeflag | PG_RW | PG_V | pmap_cache_bits(sf->m->md.pat_mode, 0)); #else *ptep = VM_PAGE_TO_PHYS(sf->m) | pgeflag | PG_RW | PG_V | pmap_cache_bits(sf->m->md.pat_mode, 0); #endif /* * Avoid unnecessary TLB invalidations: If the sf_buf's old * virtual-to-physical mapping was not used, then any processor * that has invalidated the sf_buf's virtual address from its TLB * since the last used mapping need not invalidate again. */ #ifdef SMP if ((opte & (PG_V | PG_A)) == (PG_V | PG_A)) CPU_ZERO(&sf->cpumask); sf_buf_shootdown(sf, flags); #else if ((opte & (PG_V | PG_A)) == (PG_V | PG_A)) pmap_invalidate_page(kernel_pmap, sf->kva); #endif } #ifdef SMP void sf_buf_shootdown(struct sf_buf *sf, int flags) { cpuset_t other_cpus; u_int cpuid; sched_pin(); cpuid = PCPU_GET(cpuid); if (!CPU_ISSET(cpuid, &sf->cpumask)) { CPU_SET(cpuid, &sf->cpumask); invlpg(sf->kva); } if ((flags & SFB_CPUPRIVATE) == 0) { other_cpus = all_cpus; CPU_CLR(cpuid, &other_cpus); CPU_NAND(&other_cpus, &sf->cpumask); if (!CPU_EMPTY(&other_cpus)) { CPU_OR(&sf->cpumask, &other_cpus); smp_masked_invlpg(other_cpus, sf->kva); } } sched_unpin(); } #endif /* * MD part of sf_buf_free(). */ int sf_buf_unmap(struct sf_buf *sf) { #ifdef XEN /* * Xen doesn't like having dangling R/W mappings */ pmap_qremove(sf->kva, 1); return (1); #else return (0); #endif } static void sf_buf_invalidate(struct sf_buf *sf) { vm_page_t m = sf->m; /* * Use pmap_qenter to update the pte for * existing mapping, in particular, the PAT * settings are recalculated. */ pmap_qenter(sf->kva, &m, 1); pmap_invalidate_cache_range(sf->kva, sf->kva + PAGE_SIZE, FALSE); } /* * Invalidate the cache lines that may belong to the page, if * (possibly old) mapping of the page by sf buffer exists. Returns * TRUE when mapping was found and cache invalidated. */ boolean_t sf_buf_invalidate_cache(vm_page_t m) { return (sf_buf_process_page(m, sf_buf_invalidate)); } /* * Software interrupt handler for queued VM system processing. */ void swi_vm(void *dummy) { if (busdma_swi_pending != 0) busdma_swi(); } /* * Tell whether this address is in some physical memory region. * Currently used by the kernel coredump code in order to avoid * dumping the ``ISA memory hole'' which could cause indefinite hangs, * or other unpredictable behaviour. */ int is_physical_memory(vm_paddr_t addr) { #ifdef DEV_ISA /* The ISA ``memory hole''. */ if (addr >= 0xa0000 && addr < 0x100000) return 0; #endif /* * stuff other tests for known memory-mapped devices (PCI?) * here */ return 1; } Index: head/sys/i386/include/md_var.h =================================================================== --- head/sys/i386/include/md_var.h (revision 273990) +++ head/sys/i386/include/md_var.h (revision 273991) @@ -1,120 +1,122 @@ /*- * Copyright (c) 1995 Bruce D. Evans. * All rights reserved. * * 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. Neither the name of the author nor the names of contributors * 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. * * $FreeBSD$ */ #ifndef _MACHINE_MD_VAR_H_ #define _MACHINE_MD_VAR_H_ /* * Miscellaneous machine-dependent declarations. */ extern long Maxmem; extern u_int basemem; /* PA of original top of base memory */ extern int busdma_swi_pending; extern u_int cpu_exthigh; extern u_int cpu_feature; extern u_int cpu_feature2; extern u_int amd_feature; extern u_int amd_feature2; extern u_int amd_pminfo; extern u_int via_feature_rng; extern u_int via_feature_xcrypt; extern u_int cpu_clflush_line_size; extern u_int cpu_stdext_feature; extern u_int cpu_fxsr; extern u_int cpu_high; extern u_int cpu_id; extern u_int cpu_mxcsr_mask; extern u_int cpu_procinfo; extern u_int cpu_procinfo2; extern char cpu_vendor[]; extern u_int cpu_vendor_id; extern u_int cpu_mon_mwait_flags; extern u_int cpu_mon_min_size; extern u_int cpu_mon_max_size; extern u_int cyrix_did; #if defined(I586_CPU) && !defined(NO_F00F_HACK) extern int has_f00f_bug; #endif extern u_int hv_high; extern char hv_vendor[]; extern char kstack[]; extern char sigcode[]; extern int szsigcode; #ifdef COMPAT_FREEBSD4 extern int szfreebsd4_sigcode; #endif #ifdef COMPAT_43 extern int szosigcode; #endif extern uint32_t *vm_page_dump; extern int vm_page_dump_size; extern int workaround_erratum383; +extern int _udatasel; +extern int _ucodesel; typedef void alias_for_inthand_t(u_int cs, u_int ef, u_int esp, u_int ss); struct thread; struct reg; struct fpreg; struct dbreg; struct dumperinfo; void bcopyb(const void *from, void *to, size_t len); void busdma_swi(void); void cpu_setregs(void); void cpu_switch_load_gs(void) __asm(__STRING(cpu_switch_load_gs)); void doreti_iret(void) __asm(__STRING(doreti_iret)); void doreti_iret_fault(void) __asm(__STRING(doreti_iret_fault)); void doreti_popl_ds(void) __asm(__STRING(doreti_popl_ds)); void doreti_popl_ds_fault(void) __asm(__STRING(doreti_popl_ds_fault)); void doreti_popl_es(void) __asm(__STRING(doreti_popl_es)); void doreti_popl_es_fault(void) __asm(__STRING(doreti_popl_es_fault)); void doreti_popl_fs(void) __asm(__STRING(doreti_popl_fs)); void doreti_popl_fs_fault(void) __asm(__STRING(doreti_popl_fs_fault)); void dump_add_page(vm_paddr_t); void dump_drop_page(vm_paddr_t); void finishidentcpu(void); void fillw(int /*u_short*/ pat, void *base, size_t cnt); void initializecpu(void); void initializecpucache(void); void i686_pagezero(void *addr); void sse2_pagezero(void *addr); void init_AMD_Elan_sc520(void); int is_physical_memory(vm_paddr_t addr); int isa_nmi(int cd); vm_paddr_t kvtop(void *addr); void panicifcpuunsupported(void); void ppro_reenable_apic(void); void printcpuinfo(void); void setidt(int idx, alias_for_inthand_t *func, int typ, int dpl, int selec); int user_dbreg_trap(void); int minidumpsys(struct dumperinfo *); #endif /* !_MACHINE_MD_VAR_H_ */ Index: head/sys/i386/linux/linux_sysvec.c =================================================================== --- head/sys/i386/linux/linux_sysvec.c (revision 273990) +++ head/sys/i386/linux/linux_sysvec.c (revision 273991) @@ -1,1160 +1,1159 @@ /*- * Copyright (c) 1994-1996 Søren Schmidt * All rights reserved. * * 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 * in this position and unchanged. * 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. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 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 __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include MODULE_VERSION(linux, 1); MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); #if BYTE_ORDER == LITTLE_ENDIAN #define SHELLMAGIC 0x2123 /* #! */ #else #define SHELLMAGIC 0x2321 #endif /* * Allow the sendsig functions to use the ldebug() facility * even though they are not syscalls themselves. Map them * to syscall 0. This is slightly less bogus than using * ldebug(sigreturn). */ #define LINUX_SYS_linux_rt_sendsig 0 #define LINUX_SYS_linux_sendsig 0 #define LINUX_PS_STRINGS (LINUX_USRSTACK - sizeof(struct ps_strings)) extern char linux_sigcode[]; extern int linux_szsigcode; extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); SET_DECLARE(linux_device_handler_set, struct linux_device_handler); static int linux_fixup(register_t **stack_base, struct image_params *iparams); static int elf_linux_fixup(register_t **stack_base, struct image_params *iparams); static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); static void exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack); static register_t *linux_copyout_strings(struct image_params *imgp); static boolean_t linux_trans_osrel(const Elf_Note *note, int32_t *osrel); static int linux_szplatform; const char *linux_platform; static eventhandler_tag linux_exit_tag; static eventhandler_tag linux_exec_tag; /* * Linux syscalls return negative errno's, we do positive and map them * Reference: * FreeBSD: src/sys/sys/errno.h * Linux: linux-2.6.17.8/include/asm-generic/errno-base.h * linux-2.6.17.8/include/asm-generic/errno.h */ static int bsd_to_linux_errno[ELAST + 1] = { -0, -1, -2, -3, -4, -5, -6, -7, -8, -9, -10, -35, -12, -13, -14, -15, -16, -17, -18, -19, -20, -21, -22, -23, -24, -25, -26, -27, -28, -29, -30, -31, -32, -33, -34, -11,-115,-114, -88, -89, -90, -91, -92, -93, -94, -95, -96, -97, -98, -99, -100,-101,-102,-103,-104,-105,-106,-107,-108,-109, -110,-111, -40, -36,-112,-113, -39, -11, -87,-122, -116, -66, -6, -6, -6, -6, -6, -37, -38, -9, -6, -6, -43, -42, -75,-125, -84, -95, -16, -74, -72, -67, -71 }; int bsd_to_linux_signal[LINUX_SIGTBLSZ] = { LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL, LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE, LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS, LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG, LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD, LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU, LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH, 0, LINUX_SIGUSR1, LINUX_SIGUSR2 }; int linux_to_bsd_signal[LINUX_SIGTBLSZ] = { SIGHUP, SIGINT, SIGQUIT, SIGILL, SIGTRAP, SIGABRT, SIGBUS, SIGFPE, SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2, SIGPIPE, SIGALRM, SIGTERM, SIGBUS, SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU, SIGURG, SIGXCPU, SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH, SIGIO, SIGURG, SIGSYS }; #define LINUX_T_UNKNOWN 255 static int _bsd_to_linux_trapcode[] = { LINUX_T_UNKNOWN, /* 0 */ 6, /* 1 T_PRIVINFLT */ LINUX_T_UNKNOWN, /* 2 */ 3, /* 3 T_BPTFLT */ LINUX_T_UNKNOWN, /* 4 */ LINUX_T_UNKNOWN, /* 5 */ 16, /* 6 T_ARITHTRAP */ 254, /* 7 T_ASTFLT */ LINUX_T_UNKNOWN, /* 8 */ 13, /* 9 T_PROTFLT */ 1, /* 10 T_TRCTRAP */ LINUX_T_UNKNOWN, /* 11 */ 14, /* 12 T_PAGEFLT */ LINUX_T_UNKNOWN, /* 13 */ 17, /* 14 T_ALIGNFLT */ LINUX_T_UNKNOWN, /* 15 */ LINUX_T_UNKNOWN, /* 16 */ LINUX_T_UNKNOWN, /* 17 */ 0, /* 18 T_DIVIDE */ 2, /* 19 T_NMI */ 4, /* 20 T_OFLOW */ 5, /* 21 T_BOUND */ 7, /* 22 T_DNA */ 8, /* 23 T_DOUBLEFLT */ 9, /* 24 T_FPOPFLT */ 10, /* 25 T_TSSFLT */ 11, /* 26 T_SEGNPFLT */ 12, /* 27 T_STKFLT */ 18, /* 28 T_MCHK */ 19, /* 29 T_XMMFLT */ 15 /* 30 T_RESERVED */ }; #define bsd_to_linux_trapcode(code) \ ((code)args->argc + 1); (*stack_base)--; suword(*stack_base, (intptr_t)(void *)envp); (*stack_base)--; suword(*stack_base, (intptr_t)(void *)argv); (*stack_base)--; suword(*stack_base, imgp->args->argc); return (0); } static int elf_linux_fixup(register_t **stack_base, struct image_params *imgp) { struct proc *p; Elf32_Auxargs *args; Elf32_Addr *uplatform; struct ps_strings *arginfo; register_t *pos; KASSERT(curthread->td_proc == imgp->proc, ("unsafe elf_linux_fixup(), should be curproc")); p = imgp->proc; arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform); args = (Elf32_Auxargs *)imgp->auxargs; pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2); AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); /* * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, * as it has appeared in the 2.4.0-rc7 first time. * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), * glibc falls back to the hard-coded CLK_TCK value when aux entry * is not present. * Also see linux_times() implementation. */ if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY(pos, AT_BASE, args->base); AUXARGS_ENTRY(pos, LINUX_AT_SECURE, 0); AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(uplatform)); if (args->execfd != -1) AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); imgp->auxargs = NULL; (*stack_base)--; suword(*stack_base, (register_t)imgp->args->argc); return (0); } /* * Copied from kern/kern_exec.c */ static register_t * linux_copyout_strings(struct image_params *imgp) { int argc, envc; char **vectp; char *stringp, *destp; register_t *stack_base; struct ps_strings *arginfo; struct proc *p; /* * Calculate string base and vector table pointers. * Also deal with signal trampoline code for this exec type. */ p = imgp->proc; arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; destp = (caddr_t)arginfo - SPARE_USRSPACE - linux_szplatform - roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *)); /* * install LINUX_PLATFORM */ copyout(linux_platform, ((caddr_t)arginfo - linux_szplatform), linux_szplatform); /* * If we have a valid auxargs ptr, prepare some room * on the stack. */ if (imgp->auxargs) { /* * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for * lower compatibility. */ imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size : (LINUX_AT_COUNT * 2); /* * The '+ 2' is for the null pointers at the end of each of * the arg and env vector sets,and imgp->auxarg_size is room * for argument of Runtime loader. */ vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2 + imgp->auxarg_size) * sizeof(char *)); } else { /* * The '+ 2' is for the null pointers at the end of each of * the arg and env vector sets */ vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) * sizeof(char *)); } /* * vectp also becomes our initial stack base */ stack_base = (register_t *)vectp; stringp = imgp->args->begin_argv; argc = imgp->args->argc; envc = imgp->args->envc; /* * Copy out strings - arguments and environment. */ copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); /* * Fill in "ps_strings" struct for ps, w, etc. */ suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp); suword(&arginfo->ps_nargvstr, argc); /* * Fill in argument portion of vector table. */ for (; argc > 0; --argc) { suword(vectp++, (long)(intptr_t)destp); while (*stringp++ != 0) destp++; destp++; } /* a null vector table pointer separates the argp's from the envp's */ suword(vectp++, 0); suword(&arginfo->ps_envstr, (long)(intptr_t)vectp); suword(&arginfo->ps_nenvstr, envc); /* * Fill in environment portion of vector table. */ for (; envc > 0; --envc) { suword(vectp++, (long)(intptr_t)destp); while (*stringp++ != 0) destp++; destp++; } /* end of vector table is a null pointer */ suword(vectp, 0); return (stack_base); } -extern int _ucodesel, _udatasel; extern unsigned long linux_sznonrtsigcode; static void linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_rt_sigframe *fp, frame; int sig, code; int oonstack; sig = ksi->ksi_signo; code = ksi->ksi_code; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); regs = td->td_frame; oonstack = sigonstack(regs->tf_esp); #ifdef DEBUG if (ldebug(rt_sendsig)) printf(ARGS(rt_sendsig, "%p, %d, %p, %u"), catcher, sig, (void*)mask, code); #endif /* * Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); } else fp = (struct l_rt_sigframe *)regs->tf_esp - 1; mtx_unlock(&psp->ps_mtx); /* * Build the argument list for the signal handler. */ if (p->p_sysent->sv_sigtbl) if (sig <= p->p_sysent->sv_sigsize) sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; bzero(&frame, sizeof(frame)); frame.sf_handler = catcher; frame.sf_sig = sig; frame.sf_siginfo = &fp->sf_si; frame.sf_ucontext = &fp->sf_sc; /* Fill in POSIX parts */ ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig); /* * Build the signal context to be used by sigreturn. */ frame.sf_sc.uc_flags = 0; /* XXX ??? */ frame.sf_sc.uc_link = NULL; /* XXX ??? */ frame.sf_sc.uc_stack.ss_sp = td->td_sigstk.ss_sp; frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; PROC_UNLOCK(p); bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0]; frame.sf_sc.uc_mcontext.sc_gs = rgs(); frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; frame.sf_sc.uc_mcontext.sc_edi = regs->tf_edi; frame.sf_sc.uc_mcontext.sc_esi = regs->tf_esi; frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_ebp; frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_ebx; frame.sf_sc.uc_mcontext.sc_edx = regs->tf_edx; frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_ecx; frame.sf_sc.uc_mcontext.sc_eax = regs->tf_eax; frame.sf_sc.uc_mcontext.sc_eip = regs->tf_eip; frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_eflags; frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_esp; frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; frame.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr; frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); #ifdef DEBUG if (ldebug(rt_sendsig)) printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"), frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp, td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask); #endif if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ #ifdef DEBUG if (ldebug(rt_sendsig)) printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"), fp, oonstack); #endif PROC_LOCK(p); sigexit(td, SIGILL); } /* * Build context to run handler in. */ regs->tf_esp = (int)fp; regs->tf_eip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode; regs->tf_eflags &= ~(PSL_T | PSL_VM | 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); } /* * Send an interrupt to process. * * Stack is set up to allow sigcode stored * in u. to call routine, followed by kcall * to sigreturn routine below. After sigreturn * resets the signal mask, the stack, and the * frame pointer, it returns to the user * specified pc, psl. */ static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) { struct thread *td = curthread; struct proc *p = td->td_proc; struct sigacts *psp; struct trapframe *regs; struct l_sigframe *fp, frame; l_sigset_t lmask; int sig, code; int oonstack, i; PROC_LOCK_ASSERT(p, MA_OWNED); psp = p->p_sigacts; sig = ksi->ksi_signo; code = ksi->ksi_code; mtx_assert(&psp->ps_mtx, MA_OWNED); if (SIGISMEMBER(psp->ps_siginfo, sig)) { /* Signal handler installed with SA_SIGINFO. */ linux_rt_sendsig(catcher, ksi, mask); return; } regs = td->td_frame; oonstack = sigonstack(regs->tf_esp); #ifdef DEBUG if (ldebug(sendsig)) printf(ARGS(sendsig, "%p, %d, %p, %u"), catcher, sig, (void*)mask, code); #endif /* * Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct l_sigframe *)(td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct l_sigframe)); } else fp = (struct l_sigframe *)regs->tf_esp - 1; mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(p); /* * Build the argument list for the signal handler. */ if (p->p_sysent->sv_sigtbl) if (sig <= p->p_sysent->sv_sigsize) sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; bzero(&frame, sizeof(frame)); frame.sf_handler = catcher; frame.sf_sig = sig; bsd_to_linux_sigset(mask, &lmask); /* * Build the signal context to be used by sigreturn. */ frame.sf_sc.sc_mask = lmask.__bits[0]; frame.sf_sc.sc_gs = rgs(); frame.sf_sc.sc_fs = regs->tf_fs; frame.sf_sc.sc_es = regs->tf_es; frame.sf_sc.sc_ds = regs->tf_ds; frame.sf_sc.sc_edi = regs->tf_edi; frame.sf_sc.sc_esi = regs->tf_esi; frame.sf_sc.sc_ebp = regs->tf_ebp; frame.sf_sc.sc_ebx = regs->tf_ebx; frame.sf_sc.sc_edx = regs->tf_edx; frame.sf_sc.sc_ecx = regs->tf_ecx; frame.sf_sc.sc_eax = regs->tf_eax; frame.sf_sc.sc_eip = regs->tf_eip; frame.sf_sc.sc_cs = regs->tf_cs; frame.sf_sc.sc_eflags = regs->tf_eflags; frame.sf_sc.sc_esp_at_signal = regs->tf_esp; frame.sf_sc.sc_ss = regs->tf_ss; frame.sf_sc.sc_err = regs->tf_err; frame.sf_sc.sc_cr2 = (register_t)ksi->ksi_addr; frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(ksi->ksi_trapno); for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) frame.sf_extramask[i] = lmask.__bits[i+1]; if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); } /* * Build context to run handler in. */ regs->tf_esp = (int)fp; regs->tf_eip = p->p_sysent->sv_sigcode_base; regs->tf_eflags &= ~(PSL_T | PSL_VM | 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 * psl to gain improper privileges or to cause * a machine fault. */ int linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) { struct l_sigframe frame; struct trapframe *regs; l_sigset_t lmask; sigset_t bmask; int eflags, i; ksiginfo_t ksi; regs = td->td_frame; #ifdef DEBUG if (ldebug(sigreturn)) printf(ARGS(sigreturn, "%p"), (void *)args->sfp); #endif /* * The trampoline code hands us the sigframe. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->sfp, &frame, sizeof(frame)) != 0) return (EFAULT); /* * Check for security violations. */ #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) eflags = frame.sf_sc.sc_eflags; if (!EFLAGS_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. */ #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) if (!CS_SECURE(frame.sf_sc.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); } lmask.__bits[0] = frame.sf_sc.sc_mask; for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) lmask.__bits[i+1] = frame.sf_extramask[i]; linux_to_bsd_sigset(&lmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* * Restore signal context. */ /* %gs was restored by the trampoline. */ regs->tf_fs = frame.sf_sc.sc_fs; regs->tf_es = frame.sf_sc.sc_es; regs->tf_ds = frame.sf_sc.sc_ds; regs->tf_edi = frame.sf_sc.sc_edi; regs->tf_esi = frame.sf_sc.sc_esi; regs->tf_ebp = frame.sf_sc.sc_ebp; regs->tf_ebx = frame.sf_sc.sc_ebx; regs->tf_edx = frame.sf_sc.sc_edx; regs->tf_ecx = frame.sf_sc.sc_ecx; regs->tf_eax = frame.sf_sc.sc_eax; regs->tf_eip = frame.sf_sc.sc_eip; regs->tf_cs = frame.sf_sc.sc_cs; regs->tf_eflags = eflags; regs->tf_esp = frame.sf_sc.sc_esp_at_signal; regs->tf_ss = frame.sf_sc.sc_ss; return (EJUSTRETURN); } /* * System call to cleanup state after a signal * has been taken. Reset signal mask and * stack state from context left by rt_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 * psl to gain improper privileges or to cause * a machine fault. */ int linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) { struct l_ucontext uc; struct l_sigcontext *context; sigset_t bmask; l_stack_t *lss; stack_t ss; struct trapframe *regs; int eflags; ksiginfo_t ksi; regs = td->td_frame; #ifdef DEBUG if (ldebug(rt_sigreturn)) printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp); #endif /* * The trampoline code hands us the ucontext. * It is unsafe to keep track of it ourselves, in the event that a * program jumps out of a signal handler. */ if (copyin(args->ucp, &uc, sizeof(uc)) != 0) return (EFAULT); context = &uc.uc_mcontext; /* * Check for security violations. */ #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) eflags = context->sc_eflags; if (!EFLAGS_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. */ #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) if (!CS_SECURE(context->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); } linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); /* * Restore signal context */ /* %gs was restored by the trampoline. */ regs->tf_fs = context->sc_fs; regs->tf_es = context->sc_es; regs->tf_ds = context->sc_ds; regs->tf_edi = context->sc_edi; regs->tf_esi = context->sc_esi; regs->tf_ebp = context->sc_ebp; regs->tf_ebx = context->sc_ebx; regs->tf_edx = context->sc_edx; regs->tf_ecx = context->sc_ecx; regs->tf_eax = context->sc_eax; regs->tf_eip = context->sc_eip; regs->tf_cs = context->sc_cs; regs->tf_eflags = eflags; regs->tf_esp = context->sc_esp_at_signal; regs->tf_ss = context->sc_ss; /* * call sigaltstack & ignore results.. */ lss = &uc.uc_stack; ss.ss_sp = lss->ss_sp; ss.ss_size = lss->ss_size; ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); #ifdef DEBUG if (ldebug(rt_sigreturn)) printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"), ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask); #endif (void)kern_sigaltstack(td, &ss, NULL); return (EJUSTRETURN); } static int linux_fetch_syscall_args(struct thread *td, struct syscall_args *sa) { struct proc *p; struct trapframe *frame; p = td->td_proc; frame = td->td_frame; sa->code = frame->tf_eax; sa->args[0] = frame->tf_ebx; sa->args[1] = frame->tf_ecx; sa->args[2] = frame->tf_edx; sa->args[3] = frame->tf_esi; sa->args[4] = frame->tf_edi; sa->args[5] = frame->tf_ebp; /* Unconfirmed */ if (sa->code >= p->p_sysent->sv_size) sa->callp = &p->p_sysent->sv_table[0]; else sa->callp = &p->p_sysent->sv_table[sa->code]; sa->narg = sa->callp->sy_narg; td->td_retval[0] = 0; td->td_retval[1] = frame->tf_edx; return (0); } /* * If a linux binary is exec'ing something, try this image activator * first. We override standard shell script execution in order to * be able to modify the interpreter path. We only do this if a linux * binary is doing the exec, so we do not create an EXEC module for it. */ static int exec_linux_imgact_try(struct image_params *iparams); static int exec_linux_imgact_try(struct image_params *imgp) { const char *head = (const char *)imgp->image_header; char *rpath; int error = -1; /* * The interpreter for shell scripts run from a linux binary needs * to be located in /compat/linux if possible in order to recursively * maintain linux path emulation. */ if (((const short *)head)[0] == SHELLMAGIC) { /* * Run our normal shell image activator. If it succeeds attempt * to use the alternate path for the interpreter. If an alternate * path is found, use our stringspace to store it. */ if ((error = exec_shell_imgact(imgp)) == 0) { linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc), imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, AT_FDCWD); if (rpath != NULL) imgp->args->fname_buf = imgp->interpreter_name = rpath; } } return (error); } /* * exec_setregs may initialize some registers differently than Linux * does, thus potentially confusing Linux binaries. If necessary, we * override the exec_setregs default(s) here. */ static void exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack) { struct pcb *pcb = td->td_pcb; exec_setregs(td, imgp, stack); /* Linux sets %gs to 0, we default to _udatasel */ pcb->pcb_gs = 0; load_gs(0); pcb->pcb_initial_npxcw = __LINUX_NPXCW__; } static void linux_get_machine(const char **dst) { switch (cpu_class) { case CPUCLASS_686: *dst = "i686"; break; case CPUCLASS_586: *dst = "i586"; break; case CPUCLASS_486: *dst = "i486"; break; default: *dst = "i386"; } } struct sysentvec linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_mask = 0, .sv_sigsize = LINUX_SIGTBLSZ, .sv_sigtbl = bsd_to_linux_signal, .sv_errsize = ELAST + 1, .sv_errtbl = bsd_to_linux_errno, .sv_transtrap = translate_traps, .sv_fixup = linux_fixup, .sv_sendsig = linux_sendsig, .sv_sigcode = linux_sigcode, .sv_szsigcode = &linux_szsigcode, .sv_prepsyscall = NULL, .sv_name = "Linux a.out", .sv_coredump = NULL, .sv_imgact_try = exec_linux_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_pagesize = PAGE_SIZE, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = LINUX_USRSTACK, .sv_psstrings = PS_STRINGS, .sv_stackprot = VM_PROT_ALL, .sv_copyout_strings = exec_copyout_strings, .sv_setregs = exec_linux_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_AOUT | SV_IA32 | SV_ILP32, .sv_set_syscall_retval = cpu_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, }; INIT_SYSENTVEC(aout_sysvec, &linux_sysvec); struct sysentvec elf_linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, .sv_mask = 0, .sv_sigsize = LINUX_SIGTBLSZ, .sv_sigtbl = bsd_to_linux_signal, .sv_errsize = ELAST + 1, .sv_errtbl = bsd_to_linux_errno, .sv_transtrap = translate_traps, .sv_fixup = elf_linux_fixup, .sv_sendsig = linux_sendsig, .sv_sigcode = linux_sigcode, .sv_szsigcode = &linux_szsigcode, .sv_prepsyscall = NULL, .sv_name = "Linux ELF", .sv_coredump = elf32_coredump, .sv_imgact_try = exec_linux_imgact_try, .sv_minsigstksz = LINUX_MINSIGSTKSZ, .sv_pagesize = PAGE_SIZE, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = LINUX_USRSTACK, .sv_psstrings = LINUX_PS_STRINGS, .sv_stackprot = VM_PROT_ALL, .sv_copyout_strings = linux_copyout_strings, .sv_setregs = exec_linux_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_LINUX | SV_IA32 | SV_ILP32 | SV_SHP, .sv_set_syscall_retval = cpu_set_syscall_retval, .sv_fetch_syscall_args = linux_fetch_syscall_args, .sv_syscallnames = NULL, .sv_shared_page_base = LINUX_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = linux_schedtail, }; INIT_SYSENTVEC(elf_sysvec, &elf_linux_sysvec); static char GNU_ABI_VENDOR[] = "GNU"; static int GNULINUX_ABI_DESC = 0; static boolean_t linux_trans_osrel(const Elf_Note *note, int32_t *osrel) { const Elf32_Word *desc; uintptr_t p; p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); desc = (const Elf32_Word *)p; if (desc[0] != GNULINUX_ABI_DESC) return (FALSE); /* * For linux we encode osrel as follows (see linux_mib.c): * VVVMMMIII (version, major, minor), see linux_mib.c. */ *osrel = desc[1] * 1000000 + desc[2] * 1000 + desc[3]; return (TRUE); } static Elf_Brandnote linux_brandnote = { .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), .hdr.n_descsz = 16, /* XXX at least 16 */ .hdr.n_type = 1, .vendor = GNU_ABI_VENDOR, .flags = BN_TRANSLATE_OSREL, .trans_osrel = linux_trans_osrel }; static Elf32_Brandinfo linux_brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = "/compat/linux", .interp_path = "/lib/ld-linux.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; static Elf32_Brandinfo linux_glibc2brand = { .brand = ELFOSABI_LINUX, .machine = EM_386, .compat_3_brand = "Linux", .emul_path = "/compat/linux", .interp_path = "/lib/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, .brand_note = &linux_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; Elf32_Brandinfo *linux_brandlist[] = { &linux_brand, &linux_glibc2brand, NULL }; static int linux_elf_modevent(module_t mod, int type, void *data) { Elf32_Brandinfo **brandinfo; int error; struct linux_ioctl_handler **lihp; struct linux_device_handler **ldhp; error = 0; switch(type) { case MOD_LOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_insert_brand_entry(*brandinfo) < 0) error = EINVAL; if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_register_handler(*lihp); SET_FOREACH(ldhp, linux_device_handler_set) linux_device_register_handler(*ldhp); mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF); sx_init(&emul_shared_lock, "emuldata->shared lock"); LIST_INIT(&futex_list); mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF); linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit, NULL, 1000); linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, linux_proc_exec, NULL, 1000); linux_get_machine(&linux_platform); linux_szplatform = roundup(strlen(linux_platform) + 1, sizeof(char *)); linux_osd_jail_register(); stclohz = (stathz ? stathz : hz); if (bootverbose) printf("Linux ELF exec handler installed\n"); } else printf("cannot insert Linux ELF brand handler\n"); break; case MOD_UNLOAD: for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_brand_inuse(*brandinfo)) error = EBUSY; if (error == 0) { for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; ++brandinfo) if (elf32_remove_brand_entry(*brandinfo) < 0) error = EINVAL; } if (error == 0) { SET_FOREACH(lihp, linux_ioctl_handler_set) linux_ioctl_unregister_handler(*lihp); SET_FOREACH(ldhp, linux_device_handler_set) linux_device_unregister_handler(*ldhp); mtx_destroy(&emul_lock); sx_destroy(&emul_shared_lock); mtx_destroy(&futex_mtx); EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag); EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag); linux_osd_jail_deregister(); if (bootverbose) printf("Linux ELF exec handler removed\n"); } else printf("Could not deinstall ELF interpreter entry\n"); break; default: return EOPNOTSUPP; } return error; } static moduledata_t linux_elf_mod = { "linuxelf", linux_elf_modevent, 0 }; DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); Index: head/sys/i386/svr4/svr4_machdep.c =================================================================== --- head/sys/i386/svr4/svr4_machdep.c (revision 273990) +++ head/sys/i386/svr4/svr4_machdep.c (revision 273991) @@ -1,611 +1,610 @@ /*- * Copyright (c) 1998 Mark Newton * Copyright (c) 1994 Christos Zoulas * All rights reserved. * * 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. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 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 __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #undef sigcode #undef szsigcode extern int svr4_szsigcode; extern char svr4_sigcode[]; -extern int _udatasel, _ucodesel; static void svr4_getsiginfo(union svr4_siginfo *, int, u_long, caddr_t); #if !defined(__NetBSD__) /* taken from /sys/arch/i386/include/psl.h on NetBSD-1.3 */ # define PSL_MBZ 0xffc08028 # define PSL_USERSTATIC (PSL_USER | PSL_MBZ | PSL_IOPL | PSL_NT | PSL_VM | PSL_VIF | PSL_VIP) # define USERMODE(c, f) (ISPL(c) == SEL_UPL) #endif #if defined(__NetBSD__) void svr4_setregs(td, epp, stack) struct thread *td; struct exec_package *epp; u_long stack; { register struct pcb *pcb = td->td_pcb; pcb->pcb_savefpu.sv_env.en_cw = __SVR4_NPXCW__; setregs(td, epp, stack, 0UL); } #endif /* __NetBSD__ */ void svr4_getcontext(td, uc, mask, oonstack) struct thread *td; struct svr4_ucontext *uc; sigset_t *mask; int oonstack; { struct proc *p = td->td_proc; struct trapframe *tf = td->td_frame; svr4_greg_t *r = uc->uc_mcontext.greg; struct svr4_sigaltstack *s = &uc->uc_stack; #if defined(DONE_MORE_SIGALTSTACK_WORK) struct sigacts *psp; struct sigaltstack *sf; #endif PROC_LOCK(p); #if defined(DONE_MORE_SIGALTSTACK_WORK) psp = p->p_sigacts; sf = &p->p_sigstk; #endif memset(uc, 0, sizeof(struct svr4_ucontext)); uc->uc_link = p->p_emuldata; /* * Set the general purpose registers */ #ifdef VM86 if (tf->tf_eflags & PSL_VM) { r[SVR4_X86_GS] = tf->tf_vm86_gs; r[SVR4_X86_FS] = tf->tf_vm86_fs; r[SVR4_X86_ES] = tf->tf_vm86_es; r[SVR4_X86_DS] = tf->tf_vm86_ds; r[SVR4_X86_EFL] = get_vflags(td); } else #endif { #if defined(__NetBSD__) __asm("movl %%gs,%w0" : "=r" (r[SVR4_X86_GS])); __asm("movl %%fs,%w0" : "=r" (r[SVR4_X86_FS])); #else r[SVR4_X86_GS] = rgs(); r[SVR4_X86_FS] = tf->tf_fs; #endif r[SVR4_X86_ES] = tf->tf_es; r[SVR4_X86_DS] = tf->tf_ds; r[SVR4_X86_EFL] = tf->tf_eflags; } r[SVR4_X86_EDI] = tf->tf_edi; r[SVR4_X86_ESI] = tf->tf_esi; r[SVR4_X86_EBP] = tf->tf_ebp; r[SVR4_X86_ESP] = tf->tf_esp; r[SVR4_X86_EBX] = tf->tf_ebx; r[SVR4_X86_EDX] = tf->tf_edx; r[SVR4_X86_ECX] = tf->tf_ecx; r[SVR4_X86_EAX] = tf->tf_eax; r[SVR4_X86_TRAPNO] = tf->tf_trapno; r[SVR4_X86_ERR] = tf->tf_err; r[SVR4_X86_EIP] = tf->tf_eip; r[SVR4_X86_CS] = tf->tf_cs; r[SVR4_X86_UESP] = 0; r[SVR4_X86_SS] = tf->tf_ss; /* * Set the signal stack */ #if defined(DONE_MORE_SIGALTSTACK_WORK) bsd_to_svr4_sigaltstack(sf, s); #else s->ss_sp = (void *)(((u_long) tf->tf_esp) & ~(16384 - 1)); s->ss_size = 16384; s->ss_flags = 0; #endif PROC_UNLOCK(p); /* * Set the signal mask */ bsd_to_svr4_sigset(mask, &uc->uc_sigmask); /* * Set the flags */ uc->uc_flags = SVR4_UC_SIGMASK|SVR4_UC_CPU|SVR4_UC_STACK; } /* * Set to ucontext specified. Reset signal mask and * stack state from context. * 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 * psl to gain improper privileges or to cause * a machine fault. */ int svr4_setcontext(td, uc) struct thread *td; struct svr4_ucontext *uc; { #if defined(DONE_MORE_SIGALTSTACK_WORK) struct sigacts *psp; #endif struct proc *p = td->td_proc; register struct trapframe *tf; svr4_greg_t *r = uc->uc_mcontext.greg; struct svr4_sigaltstack *s = &uc->uc_stack; struct sigaltstack *sf; sigset_t mask; PROC_LOCK(p); #if defined(DONE_MORE_SIGALTSTACK_WORK) psp = p->p_sigacts; #endif sf = &td->td_sigstk; /* * XXX: * Should we check the value of flags to determine what to restore? * What to do with uc_link? * What to do with floating point stuff? * Should we bother with the rest of the registers that we * set to 0 right now? */ if ((uc->uc_flags & SVR4_UC_CPU) == 0) { PROC_UNLOCK(p); return 0; } DPRINTF(("svr4_setcontext(%d)\n", p->p_pid)); tf = td->td_frame; /* * Restore register context. */ #ifdef VM86 #warning "VM86 doesn't work yet, please don't try to use it." if (r[SVR4_X86_EFL] & PSL_VM) { tf->tf_vm86_gs = r[SVR4_X86_GS]; tf->tf_vm86_fs = r[SVR4_X86_FS]; tf->tf_vm86_es = r[SVR4_X86_ES]; tf->tf_vm86_ds = r[SVR4_X86_DS]; set_vflags(td, r[SVR4_X86_EFL]); } else #endif { /* * Check for security violations. If we're returning to * protected mode, the CPU will validate the segment registers * automatically and generate a trap on violations. We handle * the trap, rather than doing all of the checking here. */ if (((r[SVR4_X86_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 || !USERMODE(r[SVR4_X86_CS], r[SVR4_X86_EFL])) { PROC_UNLOCK(p); return (EINVAL); } #if defined(__NetBSD__) /* %fs and %gs were restored by the trampoline. */ #else /* %gs was restored by the trampoline. */ tf->tf_fs = r[SVR4_X86_FS]; #endif tf->tf_es = r[SVR4_X86_ES]; tf->tf_ds = r[SVR4_X86_DS]; tf->tf_eflags = r[SVR4_X86_EFL]; } tf->tf_edi = r[SVR4_X86_EDI]; tf->tf_esi = r[SVR4_X86_ESI]; tf->tf_ebp = r[SVR4_X86_EBP]; tf->tf_ebx = r[SVR4_X86_EBX]; tf->tf_edx = r[SVR4_X86_EDX]; tf->tf_ecx = r[SVR4_X86_ECX]; tf->tf_eax = r[SVR4_X86_EAX]; tf->tf_trapno = r[SVR4_X86_TRAPNO]; tf->tf_err = r[SVR4_X86_ERR]; tf->tf_eip = r[SVR4_X86_EIP]; tf->tf_cs = r[SVR4_X86_CS]; tf->tf_ss = r[SVR4_X86_SS]; tf->tf_esp = r[SVR4_X86_ESP]; p->p_emuldata = uc->uc_link; /* * restore signal stack */ if (uc->uc_flags & SVR4_UC_STACK) { svr4_to_bsd_sigaltstack(s, sf); } /* * restore signal mask */ if (uc->uc_flags & SVR4_UC_SIGMASK) { #if defined(DEBUG_SVR4) { int i; for (i = 0; i < 4; i++) DPRINTF(("\tuc_sigmask[%d] = %lx\n", i, uc->uc_sigmask.bits[i])); } #endif svr4_to_bsd_sigset(&uc->uc_sigmask, &mask); SIG_CANTMASK(mask); td->td_sigmask = mask; signotify(td); } PROC_UNLOCK(p); return 0; /*EJUSTRETURN;*/ } static void svr4_getsiginfo(si, sig, code, addr) union svr4_siginfo *si; int sig; u_long code; caddr_t addr; { si->svr4_si_signo = bsd_to_svr4_sig[sig]; si->svr4_si_errno = 0; si->svr4_si_addr = addr; switch (code) { case T_PRIVINFLT: si->svr4_si_code = SVR4_ILL_PRVOPC; si->svr4_si_trap = SVR4_T_PRIVINFLT; break; case T_BPTFLT: si->svr4_si_code = SVR4_TRAP_BRKPT; si->svr4_si_trap = SVR4_T_BPTFLT; break; case T_ARITHTRAP: si->svr4_si_code = SVR4_FPE_INTOVF; si->svr4_si_trap = SVR4_T_DIVIDE; break; case T_PROTFLT: si->svr4_si_code = SVR4_SEGV_ACCERR; si->svr4_si_trap = SVR4_T_PROTFLT; break; case T_TRCTRAP: si->svr4_si_code = SVR4_TRAP_TRACE; si->svr4_si_trap = SVR4_T_TRCTRAP; break; case T_PAGEFLT: si->svr4_si_code = SVR4_SEGV_ACCERR; si->svr4_si_trap = SVR4_T_PAGEFLT; break; case T_ALIGNFLT: si->svr4_si_code = SVR4_BUS_ADRALN; si->svr4_si_trap = SVR4_T_ALIGNFLT; break; case T_DIVIDE: si->svr4_si_code = SVR4_FPE_FLTDIV; si->svr4_si_trap = SVR4_T_DIVIDE; break; case T_OFLOW: si->svr4_si_code = SVR4_FPE_FLTOVF; si->svr4_si_trap = SVR4_T_DIVIDE; break; case T_BOUND: si->svr4_si_code = SVR4_FPE_FLTSUB; si->svr4_si_trap = SVR4_T_BOUND; break; case T_DNA: si->svr4_si_code = SVR4_FPE_FLTINV; si->svr4_si_trap = SVR4_T_DNA; break; case T_FPOPFLT: si->svr4_si_code = SVR4_FPE_FLTINV; si->svr4_si_trap = SVR4_T_FPOPFLT; break; case T_SEGNPFLT: si->svr4_si_code = SVR4_SEGV_MAPERR; si->svr4_si_trap = SVR4_T_SEGNPFLT; break; case T_STKFLT: si->svr4_si_code = SVR4_ILL_BADSTK; si->svr4_si_trap = SVR4_T_STKFLT; break; default: si->svr4_si_code = 0; si->svr4_si_trap = 0; #if defined(DEBUG_SVR4) printf("sig %d code %ld\n", sig, code); /* panic("svr4_getsiginfo");*/ #endif break; } } /* * Send an interrupt to process. * * Stack is set up to allow sigcode stored * in u. to call routine. After the handler is * done svr4 will call setcontext for us * with the user context we just set up, and we * will return to the user pc, psl. */ void svr4_sendsig(catcher, ksi, mask) sig_t catcher; ksiginfo_t *ksi; sigset_t *mask; { register struct thread *td = curthread; struct proc *p = td->td_proc; register struct trapframe *tf; struct svr4_sigframe *fp, frame; struct sigacts *psp; int oonstack; int sig; int code; PROC_LOCK_ASSERT(p, MA_OWNED); sig = ksi->ksi_signo; #if defined(DEBUG_SVR4) printf("svr4_sendsig(%d)\n", sig); #endif code = ksi->ksi_trapno; /* use trap No. */ psp = p->p_sigacts; mtx_assert(&psp->ps_mtx, MA_OWNED); tf = td->td_frame; oonstack = sigonstack(tf->tf_esp); /* * Allocate space for the signal handler context. */ if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && SIGISMEMBER(psp->ps_sigonstack, sig)) { fp = (struct svr4_sigframe *)(td->td_sigstk.ss_sp + td->td_sigstk.ss_size - sizeof(struct svr4_sigframe)); td->td_sigstk.ss_flags |= SS_ONSTACK; } else { fp = (struct svr4_sigframe *)tf->tf_esp - 1; } mtx_unlock(&psp->ps_mtx); PROC_UNLOCK(p); /* * Build the argument list for the signal handler. * Notes: * - we always build the whole argument list, even when we * don't need to [when SA_SIGINFO is not set, we don't need * to pass all sf_si and sf_uc] * - we don't pass the correct signal address [we need to * modify many kernel files to enable that] */ svr4_getcontext(td, &frame.sf_uc, mask, oonstack); #if defined(DEBUG_SVR4) printf("obtained ucontext\n"); #endif svr4_getsiginfo(&frame.sf_si, sig, code, (caddr_t) tf->tf_eip); #if defined(DEBUG_SVR4) printf("obtained siginfo\n"); #endif frame.sf_signum = frame.sf_si.svr4_si_signo; frame.sf_sip = &fp->sf_si; frame.sf_ucp = &fp->sf_uc; frame.sf_handler = catcher; #if defined(DEBUG_SVR4) printf("sig = %d, sip %p, ucp = %p, handler = %p\n", frame.sf_signum, frame.sf_sip, frame.sf_ucp, frame.sf_handler); #endif if (copyout(&frame, fp, sizeof(frame)) != 0) { /* * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ PROC_LOCK(p); sigexit(td, SIGILL); /* NOTREACHED */ } #if defined(__NetBSD__) /* * Build context to run handler in. */ tf->tf_es = GSEL(GUSERLDT_SEL, SEL_UPL); tf->tf_ds = GSEL(GUSERLDT_SEL, SEL_UPL); tf->tf_eip = (int)(((char *)PS_STRINGS) - svr4_szsigcode); tf->tf_cs = GSEL(GUSERLDT_SEL, SEL_UPL); tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC|PSL_D); tf->tf_esp = (int)fp; tf->tf_ss = GSEL(GUSERLDT_SEL, SEL_UPL); #else tf->tf_esp = (int)fp; tf->tf_eip = (int)(((char *)PS_STRINGS) - *(p->p_sysent->sv_szsigcode)); tf->tf_eflags &= ~(PSL_T | PSL_D); tf->tf_cs = _ucodesel; tf->tf_ds = _udatasel; tf->tf_es = _udatasel; tf->tf_fs = _udatasel; load_gs(_udatasel); tf->tf_ss = _udatasel; PROC_LOCK(p); mtx_lock(&psp->ps_mtx); #endif } int svr4_sys_sysarch(td, v) struct thread *td; struct svr4_sys_sysarch_args *v; { struct svr4_sys_sysarch_args *uap = v; #if 0 /* USER_LDT */ #if defined(__NetBSD__) caddr_t sg = stackgap_init(p->p_emul); #else caddr_t sg = stackgap_init(); #endif int error; #endif switch (uap->op) { case SVR4_SYSARCH_FPHW: return 0; case SVR4_SYSARCH_DSCR: #if 0 /* USER_LDT */ #warning "USER_LDT doesn't work - are you sure you want this?" { struct i386_set_ldt_args sa, *sap; struct sys_sysarch_args ua; struct svr4_ssd ssd; union descriptor bsd; if ((error = copyin(uap->a1, &ssd, sizeof(ssd))) != 0) { printf("Cannot copy arg1\n"); return error; } printf("s=%x, b=%x, l=%x, a1=%x a2=%x\n", ssd.selector, ssd.base, ssd.limit, ssd.access1, ssd.access2); /* We can only set ldt's for now. */ if (!ISLDT(ssd.selector)) { printf("Not an ldt\n"); return EPERM; } /* Oh, well we don't cleanup either */ if (ssd.access1 == 0) return 0; bsd.sd.sd_lobase = ssd.base & 0xffffff; bsd.sd.sd_hibase = (ssd.base >> 24) & 0xff; bsd.sd.sd_lolimit = ssd.limit & 0xffff; bsd.sd.sd_hilimit = (ssd.limit >> 16) & 0xf; bsd.sd.sd_type = ssd.access1 & 0x1f; bsd.sd.sd_dpl = (ssd.access1 >> 5) & 0x3; bsd.sd.sd_p = (ssd.access1 >> 7) & 0x1; bsd.sd.sd_xx = ssd.access2 & 0x3; bsd.sd.sd_def32 = (ssd.access2 >> 2) & 0x1; bsd.sd.sd_gran = (ssd.access2 >> 3)& 0x1; sa.start = IDXSEL(ssd.selector); sa.desc = stackgap_alloc(&sg, sizeof(union descriptor)); sa.num = 1; sap = stackgap_alloc(&sg, sizeof(struct i386_set_ldt_args)); if ((error = copyout(&sa, sap, sizeof(sa))) != 0) { printf("Cannot copyout args\n"); return error; } ua.op = I386_SET_LDT; ua.parms = (char *) sap; if ((error = copyout(&bsd, sa.desc, sizeof(bsd))) != 0) { printf("Cannot copyout desc\n"); return error; } return sys_sysarch(td, &ua, retval); } #endif default: printf("svr4_sysarch(%d), a1 %p\n", uap->op, uap->a1); return 0; } }