diff --git a/sys/amd64/amd64/elf_machdep.c b/sys/amd64/amd64/elf_machdep.c index 4f6d178..0aea61d 100644 --- a/sys/amd64/amd64/elf_machdep.c +++ b/sys/amd64/amd64/elf_machdep.c @@ -84,7 +84,8 @@ static Elf64_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_ANY, @@ -99,7 +100,8 @@ static Elf64_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/amd64/linux32/linux.h b/sys/amd64/linux32/linux.h index e0ffcdf..94372d6 100644 --- a/sys/amd64/linux32/linux.h +++ b/sys/amd64/linux32/linux.h @@ -108,6 +108,10 @@ typedef struct { #define LINUX_CTL_MAXNAME 10 +#define LINUX_AT_COUNT 16 /* Count of used aux entry types. + * Keep this synchronized with + * elf_linux_fixup() code. + */ struct l___sysctl_args { l_uintptr_t name; diff --git a/sys/amd64/linux32/linux32_machdep.c b/sys/amd64/linux32/linux32_machdep.c index c8869e2..f3b766a 100644 --- a/sys/amd64/linux32/linux32_machdep.c +++ b/sys/amd64/linux32/linux32_machdep.c @@ -38,6 +38,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -211,10 +212,7 @@ linux_execve(struct thread *td, struct linux_execve_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(execve)) - printf(ARGS(execve, "%s"), path); -#endif + LINUX_CTR(execve); error = linux_exec_copyin_args(&eargs, path, UIO_SYSSPACE, args->argp, args->envp); @@ -309,6 +307,8 @@ linux_readv(struct thread *td, struct linux_readv_args *uap) struct uio *auio; int error; + LINUX_CTR(readv); + error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); if (error) return (error); @@ -323,6 +323,8 @@ linux_writev(struct thread *td, struct linux_writev_args *uap) struct uio *auio; int error; + LINUX_CTR(writev); + error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); if (error) return (error); @@ -340,6 +342,8 @@ int linux_ipc(struct thread *td, struct linux_ipc_args *args) { + LINUX_CTR1(ipc, "%d", args->what & 0xFFFF); + switch (args->what & 0xFFFF) { case LINUX_SEMOP: { struct linux_semop_args a; @@ -461,10 +465,7 @@ linux_old_select(struct thread *td, struct linux_old_select_args *args) struct linux_select_args newsel; int error; -#ifdef DEBUG - if (ldebug(old_select)) - printf(ARGS(old_select, "%p"), args->ptr); -#endif + LINUX_CTR(old_select); error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) @@ -485,10 +486,7 @@ linux_fork(struct thread *td, struct linux_fork_args *args) struct proc *p2; struct thread *td2; -#ifdef DEBUG - if (ldebug(fork)) - printf(ARGS(fork, "")); -#endif + LINUX_CTR(fork); if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2)) != 0) return (error); @@ -524,10 +522,7 @@ linux_vfork(struct thread *td, struct linux_vfork_args *args) struct proc *p2; struct thread *td2; -#ifdef DEBUG - if (ldebug(vfork)) - printf(ARGS(vfork, "")); -#endif + LINUX_CTR(vfork); /* Exclude RFPPWAIT */ if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2)) != 0) @@ -575,13 +570,9 @@ linux_clone(struct thread *td, struct linux_clone_args *args) int exit_signal; struct linux_emuldata *em; -#ifdef DEBUG - if (ldebug(clone)) { - printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " - "child tid: %p"), (unsigned)args->flags, - args->stack, args->parent_tidptr, args->child_tidptr); - } -#endif + LINUX_CTR4(clone, "flags %x stack %p parent tid %p child tid %p", + (unsigned)args->flags, args->stack, args->parent_tidptr, + args->child_tidptr); exit_signal = args->flags & 0x000000ff; if (LINUX_SIG_VALID(exit_signal)) { @@ -704,17 +695,16 @@ linux_clone(struct thread *td, struct linux_clone_args *args) a[1] = LINUX_LDT_entry_b(&info); memcpy(&sd, &a, sizeof(a)); -#ifdef DEBUG - if (ldebug(clone)) - printf("Segment created in clone with " - "CLONE_SETTLS: lobase: %x, hibase: %x, " - "lolimit: %x, hilimit: %x, type: %i, " - "dpl: %i, p: %i, xx: %i, long: %i, " - "def32: %i, gran: %i\n", sd.sd_lobase, - sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit, - sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx, - sd.sd_long, sd.sd_def32, sd.sd_gran); -#endif + + LINUX_CTR5(clone, "tls segment created with " + "lobase %x hibase %x lolimit %x hilimit %x type %i", + sd.sd_lobase, sd.sd_hibase, sd.sd_lolimit, + sd.sd_hilimit, sd.sd_type); + LINUX_CTR5(clone, "tls segment created with " + "dpl %i p %i xx %i long %i def32 %i", + sd.sd_dpl, sd.sd_p, sd.sd_xx, + sd.sd_long, sd.sd_def32); + td2->td_pcb->pcb_gsbase = (register_t)info.base_addr; td2->td_pcb->pcb_gs32sd = sd; td2->td_pcb->pcb_gs = GSEL(GUGS32_SEL, SEL_UPL); @@ -722,12 +712,9 @@ linux_clone(struct thread *td, struct linux_clone_args *args) } } -#ifdef DEBUG - if (ldebug(clone)) - printf(LMSG("clone: successful rfork to %d, " - "stack %p sig = %d"), (int)p2->p_pid, args->stack, - exit_signal); -#endif + LINUX_CTR3(clone, "successful rfork to %d stack %p sig = %d", + (int)p2->p_pid, args->stack, exit_signal); + if (args->flags & LINUX_CLONE_VFORK) { PROC_LOCK(p2); p2->p_flag |= P_PPWAIT; @@ -766,12 +753,9 @@ linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { struct l_mmap_argv linux_args; -#ifdef DEBUG - if (ldebug(mmap2)) - printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"), - args->addr, args->len, args->prot, - args->flags, args->fd, args->pgoff); -#endif + LINUX_CTR6(mmap2, "0x%08x %d %d 0x%08x %d %d", + args->addr, args->len, args->prot, + args->flags, args->fd, args->pgoff); linux_args.addr = PTROUT(args->addr); linux_args.len = args->len; @@ -793,12 +777,10 @@ linux_mmap(struct thread *td, struct linux_mmap_args *args) if (error) return (error); -#ifdef DEBUG - if (ldebug(mmap)) - printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"), - linux_args.addr, linux_args.len, linux_args.prot, - linux_args.flags, linux_args.fd, linux_args.pgoff); -#endif + LINUX_CTR6(mmap, "0x%08x %d %d 0x%08x %d %d", + linux_args.addr, linux_args.len, linux_args.prot, + linux_args.flags, linux_args.fd, linux_args.pgoff); + if ((linux_args.pgoff % PAGE_SIZE) != 0) return (EINVAL); linux_args.pgoff /= PAGE_SIZE; @@ -951,19 +933,11 @@ linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args) } bsd_args.pos = (off_t)linux_args->pgoff * PAGE_SIZE; -#ifdef DEBUG - if (ldebug(mmap)) - printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", - __func__, - (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot, - bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); -#endif error = mmap(td, &bsd_args); -#ifdef DEBUG - if (ldebug(mmap)) - printf("-> %s() return: 0x%x (0x%08x)\n", - __func__, error, (u_int)td->td_retval[0]); -#endif + + LINUX_CTR2(mmap, "return 0x%x (0x%08x)", + error, (u_int)td->td_retval[0]); + return (error); } @@ -972,6 +946,8 @@ linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { struct mprotect_args bsd_args; + LINUX_CTR2(mprotect, "addr %p len 0x%x", uap->addr, uap->len); + bsd_args.addr = uap->addr; bsd_args.len = uap->len; bsd_args.prot = uap->prot; @@ -985,6 +961,8 @@ linux_iopl(struct thread *td, struct linux_iopl_args *args) { int error; + LINUX_CTR1(iopl, "%d", args->level); + if (args->level < 0 || args->level > 3) return (EINVAL); if ((error = priv_check(td, PRIV_IO)) != 0) @@ -1003,10 +981,7 @@ linux_pipe(struct thread *td, struct linux_pipe_args *args) int error; int fildes[2]; -#ifdef DEBUG - if (ldebug(pipe)) - printf(ARGS(pipe, "*")); -#endif + LINUX_CTR(pipe); error = kern_pipe(td, fildes); if (error) @@ -1023,11 +998,8 @@ linux_sigaction(struct thread *td, struct linux_sigaction_args *args) l_sigaction_t act, oact; int error; -#ifdef DEBUG - if (ldebug(sigaction)) - printf(ARGS(sigaction, "%d, %p, %p"), - args->sig, (void *)args->nsa, (void *)args->osa); -#endif + LINUX_CTR3(sigaction, "%d %p %p", args->sig, + (void *)args->nsa, (void *)args->osa); if (args->nsa != NULL) { error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); @@ -1065,10 +1037,7 @@ linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) sigset_t sigmask; l_sigset_t mask; -#ifdef DEBUG - if (ldebug(sigsuspend)) - printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); -#endif + LINUX_CTR1(sigsuspend, "0x%08lx", (unsigned long)args->mask); LINUX_SIGEMPTYSET(mask); mask.__bits[0] = args->mask; @@ -1083,11 +1052,8 @@ linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) sigset_t sigmask; int error; -#ifdef DEBUG - if (ldebug(rt_sigsuspend)) - printf(ARGS(rt_sigsuspend, "%p, %d"), - (void *)uap->newset, uap->sigsetsize); -#endif + LINUX_CTR2(rt_sigsuspend, "%p %d", + (void *)uap->newset, uap->sigsetsize); if (uap->sigsetsize != sizeof(l_sigset_t)) return (EINVAL); @@ -1106,10 +1072,7 @@ linux_pause(struct thread *td, struct linux_pause_args *args) struct proc *p = td->td_proc; sigset_t sigmask; -#ifdef DEBUG - if (ldebug(pause)) - printf(ARGS(pause, "")); -#endif + LINUX_CTR(pause); PROC_LOCK(p); sigmask = td->td_sigmask; @@ -1124,10 +1087,7 @@ linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) l_stack_t lss; int error; -#ifdef DEBUG - if (ldebug(sigaltstack)) - printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); -#endif + LINUX_CTR2(sigaltstack, "%p %p", uap->uss, uap->uoss); if (uap->uss != NULL) { error = copyin(uap->uss, &lss, sizeof(l_stack_t)); @@ -1155,11 +1115,7 @@ linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) { struct ftruncate_args sa; -#ifdef DEBUG - if (ldebug(ftruncate64)) - printf(ARGS(ftruncate64, "%u, %jd"), args->fd, - (intmax_t)args->length); -#endif + LINUX_CTR2(ftruncate64, "%u %jd", args->fd, (intmax_t)args->length); sa.fd = args->fd; sa.length = args->length; @@ -1174,6 +1130,8 @@ linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap) struct timezone rtz; int error = 0; + LINUX_CTR(gettimeofday); + if (uap->tp) { microtime(&atv); atv32.tv_sec = atv.tv_sec; @@ -1196,6 +1154,8 @@ linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap) struct timezone atz, *tzp; int error; + LINUX_CTR(settimeofday); + if (uap->tp) { error = copyin(uap->tp, &atv32, sizeof(atv32)); if (error) @@ -1222,6 +1182,8 @@ linux_getrusage(struct thread *td, struct linux_getrusage_args *uap) struct rusage s; int error; + LINUX_CTR(getrusage); + error = kern_getrusage(td, uap->who, &s); if (error != 0) return (error); @@ -1257,6 +1219,8 @@ linux_sched_rr_get_interval(struct thread *td, struct l_timespec ts32; int error; + LINUX_CTR(sched_rr_get_interval); + error = kern_sched_rr_get_interval(td, uap->pid, &ts); if (error != 0) return (error); @@ -1278,15 +1242,11 @@ linux_set_thread_area(struct thread *td, if (error) return (error); -#ifdef DEBUG - if (ldebug(set_thread_area)) - printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, " - "%i, %i, %i"), info.entry_number, info.base_addr, - info.limit, info.seg_32bit, info.contents, - info.read_exec_only, info.limit_in_pages, - info.seg_not_present, info.useable); -#endif - + LINUX_CTR6(set_thread_area, "%i, %x, %x, %i, %i, %i", + info.entry_number, info.base_addr, info.limit, info.seg_32bit, + info.contents, info.read_exec_only); + LINUX_CTR3(set_thread_area, "%i, %i, %i", info.limit_in_pages, + info.seg_not_present, info.useable); /* * Semantics of Linux version: every thread in the system has array * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. @@ -1340,24 +1300,6 @@ linux_set_thread_area(struct thread *td, } memcpy(&sd, &a, sizeof(a)); -#ifdef DEBUG - if (ldebug(set_thread_area)) - printf("Segment created in set_thread_area: " - "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, " - "type: %i, dpl: %i, p: %i, xx: %i, long: %i, " - "def32: %i, gran: %i\n", - sd.sd_lobase, - sd.sd_hibase, - sd.sd_lolimit, - sd.sd_hilimit, - sd.sd_type, - sd.sd_dpl, - sd.sd_p, - sd.sd_xx, - sd.sd_long, - sd.sd_def32, - sd.sd_gran); -#endif critical_enter(); td->td_pcb->pcb_gsbase = (register_t)info.base_addr; diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c index bcb6069..9e3e095 100644 --- a/sys/amd64/linux32/linux32_sysvec.c +++ b/sys/amd64/linux32/linux32_sysvec.c @@ -42,11 +42,13 @@ __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include #include #include +#include #include #include #include @@ -76,6 +78,7 @@ __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include @@ -91,6 +94,8 @@ MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); suword32(pos++, val); \ } while (0) +#define aligned(a, t) (rounddown((unsigned long)(a), sizeof(t)) == (unsigned long)(a)) + #if BYTE_ORDER == LITTLE_ENDIAN #define SHELLMAGIC 0x2123 /* #! */ #else @@ -106,6 +111,8 @@ MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); #define LINUX_SYS_linux_rt_sendsig 0 #define LINUX_SYS_linux_sendsig 0 +const char linux_platform[] = "i686"; +static int linux_szplatform; extern char linux_sigcode[]; extern int linux_szsigcode; @@ -125,7 +132,7 @@ static void exec_linux_setregs(struct thread *td, u_long entry, static void linux32_fixlimit(struct rlimit *rl, int which); extern LIST_HEAD(futex_list, futex) futex_list; -extern struct sx futex_sx; +extern struct mtx futex_mtx; static eventhandler_tag linux_exit_tag; static eventhandler_tag linux_schedtail_tag; @@ -246,7 +253,12 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp) { Elf32_Auxargs *args; Elf32_Addr *base; - Elf32_Addr *pos; + Elf32_Addr *pos, *uplatform; + struct linux32_ps_strings *arginfo; + + arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; + uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szsigcode - + linux_szplatform); KASSERT(curthread->td_proc == imgp->proc, ("unsafe elf_linux_fixup(), should be curproc")); @@ -254,8 +266,8 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp) args = (Elf32_Auxargs *)imgp->auxargs; pos = base + (imgp->args->argc + imgp->args->envc + 2); - if (args->execfd != -1) - AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd); + AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature); + AUXARGS_ENTRY_32(pos, LINUX_AT_CLKTCK, hz); AUXARGS_ENTRY_32(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY_32(pos, AT_PHENT, args->phent); AUXARGS_ENTRY_32(pos, AT_PHNUM, args->phnum); @@ -263,10 +275,14 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp) AUXARGS_ENTRY_32(pos, AT_FLAGS, args->flags); AUXARGS_ENTRY_32(pos, AT_ENTRY, args->entry); AUXARGS_ENTRY_32(pos, AT_BASE, args->base); + AUXARGS_ENTRY_32(pos, LINUX_AT_SECURE, 0); AUXARGS_ENTRY_32(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); AUXARGS_ENTRY_32(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); AUXARGS_ENTRY_32(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); AUXARGS_ENTRY_32(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); + AUXARGS_ENTRY_32(pos, LINUX_AT_PLATFORM, PTROUT(uplatform)); + if (args->execfd != -1) + AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd); AUXARGS_ENTRY_32(pos, AT_NULL, 0); free(imgp->auxargs, M_TEMP); @@ -301,11 +317,9 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) regs = td->td_frame; oonstack = sigonstack(regs->tf_rsp); -#ifdef DEBUG - if (ldebug(rt_sendsig)) - printf(ARGS(rt_sendsig, "%p, %d, %p, %u"), - catcher, sig, (void*)mask, code); -#endif + LINUX_CTR3(rt_sendsig, "catcher %p sig %d code %u", catcher, sig, code); + LINUX_CTR2(rt_sendsig, "sp 0x%x ip 0x%x", regs->tf_rsp, regs->tf_rip); + /* * Allocate space for the signal handler context. */ @@ -371,23 +385,19 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_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%lx, 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 + LINUX_CTR4(rt_sendsig, "flags 0x%x, sp: %p, ss: 0x%lx, 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); 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 + + LINUX_CTR2(rt_sendsig, "bad stack %p, oonstack=%x", + fp, oonstack); + PROC_LOCK(p); sigexit(td, SIGILL); } @@ -447,12 +457,8 @@ linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) regs = td->td_frame; oonstack = sigonstack(regs->tf_rsp); -#ifdef DEBUG - if (ldebug(sendsig)) - printf(ARGS(sendsig, "%p, %d, %p, %u"), - catcher, sig, (void*)mask, code); -#endif - + LINUX_CTR3(sendsig, "catcher %p sig %d code %u", catcher, sig, code); + LINUX_CTR2(sendsig, "sp 0x%x ip 0x%x", regs->tf_rsp, regs->tf_rip); /* * Allocate space for the signal handler context. */ @@ -511,6 +517,10 @@ linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) * Process has trashed its stack; give it an illegal * instruction to halt it in its tracks. */ + + LINUX_CTR2(sendsig, "bad stack %p, oonstack=%x", + fp, oonstack); + PROC_LOCK(p); sigexit(td, SIGILL); } @@ -554,10 +564,7 @@ linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) regs = td->td_frame; -#ifdef DEBUG - if (ldebug(sigreturn)) - printf(ARGS(sigreturn, "%p"), (void *)args->sfp); -#endif + LINUX_CTR1(sigreturn, "%p", (void *)args->sfp); /* * The trampoline code hands us the sigframe. * It is unsafe to keep track of it ourselves, in the event that a @@ -653,10 +660,7 @@ linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) regs = td->td_frame; -#ifdef DEBUG - if (ldebug(rt_sigreturn)) - printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp); -#endif + LINUX_CTR1(rt_sigreturn, "%p", (void *)args->ucp); /* * The trampoline code hands us the ucontext. * It is unsafe to keep track of it ourselves, in the event that a @@ -732,11 +736,9 @@ linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 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%lx, mask: 0x%x"), - ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask); -#endif + LINUX_CTR4(rt_sigreturn, "flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x", + ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask); + (void)kern_sigaltstack(td, &ss, NULL); return (EJUSTRETURN); @@ -857,23 +859,27 @@ linux_copyout_strings(struct image_params *imgp) char *stringp, *destp; u_int32_t *stack_base; struct linux32_ps_strings *arginfo; - int sigcodesz; /* * Calculate string base and vector table pointers. * Also deal with signal trampoline code for this exec type. */ arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; - sigcodesz = *(imgp->proc->p_sysent->sv_szsigcode); - destp = (caddr_t)arginfo - sigcodesz - SPARE_USRSPACE - - roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *)); + destp = (caddr_t)arginfo - linux_szsigcode - SPARE_USRSPACE - + linux_szplatform - roundup((ARG_MAX - imgp->args->stringspace), + sizeof(char *)); /* * install sigcode */ - if (sigcodesz) - copyout(imgp->proc->p_sysent->sv_sigcode, - ((caddr_t)arginfo - sigcodesz), sigcodesz); + copyout(imgp->proc->p_sysent->sv_sigcode, + ((caddr_t)arginfo - linux_szsigcode), linux_szsigcode); + + /* + * Install LINUX_PLATFORM + */ + copyout(linux_platform, ((caddr_t)arginfo - linux_szsigcode - + linux_szplatform), linux_szplatform); /* * If we have a valid auxargs ptr, prepare some room @@ -885,7 +891,7 @@ linux_copyout_strings(struct image_params *imgp) * lower compatibility. */ imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size : - (AT_COUNT * 2); + (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 @@ -919,14 +925,14 @@ linux_copyout_strings(struct image_params *imgp) /* * Fill in "ps_strings" struct for ps, w, etc. */ - suword32(&arginfo->ps_argvstr, (u_int32_t)(intptr_t)vectp); + suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp); suword32(&arginfo->ps_nargvstr, argc); /* * Fill in argument portion of vector table. */ for (; argc > 0; --argc) { - suword32(vectp++, (u_int32_t)(intptr_t)destp); + suword32(vectp++, (uint32_t)(intptr_t)destp); while (*stringp++ != 0) destp++; destp++; @@ -935,14 +941,14 @@ linux_copyout_strings(struct image_params *imgp) /* a null vector table pointer separates the argp's from the envp's */ suword32(vectp++, 0); - suword32(&arginfo->ps_envstr, (u_int32_t)(intptr_t)vectp); + suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp); suword32(&arginfo->ps_nenvstr, envc); /* * Fill in environment portion of vector table. */ for (; envc > 0; --envc) { - suword32(vectp++, (u_int32_t)(intptr_t)destp); + suword32(vectp++, (uint32_t)(intptr_t)destp); while (*stringp++ != 0) destp++; destp++; @@ -999,6 +1005,54 @@ linux32_fixlimit(struct rlimit *rl, int which) } } +static const char GNULINUX_ABI_VENDOR[] = "GNU"; +static const int GNULINUX_ABI_LEN = 16; + +static int +linux32_abi_note(struct image_params *imgp, int32_t *osrel) +{ + const Elf_Note *note, *note_end; + const Elf32_Phdr *phdr, *pnote; + const Elf32_Ehdr *hdr; + const char *note_name; + int i; + + pnote = NULL; + hdr = (const Elf32_Ehdr *)imgp->image_header; + phdr = (const Elf32_Phdr *)(imgp->image_header + hdr->e_phoff); + + for (i = 0; i < hdr->e_phnum; i++) + if (phdr[i].p_type == PT_NOTE) { + pnote = &phdr[i]; + break; + } + + if (pnote == NULL || pnote->p_offset >= PAGE_SIZE || + pnote->p_offset + pnote->p_filesz >= PAGE_SIZE) + return (0); + + note = (const Elf_Note *)(imgp->image_header + pnote->p_offset); + if (!aligned(note, Elf32_Addr)) + return (0); + note_end = (const Elf_Note *)(imgp->image_header + + pnote->p_offset + pnote->p_filesz); + while (note < note_end) { + if (note->n_namesz == sizeof(GNULINUX_ABI_VENDOR) && + note->n_descsz == GNULINUX_ABI_LEN && + note->n_type == 1 /* ABI_NOTETYPE */) { + note_name = (const char *)(note + 1); + if (strncmp(GNULINUX_ABI_VENDOR, note_name, + sizeof(GNULINUX_ABI_VENDOR)) == 0) + return (1); + } + note = (const Elf_Note *)((const char *)(note + 1) + + roundup2(note->n_namesz, sizeof(Elf32_Addr)) + + roundup2(note->n_descsz, sizeof(Elf32_Addr))); + } + + return (0); +} + struct sysentvec elf_linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, @@ -1038,7 +1092,8 @@ static Elf32_Brandinfo linux_brand = { .interp_path = "/lib/ld-linux.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = linux32_abi_note, + .flags = BI_CAN_EXEC_DYN | BI_CAN_EXEC_INTERP }; static Elf32_Brandinfo linux_glibc2brand = { @@ -1049,7 +1104,8 @@ static Elf32_Brandinfo linux_glibc2brand = { .interp_path = "/lib/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = linux32_abi_note, + .flags = BI_CAN_EXEC_DYN | BI_CAN_EXEC_INTERP }; Elf32_Brandinfo *linux_brandlist[] = { @@ -1082,7 +1138,7 @@ linux_elf_modevent(module_t mod, int type, void *data) mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF); sx_init(&emul_shared_lock, "emuldata->shared lock"); LIST_INIT(&futex_list); - sx_init(&futex_sx, "futex protection lock"); + mtx_init(&futex_mtx, "futex list protection lock", NULL, MTX_DEF); linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit, NULL, 1000); linux_schedtail_tag = EVENTHANDLER_REGISTER(schedtail, @@ -1091,6 +1147,8 @@ linux_elf_modevent(module_t mod, int type, void *data) linux_proc_exec, NULL, 1000); if (bootverbose) printf("Linux ELF exec handler installed\n"); + linux_szplatform = roundup(strlen(linux_platform) + 1, + sizeof(char *)); } else printf("cannot insert Linux ELF brand handler\n"); break; @@ -1112,7 +1170,7 @@ linux_elf_modevent(module_t mod, int type, void *data) linux_device_unregister_handler(*ldhp); mtx_destroy(&emul_lock); sx_destroy(&emul_shared_lock); - sx_destroy(&futex_sx); + mtx_destroy(&futex_mtx); EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag); EVENTHANDLER_DEREGISTER(schedtail, linux_schedtail_tag); EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag); diff --git a/sys/arm/arm/elf_machdep.c b/sys/arm/arm/elf_machdep.c index 693eab1..e65f947 100644 --- a/sys/arm/arm/elf_machdep.c +++ b/sys/arm/arm/elf_machdep.c @@ -84,7 +84,8 @@ static Elf32_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_ANY, @@ -99,7 +100,8 @@ static Elf32_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/compat/ia32/ia32_sysvec.c b/sys/compat/ia32/ia32_sysvec.c index 0b32b9a..92b990a 100644 --- a/sys/compat/ia32/ia32_sysvec.c +++ b/sys/compat/ia32/ia32_sysvec.c @@ -148,6 +148,7 @@ static Elf32_Brandinfo ia32_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &ia32_freebsd_sysvec, .interp_newpath = "/libexec/ld-elf32.so.1", + .brand_abi_note = __elfN(freebsd_abi_note), .flags = BI_CAN_EXEC_DYN }; @@ -163,7 +164,8 @@ static Elf32_Brandinfo ia32_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &ia32_freebsd_sysvec, .interp_newpath = "/libexec/ld-elf32.so.1", - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oia32, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/compat/linux/linux_elf.h b/sys/compat/linux/linux_elf.h new file mode 100644 index 0000000..f601633 --- /dev/null +++ b/sys/compat/linux/linux_elf.h @@ -0,0 +1,45 @@ +/*- + * Copyright (c) 2008 Chagin Dmitry + * 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. + * + * 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. + * + * $FreeBSD$ + */ + +#ifndef _LINUX_ELF_H_ +#define _LINUX_ELF_H_ + +/* + * Non-standard aux entry types used in Linux ELF binaries. + */ + +#define LINUX_AT_PLATFORM 15 /* String identifying CPU */ +#define LINUX_AT_HWCAP 16 /* CPU capabilities */ +#define LINUX_AT_CLKTCK 17 /* frequency at which times() increments */ +#define LINUX_AT_SECURE 23 /* secure mode boolean */ +#define LINUX_AT_BASE_PLATFORM 24 /* string identifying real platform, may + * differ from AT_PLATFORM. + */ +#define LINUX_AT_EXECFN 31 /* filename of program */ + +#endif /* !_LINUX_ELF_H_ */ diff --git a/sys/compat/linux/linux_emul.c b/sys/compat/linux/linux_emul.c index 1ca48a3..51b197d 100644 --- a/sys/compat/linux/linux_emul.c +++ b/sys/compat/linux/linux_emul.c @@ -35,6 +35,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -54,6 +55,7 @@ __FBSDID("$FreeBSD$"); #include #include +#include struct sx emul_shared_lock; struct mtx emul_lock; @@ -317,10 +319,7 @@ linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args { struct linux_emuldata *em; -#ifdef DEBUG - if (ldebug(set_tid_address)) - printf(ARGS(set_tid_address, "%p"), args->tidptr); -#endif + LINUX_CTR1(set_tid_address, "%p", args->tidptr); /* find the emuldata */ em = em_find(td->td_proc, EMUL_DOLOCK); diff --git a/sys/compat/linux/linux_file.c b/sys/compat/linux/linux_file.c index 6643e71..e1d8a05 100644 --- a/sys/compat/linux/linux_file.c +++ b/sys/compat/linux/linux_file.c @@ -39,6 +39,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -72,19 +73,17 @@ __FBSDID("$FreeBSD$"); int linux_creat(struct thread *td, struct linux_creat_args *args) { - char *path; - int error; + char *path; + int error; - LCONVPATHEXIST(td, args->path, &path); + LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(creat)) - printf(ARGS(creat, "%s, %d"), path, args->mode); -#endif - error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC, - args->mode); - LFREEPATH(path); - return (error); + LINUX_CTR(creat); + + error = kern_open(td, path, UIO_SYSSPACE, O_WRONLY | O_CREAT | O_TRUNC, + args->mode); + LFREEPATH(path); + return (error); } @@ -170,10 +169,8 @@ linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mod } } -#ifdef DEBUG - if (ldebug(open)) - printf(LMSG("open returns error %d"), error); -#endif + LINUX_CTR1(open, "returns error %d", error); + LFREEPATH(path); return (error); } @@ -189,29 +186,23 @@ linux_openat(struct thread *td, struct linux_openat_args *args) LCONVPATH_AT(td, args->filename, &path, 1, dfd); else LCONVPATH_AT(td, args->filename, &path, 0, dfd); -#ifdef DEBUG - if (ldebug(openat)) - printf(ARGS(openat, "%i, %s, 0x%x, 0x%x"), args->dfd, - path, args->flags, args->mode); -#endif + + LINUX_CTR(openat); + return (linux_common_open(td, dfd, path, args->flags, args->mode)); } int linux_open(struct thread *td, struct linux_open_args *args) { - char *path; + char *path; - if (args->flags & LINUX_O_CREAT) - LCONVPATHCREAT(td, args->path, &path); - else - LCONVPATHEXIST(td, args->path, &path); + if (args->flags & LINUX_O_CREAT) + LCONVPATHCREAT(td, args->path, &path); + else + LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(open)) - printf(ARGS(open, "%s, 0x%x, 0x%x"), - path, args->flags, args->mode); -#endif + LINUX_CTR(open); return (linux_common_open(td, AT_FDCWD, path, args->flags, args->mode)); } @@ -219,25 +210,20 @@ linux_open(struct thread *td, struct linux_open_args *args) int linux_lseek(struct thread *td, struct linux_lseek_args *args) { + struct lseek_args /* { + int fd; + int pad; + off_t offset; + int whence; + } */ tmp_args; - struct lseek_args /* { - int fd; - int pad; - off_t offset; - int whence; - } */ tmp_args; - int error; - -#ifdef DEBUG - if (ldebug(lseek)) - printf(ARGS(lseek, "%d, %ld, %d"), - args->fdes, (long)args->off, args->whence); -#endif - tmp_args.fd = args->fdes; - tmp_args.offset = (off_t)args->off; - tmp_args.whence = args->whence; - error = lseek(td, &tmp_args); - return error; + LINUX_CTR3(lseek, "%d %ld %d", args->fdes, + (long)args->off, args->whence); + + tmp_args.fd = args->fdes; + tmp_args.offset = (off_t)args->off; + tmp_args.whence = args->whence; + return (lseek(td, &tmp_args)); } int @@ -247,11 +233,9 @@ linux_llseek(struct thread *td, struct linux_llseek_args *args) int error; off_t off; -#ifdef DEBUG - if (ldebug(llseek)) - printf(ARGS(llseek, "%d, %d:%d, %d"), - args->fd, args->ohigh, args->olow, args->whence); -#endif + LINUX_CTR4(llseek, "%d %d:%d %d", args->fd, args->ohigh, + args->olow, args->whence); + off = (args->olow) | (((off_t) args->ohigh) << 32); bsd_args.fd = args->fd; @@ -259,13 +243,13 @@ linux_llseek(struct thread *td, struct linux_llseek_args *args) bsd_args.whence = args->whence; if ((error = lseek(td, &bsd_args))) - return error; + return (error); if ((error = copyout(td->td_retval, args->res, sizeof (off_t)))) - return error; + return (error); td->td_retval[0] = 0; - return 0; + return (0); } int @@ -276,7 +260,7 @@ linux_readdir(struct thread *td, struct linux_readdir_args *args) lda.fd = args->fd; lda.dent = args->dent; lda.count = 1; - return linux_getdents(td, &lda); + return (linux_getdents(td, &lda)); } /* @@ -537,10 +521,7 @@ int linux_getdents(struct thread *td, struct linux_getdents_args *args) { -#ifdef DEBUG - if (ldebug(getdents)) - printf(ARGS(getdents, "%d, *, %d"), args->fd, args->count); -#endif + LINUX_CTR2(getdents, "%d %d", args->fd, args->count); return (getdents_common(td, (struct linux_getdents64_args*)args, 0)); } @@ -549,10 +530,7 @@ int linux_getdents64(struct thread *td, struct linux_getdents64_args *args) { -#ifdef DEBUG - if (ldebug(getdents64)) - printf(ARGS(getdents64, "%d, *, %d"), args->fd, args->count); -#endif + LINUX_CTR2(getdents64, "%d %d", args->fd, args->count); return (getdents_common(td, args, 1)); } @@ -573,10 +551,8 @@ linux_access(struct thread *td, struct linux_access_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(access)) - printf(ARGS(access, "%s, %d"), path, args->flags); -#endif + LINUX_CTR(access); + error = kern_access(td, path, UIO_SYSSPACE, args->flags); LFREEPATH(path); @@ -596,10 +572,7 @@ linux_faccessat(struct thread *td, struct linux_faccessat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->filename, &path, dfd); -#ifdef DEBUG - if (ldebug(access)) - printf(ARGS(access, "%s, %d"), path, args->mode); -#endif + LINUX_CTR(faccessat); error = kern_accessat(td, dfd, path, UIO_SYSSPACE, 0 /* XXX */, args->mode); @@ -617,10 +590,7 @@ linux_unlink(struct thread *td, struct linux_unlink_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(unlink)) - printf(ARGS(unlink, "%s"), path); -#endif + LINUX_CTR(unlink); error = kern_unlink(td, path, UIO_SYSSPACE); if (error == EPERM) @@ -645,10 +615,7 @@ linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); -#ifdef DEBUG - if (ldebug(unlinkat)) - printf(ARGS(unlinkat, "%s"), path); -#endif + LINUX_CTR(unlinkat); if (args->flag & LINUX_AT_REMOVEDIR) error = kern_rmdirat(td, dfd, path, UIO_SYSSPACE); @@ -671,10 +638,8 @@ linux_chdir(struct thread *td, struct linux_chdir_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(chdir)) - printf(ARGS(chdir, "%s"), path); -#endif + LINUX_CTR(chdir); + error = kern_chdir(td, path, UIO_SYSSPACE); LFREEPATH(path); return (error); @@ -688,10 +653,8 @@ linux_chmod(struct thread *td, struct linux_chmod_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(chmod)) - printf(ARGS(chmod, "%s, %d"), path, args->mode); -#endif + LINUX_CTR(chmod); + error = kern_chmod(td, path, UIO_SYSSPACE, args->mode); LFREEPATH(path); return (error); @@ -706,10 +669,7 @@ linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->filename, &path, dfd); -#ifdef DEBUG - if (ldebug(fchmodat)) - printf(ARGS(fchmodat, "%s, %d"), path, args->mode); -#endif + LINUX_CTR(fchmodat); error = kern_fchmodat(td, dfd, path, UIO_SYSSPACE, args->mode, 0); LFREEPATH(path); @@ -724,10 +684,8 @@ linux_mkdir(struct thread *td, struct linux_mkdir_args *args) LCONVPATHCREAT(td, args->path, &path); -#ifdef DEBUG - if (ldebug(mkdir)) - printf(ARGS(mkdir, "%s, %d"), path, args->mode); -#endif + LINUX_CTR(mkdir); + error = kern_mkdir(td, path, UIO_SYSSPACE, args->mode); LFREEPATH(path); return (error); @@ -742,10 +700,8 @@ linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHCREAT_AT(td, args->pathname, &path, dfd); -#ifdef DEBUG - if (ldebug(mkdirat)) - printf(ARGS(mkdirat, "%s, %d"), path, args->mode); -#endif + LINUX_CTR(mkdirat); + error = kern_mkdirat(td, dfd, path, UIO_SYSSPACE, args->mode); LFREEPATH(path); return (error); @@ -759,10 +715,8 @@ linux_rmdir(struct thread *td, struct linux_rmdir_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(rmdir)) - printf(ARGS(rmdir, "%s"), path); -#endif + LINUX_CTR(rmdir); + error = kern_rmdir(td, path, UIO_SYSSPACE); LFREEPATH(path); return (error); @@ -782,10 +736,8 @@ linux_rename(struct thread *td, struct linux_rename_args *args) return (error); } -#ifdef DEBUG - if (ldebug(rename)) - printf(ARGS(rename, "%s, %s"), from, to); -#endif + LINUX_CTR(rename); + error = kern_rename(td, from, to, UIO_SYSSPACE); LFREEPATH(from); LFREEPATH(to); @@ -808,10 +760,8 @@ linux_renameat(struct thread *td, struct linux_renameat_args *args) return (error); } -#ifdef DEBUG - if (ldebug(renameat)) - printf(ARGS(renameat, "%s, %s"), from, to); -#endif + LINUX_CTR(renameat); + error = kern_renameat(td, olddfd, from, newdfd, to, UIO_SYSSPACE); LFREEPATH(from); LFREEPATH(to); @@ -832,10 +782,8 @@ linux_symlink(struct thread *td, struct linux_symlink_args *args) return (error); } -#ifdef DEBUG - if (ldebug(symlink)) - printf(ARGS(symlink, "%s, %s"), path, to); -#endif + LINUX_CTR(symlink); + error = kern_symlink(td, path, to, UIO_SYSSPACE); LFREEPATH(path); LFREEPATH(to); @@ -857,10 +805,7 @@ linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args) return (error); } -#ifdef DEBUG - if (ldebug(symlinkat)) - printf(ARGS(symlinkat, "%s, %s"), path, to); -#endif + LINUX_CTR(symlinkat); error = kern_symlinkat(td, path, dfd, to, UIO_SYSSPACE); LFREEPATH(path); @@ -876,11 +821,8 @@ linux_readlink(struct thread *td, struct linux_readlink_args *args) LCONVPATHEXIST(td, args->name, &name); -#ifdef DEBUG - if (ldebug(readlink)) - printf(ARGS(readlink, "%s, %p, %d"), name, (void *)args->buf, - args->count); -#endif + LINUX_CTR(readlink); + error = kern_readlink(td, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE, args->count); LFREEPATH(name); @@ -896,11 +838,7 @@ linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->path, &name, dfd); -#ifdef DEBUG - if (ldebug(readlinkat)) - printf(ARGS(readlinkat, "%s, %p, %d"), name, (void *)args->buf, - args->bufsiz); -#endif + LINUX_CTR(readlinkat); error = kern_readlinkat(td, dfd, name, UIO_SYSSPACE, args->buf, UIO_USERSPACE, args->bufsiz); @@ -916,10 +854,7 @@ linux_truncate(struct thread *td, struct linux_truncate_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(truncate)) - printf(ARGS(truncate, "%s, %ld"), path, (long)args->length); -#endif + LINUX_CTR(truncate); error = kern_truncate(td, path, UIO_SYSSPACE, args->length); LFREEPATH(path); @@ -934,10 +869,7 @@ linux_truncate64(struct thread *td, struct linux_truncate64_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(truncate64)) - printf(ARGS(truncate64, "%s, %jd"), path, args->length); -#endif + LINUX_CTR(truncate64); error = kern_truncate(td, path, UIO_SYSSPACE, args->length); LFREEPATH(path); @@ -950,7 +882,9 @@ linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args) int fd; int pad; off_t length; - } */ nuap; + } */ nuap; + + LINUX_CTR(ftruncate); nuap.fd = args->fd; nuap.length = args->length; @@ -971,10 +905,8 @@ linux_link(struct thread *td, struct linux_link_args *args) return (error); } -#ifdef DEBUG - if (ldebug(link)) - printf(ARGS(link, "%s, %s"), path, to); -#endif + LINUX_CTR(link); + error = kern_link(td, path, to, UIO_SYSSPACE); LFREEPATH(path); LFREEPATH(to); @@ -1004,11 +936,7 @@ linux_linkat(struct thread *td, struct linux_linkat_args *args) return (error); } -#ifdef DEBUG - if (ldebug(linkat)) - printf(ARGS(linkat, "%i, %s, %i, %s, %i"), args->olddfd, path, - args->newdfd, to, args->flags); -#endif + LINUX_CTR(linkat); error = kern_linkat(td, olddfd, newdfd, path, to, UIO_SYSSPACE, FOLLOW); LFREEPATH(path); @@ -1023,8 +951,10 @@ linux_fdatasync(td, uap) { struct fsync_args bsd; + LINUX_CTR(fdatasync); + bsd.fd = uap->fd; - return fsync(td, &bsd); + return (fsync(td, &bsd)); } int @@ -1068,7 +998,7 @@ linux_pwrite(td, uap) bsd.buf = uap->buf; bsd.nbyte = uap->nbyte; bsd.offset = uap->offset; - return pwrite(td, &bsd); + return (pwrite(td, &bsd)); } int @@ -1092,11 +1022,7 @@ linux_mount(struct thread *td, struct linux_mount_args *args) if (error) return (error); -#ifdef DEBUG - if (ldebug(mount)) - printf(ARGS(mount, "%s, %s, %s"), - fstypename, mntfromname, mntonname); -#endif + LINUX_CTR(mount); if (strcmp(fstypename, "ext2") == 0) { strcpy(fstypename, "ext2fs"); @@ -1405,10 +1331,7 @@ linux_fcntl(struct thread *td, struct linux_fcntl_args *args) { struct linux_fcntl64_args args64; -#ifdef DEBUG - if (ldebug(fcntl)) - printf(ARGS(fcntl, "%d, %08x, *"), args->fd, args->cmd); -#endif + LINUX_CTR2(fcntl, "%d %08x", args->fd, args->cmd); args64.fd = args->fd; args64.cmd = args->cmd; @@ -1424,10 +1347,7 @@ linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args) struct flock bsd_flock; int error; -#ifdef DEBUG - if (ldebug(fcntl64)) - printf(ARGS(fcntl64, "%d, %08x, *"), args->fd, args->cmd); -#endif + LINUX_CTR2(fcntl64, "%d %08x", args->fd, args->cmd); switch (args->cmd) { case LINUX_F_GETLK64: @@ -1474,10 +1394,8 @@ linux_chown(struct thread *td, struct linux_chown_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(chown)) - printf(ARGS(chown, "%s, %d, %d"), path, args->uid, args->gid); -#endif + LINUX_CTR(chown); + error = kern_chown(td, path, UIO_SYSSPACE, args->uid, args->gid); LFREEPATH(path); return (error); @@ -1495,10 +1413,7 @@ linux_fchownat(struct thread *td, struct linux_fchownat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->filename, &path, dfd); -#ifdef DEBUG - if (ldebug(fchownat)) - printf(ARGS(fchownat, "%s, %d, %d"), path, args->uid, args->gid); -#endif + LINUX_CTR(fchownat); follow = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 : AT_SYMLINK_NOFOLLOW; @@ -1516,10 +1431,8 @@ linux_lchown(struct thread *td, struct linux_lchown_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(lchown)) - printf(ARGS(lchown, "%s, %d, %d"), path, args->uid, args->gid); -#endif + LINUX_CTR(lchown); + error = kern_lchown(td, path, UIO_SYSSPACE, args->uid, args->gid); LFREEPATH(path); return (error); diff --git a/sys/compat/linux/linux_futex.c b/sys/compat/linux/linux_futex.c index b19d290..312ec42 100644 --- a/sys/compat/linux/linux_futex.c +++ b/sys/compat/linux/linux_futex.c @@ -40,19 +40,21 @@ __KERNEL_RCSID(1, "$NetBSD: linux_futex.c,v 1.7 2006/07/24 19:01:49 manu Exp $") #include "opt_compat.h" #include -#include -#include #include -#include -#include +#include #include +#include #include #include +#include #include +#include +#include +#include #include #include -#include +#include #ifdef COMPAT_LINUX32 #include #include @@ -60,424 +62,341 @@ __KERNEL_RCSID(1, "$NetBSD: linux_futex.c,v 1.7 2006/07/24 19:01:49 manu Exp $") #include #include #endif -#include #include +#include +#include struct futex; struct waiting_proc { - struct thread *wp_t; - struct futex *wp_new_futex; + struct futex *w_ftx; + pid_t w_pid; + uint32_t w_flags; TAILQ_ENTRY(waiting_proc) wp_list; }; + struct futex { - void *f_uaddr; - int f_refcount; + struct sx lock; + uint32_t *f_uaddr; + uint32_t f_refcount; LIST_ENTRY(futex) f_list; TAILQ_HEAD(lf_waiting_proc, waiting_proc) f_waiting_proc; }; +struct fx { + uint32_t flags; + struct futex *ftx; + struct waiting_proc *wp; +}; + LIST_HEAD(futex_list, futex) futex_list; -struct sx futex_sx; /* this protects the LIST of futexes */ -#define FUTEX_LOCK sx_xlock(&futex_sx) -#define FUTEX_UNLOCK sx_xunlock(&futex_sx) +#define FUTEX_LOCK(f) sx_xlock(&f->lock) +#define FUTEX_UNLOCK(f) sx_xunlock(&f->lock) +#define FUTEX_LOCK_INIT(f, o) sx_init_flags(&f->lock, "futex lock", o) +#define FUTEX_LOCK_DESTROY(f) sx_destroy(&f->lock) +#define FUTEX_LOCK_ASSERT(f, type) sx_assert(&f->lock, type) -#define FUTEX_LOCKED 1 -#define FUTEX_UNLOCKED 0 +struct mtx futex_mtx; /* this protects the futex list */ +#define FUTEXES_LOCK mtx_lock(&futex_mtx) +#define FUTEXES_UNLOCK mtx_unlock(&futex_mtx) -#define FUTEX_SYSTEM_LOCK mtx_lock(&Giant) -#define FUTEX_SYSTEM_UNLOCK mtx_unlock(&Giant) +#define FUTEX_CREATE_WP 0x1 /* create waiting_proc */ +#define FUTEX_DONTCREATE 0x2 /* don't create futex if not exists */ -static struct futex *futex_get(void *, int); -static void futex_put(struct futex *); -static int futex_sleep(struct futex *, struct thread *, unsigned long); -static int futex_wake(struct futex *, int, struct futex *, int); -static int futex_atomic_op(struct thread *td, int encoded_op, caddr_t uaddr); +#define FUTEX_WP_REQUEUED 0x1 /* waiting_proc requeued */ +#define FUTEX_WP_REMOVED 0x2 /* waiting_proc removed from tailq*/ /* support.s */ -int futex_xchgl(int oparg, caddr_t uaddr, int *oldval); -int futex_addl(int oparg, caddr_t uaddr, int *oldval); -int futex_orl(int oparg, caddr_t uaddr, int *oldval); -int futex_andl(int oparg, caddr_t uaddr, int *oldval); -int futex_xorl(int oparg, caddr_t uaddr, int *oldval); - -int -linux_sys_futex(struct thread *td, struct linux_sys_futex_args *args) -{ - int val; - int ret; - struct l_timespec timeout = {0, 0}; - int error = 0; - struct futex *f; - struct futex *newf; - int timeout_hz; - struct timeval tv = {0, 0}; - struct futex *f2; - int op_ret; - -#ifdef DEBUG - if (ldebug(sys_futex)) - printf(ARGS(futex, "%p, %i, %i, *, %p, %i"), args->uaddr, args->op, - args->val, args->uaddr2, args->val3); -#endif - - /* - * Our implementation provides only privates futexes. Most of the apps - * should use private futexes but don't claim so. Therefore we treat - * all futexes as private by clearing the FUTEX_PRIVATE_FLAG. It works - * in most cases (ie. when futexes are not shared on file descriptor - * or between different processes.). - */ - args->op = (args->op & ~LINUX_FUTEX_PRIVATE_FLAG); - - switch (args->op) { - case LINUX_FUTEX_WAIT: - FUTEX_SYSTEM_LOCK; - - if ((error = copyin(args->uaddr, - &val, sizeof(val))) != 0) { - FUTEX_SYSTEM_UNLOCK; - return error; - } - - if (val != args->val) { - FUTEX_SYSTEM_UNLOCK; - return EWOULDBLOCK; - } - - if (args->timeout != NULL) { - if ((error = copyin(args->timeout, - &timeout, sizeof(timeout))) != 0) { - FUTEX_SYSTEM_UNLOCK; - return error; - } - } +int futex_xchgl(int oparg, uint32_t *uaddr, int *oldval); +int futex_addl(int oparg, uint32_t *uaddr, int *oldval); +int futex_orl(int oparg, uint32_t *uaddr, int *oldval); +int futex_andl(int oparg, uint32_t *uaddr, int *oldval); +int futex_xorl(int oparg, uint32_t *uaddr, int *oldval); #ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAIT %d: val = %d, uaddr = %p, " - "*uaddr = %d, timeout = %d.%09lu\n", - td->td_proc->p_pid, args->val, - args->uaddr, val, timeout.tv_sec, - (unsigned long)timeout.tv_nsec); +#define LINUX_SYS_linux_release_futexes 0 #endif - tv.tv_usec = timeout.tv_sec * 1000000 + timeout.tv_nsec / 1000; - timeout_hz = tvtohz(&tv); - - if (timeout.tv_sec == 0 && timeout.tv_nsec == 0) - timeout_hz = 0; - /* - * If the user process requests a non null timeout, - * make sure we do not turn it into an infinite - * timeout because timeout_hz gets null. - * - * We use a minimal timeout of 1/hz. Maybe it would - * make sense to just return ETIMEDOUT without sleeping. - */ - if (((timeout.tv_sec != 0) || (timeout.tv_nsec != 0)) && - (timeout_hz == 0)) - timeout_hz = 1; +static void +futex_put0(struct futex *f) +{ - f = futex_get(args->uaddr, FUTEX_UNLOCKED); - ret = futex_sleep(f, td, timeout_hz); - futex_put(f); + FUTEXES_LOCK; + if (refcount_release(&f->f_refcount)) { + LIST_REMOVE(f, f_list); + FUTEXES_UNLOCK; -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAIT %d: uaddr = %p, " - "ret = %d\n", td->td_proc->p_pid, args->uaddr, ret); -#endif + FUTEX_UNLOCK(f); + LINUX_CTR2(sys_futex, "futex_put destroy uaddr %p ref %d", + f->f_uaddr, f->f_refcount); + FUTEX_LOCK_DESTROY(f); + free(f, M_LINUX); + return; + } - FUTEX_SYSTEM_UNLOCK; - switch (ret) { - case EWOULDBLOCK: /* timeout */ - return ETIMEDOUT; - break; - case EINTR: /* signal */ - return EINTR; - break; - case 0: /* FUTEX_WAKE received */ -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAIT %d: uaddr = %p, " - "got FUTEX_WAKE\n", - td->td_proc->p_pid, args->uaddr); -#endif - return 0; - break; - default: -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAIT: unexpected ret = %d\n", - ret); -#endif - break; - } + LINUX_CTR2(sys_futex, "futex_put uaddr %p ref %d", + f->f_uaddr, f->f_refcount); + FUTEXES_UNLOCK; + FUTEX_UNLOCK(f); +} - /* NOTREACHED */ - break; +static void +futex_put(struct fx *f) +{ - case LINUX_FUTEX_WAKE: - FUTEX_SYSTEM_LOCK; + FUTEX_LOCK_ASSERT(f->ftx, SA_XLOCKED); + if (f->flags & FUTEX_CREATE_WP && f->wp != NULL) { + if ((f->wp->w_flags & FUTEX_WP_REMOVED) == 0) + TAILQ_REMOVE(&f->ftx->f_waiting_proc, f->wp, wp_list); + free(f->wp, M_LINUX); + f->wp = NULL; + } - /* - * XXX: Linux is able to cope with different addresses - * corresponding to the same mapped memory in the sleeping - * and waker process(es). - */ -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAKE %d: uaddr = %p, val = %d\n", - td->td_proc->p_pid, args->uaddr, args->val); -#endif - f = futex_get(args->uaddr, FUTEX_UNLOCKED); - td->td_retval[0] = futex_wake(f, args->val, NULL, 0); - futex_put(f); + futex_put0(f->ftx); +} - FUTEX_SYSTEM_UNLOCK; - break; - case LINUX_FUTEX_CMP_REQUEUE: - FUTEX_SYSTEM_LOCK; +static void +futex_get0(uint32_t *uaddr, struct fx *fx, struct futex **newf, + uint32_t flags, int sx_opts) +{ + struct futex *f; - if ((error = copyin(args->uaddr, - &val, sizeof(val))) != 0) { - FUTEX_SYSTEM_UNLOCK; - return error; - } + fx->flags = flags; - if (val != args->val3) { - FUTEX_SYSTEM_UNLOCK; - return EAGAIN; +retry: + FUTEXES_LOCK; + LIST_FOREACH(f, &futex_list, f_list) { + if (f->f_uaddr == uaddr) { + refcount_acquire(&f->f_refcount); + FUTEXES_UNLOCK; + + fx->ftx = f; + if (*newf != NULL) + free(*newf, M_LINUX); + + FUTEX_LOCK(f); + if (flags & FUTEX_CREATE_WP) { + TAILQ_INSERT_HEAD(&f->f_waiting_proc, fx->wp, + wp_list); + fx->wp->w_ftx = f; + } + LINUX_CTR3(sys_futex, "futex_get uaddr %p wp %p ref %d", + uaddr, fx->wp, f->f_refcount); + return; } + } - f = futex_get(args->uaddr, FUTEX_UNLOCKED); - newf = futex_get(args->uaddr2, FUTEX_UNLOCKED); - td->td_retval[0] = futex_wake(f, args->val, newf, - (int)(unsigned long)args->timeout); - futex_put(f); - futex_put(newf); - - FUTEX_SYSTEM_UNLOCK; - break; - - case LINUX_FUTEX_REQUEUE: - FUTEX_SYSTEM_LOCK; + fx->ftx = *newf; + if (flags & FUTEX_DONTCREATE) { + FUTEXES_UNLOCK; + LINUX_CTR1(sys_futex, "futex_get uaddr %p null", uaddr); + return; + } - f = futex_get(args->uaddr, FUTEX_UNLOCKED); - newf = futex_get(args->uaddr2, FUTEX_UNLOCKED); - td->td_retval[0] = futex_wake(f, args->val, newf, - (int)(unsigned long)args->timeout); - futex_put(f); - futex_put(newf); + if (fx->ftx == NULL) { + FUTEXES_UNLOCK; + *newf = malloc(sizeof(struct futex), M_LINUX, M_WAITOK | M_ZERO); + (*newf)->f_uaddr = uaddr; + refcount_init(&(*newf)->f_refcount, 1); + goto retry; + } - FUTEX_SYSTEM_UNLOCK; - break; + FUTEX_LOCK_INIT((*newf), sx_opts); + TAILQ_INIT(&(*newf)->f_waiting_proc); + LIST_INSERT_HEAD(&futex_list, *newf, f_list); + FUTEXES_UNLOCK; - case LINUX_FUTEX_FD: -#ifdef DEBUG - printf("linux_sys_futex: unimplemented op %d\n", - args->op); -#endif - return (ENOSYS); + FUTEX_LOCK((*newf)); + if (flags & FUTEX_CREATE_WP) { + TAILQ_INSERT_HEAD(&(*newf)->f_waiting_proc, fx->wp, wp_list); + fx->wp->w_ftx = *newf; + } - case LINUX_FUTEX_WAKE_OP: - FUTEX_SYSTEM_LOCK; -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX_WAKE_OP: %d: uaddr = %p, op = %d, " - "val = %x, uaddr2 = %p, val3 = %x\n", - td->td_proc->p_pid, args->uaddr, args->op, - args->val, args->uaddr2, args->val3); -#endif - f = futex_get(args->uaddr, FUTEX_UNLOCKED); - f2 = futex_get(args->uaddr2, FUTEX_UNLOCKED); + LINUX_CTR3(sys_futex, "futex_get uaddr %p wp %p ref %d new", + uaddr, fx->wp, (*newf)->f_refcount); +} - /* - * This function returns positive number as results and - * negative as errors - */ - op_ret = futex_atomic_op(td, args->val3, args->uaddr2); -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("futex_atomic_op ret %d\n", op_ret); -#endif - if (op_ret < 0) { - /* XXX: We don't handle the EFAULT yet. */ - if (op_ret != -EFAULT) { - futex_put(f); - futex_put(f2); - FUTEX_SYSTEM_UNLOCK; - return (-op_ret); - } +static struct futex * +futex_get(struct proc *p, uint32_t *uaddr, struct fx *fx, uint32_t flags) +{ + struct futex *f = NULL; - futex_put(f); - futex_put(f2); + KASSERT(flags != (FUTEX_CREATE_WP | FUTEX_DONTCREATE), + ("FUTEX_CREATE_WP|FUTEX_DONTCREATE")); - FUTEX_SYSTEM_UNLOCK; - return (EFAULT); - } + if (flags & FUTEX_CREATE_WP) { + fx->wp = malloc(sizeof(*fx->wp), M_LINUX, M_WAITOK); + fx->wp->w_pid = p->p_pid; + fx->wp->w_flags = 0; + LINUX_CTR1(sys_futex, "futex_get alloc wp %p", fx->wp); + } else + fx->wp = NULL; - ret = futex_wake(f, args->val, NULL, 0); - futex_put(f); - if (op_ret > 0) { - op_ret = 0; - /* - * Linux abuses the address of the timespec parameter - * as the number of retries. - */ - op_ret += futex_wake(f2, - (int)(unsigned long)args->timeout, NULL, 0); - ret += op_ret; - } - futex_put(f2); - td->td_retval[0] = ret; + futex_get0(uaddr, fx, &f, flags, 0); + return (fx->ftx); +} - FUTEX_SYSTEM_UNLOCK; - break; +static void +futex_get_op(uint32_t *uaddr, struct fx *fx, uint32_t *uaddr2, struct fx *fx2) +{ + struct futex *f, *f2; - case LINUX_FUTEX_LOCK_PI: - /* not yet implemented */ - return (ENOSYS); + KASSERT(uaddr != uaddr2, ("futex_get_op uaddr != uaddr2")); - case LINUX_FUTEX_UNLOCK_PI: - /* not yet implemented */ - return (ENOSYS); + fx->flags = 0; + fx2->flags = 0; + f = NULL; - case LINUX_FUTEX_TRYLOCK_PI: - /* not yet implemented */ - return (ENOSYS); + f2 = malloc(sizeof(*f2), M_LINUX, M_WAITOK | M_ZERO); + f2->f_uaddr = uaddr2; + refcount_init(&f2->f_refcount, 1); - default: - printf("linux_sys_futex: unknown op %d\n", - args->op); - return (ENOSYS); - } - return (0); + futex_get0(uaddr, fx, &f, 0, SX_DUPOK); + futex_get0(uaddr2, fx2, &f2, 0, 0); } -static struct futex * -futex_get(void *uaddr, int locked) +static int +futex_sleep(struct fx *fx, unsigned long timeout) { + struct waiting_proc *wp; struct futex *f; - - if (locked == FUTEX_UNLOCKED) - FUTEX_LOCK; - LIST_FOREACH(f, &futex_list, f_list) { - if (f->f_uaddr == uaddr) { - f->f_refcount++; - if (locked == FUTEX_UNLOCKED) - FUTEX_UNLOCK; - return f; - } - } - - f = malloc(sizeof(*f), M_LINUX, M_WAITOK); - f->f_uaddr = uaddr; - f->f_refcount = 1; - TAILQ_INIT(&f->f_waiting_proc); - LIST_INSERT_HEAD(&futex_list, f, f_list); - if (locked == FUTEX_UNLOCKED) - FUTEX_UNLOCK; - - return f; + uint32_t *uaddr; + int error; + + f = fx->ftx; + wp = fx->wp; + uaddr = f->f_uaddr; + + FUTEX_LOCK_ASSERT(f, SA_XLOCKED); + LINUX_CTR4(sys_futex, "futex_sleep enter uaddr %p wp %p timo %ld ref %d", + uaddr, wp, timeout, f->f_refcount); + + error = sx_sleep(wp, &f->lock, PCATCH, "futex", timeout); + + if (wp->w_flags & FUTEX_WP_REQUEUED) { + KASSERT(f != wp->w_ftx, ("f != w_ftx")); + LINUX_CTR5(sys_futex, "futex_sleep out error %d uaddr %p w" + " %p requeued uaddr %p ref %d", + error, uaddr, wp, wp->w_ftx->f_uaddr, wp->w_ftx->f_refcount); + futex_put0(f); + + f = wp->w_ftx; + FUTEX_LOCK(f); + } else + LINUX_CTR3(sys_futex, "futex_sleep out error %d uaddr %p wp %p", + error, uaddr, wp); + + if ((wp->w_flags & FUTEX_WP_REMOVED) == 0) + TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list); + free(wp, M_LINUX); + futex_put0(f); + return (error); } -static void -futex_put(f) - struct futex *f; +static int +futex_wake(struct fx *f, int n) { - FUTEX_LOCK; - f->f_refcount--; - if (f->f_refcount == 0) { - LIST_REMOVE(f, f_list); - free(f, M_LINUX); + struct waiting_proc *wp, *wpt; + int count = 0; + + FUTEX_LOCK_ASSERT(f->ftx, SA_XLOCKED); + TAILQ_FOREACH_SAFE(wp, &f->ftx->f_waiting_proc, wp_list, wpt) { + LINUX_CTR4(sys_futex, "futex_wake uaddr %p wp %p pid %d ref %d", + f->ftx->f_uaddr, wp, wp->w_pid, f->ftx->f_refcount); + TAILQ_REMOVE(&f->ftx->f_waiting_proc, wp, wp_list); + wp->w_flags |= FUTEX_WP_REMOVED; + wakeup_one(wp); + if (++count == n) + break; } - FUTEX_UNLOCK; - return; + return (count); } static int -futex_sleep(struct futex *f, struct thread *td, unsigned long timeout) +futex_requeue(struct fx *f, int n, struct fx *f2, int n2) { - struct waiting_proc *wp; - int ret; - - wp = malloc(sizeof(*wp), M_LINUX, M_WAITOK); - wp->wp_t = td; - wp->wp_new_futex = NULL; - FUTEX_LOCK; - TAILQ_INSERT_TAIL(&f->f_waiting_proc, wp, wp_list); - FUTEX_UNLOCK; - -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX --> %d tlseep timeout = %ld\n", - td->td_proc->p_pid, timeout); -#endif - ret = tsleep(wp, PCATCH | PZERO, "linuxfutex", timeout); -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("FUTEX -> %d tsleep returns %d\n", - td->td_proc->p_pid, ret); -#endif - - FUTEX_LOCK; - TAILQ_REMOVE(&f->f_waiting_proc, wp, wp_list); - FUTEX_UNLOCK; - - /* if we got woken up in futex_wake */ - if ((ret == 0) && (wp->wp_new_futex != NULL)) { - /* suspend us on the new futex */ - ret = futex_sleep(wp->wp_new_futex, td, timeout); - /* and release the old one */ - futex_put(wp->wp_new_futex); + struct waiting_proc *wp, *wpt; + int count = 0; + + TAILQ_FOREACH_SAFE(wp, &f->ftx->f_waiting_proc, wp_list, wpt) { + if (++count <= n) { + LINUX_CTR3(sys_futex, "futex_requeue_wake uaddr %p wp %p (%d)", + f->ftx->f_uaddr, wp, wp->w_pid); + TAILQ_REMOVE(&f->ftx->f_waiting_proc, wp, wp_list); + wp->w_flags |= FUTEX_WP_REMOVED; + wakeup_one(wp); + } else { + LINUX_CTR4(sys_futex, "futex_requeue uaddr %p wp %p (%d) to %p", + f->ftx->f_uaddr, wp, wp->w_pid, f2->ftx->f_uaddr); + wp->w_flags |= FUTEX_WP_REQUEUED; + /* Move waiting_proc to f2 futex */ + TAILQ_REMOVE(&f->ftx->f_waiting_proc, wp, wp_list); + TAILQ_INSERT_HEAD(&f2->ftx->f_waiting_proc, wp, wp_list); + wp->w_ftx = f2->ftx; + refcount_acquire(&f2->ftx->f_refcount); + if (count - n >= n2) + break; + } } - free(wp, M_LINUX); - - return ret; + return (count); } static int -futex_wake(struct futex *f, int n, struct futex *newf, int n2) +futex_wait(struct fx *f, struct l_timespec *ts) { - struct waiting_proc *wp; - int count; + struct l_timespec timeout = {0, 0}; + struct timeval tv = {0, 0}; + int timeout_hz; + int error; + + if (ts != NULL) { + error = copyin(ts, &timeout, sizeof(timeout)); + if (error) + return (error); + } + + tv.tv_usec = timeout.tv_sec * 1000000 + timeout.tv_nsec / 1000; + timeout_hz = tvtohz(&tv); + if (timeout.tv_sec == 0 && timeout.tv_nsec == 0) + timeout_hz = 0; /* - * Linux is very strange it wakes up N threads for - * all operations BUT requeue ones where its N+1 - * mimic this. + * If the user process requests a non null timeout, + * make sure we do not turn it into an infinite + * timeout because timeout_hz gets null. + * + * We use a minimal timeout of 1/hz. Maybe it would + * make sense to just return ETIMEDOUT without sleeping. */ - count = newf ? 0 : 1; + if (((timeout.tv_sec != 0) || (timeout.tv_nsec != 0)) && + (timeout_hz == 0)) + timeout_hz = 1; - FUTEX_LOCK; - TAILQ_FOREACH(wp, &f->f_waiting_proc, wp_list) { - if (count <= n) { - wakeup_one(wp); - count++; - } else { - if (newf != NULL) { - /* futex_put called after tsleep */ - wp->wp_new_futex = futex_get(newf->f_uaddr, - FUTEX_LOCKED); - wakeup_one(wp); - if (count - n >= n2) - break; - } - } + error = futex_sleep(f, timeout_hz); + + switch (error) { + case EWOULDBLOCK: + error = ETIMEDOUT; + break; + case EINTR: + /* FALLTHROUGH */ + case 0: + break; + default: + LINUX_CTR1(sys_futex, "futex_wait: unexpected ret = %d", + error); + break; } - FUTEX_UNLOCK; - return count; + return (error); } static int -futex_atomic_op(struct thread *td, int encoded_op, caddr_t uaddr) +futex_atomic_op(int encoded_op, uint32_t *uaddr) { int op = (encoded_op >> 28) & 7; int cmp = (encoded_op >> 24) & 15; @@ -488,12 +407,8 @@ futex_atomic_op(struct thread *td, int encoded_op, caddr_t uaddr) if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) oparg = 1 << oparg; -#ifdef DEBUG - if (ldebug(sys_futex)) - printf("futex_atomic_op: op = %d, cmp = %d, oparg = %x, " - "cmparg = %x, uaddr = %p\n", - op, cmp, oparg, cmparg, uaddr); -#endif + LINUX_CTR1(sys_futex, "futex_wake_op oparg %d", oparg); + /* XXX: linux verifies access here and returns EFAULT */ switch (op) { @@ -539,14 +454,184 @@ futex_atomic_op(struct thread *td, int encoded_op, caddr_t uaddr) } int +linux_sys_futex(struct thread *td, struct linux_sys_futex_args *args) +{ + struct proc *p = td->td_proc; + int op_ret, val, ret; + struct fx f, f2; + int error = 0; + + /* + * Our implementation provides only privates futexes. Most of the apps + * should use private futexes but don't claim so. Therefore we treat + * all futexes as private by clearing the FUTEX_PRIVATE_FLAG. It works + * in most cases (ie. when futexes are not shared on file descriptor + * or between different processes.). + */ + args->op = (args->op & ~LINUX_FUTEX_PRIVATE_FLAG); + + switch (args->op) { + case LINUX_FUTEX_WAIT: + + LINUX_CTR2(sys_futex, "futex_wait val %d uaddr %p", + args->val, args->uaddr); + + futex_get(p, args->uaddr, &f, FUTEX_CREATE_WP); + + error = copyin(args->uaddr, &val, sizeof(val)); + if (error) { + LINUX_CTR1(sys_futex, "futex_wait COPYIN FAILED %d", + error); + futex_put(&f); + return (error); + } + if (val != args->val) { + LINUX_CTR3(sys_futex, "futex_wait uaddr %p WHOOPS %d != %d", + args->uaddr, args->val, val); + futex_put(&f); + return (EWOULDBLOCK); + } + + error = futex_wait(&f, args->timeout); + break; + + case LINUX_FUTEX_WAKE: + + /* + * XXX: Linux is able to cope with different addresses + * corresponding to the same mapped memory in the sleeping + * and waker process(es). + */ + LINUX_CTR2(sys_futex, "futex_wake val %d uaddr %p", + args->val, args->uaddr); + + if (futex_get(p, args->uaddr, &f, FUTEX_DONTCREATE)) { + td->td_retval[0] = futex_wake(&f, args->val); + futex_put(&f); + } + break; + + case LINUX_FUTEX_CMP_REQUEUE: + + LINUX_CTR5(sys_futex, "futex_cmp_requeue uaddr %p " + "val %d val3 %d uaddr2 %p val2 %d", + args->uaddr, args->val, args->val3, args->uaddr2, + (int)(unsigned long)args->timeout); + + futex_get_op(args->uaddr, &f, args->uaddr2, &f2); + + error = copyin(args->uaddr, &val, sizeof(val)); + if (error) { + LINUX_CTR1(sys_futex, "futex_cmp_requeue " + " COPYIN FAILED %d", error); + futex_put(&f2); + futex_put(&f); + return (error); + } + if (val != args->val3) { + LINUX_CTR2(sys_futex, "futex_cmp_requeue WHOOPS" + " VAL %d != UVAL %d", args->val, val); + futex_put(&f2); + futex_put(&f); + return (EAGAIN); + } + + td->td_retval[0] = futex_requeue(&f, args->val, &f2, + (int)(unsigned long)args->timeout); + + futex_put(&f2); + futex_put(&f); + break; + + case LINUX_FUTEX_REQUEUE: + + LINUX_CTR4(sys_futex, "FUTEX_REQUEUE uaddr %p " + "val %d uaddr2 %p val2 %d", + args->uaddr, args->val, args->uaddr2, + (int)(unsigned long)args->timeout); + + futex_get_op(args->uaddr, &f, args->uaddr2, &f2); + + td->td_retval[0] = futex_requeue(&f, args->val, &f2, + (int)(unsigned long)args->timeout); + + futex_put(&f2); + futex_put(&f); + break; + + case LINUX_FUTEX_WAKE_OP: + + LINUX_CTR5(sys_futex, "futex_wake_op " + "uaddr %p op %d val %x uaddr2 %p val3 %x", + args->uaddr, args->op, args->val, + args->uaddr2, args->val3); + + futex_get_op(args->uaddr, &f, args->uaddr2, &f2); + + /* + * This function returns positive number as results and + * negative as errors + */ + op_ret = futex_atomic_op(args->val3, args->uaddr2); + + LINUX_CTR1(sys_futex, "futex_wake_op opret %d", op_ret); + + if (op_ret < 0) { + /* XXX: We don't handle the EFAULT yet. */ + if (op_ret != -EFAULT) { + futex_put(&f); + return (-op_ret); + } + futex_put(&f); + return (EFAULT); + } + + ret = futex_wake(&f, args->val); + + if (op_ret > 0) { + op_ret = 0; + /* + * Linux abuses the address of the timespec parameter + * as the number of retries. + */ + op_ret += futex_wake( &f2, + (int)(unsigned long)args->timeout); + ret += op_ret; + + } + futex_put(&f2); + futex_put(&f); + td->td_retval[0] = ret; + LINUX_CTR1(sys_futex, "futex_wake_op returns %d", ret); + break; + + case LINUX_FUTEX_LOCK_PI: + /* not yet implemented */ + return (ENOSYS); + + case LINUX_FUTEX_UNLOCK_PI: + /* not yet implemented */ + return (ENOSYS); + + case LINUX_FUTEX_TRYLOCK_PI: + /* not yet implemented */ + return (ENOSYS); + + default: + printf("linux_sys_futex: unknown op %d\n", args->op); + return (ENOSYS); + } + + return (error); +} + +int linux_set_robust_list(struct thread *td, struct linux_set_robust_list_args *args) { struct linux_emuldata *em; -#ifdef DEBUG - if (ldebug(set_robust_list)) - printf(ARGS(set_robust_list, "")); -#endif + LINUX_CTR2(set_robust_list, "head %p len %d", args->head, args->len); + if (args->len != sizeof(struct linux_robust_list_head)) return (EINVAL); @@ -565,10 +650,7 @@ linux_get_robust_list(struct thread *td, struct linux_get_robust_list_args *args l_size_t len = sizeof(struct linux_robust_list_head); int error = 0; -#ifdef DEBUG - if (ldebug(get_robust_list)) - printf(ARGS(get_robust_list, "")); -#endif + LINUX_CTR(get_robust_list); if (!args->pid) { em = em_find(td->td_proc, EMUL_DONTLOCK); @@ -601,16 +683,16 @@ linux_get_robust_list(struct thread *td, struct linux_get_robust_list_args *args } static int -handle_futex_death(void *uaddr, pid_t pid, int pi) +handle_futex_death(struct proc *p, uint32_t *uaddr, int pi) { - int uval, nval, mval; - struct futex *f; + uint32_t uval, nval, mval; + struct fx f; + LINUX_CTR1(release_futexes, "futex_death uaddr %p", uaddr); retry: if (copyin(uaddr, &uval, 4)) return (EFAULT); - - if ((uval & FUTEX_TID_MASK) == pid) { + if ((uval & FUTEX_TID_MASK) == p->p_pid) { mval = (uval & FUTEX_WAITERS) | FUTEX_OWNER_DIED; nval = casuword32(uaddr, uval, mval); @@ -621,8 +703,10 @@ retry: goto retry; if (!pi && (uval & FUTEX_WAITERS)) { - f = futex_get(uaddr, FUTEX_UNLOCKED); - futex_wake(f, 1, NULL, 0); + if (futex_get(p, uaddr, &f, FUTEX_DONTCREATE)) { + futex_wake(&f, 1); + futex_put(&f); + } } } @@ -658,6 +742,8 @@ release_futexes(struct proc *p) em = em_find(p, EMUL_DONTLOCK); head = em->robust_futexes; + LINUX_CTR1(release_futexes, "head %p", head); + if (head == NULL) return; @@ -674,8 +760,7 @@ release_futexes(struct proc *p) rc = fetch_robust_entry(&next_entry, PTRIN(&entry->next), &next_pi); if (entry != pending) - if (handle_futex_death((char *)entry + futex_offset, - p->p_pid, pi)) + if (handle_futex_death(p, (uint32_t *)entry + futex_offset, pi)) return; if (rc) @@ -691,6 +776,5 @@ release_futexes(struct proc *p) } if (pending) - handle_futex_death((char *) pending + futex_offset, - p->p_pid, pip); + handle_futex_death(p, (uint32_t *)pending + futex_offset, pip); } diff --git a/sys/compat/linux/linux_futex.h b/sys/compat/linux/linux_futex.h index d658925..ca8b768 100644 --- a/sys/compat/linux/linux_futex.h +++ b/sys/compat/linux/linux_futex.h @@ -38,7 +38,7 @@ #define LINUX_FUTEX_WAIT 0 #define LINUX_FUTEX_WAKE 1 -#define LINUX_FUTEX_FD 2 +#define LINUX_FUTEX_FD 2 /* unused */ #define LINUX_FUTEX_REQUEUE 3 #define LINUX_FUTEX_CMP_REQUEUE 4 #define LINUX_FUTEX_WAKE_OP 5 diff --git a/sys/compat/linux/linux_getcwd.c b/sys/compat/linux/linux_getcwd.c index 2ab69c5..bf23f75 100644 --- a/sys/compat/linux/linux_getcwd.c +++ b/sys/compat/linux/linux_getcwd.c @@ -47,6 +47,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -424,10 +425,7 @@ linux_getcwd(struct thread *td, struct linux_getcwd_args *args) caddr_t bp, bend, path; int error, len, lenused; -#ifdef DEBUG - if (ldebug(getcwd)) - printf(ARGS(getcwd, "%p, %ld"), args->buf, (long)args->bufsize); -#endif + LINUX_CTR2(getcwd, "%p, %ld", args->buf, (long)args->bufsize); len = args->bufsize; diff --git a/sys/compat/linux/linux_ioctl.c b/sys/compat/linux/linux_ioctl.c index 967ab71..c52dc5b 100644 --- a/sys/compat/linux/linux_ioctl.c +++ b/sys/compat/linux/linux_ioctl.c @@ -47,6 +47,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -200,12 +201,10 @@ linux_ioctl_hdio(struct thread *td, struct linux_ioctl_args *args) */ bytespercyl = (off_t) sectorsize * fwheads * fwsectors; fwcylinders = mediasize / bytespercyl; -#if defined(DEBUG) - linux_msg(td, "HDIO_GET_GEO: mediasize %jd, c/h/s %d/%d/%d, " + LINUX_CTR5(ioctl, "HDIO_GET_GEO: mediasize %jd, c/h/s %d/%d/%d, " "bpc %jd", (intmax_t)mediasize, fwcylinders, fwheads, fwsectors, (intmax_t)bytespercyl); -#endif if ((args->cmd & 0xffff) == LINUX_HDIO_GET_GEO) { struct linux_hd_geometry hdg; @@ -355,18 +354,10 @@ bsd_to_linux_termios(struct termios *bios, struct linux_termios *lios) { int i; -#ifdef DEBUG - if (ldebug(ioctl)) { - printf("LINUX: BSD termios structure (input):\n"); - printf("i=%08x o=%08x c=%08x l=%08x ispeed=%d ospeed=%d\n", - bios->c_iflag, bios->c_oflag, bios->c_cflag, bios->c_lflag, - bios->c_ispeed, bios->c_ospeed); - printf("c_cc "); - for (i=0; ic_cc[i]); - printf("\n"); - } -#endif + LINUX_CTR0(ioctl, "LINUX: BSD termios structure (input):"); + LINUX_CTR6(ioctl, "i=%08x o=%08x c=%08x l=%08x ispeed=%d ospeed=%d", + bios->c_iflag, bios->c_oflag, bios->c_cflag, bios->c_lflag, + bios->c_ispeed, bios->c_ospeed); lios->c_iflag = 0; if (bios->c_iflag & IGNBRK) @@ -477,18 +468,10 @@ bsd_to_linux_termios(struct termios *bios, struct linux_termios *lios) } lios->c_line = 0; -#ifdef DEBUG - if (ldebug(ioctl)) { - printf("LINUX: LINUX termios structure (output):\n"); - printf("i=%08x o=%08x c=%08x l=%08x line=%d\n", - lios->c_iflag, lios->c_oflag, lios->c_cflag, - lios->c_lflag, (int)lios->c_line); - printf("c_cc "); - for (i=0; ic_cc[i]); - printf("\n"); - } -#endif + LINUX_CTR0(ioctl, "LINUX: LINUX termios structure (output):"); + LINUX_CTR5(ioctl, "i=%08x o=%08x c=%08x l=%08x line=%d", + lios->c_iflag, lios->c_oflag, lios->c_cflag, + lios->c_lflag, (int)lios->c_line); } static void @@ -496,18 +479,10 @@ linux_to_bsd_termios(struct linux_termios *lios, struct termios *bios) { int i; -#ifdef DEBUG - if (ldebug(ioctl)) { - printf("LINUX: LINUX termios structure (input):\n"); - printf("i=%08x o=%08x c=%08x l=%08x line=%d\n", - lios->c_iflag, lios->c_oflag, lios->c_cflag, - lios->c_lflag, (int)lios->c_line); - printf("c_cc "); - for (i=0; ic_cc[i]); - printf("\n"); - } -#endif + LINUX_CTR0(ioctl, "LINUX: LINUX termios structure (input):"); + LINUX_CTR5(ioctl, "i=%08x o=%08x c=%08x l=%08x line=%d", + lios->c_iflag, lios->c_oflag, lios->c_cflag, + lios->c_lflag, (int)lios->c_line); bios->c_iflag = 0; if (lios->c_iflag & LINUX_IGNBRK) @@ -619,18 +594,10 @@ linux_to_bsd_termios(struct linux_termios *lios, struct termios *bios) bios->c_ispeed = bios->c_ospeed = linux_to_bsd_speed(lios->c_cflag & LINUX_CBAUD, sptab); -#ifdef DEBUG - if (ldebug(ioctl)) { - printf("LINUX: BSD termios structure (output):\n"); - printf("i=%08x o=%08x c=%08x l=%08x ispeed=%d ospeed=%d\n", - bios->c_iflag, bios->c_oflag, bios->c_cflag, bios->c_lflag, - bios->c_ispeed, bios->c_ospeed); - printf("c_cc "); - for (i=0; ic_cc[i]); - printf("\n"); - } -#endif + LINUX_CTR0(ioctl, "LINUX: BSD termios structure (output):"); + LINUX_CTR6(ioctl, "i=%08x o=%08x c=%08x l=%08x ispeed=%d ospeed=%d", + bios->c_iflag, bios->c_oflag, bios->c_cflag, bios->c_lflag, + bios->c_ispeed, bios->c_ospeed); } static void @@ -2343,10 +2310,8 @@ linux_ioctl_socket(struct thread *td, struct linux_ioctl_args *args) case LINUX_SIOCSPGRP: case LINUX_SIOCGIFCOUNT: /* these ioctls don't take an interface name */ -#ifdef DEBUG - printf("%s(): ioctl %d\n", __func__, - args->cmd & 0xffff); -#endif + LINUX_CTR2(ioctl, "%s(): ioctl %d", + __func__, args->cmd & 0xffff); break; case LINUX_SIOCGIFFLAGS: @@ -2368,10 +2333,8 @@ linux_ioctl_socket(struct thread *td, struct linux_ioctl_args *args) error = copyin((void *)args->arg, lifname, LINUX_IFNAMSIZ); if (error != 0) return (error); -#ifdef DEBUG - printf("%s(): ioctl %d on %.*s\n", __func__, - args->cmd & 0xffff, LINUX_IFNAMSIZ, lifname); -#endif + LINUX_CTR4(ioctl, "%s(): ioctl %d on %.*s", + __func__, args->cmd & 0xffff, LINUX_IFNAMSIZ, lifname); ifp = ifname_linux_to_bsd(lifname, ifname); if (ifp == NULL) return (EINVAL); @@ -2384,10 +2347,8 @@ linux_ioctl_socket(struct thread *td, struct linux_ioctl_args *args) error = copyout(ifname, (void *)args->arg, IFNAMSIZ); if (error != 0) return (error); -#ifdef DEBUG - printf("%s(): %s translated to %s\n", __func__, - lifname, ifname); -#endif + LINUX_CTR3(ioctl, "%s(): %s translated to %s", + __func__, lifname, ifname); break; default: @@ -2526,9 +2487,7 @@ linux_ioctl_socket(struct thread *td, struct linux_ioctl_args *args) /* restore the original interface name */ copyout(lifname, (void *)args->arg, LINUX_IFNAMSIZ); -#ifdef DEBUG - printf("%s(): returning %d\n", __func__, error); -#endif + LINUX_CTR2(ioctl, "%s(): returning %d", __func__, error); return (error); } @@ -2617,11 +2576,7 @@ linux_ioctl(struct thread *td, struct linux_ioctl_args *args) struct handler_element *he; int error, cmd; -#ifdef DEBUG - if (ldebug(ioctl)) - printf(ARGS(ioctl, "%d, %04lx, *"), args->fd, - (unsigned long)args->cmd); -#endif + LINUX_CTR2(ioctl, "%d %04lx", args->fd, (unsigned long)args->cmd); if ((error = fget(td, args->fd, &fp)) != 0) return (error); diff --git a/sys/compat/linux/linux_misc.c b/sys/compat/linux/linux_misc.c index bdbb5dd..fe76101 100644 --- a/sys/compat/linux/linux_misc.c +++ b/sys/compat/linux/linux_misc.c @@ -41,6 +41,7 @@ __FBSDID("$FreeBSD$"); #endif #include #include +#include #include #include #include @@ -92,10 +93,6 @@ __FBSDID("$FreeBSD$"); #include #include -#ifdef __i386__ -#include -#endif - #define BSD_TO_LINUX_SIGNAL(sig) \ (((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig) @@ -174,11 +171,8 @@ linux_alarm(struct thread *td, struct linux_alarm_args *args) u_int secs; int error; -#ifdef DEBUG - if (ldebug(alarm)) - printf(ARGS(alarm, "%u"), args->secs); -#endif - + LINUX_CTR1(alarm, "%d", args->secs); + secs = args->secs; if (secs > INT_MAX) @@ -208,10 +202,8 @@ linux_brk(struct thread *td, struct linux_brk_args *args) char * nsize; } */ tmp; -#ifdef DEBUG - if (ldebug(brk)) - printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); -#endif + LINUX_CTR1(brk, "%p", (void *)(uintptr_t)args->dsend); + old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); new = (vm_offset_t)args->dsend; tmp.nsize = (char *)new; @@ -243,10 +235,7 @@ linux_uselib(struct thread *td, struct linux_uselib_args *args) LCONVPATHEXIST(td, args->library, &library); -#ifdef DEBUG - if (ldebug(uselib)) - printf(ARGS(uselib, "%s"), library); -#endif + LINUX_CTR(uselib); a_out = NULL; vfslocked = 0; @@ -391,9 +380,7 @@ linux_uselib(struct thread *td, struct linux_uselib_args *args) * (what a waste). */ if (file_offset & PAGE_MASK) { -#ifdef DEBUG - printf("uselib: Non page aligned binary %lu\n", file_offset); -#endif + LINUX_CTR1(uselib, "non page aligned binary %lu", file_offset); /* Map text+data read/write/execute */ /* a_entry is the load address and is page aligned */ @@ -425,9 +412,7 @@ linux_uselib(struct thread *td, struct linux_uselib_args *args) if (error) goto cleanup; } else { -#ifdef DEBUG - printf("uselib: Page aligned binary %lu\n", file_offset); -#endif + LINUX_CTR1(uselib, "page aligned binary %lu", file_offset); /* * for QMAGIC, a_entry is 20 bytes beyond the load address * to skip the executable header @@ -444,10 +429,8 @@ linux_uselib(struct thread *td, struct linux_uselib_args *args) if (error) goto cleanup; } -#ifdef DEBUG - printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], - ((long *)vmaddr)[1]); -#endif + LINUX_CTR3(uselib, "mem=%08lx = %08lx %08lx", + (long)vmaddr, ((long *)vmaddr)[0], ((long *)vmaddr)[1]); if (bss_size != 0) { /* Calculate BSS start address */ vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + @@ -484,12 +467,7 @@ linux_select(struct thread *td, struct linux_select_args *args) struct timeval tv0, tv1, utv, *tvp; int error; -#ifdef DEBUG - if (ldebug(select)) - printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, - (void *)args->readfds, (void *)args->writefds, - (void *)args->exceptfds, (void *)args->timeout); -#endif + LINUX_CTR1(select, "nfds %d", args->nfds); /* * Store current time for computation of the amount of @@ -500,11 +478,9 @@ linux_select(struct thread *td, struct linux_select_args *args) goto select_out; utv.tv_sec = ltv.tv_sec; utv.tv_usec = ltv.tv_usec; -#ifdef DEBUG - if (ldebug(select)) - printf(LMSG("incoming timeout (%jd/%ld)"), - (intmax_t)utv.tv_sec, utv.tv_usec); -#endif + + LINUX_CTR2(select, "incoming timeout (%jd/%ld)", + (intmax_t)utv.tv_sec, utv.tv_usec); if (itimerfix(&utv)) { /* @@ -528,10 +504,8 @@ linux_select(struct thread *td, struct linux_select_args *args) error = kern_select(td, args->nfds, args->readfds, args->writefds, args->exceptfds, tvp); -#ifdef DEBUG - if (ldebug(select)) - printf(LMSG("real select returns %d"), error); -#endif + LINUX_CTR1(select, "real select returns %d", error); + if (error) goto select_out; @@ -550,11 +524,10 @@ linux_select(struct thread *td, struct linux_select_args *args) timevalclear(&utv); } else timevalclear(&utv); -#ifdef DEBUG - if (ldebug(select)) - printf(LMSG("outgoing timeout (%jd/%ld)"), - (intmax_t)utv.tv_sec, utv.tv_usec); -#endif + + LINUX_CTR2(select, "outgoing timeout (%jd/%ld)", + (intmax_t)utv.tv_sec, utv.tv_usec); + ltv.tv_sec = utv.tv_sec; ltv.tv_usec = utv.tv_usec; if ((error = copyout(<v, args->timeout, sizeof(ltv)))) @@ -562,10 +535,7 @@ linux_select(struct thread *td, struct linux_select_args *args) } select_out: -#ifdef DEBUG - if (ldebug(select)) - printf(LMSG("select_out -> %d"), error); -#endif + LINUX_CTR1(select, "out %d", error); return error; } @@ -578,14 +548,11 @@ linux_mremap(struct thread *td, struct linux_mremap_args *args) } */ bsd_args; int error = 0; -#ifdef DEBUG - if (ldebug(mremap)) - printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), - (void *)(uintptr_t)args->addr, - (unsigned long)args->old_len, - (unsigned long)args->new_len, - (unsigned long)args->flags); -#endif + LINUX_CTR4(mremap, "%p %08lx %08lx 08lx", + (void *)(uintptr_t)args->addr, + (unsigned long)args->old_len, + (unsigned long)args->new_len, + (unsigned long)args->flags); if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { td->td_retval[0] = 0; @@ -629,6 +596,9 @@ linux_msync(struct thread *td, struct linux_msync_args *args) { struct msync_args bsd_args; + LINUX_CTR3(msync, "%p %d %d", (void *)(uintptr_t)args->addr, + (uintptr_t)args->len, args->fl); + bsd_args.addr = (caddr_t)(uintptr_t)args->addr; bsd_args.len = (uintptr_t)args->len; bsd_args.flags = args->fl & ~LINUX_MS_SYNC; @@ -643,11 +613,6 @@ linux_time(struct thread *td, struct linux_time_args *args) l_time_t tm; int error; -#ifdef DEBUG - if (ldebug(time)) - printf(ARGS(time, "*")); -#endif - microtime(&tv); tm = tv.tv_sec; if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) @@ -675,11 +640,6 @@ linux_times(struct thread *td, struct linux_times_args *args) struct proc *p; int error; -#ifdef DEBUG - if (ldebug(times)) - printf(ARGS(times, "*")); -#endif - if (args->buf != NULL) { p = td->td_proc; PROC_LOCK(p); @@ -713,11 +673,6 @@ linux_newuname(struct thread *td, struct linux_newuname_args *args) char osrelease[LINUX_MAX_UTSNAME]; char *p; -#ifdef DEBUG - if (ldebug(newuname)) - printf(ARGS(newuname, "*")); -#endif - linux_get_osname(td, osname); linux_get_osrelease(td, osrelease); @@ -731,34 +686,8 @@ linux_newuname(struct thread *td, struct linux_newuname_args *args) *p = '\0'; break; } -#ifdef __i386__ - { - const char *class; + strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME); - switch (cpu_class) { - case CPUCLASS_686: - class = "i686"; - break; - case CPUCLASS_586: - class = "i586"; - break; - case CPUCLASS_486: - class = "i486"; - break; - default: - class = "i386"; - } - strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME); - } -#elif defined(__amd64__) /* XXX: Linux can change 'personality'. */ -#ifdef COMPAT_LINUX32 - strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME); -#else - strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME); -#endif /* COMPAT_LINUX32 */ -#else /* something other than i386 or amd64 - assume we and Linux agree */ - strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME); -#endif /* __i386__ */ mtx_lock(&hostname_mtx); strlcpy(utsname.domainname, V_domainname, LINUX_MAX_UTSNAME); mtx_unlock(&hostname_mtx); @@ -782,10 +711,7 @@ linux_utime(struct thread *td, struct linux_utime_args *args) LCONVPATHEXIST(td, args->fname, &fname); -#ifdef DEBUG - if (ldebug(utime)) - printf(ARGS(utime, "%s, *"), fname); -#endif + LINUX_CTR(utime); if (args->times) { if ((error = copyin(args->times, &lut, sizeof lut))) { @@ -815,10 +741,7 @@ linux_utimes(struct thread *td, struct linux_utimes_args *args) LCONVPATHEXIST(td, args->fname, &fname); -#ifdef DEBUG - if (ldebug(utimes)) - printf(ARGS(utimes, "%s, *"), fname); -#endif + LINUX_CTR(utimes); if (args->tptr != NULL) { if ((error = copyin(args->tptr, ltv, sizeof ltv))) { @@ -848,10 +771,7 @@ linux_futimesat(struct thread *td, struct linux_futimesat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); -#ifdef DEBUG - if (ldebug(futimesat)) - printf(ARGS(futimesat, "%s, *"), fname); -#endif + LINUX_CTR(futimesat); if (args->utimes != NULL) { if ((error = copyin(args->utimes, ltv, sizeof ltv))) { @@ -878,11 +798,8 @@ linux_waitpid(struct thread *td, struct linux_waitpid_args *args) { int error, options, tmpstat; -#ifdef DEBUG - if (ldebug(waitpid)) - printf(ARGS(waitpid, "%d, %p, %d"), - args->pid, (void *)args->status, args->options); -#endif + LINUX_CTR3(waitpid, "%d, %p, %d", args->pid, (void *)args->status, + args->options); /* * this is necessary because the test in kern_wait doesn't work * because we mess with the options here @@ -920,12 +837,8 @@ linux_wait4(struct thread *td, struct linux_wait4_args *args) struct rusage ru, *rup; struct proc *p; -#ifdef DEBUG - if (ldebug(wait4)) - printf(ARGS(wait4, "%d, %p, %d, %p"), - args->pid, (void *)args->status, args->options, - (void *)args->rusage); -#endif + LINUX_CTR4(wait4, "%d, %p, %d %p", args->pid, (void *)args->status, + args->options, (void *)args->rusage); options = (args->options & (WNOHANG | WUNTRACED)); /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ @@ -969,10 +882,7 @@ linux_mknod(struct thread *td, struct linux_mknod_args *args) LCONVPATHCREAT(td, args->path, &path); -#ifdef DEBUG - if (ldebug(mknod)) - printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev); -#endif + LINUX_CTR(mknod); switch (args->mode & S_IFMT) { case S_IFIFO: @@ -1017,10 +927,7 @@ linux_mknodat(struct thread *td, struct linux_mknodat_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHCREAT_AT(td, args->filename, &path, dfd); -#ifdef DEBUG - if (ldebug(mknodat)) - printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); -#endif + LINUX_CTR(mknodat); switch (args->mode & S_IFMT) { case S_IFIFO: @@ -1062,10 +969,9 @@ linux_mknodat(struct thread *td, struct linux_mknodat_args *args) int linux_personality(struct thread *td, struct linux_personality_args *args) { -#ifdef DEBUG - if (ldebug(personality)) - printf(ARGS(personality, "%lu"), (unsigned long)args->per); -#endif + + LINUX_CTR1(personality, "%ld", (unsigned long)args->per); + if (args->per != 0) return EINVAL; @@ -1092,11 +998,7 @@ linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) struct l_itimerval ls; struct itimerval aitv, oitv; -#ifdef DEBUG - if (ldebug(setitimer)) - printf(ARGS(setitimer, "%p, %p"), - (void *)uap->itv, (void *)uap->oitv); -#endif + LINUX_CTR2(setitimer, "%p %p", (void *)uap->itv, (void *)uap->oitv); if (uap->itv == NULL) { uap->itv = uap->oitv; @@ -1107,14 +1009,12 @@ linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) if (error != 0) return (error); B2L_ITIMERVAL(&aitv, &ls); -#ifdef DEBUG - if (ldebug(setitimer)) { - printf("setitimer: value: sec: %jd, usec: %ld\n", - (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); - printf("setitimer: interval: sec: %jd, usec: %ld\n", - (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); - } -#endif + + LINUX_CTR2(setitimer, "value: sec: %jd, usec: %ld", + (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); + LINUX_CTR2(setitimer, "interval: sec: %jd, usec: %ld", + (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); + error = kern_setitimer(td, uap->which, &aitv, &oitv); if (error != 0 || uap->oitv == NULL) return (error); @@ -1130,10 +1030,8 @@ linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) struct l_itimerval ls; struct itimerval aitv; -#ifdef DEBUG - if (ldebug(getitimer)) - printf(ARGS(getitimer, "%p"), (void *)uap->itv); -#endif + LINUX_CTR1(getitimer, "%p", (void *)uap->itv); + error = kern_getitimer(td, uap->which, &aitv); if (error != 0) return (error); @@ -1146,6 +1044,8 @@ linux_nice(struct thread *td, struct linux_nice_args *args) { struct setpriority_args bsd_args; + LINUX_CTR1(nice, "%d", args->inc); + bsd_args.which = PRIO_PROCESS; bsd_args.who = 0; /* current process */ bsd_args.prio = args->inc; @@ -1252,11 +1152,7 @@ linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) u_int which; int error; -#ifdef DEBUG - if (ldebug(setrlimit)) - printf(ARGS(setrlimit, "%d, %p"), - args->resource, (void *)args->rlim); -#endif + LINUX_CTR1(setrlimit, "%p", (void *)args->rlim); if (args->resource >= LINUX_RLIM_NLIMITS) return (EINVAL); @@ -1282,11 +1178,7 @@ linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) struct rlimit bsd_rlim; u_int which; -#ifdef DEBUG - if (ldebug(old_getrlimit)) - printf(ARGS(old_getrlimit, "%d, %p"), - args->resource, (void *)args->rlim); -#endif + LINUX_CTR1(old_getrlimit, "%p", (void *)args->rlim); if (args->resource >= LINUX_RLIM_NLIMITS) return (EINVAL); @@ -1325,11 +1217,7 @@ linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) struct rlimit bsd_rlim; u_int which; -#ifdef DEBUG - if (ldebug(getrlimit)) - printf(ARGS(getrlimit, "%d, %p"), - args->resource, (void *)args->rlim); -#endif + LINUX_CTR2(getrlimit, "%d %p", args->resource, (void *)args->rlim); if (args->resource >= LINUX_RLIM_NLIMITS) return (EINVAL); @@ -1353,11 +1241,8 @@ linux_sched_setscheduler(struct thread *td, { struct sched_setscheduler_args bsd; -#ifdef DEBUG - if (ldebug(sched_setscheduler)) - printf(ARGS(sched_setscheduler, "%d, %d, %p"), - args->pid, args->policy, (const void *)args->param); -#endif + LINUX_CTR3(sched_setscheduler, "%d %d %p", args->pid, args->policy, + (const void *)args->param); switch (args->policy) { case LINUX_SCHED_OTHER: @@ -1385,10 +1270,7 @@ linux_sched_getscheduler(struct thread *td, struct sched_getscheduler_args bsd; int error; -#ifdef DEBUG - if (ldebug(sched_getscheduler)) - printf(ARGS(sched_getscheduler, "%d"), args->pid); -#endif + LINUX_CTR1(sched_getscheduler, "%d", args->pid); bsd.pid = args->pid; error = sched_getscheduler(td, &bsd); @@ -1414,10 +1296,7 @@ linux_sched_get_priority_max(struct thread *td, { struct sched_get_priority_max_args bsd; -#ifdef DEBUG - if (ldebug(sched_get_priority_max)) - printf(ARGS(sched_get_priority_max, "%d"), args->policy); -#endif + LINUX_CTR1(sched_get_priority_max, "%d", args->policy); switch (args->policy) { case LINUX_SCHED_OTHER: @@ -1441,10 +1320,7 @@ linux_sched_get_priority_min(struct thread *td, { struct sched_get_priority_min_args bsd; -#ifdef DEBUG - if (ldebug(sched_get_priority_min)) - printf(ARGS(sched_get_priority_min, "%d"), args->policy); -#endif + LINUX_CTR1(sched_get_priority_min, "%d", args->policy); switch (args->policy) { case LINUX_SCHED_OTHER: @@ -1478,10 +1354,7 @@ linux_reboot(struct thread *td, struct linux_reboot_args *args) { struct reboot_args bsd_args; -#ifdef DEBUG - if (ldebug(reboot)) - printf(ARGS(reboot, "0x%x"), args->cmd); -#endif + LINUX_CTR1(reboot, "0x%x", args->cmd); if (args->magic1 != REBOOT_MAGIC1) return EINVAL; @@ -1532,11 +1405,6 @@ linux_getpid(struct thread *td, struct linux_getpid_args *args) { struct linux_emuldata *em; -#ifdef DEBUG - if (ldebug(getpid)) - printf(ARGS(getpid, "")); -#endif - if (linux_use26(td)) { em = em_find(td->td_proc, EMUL_DONTLOCK); KASSERT(em != NULL, ("getpid: emuldata not found.\n")); @@ -1552,11 +1420,6 @@ int linux_gettid(struct thread *td, struct linux_gettid_args *args) { -#ifdef DEBUG - if (ldebug(gettid)) - printf(ARGS(gettid, "")); -#endif - td->td_retval[0] = td->td_proc->p_pid; return (0); } @@ -1568,11 +1431,6 @@ linux_getppid(struct thread *td, struct linux_getppid_args *args) struct linux_emuldata *em; struct proc *p, *pp; -#ifdef DEBUG - if (ldebug(getppid)) - printf(ARGS(getppid, "")); -#endif - if (!linux_use26(td)) { PROC_LOCK(td->td_proc); td->td_retval[0] = td->td_proc->p_pptr->p_pid; @@ -1588,9 +1446,7 @@ linux_getppid(struct thread *td, struct linux_getppid_args *args) p = pfind(em->shared->group_pid); if (p == NULL) { -#ifdef DEBUG - printf(LMSG("parent process not found.\n")); -#endif + LINUX_CTR0(getppid, "parent process not found"); return (0); } @@ -1616,11 +1472,6 @@ int linux_getgid(struct thread *td, struct linux_getgid_args *args) { -#ifdef DEBUG - if (ldebug(getgid)) - printf(ARGS(getgid, "")); -#endif - td->td_retval[0] = td->td_ucred->cr_rgid; return (0); } @@ -1629,11 +1480,6 @@ int linux_getuid(struct thread *td, struct linux_getuid_args *args) { -#ifdef DEBUG - if (ldebug(getuid)) - printf(ARGS(getuid, "")); -#endif - td->td_retval[0] = td->td_ucred->cr_ruid; return (0); } @@ -1644,11 +1490,6 @@ linux_getsid(struct thread *td, struct linux_getsid_args *args) { struct getsid_args bsd; -#ifdef DEBUG - if (ldebug(getsid)) - printf(ARGS(getsid, "%i"), args->pid); -#endif - bsd.pid = args->pid; return getsid(td, &bsd); } @@ -1666,10 +1507,7 @@ linux_getpriority(struct thread *td, struct linux_getpriority_args *args) struct getpriority_args bsd_args; int error; -#ifdef DEBUG - if (ldebug(getpriority)) - printf(ARGS(getpriority, "%i, %i"), args->which, args->who); -#endif + LINUX_CTR2(getpriority, "%i %i", args->which, args->who); bsd_args.which = args->which; bsd_args.who = args->who; @@ -1683,11 +1521,6 @@ linux_sethostname(struct thread *td, struct linux_sethostname_args *args) { int name[2]; -#ifdef DEBUG - if (ldebug(sethostname)) - printf(ARGS(sethostname, "*, %i"), args->len); -#endif - name[0] = CTL_KERN; name[1] = KERN_HOSTNAME; return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, @@ -1699,11 +1532,6 @@ linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) { int name[2]; -#ifdef DEBUG - if (ldebug(setdomainname)) - printf(ARGS(setdomainname, "*, %i"), args->len); -#endif - name[0] = CTL_KERN; name[1] = KERN_NISDOMAINNAME; return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, @@ -1716,10 +1544,7 @@ linux_exit_group(struct thread *td, struct linux_exit_group_args *args) struct linux_emuldata *em, *td_em, *tmp_em; struct proc *sp; -#ifdef DEBUG - if (ldebug(exit_group)) - printf(ARGS(exit_group, "%i"), args->error_code); -#endif + LINUX_CTR1(exit_group, "%i", args->error_code); if (linux_use26(td)) { td_em = em_find(td->td_proc, EMUL_DONTLOCK); @@ -1734,9 +1559,8 @@ linux_exit_group(struct thread *td, struct linux_exit_group_args *args) sp = pfind(em->pid); psignal(sp, SIGKILL); PROC_UNLOCK(sp); -#ifdef DEBUG - printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid); -#endif + + LINUX_CTR1(exit_group, "kill %d", em->pid); } EMUL_SHARED_RUNLOCK(&emul_shared_lock); @@ -1760,11 +1584,8 @@ linux_prctl(struct thread *td, struct linux_prctl_args *args) struct linux_emuldata *em; int pdeath_signal; -#ifdef DEBUG - if (ldebug(prctl)) - printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option, - args->arg2, args->arg3, args->arg4, args->arg5); -#endif + LINUX_CTR5(prctl, "%d %d %d %d %d", args->option, args->arg2, + args->arg3, args->arg4, args->arg5); switch (args->option) { case LINUX_PR_SET_PDEATHSIG: @@ -1839,11 +1660,8 @@ linux_sched_getaffinity(struct thread *td, int error; struct cpuset_getaffinity_args cga; -#ifdef DEBUG - if (ldebug(sched_getaffinity)) - printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, - args->len); -#endif + LINUX_CTR2(sched_getaffinity, "%d %d", args->pid, args->len); + if (args->len < sizeof(cpuset_t)) return (EINVAL); @@ -1868,11 +1686,8 @@ linux_sched_setaffinity(struct thread *td, { struct cpuset_setaffinity_args csa; -#ifdef DEBUG - if (ldebug(sched_setaffinity)) - printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, - args->len); -#endif + LINUX_CTR2(sched_setaffinity, "%d %d", args->pid, args->len); + if (args->len < sizeof(cpuset_t)) return (EINVAL); diff --git a/sys/compat/linux/linux_misc.h b/sys/compat/linux/linux_misc.h index c80a432..2cdb3c3 100644 --- a/sys/compat/linux/linux_misc.h +++ b/sys/compat/linux/linux_misc.h @@ -45,4 +45,6 @@ #define LINUX_MREMAP_MAYMOVE 1 #define LINUX_MREMAP_FIXED 2 +extern const char linux_platform[]; + #endif /* _LINUX_MISC_H_ */ diff --git a/sys/compat/linux/linux_signal.c b/sys/compat/linux/linux_signal.c index 9bbc268..7948d48 100644 --- a/sys/compat/linux/linux_signal.c +++ b/sys/compat/linux/linux_signal.c @@ -31,6 +31,7 @@ __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include @@ -178,11 +179,7 @@ linux_signal(struct thread *td, struct linux_signal_args *args) l_sigaction_t nsa, osa; int error; -#ifdef DEBUG - if (ldebug(signal)) - printf(ARGS(signal, "%d, %p"), - args->sig, (void *)(uintptr_t)args->handler); -#endif + LINUX_CTR2(signal, "%d %p", args->sig, (void *)(uintptr_t)args->handler); nsa.lsa_handler = args->handler; nsa.lsa_flags = LINUX_SA_ONESHOT | LINUX_SA_NOMASK; @@ -200,12 +197,8 @@ linux_rt_sigaction(struct thread *td, struct linux_rt_sigaction_args *args) l_sigaction_t nsa, osa; int error; -#ifdef DEBUG - if (ldebug(rt_sigaction)) - printf(ARGS(rt_sigaction, "%ld, %p, %p, %ld"), - (long)args->sig, (void *)args->act, - (void *)args->oact, (long)args->sigsetsize); -#endif + LINUX_CTR4(rt_sigaction, "%d %p %p %ld", args->sig, (void *)args->act, + (void *)args->oact, (long)args->sigsetsize); if (args->sigsetsize != sizeof(l_sigset_t)) return (EINVAL); @@ -269,10 +262,7 @@ linux_sigprocmask(struct thread *td, struct linux_sigprocmask_args *args) l_sigset_t set, oset; int error; -#ifdef DEBUG - if (ldebug(sigprocmask)) - printf(ARGS(sigprocmask, "%d, *, *"), args->how); -#endif + LINUX_CTR1(sigprocmask, "%d", args->how); if (args->mask != NULL) { error = copyin(args->mask, &mask, sizeof(l_osigset_t)); @@ -300,12 +290,7 @@ linux_rt_sigprocmask(struct thread *td, struct linux_rt_sigprocmask_args *args) l_sigset_t set, oset; int error; -#ifdef DEBUG - if (ldebug(rt_sigprocmask)) - printf(ARGS(rt_sigprocmask, "%d, %p, %p, %ld"), - args->how, (void *)args->mask, - (void *)args->omask, (long)args->sigsetsize); -#endif + LINUX_CTR1(rt_sigprocmask, "%d", args->how); if (args->sigsetsize != sizeof(l_sigset_t)) return EINVAL; @@ -333,11 +318,6 @@ linux_sgetmask(struct thread *td, struct linux_sgetmask_args *args) struct proc *p = td->td_proc; l_sigset_t mask; -#ifdef DEBUG - if (ldebug(sgetmask)) - printf(ARGS(sgetmask, "")); -#endif - PROC_LOCK(p); bsd_to_linux_sigset(&td->td_sigmask, &mask); PROC_UNLOCK(p); @@ -352,11 +332,6 @@ linux_ssetmask(struct thread *td, struct linux_ssetmask_args *args) l_sigset_t lset; sigset_t bset; -#ifdef DEBUG - if (ldebug(ssetmask)) - printf(ARGS(ssetmask, "%08lx"), (unsigned long)args->mask); -#endif - PROC_LOCK(p); bsd_to_linux_sigset(&td->td_sigmask, &lset); td->td_retval[0] = lset.__bits[0]; @@ -381,11 +356,6 @@ linux_sigpending(struct thread *td, struct linux_sigpending_args *args) l_sigset_t lset; l_osigset_t mask; -#ifdef DEBUG - if (ldebug(sigpending)) - printf(ARGS(sigpending, "*")); -#endif - PROC_LOCK(p); bset = p->p_siglist; SIGSETOR(bset, td->td_siglist); @@ -409,12 +379,6 @@ linux_rt_sigpending(struct thread *td, struct linux_rt_sigpending_args *args) if (args->sigsetsize > sizeof(lset)) return EINVAL; /* NOT REACHED */ - -#ifdef DEBUG - if (ldebug(rt_sigpending)) - printf(ARGS(rt_sigpending, "*")); -#endif - PROC_LOCK(p); bset = p->p_siglist; SIGSETOR(bset, td->td_siglist); @@ -440,10 +404,6 @@ linux_rt_sigtimedwait(struct thread *td, l_siginfo_t linfo; ksiginfo_t info; -#ifdef DEBUG - if (ldebug(rt_sigtimedwait)) - printf(ARGS(rt_sigtimedwait, "*")); -#endif if (args->sigsetsize != sizeof(l_sigset_t)) return (EINVAL); @@ -455,11 +415,10 @@ linux_rt_sigtimedwait(struct thread *td, if (args->timeout) { if ((error = copyin(args->timeout, <v, sizeof(ltv)))) return (error); -#ifdef DEBUG - if (ldebug(rt_sigtimedwait)) - printf(LMSG("linux_rt_sigtimedwait: incoming timeout (%d/%d)\n"), - ltv.tv_sec, ltv.tv_usec); -#endif + + LINUX_CTR2(rt_sigtimedwait, "incoming timeout (%d/%d)", + ltv.tv_sec, ltv.tv_usec); + tv.tv_sec = (long)ltv.tv_sec; tv.tv_usec = (suseconds_t)ltv.tv_usec; if (itimerfix(&tv)) { @@ -475,20 +434,17 @@ linux_rt_sigtimedwait(struct thread *td, } if (tv.tv_sec < 0) timevalclear(&tv); -#ifdef DEBUG - if (ldebug(rt_sigtimedwait)) - printf(LMSG("linux_rt_sigtimedwait: converted timeout (%jd/%ld)\n"), - (intmax_t)tv.tv_sec, tv.tv_usec); -#endif + + LINUX_CTR2(rt_sigtimedwait, "converted (%jd/%ld)", + (intmax_t)tv.tv_sec, tv.tv_usec); } TIMEVAL_TO_TIMESPEC(&tv, &ts); tsa = &ts; } error = kern_sigtimedwait(td, bset, &info, tsa); -#ifdef DEBUG - if (ldebug(rt_sigtimedwait)) - printf(LMSG("linux_rt_sigtimedwait: sigtimedwait returning (%d)\n"), error); -#endif + + LINUX_CTR1(rt_sigtimedwait, "returns %d", error); + if (error) return (error); @@ -517,11 +473,7 @@ linux_kill(struct thread *td, struct linux_kill_args *args) int signum; } */ tmp; -#ifdef DEBUG - if (ldebug(kill)) - printf(ARGS(kill, "%d, %d"), args->pid, args->signum); -#endif - + LINUX_CTR2(kill, "%d %d", args->pid, args->signum); /* * Allow signal 0 as a means to check for privileges */ @@ -598,10 +550,8 @@ int linux_tgkill(struct thread *td, struct linux_tgkill_args *args) { -#ifdef DEBUG - if (ldebug(tgkill)) - printf(ARGS(tgkill, "%d, %d, %d"), args->tgid, args->pid, args->sig); -#endif + LINUX_CTR3(tgkill, "%d %d %d", args->tgid, args->pid, args->sig); + if (args->pid <= 0 || args->tgid <=0) return (EINVAL); @@ -611,10 +561,9 @@ linux_tgkill(struct thread *td, struct linux_tgkill_args *args) int linux_tkill(struct thread *td, struct linux_tkill_args *args) { -#ifdef DEBUG - if (ldebug(tkill)) - printf(ARGS(tkill, "%i, %i"), args->tid, args->sig); -#endif + + LINUX_CTR2(tkill, "%i %i", args->tid, args->sig); + if (args->tid <= 0) return (EINVAL); diff --git a/sys/compat/linux/linux_socket.c b/sys/compat/linux/linux_socket.c index af133d9..a6bfd90 100644 --- a/sys/compat/linux/linux_socket.c +++ b/sys/compat/linux/linux_socket.c @@ -39,6 +39,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -1361,6 +1362,8 @@ linux_socketcall(struct thread *td, struct linux_socketcall_args *args) { void *arg = (void *)(intptr_t)args->args; + LINUX_CTR1(socketcall, "%d", args->what); + switch (args->what) { case LINUX_SOCKET: return (linux_socket(td, arg)); diff --git a/sys/compat/linux/linux_stats.c b/sys/compat/linux/linux_stats.c index 5f1ce53..dae050e 100644 --- a/sys/compat/linux/linux_stats.c +++ b/sys/compat/linux/linux_stats.c @@ -36,6 +36,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -182,10 +183,7 @@ linux_newstat(struct thread *td, struct linux_newstat_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(newstat)) - printf(ARGS(newstat, "%s, *"), path); -#endif + LINUX_CTR(newstat); error = kern_stat(td, path, UIO_SYSSPACE, &buf); if (!error) @@ -205,10 +203,7 @@ linux_newlstat(struct thread *td, struct linux_newlstat_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(newlstat)) - printf(ARGS(newlstat, "%s, *"), path); -#endif + LINUX_CTR(newlstat); error = kern_lstat(td, path, UIO_SYSSPACE, &sb); if (!error) @@ -225,10 +220,7 @@ linux_newfstat(struct thread *td, struct linux_newfstat_args *args) struct stat buf; int error; -#ifdef DEBUG - if (ldebug(newfstat)) - printf(ARGS(newfstat, "%d, *"), args->fd); -#endif + LINUX_CTR1(newfstat, "%d", args->fd); error = kern_fstat(td, args->fd, &buf); translate_fd_major_minor(td, args->fd, &buf); @@ -275,10 +267,8 @@ linux_stat(struct thread *td, struct linux_stat_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(stat)) - printf(ARGS(stat, "%s, *"), path); -#endif + LINUX_CTR(stat); + error = kern_stat(td, path, UIO_SYSSPACE, &buf); if (error) { LFREEPATH(path); @@ -298,10 +288,8 @@ linux_lstat(struct thread *td, struct linux_lstat_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(lstat)) - printf(ARGS(lstat, "%s, *"), path); -#endif + LINUX_CTR(lstat); + error = kern_lstat(td, path, UIO_SYSSPACE, &buf); if (error) { LFREEPATH(path); @@ -389,10 +377,8 @@ linux_statfs(struct thread *td, struct linux_statfs_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(statfs)) - printf(ARGS(statfs, "%s, *"), path); -#endif + LINUX_CTR(statfs); + error = kern_statfs(td, path, UIO_SYSSPACE, &bsd_statfs); LFREEPATH(path); if (error) @@ -430,10 +416,8 @@ linux_statfs64(struct thread *td, struct linux_statfs64_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(statfs64)) - printf(ARGS(statfs64, "%s, *"), path); -#endif + LINUX_CTR(statfs64); + error = kern_statfs(td, path, UIO_SYSSPACE, &bsd_statfs); LFREEPATH(path); if (error) @@ -449,10 +433,8 @@ linux_fstatfs(struct thread *td, struct linux_fstatfs_args *args) struct statfs bsd_statfs; int error; -#ifdef DEBUG - if (ldebug(fstatfs)) - printf(ARGS(fstatfs, "%d, *"), args->fd); -#endif + LINUX_CTR1(fstatfs, "%d", args->fd); + error = kern_fstatfs(td, args->fd, &bsd_statfs); if (error) return error; @@ -471,10 +453,8 @@ struct l_ustat int linux_ustat(struct thread *td, struct linux_ustat_args *args) { -#ifdef DEBUG - if (ldebug(ustat)) - printf(ARGS(ustat, "%d, *"), args->dev); -#endif + + LINUX_CTR1(ustat, "%d", args->dev); return (EOPNOTSUPP); } @@ -521,10 +501,7 @@ linux_stat64(struct thread *td, struct linux_stat64_args *args) LCONVPATHEXIST(td, args->filename, &filename); -#ifdef DEBUG - if (ldebug(stat64)) - printf(ARGS(stat64, "%s, *"), filename); -#endif + LINUX_CTR(stat64); error = kern_stat(td, filename, UIO_SYSSPACE, &buf); if (!error) @@ -544,10 +521,7 @@ linux_lstat64(struct thread *td, struct linux_lstat64_args *args) LCONVPATHEXIST(td, args->filename, &filename); -#ifdef DEBUG - if (ldebug(lstat64)) - printf(ARGS(lstat64, "%s, *"), args->filename); -#endif + LINUX_CTR(lstat64); error = kern_lstat(td, filename, UIO_SYSSPACE, &sb); if (!error) @@ -564,10 +538,7 @@ linux_fstat64(struct thread *td, struct linux_fstat64_args *args) struct stat buf; int error; -#ifdef DEBUG - if (ldebug(fstat64)) - printf(ARGS(fstat64, "%d, *"), args->fd); -#endif + LINUX_CTR1(fstat64, "%d", args->fd); error = kern_fstat(td, args->fd, &buf); translate_fd_major_minor(td, args->fd, &buf); @@ -592,10 +563,7 @@ linux_fstatat64(struct thread *td, struct linux_fstatat64_args *args) dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; LCONVPATHEXIST_AT(td, args->pathname, &path, dfd); -#ifdef DEBUG - if (ldebug(fstatat64)) - printf(ARGS(fstatat64, "%i, %s, %i"), args->dfd, path, args->flag); -#endif + LINUX_CTR(fstatat64); error = kern_statat(td, flag, dfd, path, UIO_SYSSPACE, &buf); translate_path_major_minor_at(td, args->pathname, &buf, dfd); diff --git a/sys/compat/linux/linux_time.c b/sys/compat/linux/linux_time.c index e9bc71b..6ec2507 100644 --- a/sys/compat/linux/linux_time.c +++ b/sys/compat/linux/linux_time.c @@ -46,6 +46,7 @@ __KERNEL_RCSID(0, "$NetBSD: linux_time.c,v 1.14 2006/05/14 03:40:54 christos Exp #include #include +#include #include #include #include @@ -63,6 +64,8 @@ __KERNEL_RCSID(0, "$NetBSD: linux_time.c,v 1.14 2006/05/14 03:40:54 christos Exp #include #endif +#include + static void native_to_linux_timespec(struct l_timespec *, struct timespec *); static int linux_to_native_timespec(struct timespec *, @@ -117,6 +120,8 @@ linux_clock_gettime(struct thread *td, struct linux_clock_gettime_args *args) clockid_t nwhich = 0; /* XXX: GCC */ struct timespec tp; + LINUX_CTR(clock_gettime); + error = linux_to_native_clockid(&nwhich, args->which); if (error != 0) return (error); @@ -136,6 +141,8 @@ linux_clock_settime(struct thread *td, struct linux_clock_settime_args *args) int error; clockid_t nwhich = 0; /* XXX: GCC */ + LINUX_CTR(clock_settime); + error = linux_to_native_clockid(&nwhich, args->which); if (error != 0) return (error); @@ -157,6 +164,8 @@ linux_clock_getres(struct thread *td, struct linux_clock_getres_args *args) int error; clockid_t nwhich = 0; /* XXX: GCC */ + LINUX_CTR(clock_getres); + if (args->tp == NULL) return (0); @@ -179,6 +188,8 @@ linux_nanosleep(struct thread *td, struct linux_nanosleep_args *args) struct timespec rqts, rmts; int error; + LINUX_CTR(nanosleep); + error = copyin(args->rqtp, &lrqts, sizeof lrqts); if (error != 0) return (error); @@ -213,6 +224,8 @@ linux_clock_nanosleep(struct thread *td, struct linux_clock_nanosleep_args *args struct timespec rqts, rmts; int error; + LINUX_CTR(clock_nanosleep); + if (args->flags != 0) return (EINVAL); /* XXX deal with TIMER_ABSTIME */ diff --git a/sys/compat/linux/linux_uid16.c b/sys/compat/linux/linux_uid16.c index 2685eaa..0459efe 100644 --- a/sys/compat/linux/linux_uid16.c +++ b/sys/compat/linux/linux_uid16.c @@ -31,6 +31,7 @@ __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include @@ -65,10 +66,8 @@ linux_chown16(struct thread *td, struct linux_chown16_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(chown16)) - printf(ARGS(chown16, "%s, %d, %d"), path, args->uid, args->gid); -#endif + LINUX_CTR2(chown16, "%d, %d", args->uid, args->gid); + error = kern_chown(td, path, UIO_SYSSPACE, CAST_NOCHG(args->uid), CAST_NOCHG(args->gid)); LFREEPATH(path); @@ -83,11 +82,8 @@ linux_lchown16(struct thread *td, struct linux_lchown16_args *args) LCONVPATHEXIST(td, args->path, &path); -#ifdef DEBUG - if (ldebug(lchown16)) - printf(ARGS(lchown16, "%s, %d, %d"), path, args->uid, - args->gid); -#endif + LINUX_CTR2(lchown16, "%d, %d", args->uid, args->gid); + error = kern_lchown(td, path, UIO_SYSSPACE, CAST_NOCHG(args->uid), CAST_NOCHG(args->gid)); LFREEPATH(path); @@ -103,10 +99,7 @@ linux_setgroups16(struct thread *td, struct linux_setgroups16_args *args) int ngrp, error; struct proc *p; -#ifdef DEBUG - if (ldebug(setgroups16)) - printf(ARGS(setgroups16, "%d, *"), args->gidsetsize); -#endif + LINUX_CTR1(setgroups16, "%d", args->gidsetsize); ngrp = args->gidsetsize; if (ngrp < 0 || ngrp >= NGROUPS) @@ -160,10 +153,7 @@ linux_getgroups16(struct thread *td, struct linux_getgroups16_args *args) gid_t *bsd_gidset; int bsd_gidsetsz, ngrp, error; -#ifdef DEBUG - if (ldebug(getgroups16)) - printf(ARGS(getgroups16, "%d, *"), args->gidsetsize); -#endif + LINUX_CTR1(getgroups16, "%d", args->gidsetsize); cred = td->td_ucred; bsd_gidset = cred->cr_groups; diff --git a/sys/compat/linux/linux_util.h b/sys/compat/linux/linux_util.h index 4caad9d..306b7b0 100644 --- a/sys/compat/linux/linux_util.h +++ b/sys/compat/linux/linux_util.h @@ -106,4 +106,33 @@ int linux_driver_get_major_minor(char *node, int *major, int *minor); char *linux_get_char_devices(void); void linux_free_get_char_devices(char *string); +#if defined(DEBUG) && defined(KTR) + +#define KTR_LINUX KTR_SUBSYS +#define LINUX_CTRFMT(nm, fmt) #nm"("fmt")" + +#define LINUX_CTR6(f, m, p1, p2, p3, p4, p5, p6) do { \ + if (ldebug(f)) \ + CTR6(KTR_LINUX, LINUX_CTRFMT(f, m), \ + p1, p2, p3, p4, p5, p6); \ +} while (0) + +#define LINUX_CTR(f) LINUX_CTR6(f, "", 0, 0, 0, 0, 0, 0) +#define LINUX_CTR0(f, m) LINUX_CTR6(f, m, 0, 0, 0, 0, 0, 0) +#define LINUX_CTR1(f, m, p1) LINUX_CTR6(f, m, p1, 0, 0, 0, 0, 0) +#define LINUX_CTR2(f, m, p1, p2) LINUX_CTR6(f, m, p1, p2, 0, 0, 0, 0) +#define LINUX_CTR3(f, m, p1, p2, p3) LINUX_CTR6(f, m, p1, p2, p3, 0, 0, 0) +#define LINUX_CTR4(f, m, p1, p2, p3, p4) LINUX_CTR6(f, m, p1, p2, p3, p4, 0, 0) +#define LINUX_CTR5(f, m, p1, p2, p3, p4, p5) LINUX_CTR6(f, m, p1, p2, p3, p4, p5, 0) +#else +#define LINUX_CTR(f) +#define LINUX_CTR0(f, m) +#define LINUX_CTR1(f, m, p1) +#define LINUX_CTR2(f, m, p1, p2) +#define LINUX_CTR3(f, m, p1, p2, p3) +#define LINUX_CTR4(f, m, p1, p2, p3, p4) +#define LINUX_CTR5(f, m, p1, p2, p3, p4, p5) +#define LINUX_CTR6(f, m, p1, p2, p3, p4, p5, p6) +#endif + #endif /* !_LINUX_UTIL_H_ */ diff --git a/sys/compat/svr4/svr4_sysvec.c b/sys/compat/svr4/svr4_sysvec.c index 0030e3a..a406576 100644 --- a/sys/compat/svr4/svr4_sysvec.c +++ b/sys/compat/svr4/svr4_sysvec.c @@ -204,6 +204,7 @@ Elf32_Brandinfo svr4_brand = { .interp_path = "/lib/libc.so.1", .sysvec = &svr4_sysvec, .interp_newpath = NULL, + .brand_abi_note = NULL, .flags = 0 }; diff --git a/sys/i386/i386/elf_machdep.c b/sys/i386/i386/elf_machdep.c index 19eddd0..593f9bd 100644 --- a/sys/i386/i386/elf_machdep.c +++ b/sys/i386/i386/elf_machdep.c @@ -84,7 +84,8 @@ static Elf32_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_ANY, @@ -99,7 +100,8 @@ static Elf32_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/i386/linux/linux.h b/sys/i386/linux/linux.h index f9c7ee5..ca01c89 100644 --- a/sys/i386/linux/linux.h +++ b/sys/i386/linux/linux.h @@ -102,6 +102,10 @@ typedef struct { #define LINUX_CTL_MAXNAME 10 +#define LINUX_AT_COUNT 16 /* Count of used aux entry types. + * Keep this synchronized with + * elf_linux_fixup() code. + */ struct l___sysctl_args { l_int *name; diff --git a/sys/i386/linux/linux_machdep.c b/sys/i386/linux/linux_machdep.c index 7f727ac..0093898 100644 --- a/sys/i386/linux/linux_machdep.c +++ b/sys/i386/linux/linux_machdep.c @@ -34,6 +34,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -126,10 +127,7 @@ linux_execve(struct thread *td, struct linux_execve_args *args) LCONVPATHEXIST(td, args->path, &newpath); -#ifdef DEBUG - if (ldebug(execve)) - printf(ARGS(execve, "%s"), newpath); -#endif + LINUX_CTR(execve); error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, args->argp, args->envp); @@ -277,10 +275,7 @@ linux_old_select(struct thread *td, struct linux_old_select_args *args) struct linux_select_args newsel; int error; -#ifdef DEBUG - if (ldebug(old_select)) - printf(ARGS(old_select, "%p"), args->ptr); -#endif + LINUX_CTR1(old_select, "%p", args->ptr); error = copyin(args->ptr, &linux_args, sizeof(linux_args)); if (error) @@ -301,10 +296,7 @@ linux_fork(struct thread *td, struct linux_fork_args *args) struct proc *p2; struct thread *td2; -#ifdef DEBUG - if (ldebug(fork)) - printf(ARGS(fork, "")); -#endif + LINUX_CTR(fork); if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2)) != 0) return (error); @@ -340,10 +332,7 @@ linux_vfork(struct thread *td, struct linux_vfork_args *args) struct proc *p2; struct thread *td2; -#ifdef DEBUG - if (ldebug(vfork)) - printf(ARGS(vfork, "")); -#endif + LINUX_CTR(vfork); /* exclude RFPPWAIT */ if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2)) != 0) @@ -391,13 +380,9 @@ linux_clone(struct thread *td, struct linux_clone_args *args) int exit_signal; struct linux_emuldata *em; -#ifdef DEBUG - if (ldebug(clone)) { - printf(ARGS(clone, "flags %x, stack %x, parent tid: %x, child tid: %x"), - (unsigned int)args->flags, (unsigned int)args->stack, - (unsigned int)args->parent_tidptr, (unsigned int)args->child_tidptr); - } -#endif + LINUX_CTR4(clone, "flags %x, stack %x, parent tid: %x, child tid: %x", + (unsigned int)args->flags, (unsigned int)args->stack, + (unsigned int)args->parent_tidptr, (unsigned int)args->child_tidptr); exit_signal = args->flags & 0x000000ff; if (LINUX_SIG_VALID(exit_signal)) { @@ -535,19 +520,6 @@ linux_clone(struct thread *td, struct linux_clone_args *args) a[1] = LINUX_LDT_entry_b(&info); memcpy(&sd, &a, sizeof(a)); -#ifdef DEBUG - if (ldebug(clone)) - printf("Segment created in clone with CLONE_SETTLS: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, - sd.sd_hibase, - sd.sd_lolimit, - sd.sd_hilimit, - sd.sd_type, - sd.sd_dpl, - sd.sd_p, - sd.sd_xx, - sd.sd_def32, - sd.sd_gran); -#endif /* set %gs */ td2->td_pcb->pcb_gsd = sd; @@ -555,11 +527,9 @@ linux_clone(struct thread *td, struct linux_clone_args *args) } } -#ifdef DEBUG - if (ldebug(clone)) - printf(LMSG("clone: successful rfork to %ld, stack %p sig = %d"), - (long)p2->p_pid, args->stack, exit_signal); -#endif + LINUX_CTR3(clone, "successful rfork to %ld, stack %p sig = %d", + (long)p2->p_pid, args->stack, exit_signal); + if (args->flags & LINUX_CLONE_VFORK) { PROC_LOCK(p2); p2->p_flag |= P_PPWAIT; @@ -598,12 +568,9 @@ linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { struct l_mmap_argv linux_args; -#ifdef DEBUG - if (ldebug(mmap2)) - printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), - (void *)args->addr, args->len, args->prot, - args->flags, args->fd, args->pgoff); -#endif + LINUX_CTR6(mmap2, "%p, %d, %d, 0x%08x, %d, %d", + (void *)args->addr, args->len, args->prot, + args->flags, args->fd, args->pgoff); linux_args.addr = args->addr; linux_args.len = args->len; @@ -625,12 +592,9 @@ linux_mmap(struct thread *td, struct linux_mmap_args *args) if (error) return (error); -#ifdef DEBUG - if (ldebug(mmap)) - printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), - (void *)linux_args.addr, linux_args.len, linux_args.prot, - linux_args.flags, linux_args.fd, linux_args.pgoff); -#endif + LINUX_CTR6(mmap, "%p, %d, %d, 0x%08x, %d, %d", + linux_args.addr, linux_args.len, linux_args.prot, + linux_args.flags, linux_args.fd, linux_args.pgoff); return (linux_mmap_common(td, &linux_args)); } @@ -780,19 +744,10 @@ linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args) } bsd_args.pos = linux_args->pgoff; -#ifdef DEBUG - if (ldebug(mmap)) - printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", - __func__, - (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, - bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); -#endif error = mmap(td, &bsd_args); -#ifdef DEBUG - if (ldebug(mmap)) - printf("-> %s() return: 0x%x (0x%08x)\n", - __func__, error, (u_int)td->td_retval[0]); -#endif + + LINUX_CTR3(mmap, "-> %s() return: 0x%x (0x%08x)\n", + __func__, error, (u_int)td->td_retval[0]); return (error); } @@ -815,10 +770,7 @@ linux_pipe(struct thread *td, struct linux_pipe_args *args) int error; int fildes[2]; -#ifdef DEBUG - if (ldebug(pipe)) - printf(ARGS(pipe, "*")); -#endif + LINUX_CTR(pipe); error = kern_pipe(td, fildes); if (error) @@ -930,11 +882,8 @@ linux_sigaction(struct thread *td, struct linux_sigaction_args *args) l_sigaction_t act, oact; int error; -#ifdef DEBUG - if (ldebug(sigaction)) - printf(ARGS(sigaction, "%d, %p, %p"), - args->sig, (void *)args->nsa, (void *)args->osa); -#endif + LINUX_CTR3(sigaction, "%d, %p, %p", + args->sig, (void *)args->nsa, (void *)args->osa); if (args->nsa != NULL) { error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); @@ -972,10 +921,7 @@ linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) sigset_t sigmask; l_sigset_t mask; -#ifdef DEBUG - if (ldebug(sigsuspend)) - printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); -#endif + LINUX_CTR1(sigsuspend, "%08lx", (unsigned long)args->mask); LINUX_SIGEMPTYSET(mask); mask.__bits[0] = args->mask; @@ -990,11 +936,8 @@ linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) sigset_t sigmask; int error; -#ifdef DEBUG - if (ldebug(rt_sigsuspend)) - printf(ARGS(rt_sigsuspend, "%p, %d"), - (void *)uap->newset, uap->sigsetsize); -#endif + LINUX_CTR2(rt_sigsuspend, "%p, %d", + (void *)uap->newset, uap->sigsetsize); if (uap->sigsetsize != sizeof(l_sigset_t)) return (EINVAL); @@ -1013,10 +956,7 @@ linux_pause(struct thread *td, struct linux_pause_args *args) struct proc *p = td->td_proc; sigset_t sigmask; -#ifdef DEBUG - if (ldebug(pause)) - printf(ARGS(pause, "")); -#endif + LINUX_CTR(pause); PROC_LOCK(p); sigmask = td->td_sigmask; @@ -1031,10 +971,7 @@ linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) l_stack_t lss; int error; -#ifdef DEBUG - if (ldebug(sigaltstack)) - printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); -#endif + LINUX_CTR2(sigaltstack, "%p, %p", uap->uss, uap->uoss); if (uap->uss != NULL) { error = copyin(uap->uss, &lss, sizeof(l_stack_t)); @@ -1062,11 +999,8 @@ linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) { struct ftruncate_args sa; -#ifdef DEBUG - if (ldebug(ftruncate64)) - printf(ARGS(ftruncate64, "%u, %jd"), args->fd, - (intmax_t)args->length); -#endif + LINUX_CTR2(ftruncate64, "%u, %jd", + args->fd, (intmax_t)args->length); sa.fd = args->fd; sa.length = args->length; @@ -1086,19 +1020,18 @@ linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args if (error) return (error); -#ifdef DEBUG - if (ldebug(set_thread_area)) - printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), - info.entry_number, - info.base_addr, - info.limit, - info.seg_32bit, - info.contents, - info.read_exec_only, - info.limit_in_pages, - info.seg_not_present, - info.useable); -#endif + LINUX_CTR6(set_thread_area, "%i, %x, %x, %i, %i, %i", + info.entry_number, + info.base_addr, + info.limit, + info.seg_32bit, + info.contents, + info.read_exec_only); + + LINUX_CTR3(set_thread_area, "%i, %i, %i", + info.limit_in_pages, + info.seg_not_present, + info.useable); idx = info.entry_number; /* @@ -1148,19 +1081,20 @@ linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args } memcpy(&sd, &a, sizeof(a)); -#ifdef DEBUG - if (ldebug(set_thread_area)) - printf("Segment created in set_thread_area: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, - sd.sd_hibase, - sd.sd_lolimit, - sd.sd_hilimit, - sd.sd_type, - sd.sd_dpl, - sd.sd_p, - sd.sd_xx, - sd.sd_def32, - sd.sd_gran); -#endif + + LINUX_CTR6(set_thread_area, "Segment created %x %x %x %x %i %i", + sd.sd_lobase, + sd.sd_hibase, + sd.sd_lolimit, + sd.sd_hilimit, + sd.sd_type, + sd.sd_dpl); + + LINUX_CTR4(set_thread_area, "Segment created %i %i %i %i", + sd.sd_p, + sd.sd_xx, + sd.sd_def32, + sd.sd_gran); /* this is taken from i386 version of cpu_set_user_tls() */ critical_enter(); @@ -1183,10 +1117,7 @@ linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args struct l_desc_struct desc; struct segment_descriptor sd; -#ifdef DEBUG - if (ldebug(get_thread_area)) - printf(ARGS(get_thread_area, "%p"), args->desc); -#endif + LINUX_CTR1(get_thread_area, "%p", args->desc); error = copyin(args->desc, &info, sizeof(struct l_user_desc)); if (error) diff --git a/sys/i386/linux/linux_sysvec.c b/sys/i386/linux/linux_sysvec.c index 42365fb..ce1b995 100644 --- a/sys/i386/linux/linux_sysvec.c +++ b/sys/i386/linux/linux_sysvec.c @@ -37,6 +37,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include @@ -58,11 +59,13 @@ __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include #include +#include #include #include #include @@ -89,6 +92,7 @@ MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); #define fldcw(addr) __asm("fldcw %0" : : "m" (*(addr))) #define __LINUX_NPXCW__ 0x37f +#define aligned(a, t) (rounddown((unsigned long)(a), sizeof(t)) == (unsigned long)(a)) extern char linux_sigcode[]; extern int linux_szsigcode; @@ -107,9 +111,13 @@ static void linux_prepsyscall(struct trapframe *tf, int *args, u_int *code, static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); static void exec_linux_setregs(struct thread *td, u_long entry, u_long stack, u_long ps_strings); +static register_t *linux_copyout_strings(struct image_params *imgp); + +static int linux_szplatform; +const char *linux_platform; extern LIST_HEAD(futex_list, futex) futex_list; -extern struct sx futex_sx; +extern struct mtx futex_mtx; static eventhandler_tag linux_exit_tag; static eventhandler_tag linux_schedtail_tag; @@ -231,22 +239,30 @@ linux_fixup(register_t **stack_base, struct image_params *imgp) **stack_base = (intptr_t)(void *)argv; (*stack_base)--; **stack_base = imgp->args->argc; - return 0; + 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_szsigcode - + linux_szplatform); args = (Elf32_Auxargs *)imgp->auxargs; pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2); - if (args->execfd != -1) - AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); + AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); + AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, hz); AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); AUXARGS_ENTRY(pos, AT_PHENT, args->phent); AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); @@ -254,10 +270,14 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp) 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); @@ -265,9 +285,125 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp) (*stack_base)--; **stack_base = (register_t)imgp->args->argc; - return 0; + 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 - linux_szsigcode - SPARE_USRSPACE - + linux_szplatform - roundup((ARG_MAX - imgp->args->stringspace), + sizeof(char *)); + + /* + * install sigcode + */ + copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo - + linux_szsigcode), linux_szsigcode); + + /* + * install LINUX_PLATFORM + */ + copyout(linux_platform, ((caddr_t)arginfo - linux_szsigcode - + 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; @@ -290,11 +426,8 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 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 + LINUX_CTR4(rt_sendsig, "%p, %d, %p, %u", + catcher, sig, (void*)mask, code); /* * Allocate space for the signal handler context. */ @@ -358,23 +491,13 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 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 + LINUX_CTR2(rt_sendsig, "bad stack %p, oonstack=%x", + fp, oonstack); PROC_LOCK(p); sigexit(td, SIGILL); } @@ -395,6 +518,53 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) mtx_lock(&psp->ps_mtx); } +static const char GNULINUX_ABI_VENDOR[] = "GNU"; +static const int GNULINUX_ABI_LEN = 16; + +static int +linux_abi_note(struct image_params *imgp, int32_t *osrel) +{ + const Elf_Note *note, *note_end; + const Elf32_Phdr *phdr, *pnote; + const Elf32_Ehdr *hdr; + const char *note_name; + int i; + + pnote = NULL; + hdr = (const Elf32_Ehdr *)imgp->image_header; + phdr = (const Elf32_Phdr *)(imgp->image_header + hdr->e_phoff); + + for (i = 0; i < hdr->e_phnum; i++) + if (phdr[i].p_type == PT_NOTE) { + pnote = &phdr[i]; + break; + } + if (pnote == NULL || pnote->p_offset >= PAGE_SIZE || + pnote->p_offset + pnote->p_filesz >= PAGE_SIZE) + return (0); + + note = (const Elf_Note *)(imgp->image_header + pnote->p_offset); + if (!aligned(note, Elf32_Addr)) + return (0); + note_end = (const Elf_Note *)(imgp->image_header + + pnote->p_offset + pnote->p_filesz); + while (note < note_end) { + if (note->n_namesz == sizeof(GNULINUX_ABI_VENDOR) && + note->n_descsz == GNULINUX_ABI_LEN && + note->n_type == 1 /* ABI_NOTETYPE */) { + note_name = (const char *)(note + 1); + if (strncmp(GNULINUX_ABI_VENDOR, note_name, + sizeof(GNULINUX_ABI_VENDOR)) == 0) + return (1); + } + note = (const Elf_Note *)((const char *)(note + 1) + + roundup2(note->n_namesz, sizeof(Elf32_Addr)) + + roundup2(note->n_descsz, sizeof(Elf32_Addr))); + } + + return (0); +} + /* * Send an interrupt to process. @@ -431,12 +601,8 @@ linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 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 - + LINUX_CTR4(sendsig, "%p, %d, %p, %u", + catcher, sig, (void*)mask, code); /* * Allocate space for the signal handler context. */ @@ -536,10 +702,7 @@ linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) regs = td->td_frame; -#ifdef DEBUG - if (ldebug(sigreturn)) - printf(ARGS(sigreturn, "%p"), (void *)args->sfp); -#endif + LINUX_CTR1(sigreturn, "%p", args->sfp); /* * The trampoline code hands us the sigframe. * It is unsafe to keep track of it ourselves, in the event that a @@ -638,10 +801,7 @@ linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) regs = td->td_frame; -#ifdef DEBUG - if (ldebug(rt_sigreturn)) - printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp); -#endif + LINUX_CTR1(rt_sigreturn, "%p", args->ucp); /* * The trampoline code hands us the ucontext. * It is unsafe to keep track of it ourselves, in the event that a @@ -720,11 +880,6 @@ linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 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); @@ -808,6 +963,28 @@ exec_linux_setregs(struct thread *td, u_long entry, fldcw(&control); } +static int +linux_get_machine(const char **dst) +{ + const char *class; + + switch (cpu_class) { + case CPUCLASS_686: + class = "i686"; + break; + case CPUCLASS_586: + class = "i586"; + break; + case CPUCLASS_486: + class = "i486"; + break; + default: + class = "i386"; + } + *dst = class; + return (0); +} + struct sysentvec linux_sysvec = { .sv_size = LINUX_SYS_MAXSYSCALL, .sv_table = linux_sysent, @@ -863,7 +1040,7 @@ struct sysentvec elf_linux_sysvec = { .sv_usrstack = USRSTACK, .sv_psstrings = PS_STRINGS, .sv_stackprot = VM_PROT_ALL, - .sv_copyout_strings = exec_copyout_strings, + .sv_copyout_strings = linux_copyout_strings, .sv_setregs = exec_linux_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, @@ -878,7 +1055,8 @@ static Elf32_Brandinfo linux_brand = { .interp_path = "/lib/ld-linux.so.1", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = linux_abi_note, + .flags = BI_CAN_EXEC_DYN | BI_CAN_EXEC_INTERP }; static Elf32_Brandinfo linux_glibc2brand = { @@ -889,7 +1067,8 @@ static Elf32_Brandinfo linux_glibc2brand = { .interp_path = "/lib/ld-linux.so.2", .sysvec = &elf_linux_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = linux_abi_note, + .flags = BI_CAN_EXEC_DYN | BI_CAN_EXEC_INTERP }; Elf32_Brandinfo *linux_brandlist[] = { @@ -922,13 +1101,16 @@ linux_elf_modevent(module_t mod, int type, void *data) mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF); sx_init(&emul_shared_lock, "emuldata->shared lock"); LIST_INIT(&futex_list); - sx_init(&futex_sx, "futex protection lock"); + mtx_init(&futex_mtx, "futex list protection lock", NULL, MTX_DEF); linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, linux_proc_exit, NULL, 1000); linux_schedtail_tag = EVENTHANDLER_REGISTER(schedtail, linux_schedtail, 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 *)); if (bootverbose) printf("Linux ELF exec handler installed\n"); } else @@ -952,7 +1134,7 @@ linux_elf_modevent(module_t mod, int type, void *data) linux_device_unregister_handler(*ldhp); mtx_destroy(&emul_lock); sx_destroy(&emul_shared_lock); - sx_destroy(&futex_sx); + mtx_destroy(&futex_mtx); EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag); EVENTHANDLER_DEREGISTER(schedtail, linux_schedtail_tag); EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag); diff --git a/sys/ia64/ia64/elf_machdep.c b/sys/ia64/ia64/elf_machdep.c index a3a6e57..a0289de 100644 --- a/sys/ia64/ia64/elf_machdep.c +++ b/sys/ia64/ia64/elf_machdep.c @@ -92,7 +92,8 @@ static Elf64_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t)elf64_insert_brand_entry, &freebsd_brand_info); @@ -105,7 +106,8 @@ static Elf64_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t)elf64_insert_brand_entry, &freebsd_brand_oinfo); diff --git a/sys/kern/imgact_elf.c b/sys/kern/imgact_elf.c index ec96974..176ec84 100644 --- a/sys/kern/imgact_elf.c +++ b/sys/kern/imgact_elf.c @@ -78,8 +78,8 @@ __FBSDID("$FreeBSD$"); #define OLD_EI_BRAND 8 static int __elfN(check_header)(const Elf_Ehdr *hdr); -static Elf_Brandinfo *__elfN(get_brandinfo)(const Elf_Ehdr *hdr, - const char *interp); +static Elf_Brandinfo *__elfN(get_brandinfo)(struct image_params *imgp, + const char *interp, int32_t *osrel); static int __elfN(load_file)(struct proc *p, const char *file, u_long *addr, u_long *entry, size_t pagesize); static int __elfN(load_section)(struct vmspace *vmspace, vm_object_t object, @@ -158,19 +158,31 @@ __elfN(brand_inuse)(Elf_Brandinfo *entry) } static Elf_Brandinfo * -__elfN(get_brandinfo)(const Elf_Ehdr *hdr, const char *interp) +__elfN(get_brandinfo)(struct image_params *imgp, const char *interp, + int32_t *osrel) { + const Elf_Ehdr *hdr = (const Elf_Ehdr *)imgp->image_header; Elf_Brandinfo *bi; + int ret, fname_len, interp_len; int i; /* - * We support three types of branding -- (1) the ELF EI_OSABI field + * We support four types of branding -- (1) the ELF EI_OSABI field * that SCO added to the ELF spec, (2) FreeBSD 3.x's traditional string - * branding w/in the ELF header, and (3) path of the `interp_path' - * field. We should also look for an ".note.ABI-tag" ELF section now - * in all Linux ELF binaries, FreeBSD 4.1+, and some NetBSD ones. + * branding w/in the ELF header, (3) path of the `interp_path' + * field, and (4) the ".note.ABI-tag" ELF section. */ + /* Look for an ".note.ABI-tag" ELF section */ + for (i = 0; i < MAX_BRANDS; i++) { + bi = elf_brand_list[i]; + if (bi != NULL && hdr->e_machine == bi->machine && bi->brand_abi_note) { + ret = (*bi->brand_abi_note)(imgp, osrel); + if (ret) + return (bi); + } + } + /* If the executable has a brand, search for it in the brand list. */ for (i = 0; i < MAX_BRANDS; i++) { bi = elf_brand_list[i]; @@ -191,6 +203,23 @@ __elfN(get_brandinfo)(const Elf_Ehdr *hdr, const char *interp) } } + /* Some ABI (Linux) allow to run the interpreter. */ + for (i = 0; i < MAX_BRANDS; i++) { + bi = elf_brand_list[i]; + if (bi == NULL || hdr->e_machine != bi->machine || + (bi->flags & BI_CAN_EXEC_INTERP) == 0) + continue; + + fname_len = strlen(imgp->args->fname); + interp_len = strlen(bi->interp_path); + if (fname_len < interp_len) + continue; + ret = strncmp(imgp->args->fname + (fname_len - interp_len), + bi->interp_path, interp_len); + if (ret == 0) + return (bi); + } + /* Lacking a recognized interpreter, try the default brand */ for (i = 0; i < MAX_BRANDS; i++) { bi = elf_brand_list[i]; @@ -590,13 +619,11 @@ fail: return (error); } -static const char FREEBSD_ABI_VENDOR[] = "FreeBSD"; - static int __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp) { const Elf_Ehdr *hdr = (const Elf_Ehdr *)imgp->image_header; - const Elf_Phdr *phdr, *pnote = NULL; + const Elf_Phdr *phdr; Elf_Auxargs *elf_auxargs; struct vmspace *vmspace; vm_prot_t prot; @@ -604,12 +631,11 @@ __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp) u_long text_addr = 0, data_addr = 0; u_long seg_size, seg_addr; u_long addr, entry = 0, proghdr = 0; + int32_t osrel = 0; int error = 0, i; const char *interp = NULL, *newinterp = NULL; Elf_Brandinfo *brand_info; - const Elf_Note *note, *note_end; char *path; - const char *note_name; struct sysentvec *sv; /* @@ -646,7 +672,7 @@ __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp) } } - brand_info = __elfN(get_brandinfo)(hdr, interp); + brand_info = __elfN(get_brandinfo)(imgp, interp, &osrel); if (brand_info == NULL) { uprintf("ELF binary type \"%u\" not known.\n", hdr->e_ident[EI_OSABI]); @@ -750,9 +776,6 @@ __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp) case PT_PHDR: /* Program header table info */ proghdr = phdr[i].p_vaddr; break; - case PT_NOTE: - pnote = &phdr[i]; - break; default: break; } @@ -840,41 +863,7 @@ __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp) imgp->auxargs = elf_auxargs; imgp->interpreted = 0; - - /* - * Try to fetch the osreldate for FreeBSD binary from the ELF - * OSABI-note. Only the first page of the image is searched, - * the same as for headers. - */ - if (pnote != NULL && pnote->p_offset < PAGE_SIZE && - pnote->p_offset + pnote->p_filesz < PAGE_SIZE ) { - note = (const Elf_Note *)(imgp->image_header + pnote->p_offset); - if (!aligned(note, Elf32_Addr)) { - free(imgp->auxargs, M_TEMP); - imgp->auxargs = NULL; - return (ENOEXEC); - } - note_end = (const Elf_Note *)(imgp->image_header + pnote->p_offset + - pnote->p_filesz); - while (note < note_end) { - if (note->n_namesz == sizeof(FREEBSD_ABI_VENDOR) && - note->n_descsz == sizeof(int32_t) && - note->n_type == 1 /* ABI_NOTETYPE */) { - note_name = (const char *)(note + 1); - if (strncmp(FREEBSD_ABI_VENDOR, note_name, - sizeof(FREEBSD_ABI_VENDOR)) == 0) { - imgp->proc->p_osrel = *(const int32_t *) - (note_name + - round_page_ps(sizeof(FREEBSD_ABI_VENDOR), - sizeof(Elf32_Addr))); - break; - } - } - note = (const Elf_Note *)((const char *)(note + 1) + - round_page_ps(note->n_namesz, sizeof(Elf32_Addr)) + - round_page_ps(note->n_descsz, sizeof(Elf32_Addr))); - } - } + imgp->proc->p_osrel = osrel; return (error); } @@ -1335,6 +1324,63 @@ __elfN(putnote)(void *dst, size_t *off, const char *name, int type, *off += roundup2(note.n_descsz, sizeof(Elf_Size)); } +static const char FREEBSD_ABI_VENDOR[] = "FreeBSD"; + +int +__elfN(freebsd_abi_note)(struct image_params *imgp, int32_t *osrel) +{ + const Elf_Note *note, *note_end; + const Elf32_Phdr *phdr, *pnote; + const Elf32_Ehdr *hdr; + const char *note_name; + int i; + + pnote = NULL; + hdr = (const Elf32_Ehdr *)imgp->image_header; + phdr = (const Elf32_Phdr *)(imgp->image_header + hdr->e_phoff); + + for (i = 0; i < hdr->e_phnum; i++) + if (phdr[i].p_type == PT_NOTE) { + pnote = &phdr[i]; + break; + } + + if (pnote == NULL || pnote->p_offset >= PAGE_SIZE || + pnote->p_offset + pnote->p_filesz >= PAGE_SIZE) + return (0); + + note = (const Elf_Note *)(imgp->image_header + pnote->p_offset); + if (!aligned(note, Elf32_Addr)) + return (0); + note_end = (const Elf_Note *)(imgp->image_header + + pnote->p_offset + pnote->p_filesz); + while (note < note_end) { + if (note->n_namesz == sizeof(FREEBSD_ABI_VENDOR) && + note->n_descsz == sizeof(int32_t) && + note->n_type == 1 /* ABI_NOTETYPE */) { + note_name = (const char *)(note + 1); + if (strncmp(FREEBSD_ABI_VENDOR, note_name, + sizeof(FREEBSD_ABI_VENDOR)) != 0) + continue; + /* + * Fetch the osreldate for FreeBSD binary + * from the ELF OSABI-note. + */ + if (osrel != NULL) + *osrel = *(const int32_t *) + (note_name + + round_page_ps(sizeof(FREEBSD_ABI_VENDOR), + sizeof(Elf32_Addr))); + return (1); + } + note = (const Elf_Note *)((const char *)(note + 1) + + round_page_ps(note->n_namesz, sizeof(Elf32_Addr)) + + round_page_ps(note->n_descsz, sizeof(Elf32_Addr))); + } + + return (0); +} + /* * Tell kern_execve.c about it, with a little help from the linker. */ diff --git a/sys/mips/mips/elf_machdep.c b/sys/mips/mips/elf_machdep.c index 163d0ee..807ae67 100644 --- a/sys/mips/mips/elf_machdep.c +++ b/sys/mips/mips/elf_machdep.c @@ -86,6 +86,8 @@ static Elf32_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, + .brand_abi_note = NULL, + .brand_abi_note = NULL, .flags = 0 }; diff --git a/sys/modules/linux/Makefile b/sys/modules/linux/Makefile index fbf5a58..280cb09 100644 --- a/sys/modules/linux/Makefile +++ b/sys/modules/linux/Makefile @@ -17,7 +17,7 @@ SRCS= linux${SFX}_dummy.c linux_emul.c linux_file.c \ device_if.h bus_if.h assym.s # XXX: for assym.s -SRCS+= opt_kstack_pages.h opt_nfs.h opt_apic.h opt_compat.h +SRCS+= opt_kstack_pages.h opt_nfs.h opt_apic.h opt_compat.h opt_hwpmc_hooks.h OBJS= linux${SFX}_locore.o linux${SFX}_support.o @@ -55,6 +55,18 @@ linux${SFX}_genassym.o: linux${SFX}_genassym.c linux.h @ machine .if !defined(KERNBUILDDIR) opt_inet6.h: echo "#define INET6 1" > ${.TARGET} + +.if defined(DEBUG) +CFLAGS+=-DDEBUG +.endif + +.if defined(KTR) +CFLAGS+=-DKTR +.endif + +.if defined(INVARIANTS) +CFLAGS+=-DINVARIANTS +.endif .endif .include diff --git a/sys/powerpc/powerpc/elf_machdep.c b/sys/powerpc/powerpc/elf_machdep.c index 69ac55b..f0dc474 100644 --- a/sys/powerpc/powerpc/elf_machdep.c +++ b/sys/powerpc/powerpc/elf_machdep.c @@ -87,7 +87,8 @@ static Elf32_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_ANY, @@ -102,7 +103,8 @@ static Elf32_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/sparc64/sparc64/elf_machdep.c b/sys/sparc64/sparc64/elf_machdep.c index a956c5c..03b9056 100644 --- a/sys/sparc64/sparc64/elf_machdep.c +++ b/sys/sparc64/sparc64/elf_machdep.c @@ -99,7 +99,8 @@ static Elf64_Brandinfo freebsd_brand_info = { .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(elf64, SI_SUB_EXEC, SI_ORDER_ANY, @@ -114,7 +115,8 @@ static Elf64_Brandinfo freebsd_brand_oinfo = { .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec, .interp_newpath = NULL, - .flags = BI_CAN_EXEC_DYN, + .brand_abi_note = __elfN(freebsd_abi_note), + .flags = BI_CAN_EXEC_DYN }; SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY, diff --git a/sys/sys/imgact_elf.h b/sys/sys/imgact_elf.h index deb5b10..d290185 100644 --- a/sys/sys/imgact_elf.h +++ b/sys/sys/imgact_elf.h @@ -62,8 +62,10 @@ typedef struct { const char *interp_path; struct sysentvec *sysvec; const char *interp_newpath; + int (*brand_abi_note)(struct image_params *, int32_t *); int flags; -#define BI_CAN_EXEC_DYN 0x0001 +#define BI_CAN_EXEC_DYN 0x0001 +#define BI_CAN_EXEC_INTERP 0x0002 } __ElfN(Brandinfo); __ElfType(Auxargs); @@ -76,6 +78,7 @@ int __elfN(insert_brand_entry)(Elf_Brandinfo *entry); int __elfN(remove_brand_entry)(Elf_Brandinfo *entry); int __elfN(freebsd_fixup)(register_t **, struct image_params *); int __elfN(coredump)(struct thread *, struct vnode *, off_t); +int __elfN(freebsd_abi_note)(struct image_params *, int32_t *); /* Machine specific function to dump per-thread information. */ void __elfN(dump_thread)(struct thread *, void *, size_t *);